Hi,

I recently purchased some Binaural headphones and I noticed that deep bass will cause what sounds like a crackling noise during the bass sounds.

I am using an R-09 (http://www.rolandus.com/products/productdetails.php?ProductId=757) and an MM-BSM-8 "Sennheiser" (http://www.microphonemadness.com/products/mmtrunathook.htm).

The crackling really put a damper on one of my last audio captures... I think these were my R-09 settings:

AGC = On
Low Cut = On (it could have been off)
Plug-in power = off
Mic Gain = H (it could have been set to "L")

Any tips as to what could be causing the crackling noise?

Would low cut on/off cause this problem? Maybe a combination of AGC and Low Cut? Maybe the mic gain?

Thanks!
Micky

Without a sample it's hard to say just what is going on but "crackling" on loud bass is very often a sign of clipping somewhere in the recording chain. For a variety of reasons, somewhere something is being driven beyond its maximum capacity.

It appears the mics in your setup need plugin power so that should be turned ON.

AGC leads to "pumping" where the gain is turned up when things are quiet as it looks for sounds that aren't there. When sound suddently appears, like a song starting, the gain is all the way up and has to be clamped down to normal levels again, which takes a few beats. The high gain could mean the initial sounds are recorded so loud it is driven into clipping until the level is pushed back down again. Turn AGC OFF.

Low Cut reduces low bass response ... experiment

It appears you don't know or pay close attention to the setting that is most important in controlling clipping, whixh is the mic gain setting. Pay closer attention and again, experiment. The correct setting depends on the loudness of the sounds you want to record and the sensitivity of the mics you're using.

Finally, you didn't say what your recording format was but FYI, MP3 is a distribution format, not a recording format. If you care about the sound quality, record to uncompressed WAVE files and convert to MP3 later after editing.

Hi Steve! Many thanks for your help, I really appreciate it! :)

Without a sample it's hard to say just what is going on but "crackling" on loud bass is very often a sign of clipping somewhere in the recording chain. For a variety of reasons, somewhere something is being driven beyond its maximum capacity.

Ah, very interesting!

That is a good thing to know, because I was kinda worried that it could have been a physical problem with the microphones and/or cabling.

So, in other words, it sounds like user error vs. a physical problem with equipment.

Sorry that I did not provide a sample... I will post a sample as soon as I get home later today.

It appears the mics in your setup need plugin power so that should be turned ON.

Doh! I forgot to mention that I am using a mini battery pack:

MM-CBM-Mini (http://microphonemadness.com/products/mmcbmminminc.htm)

To be truthfully honest, I don't even know how to open this sucker to replace the 3- 3volt lithium batteries. Hmm, do you think the battery pack could be the problem?

I don't think my audio levels were clipping during record, and they look good during playback on my computer.

I wonder if I need to spend more money on a better battery pack? There are others on this page:

Battery Filter Modules (Stereo) (http://microphonemadness.com/categories/battery_filter_module_stereo.html)

Or, should I ditch the battery pack altogether and use plug-in power? The only reason why I got the battery pack was to save the battery life on my R-09.

AGC leads to "pumping" where the gain is turned up when things are quiet as it looks for sounds that aren't there. When sound suddently appears, like a song starting, the gain is all the way up and has to be clamped down to normal levels again, which takes a few beats. The high gain could mean the initial sounds are recorded so loud it is driven into clipping until the level is pushed back down again. Turn AGC OFF.

Great advice! Thanks for explanation too!

I normally try to shoot/record all manual when working with video/audio, but I was nervous that I would clip and I was not able to ride levels during this last record session (vacation in S.F., and I had the R-09 in my camera bag with headphones on while I took pics). On the other hand, I had a friend tell me to record at a lower level and then, if I need to, boost audio levels in the editing application.

Either way, I am turning AGC off from this point forward.

Low Cut reduces low bass response ... experiment.
It appears you don't know or pay close attention to the setting that is most important in controlling clipping, whixh is the mic gain setting. Pay closer attention and again, experiment. The correct setting depends on the loudness of the sounds you want to record and the sensitivity of the mics you're using.

Thank you for more excellent and pro advice!

I think the reason why I can't remember if low cut and gain were on/off was because I was recording for hours and remember flipping these settings on/off a few times... Unfroturnately, I don't remember when I changed these settings on the R-09. :(

Finally, you didn't say what your recording format was but FYI, MP3 is a distribution format, not a recording format. If you care about the sound quality, record to uncompressed WAVE files and convert to MP3 later after editing.

Doh! Sorry about that. I was recording at the highest quality setting possible. I am amazed at how much an 8gb memory card can hold... It seemed like the R-09 lasted all day. :)

Thanks again Steve, I really appreciate your pro help!

I will post an audio sample later today.

Have a great day!

Cheers,
Micky

Just a couple of note ... haven't used either of your pieces of equipment so my previous post was based on general knowledge, in other words, an educated guess. "Crackle" is often clipping and it can occur even if levels look normal, depending on where it happens. Take a look at the waveform in your computer - if the peaks are flattened off like a plane has been run across them, that's clipping. Damaged hardware is another possibility but I'd look to clipping first. Which headset do you have, the standard version or the deluxe version with the ME-2 mic capsules?

Hi Steve! Sorry for my late reply... Work has been kicking my butt for the last day or so.

Thanks again for your help, I greatly appreciate it! :)

Just a couple of note ... haven't used either of your pieces of equipment so my previous post was based on general knowledge, in other words, an educated guess.

I completely understand. I totally appreciate the guesses you have provided. :)

"Crackle" is often clipping and it can occur even if levels look normal, depending on where it happens. Take a look at the waveform in your computer - if the peaks are flattened off like a plane has been run across them, that's clipping. Damaged hardware is another possibility but I'd look to clipping first. Which headset do you have, the standard version or the deluxe version with the ME-2 mic capsules?

I purchased the deluxe version (i.e. "Sennheiser"). I figured that the deluxe model would give me the best quality sound possible... That is, if I can get my R-09 settings figured out.

Here is a sample audio file (http://sandbox.hulse.me/planetarium.mp3).

And here is a screen grab of a close-up of the waveform (http://sandbox.hulse.me/waveform.png) in question (the red arrow is where the first click happens in the sample I provided).

I hope that helps fill in some of the blanks.

I will have to experiment with my R-09 settings... Part of me wonders if the battery pack I have is to blame, or is this a clear cut case of "auto" settings ruining an audio capture?

I am hoping it is not a hardware problem. :(

If you playback the orginal file on your recorder before importing it into your computer do you still hear the 'crackles.' I'm wondering if the harshness you're describing as 'crackles' might have been in the speaker system of the venue where you recorded and you just didn't notice it at the time. Did you record your original in wav or mp3 file format?

If you playback the orginal file on your recorder before importing it into your computer do you still hear the 'crackles.' I'm wondering if the harshness you're describing as 'crackles' might have been in the speaker system of the venue where you recorded and you just didn't notice it at the time. Did you record your original in wav or mp3 file format?

Ah, good idea!

You know, now that you mention it, and this might be totally stupid to say, but... I do remember noticing that morning that when I swallowed my ears were popping/crackling (normally when I swallow my ears do not pop... Not sure why, but it was very noticeable that morning). Do you think inner-ear pops would be recorded by the mics?

Anyway, I will check the playback off of my R-09 and get back to you (I think I still have the raw/source audio on my memory card).

Also, I recorded the audio in the highest quality possible (wav).

Thanks so much for you help Steve, I really owe you one! ;)

Have a great day!
Cheers,
Micky

Micky, if you're still around, I would be curious to see/hear the original WAV file rather than a file that has been encoded into MP3 format. It doesn't need to be very long, just a few seconds containing the clicks and pops.

Thanks!

Hi Greg!

I will post a few more samples later tonight.

Thanks!
Micky

Hehe, so here is what I got. :)

First, here is the clip with the "pops":

clicks.WAV (http://hulse.me/sandbox/R-09_BSM-8/clicks.WAV)

I was not sure of the best way to save an edit of the original clip other than using QTPro and saving as "source", which saves as .mov extension... I manually changed the extension to .WAV.

Here are some other test clips... Please ignore my dorky voice! :)

#1


Battery pack: Connected
Limiter: off
Low Cut: off
Mic Gain: Low
Plug-in power: Off
Input Level: 50
(http://hulse.me/sandbox/R-09_BSM-8/R09_0001.WAV)

#2


Battery pack: Connected
Limiter: On
Low Cut: off
Mic Gain: Low
Plug-in power: Off
Input Level: 50
(http://hulse.me/sandbox/R-09_BSM-8/R09_0002.WAV)

#3


Battery pack: Connected
Limiter: On
Low Cut: On
Mic Gain: Low
Plug-in power: Off
Input Level: 50
(http://hulse.me/sandbox/R-09_BSM-8/R09_0003.WAV)

#4


Battery pack: Connected
Limiter: On
Low Cut: On
Mic Gain: High
Plug-in power: Off
Input Level: 50
(http://hulse.me/sandbox/R-09_BSM-8/R09_0004.WAV)

#5


Battery pack: No
Limiter: Off
Low Cut: Off
Mic Gain: Low
Plug-in power: On
Input Level: 50
(http://hulse.me/sandbox/R-09_BSM-8/R09_0005.WAV)

#6


Battery pack: No
Limiter: On
Low Cut: Off
Mic Gain: Low
Plug-in power: On
Input Level: 50
(http://hulse.me/sandbox/R-09_BSM-8/R09_0006.WAV)

Surprisingly, I did not get any clicking and/or popping in any of the above 6 tests. Of course, my tests were not scientific and the bass was probably not as deep as the bass was in "clicks.WAV" recording.

Also, I recently upgraded the firmware on my R-09, so I am wondering if that was of some help?

Well, let me know if you would like to hear other samples.

Thanks!
Micky

Hi Micky,

So far I have listened only to the R09_0001.wav which sounds absolutely fine: good levels, no clicks or other extraneous noises.

To try to figure out what went wrong, I really want to listen to the portion of your original file which contains the "clicks." I want to analyze it in a fairly good audio program, something like CoolEdit or Audition. Unfortunately, the file you sent ("clicks.wav") won't open in those programs. I think that's because it's not really a WAV file. If I understand what you did, you saved it in the MOV format, then just renamed the extension. That's not a valid thing to do.

Besides, when a program automatically converts a file like that (from the original WAV file from your R-09, to a MOV file) it could introduce some other problems, as a result of sampling rate conversion, gain change, normalization, etc... maybe some change you didn't intend or even realize was happening.

I really want to look at an *unchanged* portion of the original file. I think the simplest way to do that would be for you to open it in some audio editing program, such as Audition, CoolEdit, Audacity, etc. Then just select (highlight) the portion of the file with the clicks, and use a function with a name like "Save Selection As" to immediately save that portion of the original file, keeping it in WAV format, without any changes to sample rate, gain, EQ, normalization, etc. The key here is to end up with a file in the exact same format as the original problem file, with absolutely no changes! (Of course I could open the entire original file, except it's probably much too big to upload and download conveniently.)

Can you manage to do what I've described above?

Meanwhile, I'll read through the R-09 manual to look for possible trouble spots.

Thanks!

Hi Greg! Thanks so much for the help, I greatly appreciate the professional advice and assistance. :)

So far I have listened only to the R09_0001.wav which sounds absolutely fine: good levels, no clicks or other extraneous noises.

Pheww. I am really happy that I was able to get a decent recording of something with bass... I had my head right next to the speaker and the recording sounded pretty good.

To try to figure out what went wrong, I really want to listen to the portion of your original file which contains the "clicks." I want to analyze it in a fairly good audio program, something like CoolEdit or Audition. Unfortunately, the file you sent ("clicks.wav") won't open in those programs. I think that's because it's not really a WAV file. If I understand what you did, you saved it in the MOV format, then just renamed the extension. That's not a valid thing to do.

Doh! Sorry about that!

Using QT is typically what I do for videos when I need to quickly save as the same format. I ignorantly thought this would work for my R-09's wav files. :(

Besides, when a program automatically converts a file like that (from the original WAV file from your R-09, to a MOV file) it could introduce some other problems, as a result of sampling rate conversion, gain change, normalization, etc... maybe some change you didn't intend or even realize was happening.

That makes total sense now that you point it out! I will be sure to Audacity from now on. Thanks again for the tips and clarification, I really appreciate it. :)

I really want to look at an *unchanged* portion of the original file. I think the simplest way to do that would be for you to open it in some audio editing program, such as Audition, CoolEdit, Audacity, etc. Then just select (highlight) the portion of the file with the clicks, and use a function with a name like "Save Selection As" to immediately save that portion of the original file, keeping it in WAV format, without any changes to sample rate, gain, EQ, normalization, etc. The key here is to end up with a file in the exact same format as the original problem file, with absolutely no changes!

Here is another version. (http://hulse.me/sandbox/R-09_BSM-8/R09_0006.snippet.wav) This time I used Audacity and selected the "Export Selection as WAV" from the file menu. I hope this does the trick! Thanks for the tip on how to do it. :)

(Of course I could open the entire original file, except it's probably much too big to upload and download conveniently.)

I don't mind posting it if you want to hear the full version. It is a little over a gig, but I have the server space if you think that would be best. :)

Meanwhile, I'll read through the R-09 manual to look for possible trouble spots.

Wow! Thanks so much Greg!

I will do the same.

I was kinda wondering if the battery pack was to blame, but my tests yesterday did not yield any negative results when it was being used.

On the other hand, I did upgrade the R-09 firmware recently, and I am wondering if that was the fix I needed.

One last thought... Do you think the mics could pick up inner-ear pops? I know this might sounds silly, but that morning my ears were popping (more than usual) when I swallowed. When I was recording, I remember thinking to myself "I wonder if my ear pops will show up the recording?".

Anyway, thanks so much for your assistance. I owe you one! :)

Have a great day/night.

Cheers,
Micky

I'm still not ruling out that there might have been clipping or speaker crackles from overloading at high volumes in the PA system that you were recording rather than in the recording you yourself made of it. Perhaps you were recording what actually was there. I've been in many venues and movie theatres, etc, where the PA wasn't up to par and some idiot who thought "louder is better" cranked it up beyond what the system could properly handle. You might not notice it at first hearing on-site but in the leisure of listening again it becomes apparent. In your first recording there's so much bass it sounds like whoever setup the PA thought "good means as much bass as you can get" and cranked the amplifier's bass control all the way to 11. I loaded your clip into Soundforge and looked at the waveform. There are spikes at the crackles but they're not showing the flattened peaks indicating clipping or going so high as to threaten clipping, making me wonder if they might have already been in the original source sounds. Of course the waveform I'm seeing may be an artifact of your file conversion processing

Steve, I certainly agree that we can't rule out anything yet. At first glance those little clicks are interesting, in that they are so brief... only 3 or 4 samples in duration. And nowhere near 100% FS.

But then I notice that the latest file has a 96kHz sampling frequency. And according to the Owner's Manual, the R-09 has a maximum sampling frequency of 48kHz. That means that this is *not* a snippet of the original file, but rather a new file that is apparently a resampled version of the original. That resampling could easily obscure whatever went on in the original file.

So, Micky, I am going to ask you to post the entire original WAV file, just as it came from the recorder. We are just wasting time looking at resampled (and otherwise possibly altered) versions of the audio.

[Steve, do you know of any freeware or utility that Micky could use to actually just snip and send a short portion of the file, guaranteed to be *unaltered*? I don't believe he has Audition.]

Micky, when you post again here, please list some of the actual times where those clicks occur in the original file, so we do not need to listen to the entire thing.

Carry on.....

...

So, Micky, I am going to ask you to post the entire original WAV file, just as it came from the recorder. We are just wasting time looking at resampled (and otherwise possibly altered) versions of the audio.

[Steve, do you know of any freeware or utility that Micky could use to actually just snip and send a short portion of the file, guaranteed to be *unaltered*? I don't believe he has Audition.]

...

How about freeware Audacity? Available for both Windows and Mac. All it'll take Miicky to get it up and running is about 10 minutes of his time for the download and install.

Thanks, Steve.

Micky said he did use Audacity the last time around. Since the resulting file was 96kHz, and the R-09 only goes up to 48kHz, Audacity apparently changed the sample rate without Micky's conscious consent to do so.

So my fear is that it may have made some other unknown changes, too... level, normalization, who knows what all? (And once the software has changed the samples, it will be harder for us to see what was actually recorded.)

I'm looking for something essentially foolproof, so that he can send a short snip, without our worrying that the software has made changes in the actual sample values.

Meanwhile, as a fallback, Micky can upload the entire 1GB file... but IMHO a nice *unaltered* snippet would be a lot easier to deal with.

Hi Greg and Steve! I can't thank ya'll enough for the help! I am learning a ton, and I totally appreciate all of the professional help here!

@Steve:

Your theory would definitely explain a lot of things. I wish I could re-visit the Planetarium to find out! I probably did not notice the "pops" because I had the mics in my ears (and I was having some inner-ear popping myself when I swallowed.)

Are any of you near S.F.? The audio recording was made during the "Life: A Cosmic Story (http://www.calacademy.org/academy/exhibits/planetarium/)" exhibition at Morrison Planetarium, California Academy of Sciences. The story line was not the best, but the visuals were pretty amazing. It seemed like they had spent a lot of money producing the sucker... Wouldn't that be a shame if someone fudged the sound coming from the PA system! :D

This would explain why the waveform looks decent.

@Greg:

Doh! I am so sorry that I did not specify the model of the R-09! :(

I was actually using the R-09HR... A slightly newer model than the original R-09.

God, I see now that I accidentally linked to the older model in my first post! :: Slaps self on forehead ::

For all my recordings I use the 24-bit/96kHz settings.

Sorry that I did not specify earlier. I hope that I did not waste any of your time. :(

Thanks a billion for all your pro help Greg and Steve! I can't thank you enough. I definitely owe ya'll one.

Have a great day!

Cheers,
Micky

Micky,

Ouch, I wish you had been more specific, earlier, about your recorder model number. It would have avoided some confusion on my part.

However, I still think your use of Audacity did, in fact, change the file. You state that your original recording was 24bit/96kHz. However, the "snippet" you posted was only 16bit depth!

So please, before we spend any more time on questionable files, post the original, unaltered, file!!! And post a timeline of where we can find some of the clicks (a cluster of at least a dozen clicks, close together, would be really helpful).

No, I do *not* think your binaural mics recorded your ears popping! The anomalies in the file (actually I see abrupt jumps in level) occur exactly simultaneously on both channels. There is no chance in hades that both of your ears popped simultaneously, time after time after time. Forget that "ear popping" concept.

Speaker pops? I'm not quite ready to rule that out, although I the more I look at the file, the less likely I am to name that as the culprit.

Steve, by a strange coincidence, I actually worked as a tech at the "Albert Einstein Spacearium" which is the bureaucratic name of the planetarium in the National Air and Space Museum in Washington, DC, part of the Smithsonian Institution. When I was there, we did have some wretched Altec Lansing speaker cabs, which bottomed out drastically whenever the track contained any substantial amount of LF info... we had to roll off the LF effects to get a clean playback. In fact the track had been mixed at a studio with JBLs which were, of course, much cleaner. (It was embarassing... especially since the playback system in our IMAX theatre was so much better.) But that's not quite what a hear/see in Micky's files... it does not sound like speakers bottoming, and may or may not be in the playback system.

OK, Micky, waiting for a complete original file from you...

Ouch, I wish you had been more specific, earlier, about your recorder model number. It would have avoided some confusion on my part.

Ack! I know. :(

I feel pretty lame about not being more specific. Sorry about that. :(

However, I still think your use of Audacity did, in fact, change the file. You state that your original recording was 24bit/96kHz. However, the "snippet" you posted was only 16bit depth!

I am not starting to feel pretty ADD... I could swear that I recorded that audio in the higher bit setting, but there could be a possibility that I was in 16bit depth mode.

Maybe before I post the full gig+ file, I will check the bit depth via Audacity? I know for sure that I have not altered the original file, and I could have accidentally chose 16bit mode instead of 24 while I was on vacation in S.F. (I might have done this to save space on the card(s)... In hindsight, I probably should have used 24bit mode because I had multiple 8gig memory cards handy).

So please, before we spend any more time on questionable files, post the original, unaltered, file!!! And post a timeline of where we can find some of the clicks (a cluster of at least a dozen clicks, close together, would be really helpful).

Sounds good to me. Sorry about all the mix ups. :(

As soon as I get home from work, I will upload the source file and post a timeline.

No, I do *not* think your binaural mics recorded your ears popping! The anomalies in the file (actually I see abrupt jumps in level) occur exactly simultaneously on both channels. There is no chance in hades that both of your ears popped simultaneously, time after time after time. Forget that "ear popping" concept.

Lol! Thanks for clarifying! :D

I knew that theory sounded pretty silly.

Speaker pops? I'm not quite ready to rule that out, although I the more I look at the file, the less likely I am to name that as the culprit.

Interesting. Like I said, I wish I could just re-record the planetarium again... I do plan on going back down there in a few months... Maybe I will get a chance to record again! :)

OK, Micky, waiting for a complete original file from you...

Thanks so much Greg! I really appreciate the help!!!! Sorry again for all the confusion. :(

I will post back asap after work.

Have a great day!

Cheers,
Micky

Just a thought - but did you use the battery module? The reason for asking is that if the bias voltage is supplied by the recorder, then the dynamic range of the mic is quite low - 105dB SPL (120 db when used with one of our battery modules). With so much bass, it's quite possible that the distortion are the mics itself - the recorder recording the distortion faithfully, which might also explain why the file, without being tinkered with, is at a modest level, but showing these clicks. The waveform, as mentioned hasn't flat topped, but is just spiky and rough - there's no audio peaks that cause it like snares or other percussive stuff - so could be the heavy bass components below 50Hz - which on a spectrum display seem very prominent.

Hi Paul,

Your questions are in line with my suspicions, which is why I want to see the original file.

In post #1, Micky said he had the recorder's "in line" power turned off.

In post #3, he said he was using an external battery module.

With a battery module, the mics themselves probably weren't clipping at the moderate levels one would encounter in a planetarium (lots lower than a live rock band, for example). And anyway, both mics would not clip at exactly the same time in every instance. But if he was running them into a mic input (rather than a line level input, as suggested in the instructions for that mic/module combo) then he might have overloaded the mic preamps. That would show up as some sort of anomaly in the waveform, but might have been much lower than 0dBFS depending on the configuration of the recorder.

It still strikes me as odd, though, because some of the clicks are *not* at program peaks.

For example, at about 28.707 there is a click with a level of about -5dB, but then after the click the level continues to increase to nearly -3dB without any further clipping.

Then at about 28.746 there is a click with a level of about -6dB.

So the clicks do not occur at a consistent level, and the waveform "voltage" sometimes increases after the time when the click occurs. Furthermore, the clicks occur at exactly the same *time* on both channels, even though the levels on the two channels are different. This isn't really what I'd expect if the electronics were clipping... I would expect that to happen at exactly the same *level*.

Yet the clicks occur at exactly the same time on both channels, each and every time. Could it be some sort of malfunction of the AGC circuitry in the recorder? Seems unlikely... the AGC circuitry should have some sort of reasonable attack time and should not generate audible clicks when it operates.

I really want to see the original file!

------------------

UPDATE:

A pro who really knows what he's talking about assures me that if you import a WAV file into QTPro, and trim it to the size you want, then the "export" function can save a WAV format file with the data entirely unchanged from the original file.

I don't use QTPro, so you'll have to look through your menus, settings, options, preferences, etc. and try to find the correct way to do this.

Meanwhile, why don't you tell us the exact length (hrs/mins/secs) of the original file, and the exact file size (Bytes, kBytes, MBytes, or whatever). Then we can do the math and confirm what the sample rate and bit depth were. That will help us verify whether or not your editing procedure is changing bitrate or depth.

ALSO: Are you running Windows or Mac platform?

Just a thought - but did you use the battery module? The reason for asking is that if the bias voltage is supplied by the recorder, then the dynamic range of the mic is quite low - 105dB SPL (120 db when used with one of our battery modules). With so much bass, it's quite possible that the distortion are the mics itself - the recorder recording the distortion faithfully, which might also explain why the file, without being tinkered with, is at a modest level, but showing these clicks. The waveform, as mentioned hasn't flat topped, but is just spiky and rough - there's no audio peaks that cause it like snares or other percussive stuff - so could be the heavy bass components below 50Hz - which on a spectrum display seem very prominent.

Interesting! I did use the battery module and the plug-in power on the R-09 was turned off.

This might be a silly question, but should I be using a battery module or is it better to use the plug-in power of the R-09HR? I assume it is better to use the battery pack (120 dB dynamic range is better than 105dB without a battery pack, no?)

I purchased the MM-CBM-Mini... This was the cheapest battery pack, but it looked like the most simple and compact. Would another style of battery pack help in this situation, or are the mics the limiting factor for when it comes to over-the-top bass?

I would be willing to buy another style battery pack if it would help me get better audio recordings.

Sorry for my lack of technical terms here, I am still learning all the lingo. :)

With a battery module, the mics themselves probably weren't clipping at the moderate levels one would encounter in a planetarium (lots lower than a live rock band, for example). And anyway, both mics would not clip at exactly the same time in every instance. But if he was running them into a mic input (rather than a line level input, as suggested in the instructions for that mic/module combo) then he might have overloaded the mic preamps. That would show up as some sort of anomaly in the waveform, but might have been much lower than 0dBFS depending on the configuration of the recorder.

If I am understanding correctly, that is very interesting!

Are you saying that I might have accidentally had the R-09HR's "plug-in power" turned on? Gosh, that sure would explain a lot of things.

I am pretty positive that I kept the "plug-in power" turned off for my recordings, but I could have accidentally clicked it on. I guess that is why there is the "lock" switch. :)

Would having "plug-in power" and the battery pack on at the same time do any damage to either my mics and/or the recorder?

It still strikes me as odd, though, because some of the clicks are *not* at program peaks.

For example, at about 28.707 there is a click with a level of about -5dB, but then after the click the level continues to increase to nearly -3dB without any further clipping.

Then at about 28.746 there is a click with a level of about -6dB.

So the clicks do not occur at a consistent level, and the waveform "voltage" sometimes increases after the time when the click occurs. Furthermore, the clicks occur at exactly the same *time* on both channels, even though the levels on the two channels are different. This isn't really what I'd expect if the electronics were clipping... I would expect that to happen at exactly the same *level*.

Yet the clicks occur at exactly the same time on both channels, each and every time. Could it be some sort of malfunction of the AGC circuitry in the recorder? Seems unlikely... the AGC circuitry should have some sort of reasonable attack time and should not generate audible clicks when it operates.

I really want to see the original file!

Wow, that is very curious!

Sorry that I did not get the file to you sooner:

R09_0006.WAV (1.71GB) (http://media.hulse.me/audio/temp/R09_0006.WAV)

Here is a spot with clicks:

21:11 - 21:22

Thanks so much for all of the help Greg, you are a life saver! I am learning a ton!

UPDATE:

A pro who really knows what he's talking about assures me that if you import a WAV file into QTPro, and trim it to the size you want, then the "export" function can save a WAV format file with the data entirely unchanged from the original file.

I don't use QTPro, so you'll have to look through your menus, settings, options, preferences, etc. and try to find the correct way to do this.

Meanwhile, why don't you tell us the exact length (hrs/mins/secs) of the original file, and the exact file size (Bytes, kBytes, MBytes, or whatever). Then we can do the math and confirm what the sample rate and bit depth were. That will help us verify whether or not your editing procedure is changing bitrate or depth.

ALSO: Are you running Windows or Mac platform?

I am using a Mac.

Ooh, that's good to know about Quicktime.

Here is an export to WAV from qtpro:

http://sandbox.hulse.me/R-09_BSM-8/new/R09_0006.WAV

Here is a screen shot of the QTPro settings that I was using:

http://sandbox.hulse.me/R-09_BSM-8/new/qtpro-settings.png

Based on the Quicktime clip info:

HH:MM:SS = 00:49:24.18

Size: 1.71GB, or, 1,707,370,028 bytes

Here's a screen shot of the info window:

http://sandbox.hulse.me/R-09_BSM-8/new/qtpro.png

Thanks Greg! Thanks everyone! You folks are AMAZING! :)

Have an excellent day.

Cheers,
Micky

Hi Micky,

Thanks for sending the additional info and files.

It's a lot to tackle, as this is getting deeper and deeper. But it's interesting. I'll at least start to reply now, and will post again later with more info.

1.) An electret condenser mic is a type of transducer that converts audio energy to electrical energy. All mics have two terminals where the signal comes out. Professional mics, with a "balanced" electrical connection, have two signal terminals, plus a separate ground/shield. Consumer electret mics are "unbalanced" and one of the signal terminals is exactly the same terminal as the ground/shield. In addition, there is a "hot" terminal which does two things: it accepts DC voltage ("plug in power") coming from the recorder or from the battery box, and it supplies the audio signal back to the recorder.

Within certain limits, the signal voltage coming out of the mic is directly related to the acoustical energy going into the mic. More acoustical level produces more signal voltage. BUT there are some limits to this transfer function. For one thing, the physical reality of the mic limits the maximum physical excursion of the diaphragm. If, for example, the diaphragm can only move 0.01mm (which will be caused by some specific Sound Pressure Level) then a louder sound cannot move the diaphragm any farther than that. Another limiting factor is the DC voltage level being supplied to the mic (from the recorder or battery box). The mic output voltage cannot go any higher than the DC supply voltage (actually a bit less). So once the SPL reaches that level, the mic cannot produce any more output, even if the sound gets louder.

For most "plug in power" condenser mics, the physical limit of the mic will be higher than the electrical limit (which is related to the DC voltage). Thus, you see specs for your mic that say the maximum level it can reproduce is 105dB SPL when powered by a typical recorder (which probably provides somewhere between 1.5 volts and 3 or 4 volts DC, depending on the model recorder); or the mic can reproduce 120dB SPL when powered by the battery box which provides 9 volts DC. So when you're recording anything that might be fairly loud, you definitely should use the battery box. (The upper limit for most "plug in power" mics is around 9 volts... you probably won't find battery boxes that have a higher voltage than that.)

But now when the SPL gets loud, the voltage coming out of the mic can be pretty high... perhaps approaching 1 volt RMS. (I'm checking the specs for your specific mics.) That is too much level for most mic inputs to handle! That's why the mic's instructions say that you should always use the recorder's line input when using the battery box. And of course the line input does NOT have any "in line power" to worry about.

2.) Unless there's something really odd/bad about your present battery box, I don't think a different one would make the recordings any better.

(Actually there are some purists who say that re-wiring the mic capsules, and using a custom battery box with the positive terminal grounded, rather than the usual negative ground, will provide cleaner audio. That's way far beyond the options that are available to you. Forget I mentioned it.)

3.) I was not really suggesting that you might have had the "plug in power" turned on, at the same time as the battery box. But yet, you might have. (see #4 below) In that case, it's hard to predict what might have happened. The output voltage from the mic might have risen higher than the recorder's "plug in power" voltage. Now, both the battery box and the recorder should have resistors (at least 1k ohms, maybe as high as 10k ohms) between the DC supply and the mic connector. Those resistors should hopefully protect against damage in a situation like this. But even if there was no *damage* per se, perhaps having the recorder's "plug in power" and the battery box both connected at the same time, might cause some unexpected behavior... such as strange clicks. Perhaps we'll never know for certain.

4.) You see by now that this is in the nature of a scientific investigation. It's hard to draw conclusions without knowing all the conditions of the experiment. And you've been very openly uncertain about what you did when you made this recording. (Was the gain high or low? Was "plug in power" on or off? Etc.) Obviously since we don't know these things for certain, we may never know for certain what caused the problem. One lesson here is that you should always be sure how everything is set and connected, and hopefully make a few notes for later reference!

5.) Your file was 1,707,370,028 Bytes / 2964 seconds. That's 576,036 Bytes/second = 4,608,286 bits/second.

(96,000 samples/second) * (24 bits/sample) * (2 channels) = 4,608,000 bits/second

So yes, your original recording was 96kHz/24bits, as you originally recalled.

And that means that your previous "snip" operation did change the bit depth. So I definitely want to look at the original unchanged data, which hopefully you have sent today.

-----

I think that answers most of the hypothetical questions so far. I think this is a good stopping point for this particular post. After I have time to check the new files, I'll post again with some more info.

Greg... One word:

WOW!

THANKS so much for all the pro information!

I am learning a ton. Awesome stuff!

I definitely owe you one. Next time you are in Eugene I will buy you a beer! :)

Thanks!!!!!

Have an excellent day!

Cheers,
Micky :)

Micky,

I have another installment for you. It includes 5 .gif files to illustrate what I'm talking about. I'm trying to find out how I can upload those files to the forum, so you can see them. As soon as I figure that out, I will post the stuff.

Micky,

I have another installment for you. It includes 5 .gif files to illustrate what I'm talking about. I'm trying to find out how I can upload those files to the forum, so you can see them. As soon as I figure that out, I will post the stuff.

Ooooh, awesome! Looking forward to seeing. :)

Last time I tried to post images unfortunately I could only link to images hosted remotely. If you want to e-mail the images to me, I can post them on my web server and link to them... Let me know.

Thanks Greg!!!!!!

Cheers,
Micky

Ah, OK. That was the only option I could find.

The forum's FAQ says that, after I click "Reply" the reply page has a "Manage Attachments" area at the bottom of the page. I don't see it; can't find it. I thought either (1.) I was missing something obvious, or (2.) I hadn't generated enough posts yet to "earn" the right to post files.

I EMailed the administrator about this question, but never got a reply.

I will see about hosting them locally and then try linking to them.

Film at eleven...

Hey Greg!

Ah, OK. That was the only option I could find.
The forum's FAQ says that, after I click "Reply" the reply page has a "Manage Attachments" area at the bottom of the page. I don't see it; can't find it. I thought either (1.) I was missing something obvious, or (2.) I hadn't generated enough posts yet to "earn" the right to post files.
I EMailed the administrator about this question, but never got a reply.
I will see about hosting them locally and then try linking to them.
Film at eleven...

Yah, I remember thinking some of the same things (missing the obvious)... I actually posted a message here:

http://www.dvinfo.net/forum/open-dv-discussion/489796-bbcode-buttons.html

Back then, I was trying to post an image and have it display on these forums (vs. a link). The FAQ says "that you can wrap ... around an image link and have it appear as an image," but all it really does (FWIK) is make the image URL a link. :(

Links become links automatically; no button or BB code needed.
Vimeo clips become embedded in a player just by posting the URL.
If you really feel like making something bold, make the effort and do
the extra keystrokes. We're not making BB code easy or accessible
because we really prefer readable posts.

I can understand why they made these forums very vanilla for when it comes to "readable", but I think there is a good reason to have at least the basic BBCode buttons and file upload features.

Anyway, sorry if totally OT... Let me know if you need any help posting those images. :)

If you are looking for a nice way to share files between computers (and with the public via the www) I recomend Dropbox (http://www.dropbox.com/). They give you a few gigs of space for free. I love using it for sharing things like my actionscript classes, premiere presets/templates, and/or fonts across the different computers that I use.

Well, have a great day!

Cheers,
Micky

Ahhh, you know what, I just noticed that I can upload files! I guess you do have to have XX many posts before you can use this feature. :(

Too bad I didn't see that posted anywhere in the FAQ. Maybe I need new glasses (if it's in there).

Micky, I've posted the image files elsewhere, hopefully I'll be able to link to them, so here goes...


Distortion, of course, can take many forms. A very common form of distortion is clipping, caused when the signal amplitude is greater at some point than the signal chain can handle; the signal chain "runs out of headroom" for that signal.

In your demo file, we could call the "clicks" a type of distortion (although we could also call them a type of noise). We ultimately will try to inspect the waveform of your file visually, in order to determine whether the "clicks" are typical clipping or not. That might give us a clue as to how to avoid this problem in the future.

First let's have a quick demo of clipping.

demo1-1.gif (http://www.centre.ws/DVinfo/demo1-1.gif)
Demo1-1.gif shows a few syllables of speech, without distortion (this is from a file I had handy). The top green waveform is the left channel; the bottom waveform is the right channel. I've marked a few reference levels (on the right channel only): 100% (0dB), and 50% (-6dB). Note, too, that I've highlighted a small portion in the second syllable.

demo1-2.gif (http://www.centre.ws/DVinfo/demo1-2.gif)
Demo1-2.gif is the same file, but now I've zoomed in on the portion that was highlighted. Note that the peaks (the loudest levels in this part of the file) have a rather smooth, rounded shape.

demo1-3.gif (http://www.centre.ws/DVinfo/demo1-3.gif)
Demo1-3.gif shows the same timeframe as the first image. Now, however, I have raised the gain by +6dB, so the signal amplitude is twice as loud. Compare the second syllable here (with the highlighted area) with the second syllable on the original image (Demo1-1.gif). Here, all the peaks stop at a level of 100%, because, by definition, the signal chain can not handle more than 100%.

demo1-4.gif (http://www.centre.ws/DVinfo/demo1-4.gif)
Demo1-4.gif zooms in on the area highlighted in the previous image. This clearly shows an example of clipping. When the waveform reaches 100% (the thin horizontal gray line), it can't go any higher. So, although the peaks originally were rounded (Demo1-2.gif), they now become flat-topped. This is clipping.

demo1-5.gif (http://www.centre.ws/DVinfo/demo1-5.gif)
Demo1-5.gif shows that you can't easily fix a clipped signal. Here I have taken the clipped signal from Demo1-4.gif, and lowered the gain by -6dB. Thus, most of the waveform is back to the original level. But compare the peaks here to the nice rounded peaks in Demo1-2.gif. Now they are flat-topped, even though their peak level is -6dB. That's because they were clipped previously (Demo1-4.gif) at 100%, and when we reduce the gain by -6dB (50%), they stay flat-topped but at a lower level. They are still clipped, and if you have a significant number of clipped peaks like these, the clipping will be quite audible.

In fact there is some software that will try to restore clipped peaks, and it will more or less work in some cases. It would probably work for the small amount of clipping in this demo. But that's beside the point for the sake of this discussion. The point of this demonstration is simply to show what a clipped waveform looks like.

In the next installment we'll take a look at your waveform and compare your audible clicks to this illustration of clipping.

---

PS: Micky, I'm sorry, I cannot get those images to display here. I did put the full URL in the body of my message, and the BBS software converted it to a "url=" format, so you have clickable links on the page. I guess you'll have to click the link, then <alt><tab> back and forth between my text and the relevant image. ... Anyway, I hope this description and visual depiction of clipping provides an introduction for you. As I said, the next installment will deal with the file you sent that has the strange clicks in it. Meanwhile, I will continue trying to figure out how to display images within the body of the message.

Greg....

Awsome!!!!!

Thanks soooo much for all of your help! Very very very informative! I am learning a ton. This is great because I do plan on recording a bunch more with my binaural mics... This will definitely help me out for when it comes to editing (same with audio from my video clips!)

Thanks man! I really owe you one!!!! :)


...<snip>...
demo1-5.gif (http://www.centre.ws/dvinfo/demo1-5.gif)
demo1-5.gif shows that you can't easily fix a clipped signal. Here i have taken the clipped signal from demo1-4.gif, and lowered the gain by -6db. Thus, most of the waveform is back to the original level. But compare the peaks here to the nice rounded peaks in demo1-2.gif. Now they are flat-topped, even though their peak level is -6db. That's because they were clipped previously (demo1-4.gif) at 100%, and when we reduce the gain by -6db (50%), they stay flat-topped but at a lower level. They are still clipped, and if you have a significant number of clipped peaks like these, the clipping will be quite audible.

That's all so interesting!

What program are you using?

Just to clarify, in demo1-5, did you save the clip from demo1-4, and then re-open it? I assume this is the case, I just wanted to make sure this is the case.

In fact there is some software that will try to restore clipped peaks, and it will more or less work in some cases. It would probably work for the small amount of clipping in this demo. But that's beside the point for the sake of this discussion. The point of this demonstration is simply to show what a clipped waveform looks like.

Interesting! Of course, my goal would be to capture the best sounding audio from the source, but that is still good to know that there are softwares that can fix stuff like that. Do you happen to know of a Mac software that does this?

In the next installment we'll take a look at your waveform and compare your audible clicks to this illustration of clipping.

Man, I am totally stoked! This is soo cool. Thanks so much for taking the time to school a noob like me! :)

ps: Micky, i'm sorry, i cannot get those images to display here. I did put the full url in the body of my message, and the bbs software converted it to a "url=" format, so you have clickable links on the page. I guess you'll have to click the link, then <alt><tab> back and forth between my text and the relevant image. ... Anyway, i hope this description and visual depiction of clipping provides an introduction for you. As i said, the next installment will deal with the file you sent that has the strange clicks in it. Meanwhile, i will continue trying to figure out how to display images within the body of the message.

Argh, I know! I wish this forum would at least let us embed images... I think the "url=" format is the best we are gonna get. :(

Actually, I do have a question about what Paul said:


Just a thought - but did you use the battery module? The reason for asking is that if the bias voltage is supplied by the recorder, then the dynamic range of the mic is quite low - 105dB SPL (120 db when used with one of our battery modules). With so much bass, it's quite possible that the distortion are the mics itself - the recorder recording the distortion faithfully, which might also explain why the file, without being tinkered with, is at a modest level, but showing these clicks. The waveform, as mentioned hasn't flat topped, but is just spiky and rough - there's no audio peaks that cause it like snares or other percussive stuff - so could be the heavy bass components below 50Hz - which on a spectrum display seem very prominent.


I hope this does not sound like a silly question, but what does he mean by "bias voltage"? Is that phantom power?

How often in day-to-day life do we encounter bass below 50Hz? Could wind rumble reach such frequencies?

Also, I will have to compare my waveform to your example images... Paul mentions "spiky" and I noticed your non-clipping examples show a smother/rounded waveform at the peaks...

Hrmm, sorry, I am rambling! When I get home from work later today I will have to piece all of this together. This is all very interesting!

Thanks again Greg!!!!!!! Much appreciated!!!!

Have a great day!

Cheers,
Micky :)

Hi Micky,

To answer one of your first questions: Actually I never saved that demo file at all. The original file was something that another member (Syed Junaid) had posted on a different website's forum. (He posted a lot of the same questions on that forum and on this one.) It was speech with a lot of background noise, but the speech was *not* distorted. I happened to have that file handy so that's what I used.

So I opened it, made a few screen captures, applied +6dB of gain, captured again, applied -6dB of gain (in other words, 6dB of gain reduction), and made the final capture. Then I just closed the file without saving it, because I had no reason to save it.

--

I work on the PC platform exclusively. For this demo, I used Cool Edit Pro (v. 2.something). That's no longer on the market. Adobe bought the company (Syntrillium) who wrote Cool Edit, and re-badged it as Adobe Audition. Audition has evolved through a few versions, and has some additional features not found in Cool Edit. You might occasionally find an old version of Cool Edit for sale on eBay, but AFAIK it's PC only. I don't know whether Audition is also available for Mac, or not... it should be easy for you to find out. I don't know anything about any Mac software.

--

"Plug in power" is sometimes called "bias voltage" although that's really inaccurate... bias voltage is actually something entirely different. This system is used with consumer equipment that has unbalanced mic connections. Unbalanced means there are two terminals for a given mono mic, and two conductors in a given mono mic line: hot, and return/ground/shield.

In order for electrical current to flow in a simple circuit (one channel of audio, a flashlight, a table lamp, etc.), you must have two terminals. A loudspeaker is a good example of this, it needs two terminals, and you normally use two-conductor zip cord (lamp cord) to get the signal from your power amp to the speaker.

In theory, you could use the same type of zip cord with a dynamic microphone; but there would be problems. The microphone preamplifier in your recorder (or mixer, etc.) has very high gain, in order to boost the very weak signal coming from the mic. If you used zip cord on your mic input, that wire would act as an antenna and would pick up all sorts of electrical noise... strong hum from your house's power wiring, clicks and pops from appliances turning on and off, maybe strong nearby radio stations, too. Therefore, for a mic input you need to use "shielded" cable... look it up on Wikipedia. The common ground circuit of the preamp is connected to the shield at the mic jack, and the common ground end of the mic is connected to the shield at the other end. So in this case, the shield of the wire also acts as a ground connection, and is also one of the two conductors carrying the mic signal to the preamp. This is called "unbalanced" because one of the two conductors (the "hot" one) is **not** grounded, but the other one (the "return") one **is** grounded. They are not the same, hence the circuit is UNbalanced.

Condensor mics have active electronics in them... at the very least a single FET that acts as a type of preamp (actually an impedance converter). This electronics needs DC voltage to operate. Of course it could come from a battery inside the mic housing, and some mics work that way. But it's also possible to send the DC voltage to the mic, from the preamp, over the mic wiring. Consumer equipment uses the "plug in power" scheme. This is done by connecting a low DC voltage source (usually between 1.5 and 9 volts) from the preamp, through a resistor (usually around 1,000 ohms or a bit more) in the preamp, to the "hot" terminal of the mic wiring. Then that DC voltage uses the "hot" mic wire to get to the mic, where it powers the electronics. Eventually the audio voltage flows back to the preamp using the same "hot" wire. The ground/shield is the return conductor the the audio, and also the return conductor for the DC current.

Professional mics use a different scheme, which is "balanced" wiring. A dynamic element (which generates the mic signal voltage) has two terminals on it. In a professional dynamic mic, each terminal goes to an insulated conductor in the mic cable, and then on to two input terminals on the preamp (pins 2 and 3 of the XLR connector). In addition, there is a shield, which connects the ground of the preamp to the metal body of the mic. But in this case, there is NO connection between the mic element and the shield... there are three entirely separate conductors (two inside, plus the shield). And NO audio current flows through the shield. The braid of the shield protects the two audio conductors (which are inside) from electrical and RF noise in your house, and that's all it does. The two audio conductors are both similar, in that neither one of them is connected to shield or ground at any point. They are the same, hence the audio signal is "balanced."

With a professional condensor mic, a higher DC voltage (sometimes as high as +48 volts) is fed from the mixer, equally on both of the audio conductors (pins 2 and 3), to the mic. The negative side of the DC power supply is connected to the mixer's ground (same as the negative terminal on your car battery is connected to the chassis), and the DC current from the mic's internal electronics returns to the mixer through the shield of the mic cable (pin 1). This is "phantom" powering.

As you see now, "plug in power" and "phantom" are entirely different. The former applies only to UNbalanced mics, the latter applies only to balanced mics. The former uses a fairly low voltage (+9 volts or less) while the latter uses a higher voltage (usually at least +12 volts, and often as high as +48). "Plug in power" and "phantom" are entirely non-compatible. Mis-connecting something could easily damage a mic. So you need to always keep in mind which one you're using. A consumer-style recorder, with a 3.5mm mini jack for the mic connector, would be "plug in power." A professional recorder or mixer would be phantom. DO NOT MIX!!!

--

We encounter very low frequencies all the time in day-to-day life. If you're in a room and someone opens or closes the door, that will produce a positive or negative air pressure peak, and, depending on the room dimensions, there may be a few cycles of decaying resonance. It's below the range of audible frequencies, and you don't need or want to record it. Wave your hand past your ear, and you'll generate a very low frequency pressure wave, but again it's sub-audible and you don't need or want to record it.

I used to work at a radio station that had some awesome RCA ribbon mics. When it was time for the station to sign off at midnight (this was *many* years ago), I would hold a broom out at arm's length and swing it past the mic (with the mic turned on). The low-frequency voltage pulse from the mic made it through the mixing board, through the audio processing, and into the transmitter, where it caused such a huge overload that the transmitter's safety circuits would shut down the transmitter! Sub-audible energy at work!

Music contains a fair amount of energy below 50 Hz, and if you need to roll off it's preferable to do so below 30 Hz. Put on your some headphones (a Sennheiser 280 or better) and listen to a well-recorded classical CD. Listen to the tympani... you can feel them. Listen to the "room tone" between tracks, when nobody is playing their instruments; you will hear and almost feel the low frequency resonance of the room. On the other hand, for dialog recording, you can easily roll off at 50 Hz, or perhaps even higher if you have a lot of background noise to deal with. But by the time you get up to 100 Hz you're starting to affect the voice quality... it will still be completely intelligible, but it will start to become unnatural. Of course with higher-pitched female voices you can move the rolloff up a little higher in frequency.

--

The only waveform that is completely perfectly rounded is a pure sine wave. Such a thing rarely exists in nature. How a waveform looks depends on how far you zoom in. ***SPOILER ALERT*** If you zoom in enough on your own track, it will start to look more and more rounded. The key is that the peaks are not flat-topped. Zoomed out they look "spiky" and zoomed in they look rounded... they are not clipped. If they were clipped there would be a huge number of flat tops, and your track doesn't have that.

Is Micky's track clipped?

What are those clicks?

Tune in again, same time, same station, for another episode of "Cosmic Clicks"!

Greg, this is like a master class in audio. Amazing stuff!!!!

Do you have an Amazon wishlist? I would love to get you a book or something. :)

...<snip>... So I opened it, made a few screen captures, applied +6dB of gain, captured again, applied -6dB of gain (in other words, 6dB of gain reduction), and made the final capture. Then I just closed the file without saving it, because I had no reason to save it.

Ahhh, great! Thanks for the clarification! I was just not sure if you saved before/after applying/removing the +/-6DdB... Thank you for the additional info. :)

I work on the PC platform exclusively. For this demo, I used Cool Edit Pro (v. 2.something). That's no longer on the market. Adobe bought the company (Syntrillium) who wrote Cool Edit, and re-badged it as Adobe Audition. Audition has evolved through a few versions, and has some additional features not found in Cool Edit. You might occasionally find an old version of Cool Edit for sale on eBay, but AFAIK it's PC only. I don't know whether Audition is also available for Mac, or not... it should be easy for you to find out. I don't know anything about any Mac software.

Looks like I could run Audition via an emulated windows environment... I will have to find a demo and give Audition a whirl on one of the PCs at my work. :)

"Plug in power" is sometimes called "bias voltage" although that's really inaccurate... bias voltage is actually something entirely different. This system is used with consumer equipment that has unbalanced mic connections. Unbalanced means there are two terminals for a given mono mic, and two conductors in a given mono mic line: hot, and return/ground/shield. ...<snip>... As you see now, "plug in power" and "phantom" are entirely different. The former applies only to UNbalanced mics, the latter applies only to balanced mics. The former uses a fairly low voltage (+9 volts or less) while the latter uses a higher voltage (usually at least +12 volts, and often as high as +48). "Plug in power" and "phantom" are entirely non-compatible. Mis-connecting something could easily damage a mic. So you need to always keep in mind which one you're using. A consumer-style recorder, with a 3.5mm mini jack for the mic connector, would be "plug in power." A professional recorder or mixer would be phantom. DO NOT MIX!!!

Wow! Like I said, master class!!!!

Tons of excellent information Greg! THANK YOU!!!!

Sorry if I mixed up my lingo. When shooting video, I use an XHA1 (Canon) and, IIRC, I think the manual calls the XLR +48 switch "phantom" power. I just assumed the "plug in power" on my R-09hr was the same thing. Thank you for taking the time to teach me the difference. :)


...<snip>... Is Micky's track clipped?

What are those clicks?

Tune in again, same time, same station, for another episode of "Cosmic Clicks"!

Lol! I am looking forward to the next installation!!!

Let me know if there is a book I can get you off of Amazon or something... If you setup an Amazon wishlist then I could have it shipped to you quick and easy. ;)

Thanks a billion Greg!!!!!

Have an excellent weekend. :)

Cheers,
Micky

Micky,

Thanks for the kind words. I'm glad some of this info is useful to you.

Thank you also for the generous offer of a book. Nothing comes to mind, so don't worry about it. Besides, I'm not doing this with the thought of receiving any sort of "earthly goods." It's reward enough to know that what I'm doing is helpful to someone... in this case, to you. Besides, if I ever get back to Portland, you owe me a beer! ;)

Seriously, I believe in some sort of "pass it along" philosophy... hopefully some day you will be able to spend a few minutes and pass along some knowledge to someone else who needs some help.

Since you mentioned books, it occurs to me that you, yourself, might benefit from a few good books. I can recommend Jay Rose as a very knowledgeable audio guy and author. You can get some information about his books on his website: How-to books about Sound for Digital Filmmakers (http://www.dplay.com/book/) and of course if you back up to his homepage you will find more useful information. I have a copy of his Postproduction book and I think the information is pretty good. (Jay posts frequently on another Digital Video forum, but I haven't seen him on this one.)

Now as to your original question... As you might imagine, it takes a bit of time and concentration to put together a comprehensive post about this stuff. All well and good, I'm not complaining, it's good mental exercise for me. However, the last few days I've been fighting a pretty nasty toothache, and the Percoset has my head spinning... literally. So I don't think I will be composing any lengthy technical comments in the next few days, until the tooth situation is resolved. Sorry to ask you to wait any longer, but please be patient. I've not abandoned the question, I just think I'll do a better job with it when my head is clear.

For now... Happy Trails!

Gentleman,
I will admit that I have not read the whole post, but I listen to your sample and found your ‘clicks’.
To me it looks like hardware, maybe power supply problem.
Take a look at the first picture and see where the click is in that waveform.

21381

Then look at the second one and see if you agree that the waveform should look like the blue line in the second picture.

21382

The whole ‘signal rise’ (the correction) in the first picture (the top waveform) is only over 7 samples @ 96Ksps, hence the click.
The difference over these 7 samples is 68dB.

I don’t have a solution for you, just thought I’ll share my findings. Interesting problem.

@Rob

Yes, that's the pattern that we're seeing. It is not always 7 samples, sometimes it seems to be as short as 3 samples.


The difference over these 7 samples is 68dB

I think you're way off. For example, in one "click" I analyzed, the "pre-click" sample value was 3959127, the "post-click" sample value was 3337707. Dividing those numbers gives a result of 0.843, which corresponds to a change of -1.48dB. If you expand whatever waveform display you're looking at, the vertical (amplitude) scale of the display will confirm that.

Cause still to be determined...

I think you're way off. For example, in one "click" I analyzed, the "pre-click" sample value was 3959127, the "post-click" sample value was 3337707. Dividing those numbers gives a result of 0.843, which corresponds to a change of -1.48dB. If you expand whatever waveform display you're looking at, the vertical (amplitude) scale of the display will confirm that.


21384

Funny thing about dB's, they are very misunderstood creatures.

dB is a ratios of two (one known, the other measured) quantities expressed in logarithmic units.
If you ware referring to full scale (the known value), compared to distorted samples (pre-click & post-click = measured value) then you are absolutely right, the difference is less than 1.5dBFS.
But I didn’t say full scale, I said ‘…the difference over these 7 samples…’ implying same as you pre-click & post-click:

-14919 - -17513 = 2594

Then to get the voltage ratio of the difference: 20 * log(2594) = 68.28dB

I hope this explains how I arrived at that number.
Remember, dB’s are not absolute numbers, they are always in reference to something.
Your numbers are in reference to full scale, my numbers are in reference to each other (pre-click vs. post-click).
Both results are correct.

Sorry, no, both results can't be correct, because 68 does not equal 1.5.

Funny thing about dB's, they are very misunderstood creatures
Funny thing is, I've been using dBs for 40 years, both in terms of circuit voltages, sample values, and in terms of antenna power. And I do understand the proper way to use dB.

When comparing two signals, you either divide the arithmetic sample values, or you subtract the dB values. The two operations, dividing the arithmetic values, or subtracting the dB values, are mathematically equivalent.

You're getting confused because you subtracted the two sample values, rather than dividing them to get their ratio.

As you correctly state,
dB is a ratios of two --<snip>-- quantities expressed in logarithmic units
dB always represents a ratio, which is then converted to a logarithmic scale.

Since your goal is to express how the second sample is related to the first sample, and express this relationship in dB, you need to start out with the ratio of the sample values. To get the ratio, you divide the two values. (Your mistake was that you did not find the ratio between the two values, because you did not divide them... you erroneously subtracted them. Subtracting does not give you a ratio.)

Then, as you state, you convert that ratio to a log, and multiply that log times 20.

The arithmetic sample values are absolute values. 3,959,127 and 3,337,707 are the actual value of those samples (converted from the binary data stored in the file). Full scale doesn't matter. Full scale could be any number bigger than 3,959,127. (Of course the actual value of "full scale" depends on the bit depth of the samples.) Regardless of what full scale is, those sample values are what they are.

In this case, to find the ratio of the two samples, you divide the second value by the first value, and you find that the second value is 84.3% of the first number. The log of 0.843 is -0.074. Multiply that by 20, and the result is -1.48dB. Therefore, the second sample is -1.48 lower than the first sample.

--

Don't take my word for it. (Apparently you do not.) You probably don't have the same books that I have in my reference library, so let's go to a readily available source: Wikipedia. Look up "decibel." Scroll down to the section titled "Field Intensities." (Note that the dB originally was used to measure power ratios, where the formula is 10*log(base10) of P1/P0.) Voltage, or in this case digital samples, is not a measure of power, it's a measure of intensity. So in this case the formula is 20*log(base10) of A1/A0.

Again, you did not divide A1/A0, you subtracted A1-A0. That does not conform to the correct formula, as shown above (from Wikipedia).

--

Let's look at it graphically. We'll use the file you posted. The "pre-click" sample you have circled is at -5.45dB on your software's scale. The "post-click" sample you have circled is at roughly -6.8dB. The second sample is about 1.35dB softer than the first sample. Incidentally, that's fairly close to the two samples that I analyzed... quite close, considering we analyzed two different samples at two different points in the file. In fact, the distortion mechanism here is quite consistent from one click to the next.

--

Another way to think about this: If our two "click" samples were 68dB different, that wouldn't be just a little glitch on the graph, it would be a huge jump. If the first sample was -5.45dB, the second sample would be -73.45dB. That wouldn't be just a slightly audible click, that would be a huge speaker excursion!

--

Look at it mathematically, look at it graphically to confirm it. Either way, the two samples differ by roughly 1.5dB, and that is not the same as 68dB. -68 is an entirely spurious number.

Thank you also for the generous offer of a book. Nothing comes to mind, so don't worry about it. Besides, I'm not doing this with the thought of receiving any sort of "earthly goods." It's reward enough to know that what I'm doing is helpful to someone... in this case, to you. Besides, if I ever get back to Portland, you owe me a beer! ;)

Hehe! Well, let me know if you change your mind. I totally don't mind getting you a treat off of Amazon... It's the least I could do to say thanks for all of your pro help. :)

Beer sounds good though!

Seriously, I believe in some sort of "pass it along" philosophy... hopefully some day you will be able to spend a few minutes and pass along some knowledge to someone else who needs some help.

That's a great philosophy! I believe in the same thing (I try to be as helpful as I can in the areas that have more experience in). It's true what they say: What goes around comes around. :)

Since you mentioned books, it occurs to me that you, yourself, might benefit from a few good books. I can recommend Jay Rose as a very knowledgeable audio guy and author. You can get some information about his books on his website: How-to books about Sound for Digital Filmmakers (http://www.dplay.com/book/) and of course if you back up to his homepage you will find more useful information. I have a copy of his Postproduction book and I think the information is pretty good. (Jay posts frequently on another Digital Video forum, but I haven't seen him on this one.)

Awsome! Ordering both of Jay's books now. Thanks for linkage. :)

Now as to your original question... As you might imagine, it takes a bit of time and concentration to put together a comprehensive post about this stuff. All well and good, I'm not complaining, it's good mental exercise for me. However, the last few days I've been fighting a pretty nasty toothache, and the Percoset has my head spinning... literally. So I don't think I will be composing any lengthy technical comments in the next few days, until the tooth situation is resolved. Sorry to ask you to wait any longer, but please be patient. I've not abandoned the question, I just think I'll do a better job with it when my head is clear.

No worries! I completely understand. No rush at all. You could stop now and I would still feel like repaying you with an Amazon treat! :D

For now... Happy Trails!

Ditto! Thanks Greg!!!!

Have an awsome day!

Cheers,
Micky

Hi Robert! Many thanks for the help! I really appreciate it. :)

First, how the heck did you get embeded images inline with your your post!!! I see you are using "attach" and then the ID of the attachment. Pretty cool! I will have to play around with this the next time I want to post images inline with my comments. :)

Gentleman,
I will admit that I have not read the whole post, but I listen to your sample and found your ‘clicks’.
To me it looks like hardware, maybe power supply problem.
Take a look at the first picture and see where the click is in that waveform.

Ugh, I would hate for it to be hardware related. :(

I can't really afford to re-buy a recorder and/or binaural mics.

Do you think it could be the battery pack?

In the last tests I did at home I did not experience any cracks/pops, but then again the bass was not as deep, although I did have my head right up next to the speaker (and it was turned up loud... My wife kinda got pissed at me that day). :D

I don’t have a solution for you, just thought I’ll share my findings. Interesting problem.

Thanks so much for helping out!

I think the next time I am down in S.F. (should be a couple months) I will go back to the planetarium to re-record; I will be sure to note all of my settings this time.

Thanks Robert!!! Much appreciated! :)

Have a great day!

Cheers,
Micky

..

I think the next time I am down in S.F. (should be a couple months) I will go back to the planetarium to re-record; I will be sure to note all of my settings this time.
..Micky

Why re-record? The clicks are not all that objectionable if you're simply listening for your own enjoyment and since I'm sure the program is copyright, personal listening is all you'll ever be able to do with it legally.

Why re-record? The clicks are not all that objectionable if you're simply listening for your own enjoyment and since I'm sure the program is copyright, personal listening is all you'll ever be able to do with it legally.

Good points! :)

I personally don't want to see it again, but it's the only situation that I can think of with bass that deep.

Also, I was thinking that it would be nice to compare the same audio, but with the second one I would have noted all of my recorder settings. My main concern is avoiding pops and clicks for future recordings that are NOT in a planetarium! :D

Last time I was in California, I spent every day photographing, gps logging and audio recording (almost) every day of my vacation. This may sound cheezy, but I was doing it in the name of "art" and "journalism" (i.e. documenting my vacation). I ended up with a ton of cool audio (the Dim Sum and F-line Streetcar recordings were some of the most interesting audio captures).

Unfortunately, the only time I noticed the clicks was during the planetarium event, otherwise I would have chosen another audio clip to share on these forums. :)

Thanks for the reply! Much appreciated!

Cheers,
Micky

...Last time I was in California, I spent every day photographing, gps logging and audio recording (almost) every day of my vacation. This may sound cheezy, but I was doing it in the name of "art" and "journalism" (i.e. documenting my vacation). I ended up with a ton of cool audio (the Dim Sum and F-line Streetcar recordings were some of the most interesting audio captures).

...

I lived many years in San Francisco in the days when the green and yellow PCC cars were the backbone of the Muni trolly lines, worked right at the corner of Market and Van Ness were a number of the inbound cars reversed on the wye at Market and 11th. I lived for a while on Church Street with the streetcars running right past my front door. Their sounds were a constant accompaniment to my day, both at home and at work, and when I hear them now it makes me very nostaligic. The same with the sound of the foghorns (shut down back in the 80's) echoing in from the bay late on a damp and foggy night.

I lived many years in San Francisco in the days when the green and yellow PCC cars were the backbone of the Muni trolly lines, worked right at the corner of Market and Van Ness were a number of the inbound cars reversed on the wye at Market and 11th. I lived for a while on Church Street with the streetcars running right past my front door. Their sounds were a constant accompaniment to my day, both at home and at work, and when I hear them now it makes me very nostaligic. The same with the sound of the foghorns (shut down back in the 80's) echoing in from the bay late on a damp and foggy night.

Oh, man, that sounds awesome!!!! I am jealous! I would love to have the experience of living there.

The street cars are so cool! Last time I was there I got to ride on one of the oldschool trolly cars you mention (I think it was called the F-Line along the Embarcadaro (http://www.sfmta.com/cms/mfleet/histcars.php)).

San Francisco is an amazing place to visit! So much history and diversity. I love all the old buildings/sights/sounds. :)

I would love to capture audio of the fog horns! That would make for some awesome binaural audio. :)

I wish my hometown had interesting stuff like that to record... Oh, well, I guess the grass is always greener on the other side. :(

Thanks again for the reply! Have a great day.

Cheers
Micky

I was thinking that it would be nice to compare the same audio, but with the second one I would have noted all of my recorder settings. My main concern is avoiding pops and clicks for future recordings that are NOT in a planetarium!

Good idea, Micky. We'll try to figure out this "clicking" problem, or at least come up with some good theories. Then if you can go back and record exactly the same source material, using different (and documented) procedures, that should be very useful to you, in terms of confirming our theories and hopefully avoiding the same problem in the future.

@Steve:

Why did they shut down the foghorns? Availability of GPS and RADAR? I'm sure the fog is still there.

@Greg

…Sorry, no, both results can't be correct, because 68 does not equal 1.5….

Well, the way you wrote it, it does not. But if you write it this way “1.5dBFS = 68dB” then they do.

A term dB means nothing without including the reference.
The only exception is if your reference is 1. Then you don’t need to express it, it is assumed (it would look silly if you wrote +6dB1).
How then two or more different dB values can be equal?
The answer is: If the reference is different.
Let me give you example. Let say, I would like to express voltage in dBs.
To make it simple, let pick 5 volts:


When my reference is 1mV (0.001V), then the value becomes +73.98dBmV.
When my reference is 1V, then the value becomes +13.98dBV.
When my reference is 5V, then the value becomes 0dB5V.
When my reference is 1000V (1KV), then the value becomes -46.02dBKV.
When my reference is 1,000,000V (1MV), then the value becomes -106.02dBMV.


You see, they all represent the same 5V.
Therefore it is safe to say that: +73.98dBmV = +13.98dBV = 0dB5V = -46.02dBKV = -106.02dBMV.
On the other hand, it would make no sense whatsoever to say that: +73.98dB = +13.98dB = 0dB = -46.02dB = -106.02dB because we don’t know what the references are and if we assume that the reference is 1 then mathematically it is false.

…Funny thing is, I've been using dBs for 40 years, both in terms of circuit voltages, sample values, and in terms of antenna power. And I do understand the proper way to use dB. …

I only have 15 years of experience. It makes no difference…

…When comparing two signals, you either divide the arithmetic sample values, or you subtract the dB values. The two operations, dividing the arithmetic values, or subtracting the dB values, are mathematically equivalent.

You're getting confused because you subtracted the two sample values, rather than dividing them to get their ratio. …

I subtracted to get the difference, the amount of error and I didn’t divide it because my reference was 1.
So if I divide 2594 by 1 it still is 2594, a simple reduction in number of calculations necessary to get the same result.

…The arithmetic sample values are absolute values. 3,959,127 and 3,337,707 are the actual value of those samples (converted from the binary data stored in the file). …

Where do you get those numbers from? The file I looked at is a 16bit file. Maximum values for a 16bit variable is 2^16 = 65535 or ±32767 (if sign is used).

…Full scale doesn't matter. Full scale could be any number bigger than 3,959,127. (Of course the actual value of "full scale" depends on the bit depth of the samples.) Regardless of what full scale is, those sample values are what they are. …

If you are reading the numbers on the right side of the Audition (or Cool Edit) screen, then I’m sorry to say, but all these numbers are in reference to Full Scale. Notice that 0dB is on the outer edge of the scale and the numbers become smaller (negative) closer you get to the center. A sine wave tone signal of 0dBFS amplitude (assuming 16bits) would have sample values spanning the full range ±32767, the same 0dBFS signal in 24bit file would have range ±8388352. So to say “…Full scale doesn't matter. …” is nonsense, full scale is the measuring stick for the scale you are reading.

Here is a simple example you can try: Open your Cool Edit and create a 16bit file. Let the program generate a 440Hz tone at -18dBFS (duration and sampling makes no difference). Zoom in on top of one of the waveforms and right click on the top most sample. In a popup window, you will have a sample value. If you selected the top most sample and you are in the 16bit range, the value should be 4125. Now select 24bit range, the value of the same sample will be 1056000. So how can you have -18dBFS signal with two different values? Again, because of the reference, your measuring stick is different and that’s what dB is all about.

…Look at it mathematically, look at it graphically to confirm it. Either way, the two samples differ by roughly 1.5dB, and that is not the same as 68dB. -68 is an entirely spurious number. …

I have already showed you how I arrived at 68dB. Again, 1.5dB does not equal 68dB, but 1.5dBFS does equals 68dB because the 68dB represents the sample error of 2594 units expressed logarithmically with reference as 1 (hence no reference indicated), therefore the number is independent of scale. Your 1.5dBFS is in reference to full scale (the actual error amount will become a different number depending on the number of bits used as a full scale).

If you wish to continue this, I suggest you create a new thread so others will be able to comment on this subject. Right now we are hijacking this thread.

Let me give you example. Let say, I would like to express voltage in dBs.
To make it simple, let pick 5 volts:


When my reference is 1mV (0.001V), then the value becomes +79.98dBmV.



Actually, Robert, even those calculations are wrong. You want to express 5V, relative to a reference of 1mV, in dB.

The correct calculations are as follows: 5.000V/0.001V = 5000. The log of 5000 is 3.699. To get dB, multiply the log by 20: 3.699 x 20 = 73.98dB. Not 79.98 as you stated.

--

However, your other values are correct.

The key is that you are performing the correct operation here. You are dividing the two values, to optain a ratio; then finding the log of that ratio, then multiplying the log by 20.

But when working with Micky's clicks you did not divide the two values; you subtracted them. That's where you went astray.

--


To answer your question, the sample values I used were from Micky's original 24-bit file. If you compare "pre-click" to "post-click" in his 24-bit original, or compare "pre-click" to "post-click" in the 16-bit version, either way, the amplitude ratio between a given "post-click" sample and a given "pre-click" sample is usually in the range of 90% to 84%, and when you convert that to dB that's 0.93dB to 1.5dB. These are irrespective of the number of bits, and irrespective of the full-scale value. These are just the ratio between those two samples at any given "click" event.

I would refer you (or anyone interested) to such well-established reference sources as The Audio Cyclopedia (Tremaine) or Reference Data for Radio Engineers (Howard Sams Co., International Telephone & Telegraph Corp.) or even something simple like "Handbook of Electronic Tables & Formulas" (also Howard Sams Co.). For example, the values you give for the two samples you used are -14719 and -17513. Those sample values represent amplitudes. dB represents a ratio of amplitudes, and the ratio of your sample values is found by dividing -14719/-17513; the result is roughly 84%. (Note that I am not using the full scale value, or any other value in my calculations... just the values of the two samples.) Just use the table in the last reference above, and you'll find that 84% is quite close to -1.5dB. That number (and only that number) of decibels represents the ratio of the amplitude of those two samples (to each other). It really is that simple.

You suggest starting a new thread, but I really don't think there is any point to continuing this discussion. I've cited my references and explained my methodology. As you yourself stated, dB represents a ratio and a ratio is found by division. Your methodology used subtraction which does not find a ratio.

I think there's a real risk in continuing this discussion, because the more time you spend explaining your calculations, the more risk there is that you will confuse people who don't have a thorough understanding of dB calculations.

Anyone who understands the correct way to use dB will know, and can confirm visually or from my given references, that my explanation corresponds to the published procedure and calculations.

--

Indeed, this is a very long tangent to the original question. When I get into the next step -- the analysis and repair of Micky's clicks -- you'll see why these numbers I've found -- from 0.93dB to 1.5dB -- are an important key to the puzzle. The clicks can be fixed using those dB values that I calculated. (As we'll see, the number is slightly different for each click in the file -- but nowhere near 68dB).

Meanwhile, yes, I think if you repeat your explanation (of how to use amplitude subtraction, rather than amplitude division, to arrive at an erroneous dB value), that will, indeed, just hijack the thread off in another direction.

OK, Micky, thank you for been very patient.

Today we are going to analyze and repair the clicks, so we can try to understand the original cause of the problem.

You may recall, long ago, that you uploaded a file called "R09_0006.snippet.wav" which was 16-bit. I then asked you to upload the original 24-bit file. After analyzing some clicks in those two files, I've concluded that we can use the "snippet" file for this demonstration. That will confine us to a smaller file and make it easier to refer to times within the file; also, the sample values will be more manageable. So we are going to work with "R09_0006.snippet.wav" in the following exercise.

(To make things easy for myself, I will use my online PC, which has CoolEdit Pro v.2.1. I'm going to ask you to flip back and forth between this text and a number of screen captures. I'm sorry if this will be inconvenient. It would be simpler if I could just project my computer on a screen and explain this in person; but that's not going to happen. When we get to the end, I hope you'll feel this was worthwhile.)

When I played the file, I heard numerous clicks. I decided to pick a group of eight clicks for the experiment. Here CUE LIST (http://www.centre.ws/DVinfo/ClickFix-01.gif) is a list of the times of the clicks I isolated.

Let's take a look at the first of those clicks. CUE 1 RAW (http://www.centre.ws/DVinfo/ClickFix-02.gif) The first thing we need to notice is that this click occurs when the waveform is below the axis of zero amplitude (marked -infinity dB) which is white on my display. In fact, all the clicks we find will occur when the waveform is below the zero axis. Therefore the numeric values of all the samples we're going to analyze are negative.

Now let's zoom in a bit, both timewise (the horizontal scale) and amplitude-wise (the vertical scale).
CUE 1 RAW - ZOOMED IN (http://www.centre.ws/DVinfo/ClickFix-03.gif) Although the two channels are somewhat different, we can describe them both in the same way.

• Before the click, we see a series of negative sample values, and the waveform is becoming increasingly negative with time. (In other words, the samples are becoming farther and farther below the zero axis with each successive sample in time.)

• Next, suddenly, we find a group of five samples which are going steeply in the opposite direction: each successive sample is less negative. (In other words, the samples are becoming closer and closer to the zero axis with each successive sample in time.)

• Finally, we see another series of samples which are becoming increasingly negative with time... although note that the left channel soon reverses and starts trending upward on the graph (decreasingly negative) pretty soon.

Before we repair this click, we need to make an assumption. We will assume that the samples before the click are the correct values, and that the samples after the click are the wrong value. (Either way, the five samples in the middle -- the ones with the reverse slope to their waveform -- are obviously incorrect values.)

Having made that assumption, we need to choose between two options. If we eventually need to sync this audio to some video, then we should keep every single sample in the file. In that case, we will have to correct the five samples with the "reverse" slope, as well as the samples after the click. But, if we do not need to sync this later, it will be half as much work if we just discard those five bad samples; then we will need to repair only the samples after the click. Since this is an "audio only" exercise, we're going to choose the second option.

OK, so here CUE 1 RAW - SNIPPED (http://www.centre.ws/DVinfo/ClickFix-04.gif) is the file again, after we discard the five intermediate samples. I suspect that the part of the file to the left of the "Cue 1" line would fit nicely together with the part to the right of the "Cue 1" line, if only we could shift them somehow. The values to the left of the Cue line are more negative than the values to the right of the Cue line. That means the absolute amplitude to the left of the Cue line is greater than the amplitude to the right of the cue line. (Remember, zero amplitude -- a sample value of 0 -- is up above what we can see on the graph, because we've zoomed in. And 0dBFS - 100% full scale, the largest possible negative sample value, is down below what we can see.) So I'm going to start with the premise that the gain of the recorded signal was somehow reduced at the point where the click occurred.

To re-align these waveforms, I'm going to increase the gain to the right of the Cue line. Of course I don't want to raise the amplitude of the entire file after this point, so I will immediately raise the gain, then gradually lower it back down to unity (the original level). In other words, this will be an immediate gain increase, with a fade down to unity gain (0dB).

To figure out how much we need to increase the gain, we need to find the level difference at the "Cue 1" line, where the abrupt level change occurs in the file. The first step is to find the level of the two samples in the left channel, and then find the level of the two samples in the right channel.

CUE 1 RAW - VALUES (http://www.centre.ws/DVinfo/ClickFix-05.gif) The first left channel sample is -4483328; the second left channel sample is -3819264. We need to find, in dB, how the second sample differs from the first sample. Then we can increase the gain by that amount. To calculate this, we first divide -3819264/-4483328, which gives a ratio of 0.852. (In other words, the second left channel sample is 85.2% as large as the first left channel sample.) Next we take the log of 0.852 (easily done on a scientific calculator) and we find that it's -0.0696. Finally, we multiply that log by 20, and our result is -1.39dB. So the second left channel sample is -1.39dB lower than the first left channel sample.

Now we'll repeat the procedure with the right channel. The first right channel sample is -3933440; the second right channel sample is -3504896. The ratio is -3504896/-3933440 = 0.891. The log of 0.891 is -0.050. Multiply that log by 20, and the result is -1.00dB. So the second right channel sample is -1.00 lower than the first right channel sample.

Now we'll set up our gain adjustment. The left channel gain will initially be +1.39dB. The right channel gain will initially be +1.00dB. Both channels will return to 0dB. Let's make the total length of the adjustment roughly 0.1 seconds. Here CUE 1 RAW - REPAIR REGION (http://www.centre.ws/DVinfo/ClickFix-06.gif) is the area we are going to adjust.

Here CUE 1 - GAIN SETTINGS (http://www.centre.ws/DVinfo/ClickFix-07.gif) is the setup of the gain adjustment window.

Now let's zoom back in on the waveforms after the adjustment. CUE 1 - REPAIRED (http://www.centre.ws/DVinfo/ClickFix-08.gif) We see that the alignment is fairly good, in general, which tells us that our gain figures were correct. We notice a little bit of roughness is the waveform, for a few samples before and after the Cue line. That represents some high frequency energy just before and after the Cue line. That tells us that the malfunction that caused the click actually started slightly before the big level jump, and ended slightly after the level jump.

When I play over this point in the file, the click seems to be gone! Since this was successful, let's continue with our exercise.

Let's move on to Cue 2. Here CUE 2 RAW (http://www.centre.ws/DVinfo/ClickFix-09.gif) it appears there are only three spurious samples between the two parts of the waveform that we want to join. So we'll remove just three samples this time.

Here's Cue 2 after removing three samples CUE 2 RAW - SNIPPED (http://www.centre.ws/DVinfo/ClickFix-10.gif). The first left channel sample = -4785152. The second left channel sample = -4310016. The ratio of the second sample, in relation to the first sample, is 0.901. The log of that ratio is -0.045. Multiply that by 20, and we find that the second sample is -0.91dB lower than the first sample.

The first right channel sample = -4096000. The second right channel sample = -3668736. The ratio of the second sample, in relation to the first sample, is 0.896. The log of that ratio is -0.048. Multiply that by 20, and we find that the second sample is -0.96dB lower than the first sample.

Again, we'll set up a gain adjustment. The initial left channel gain will be +0.91dB. The initial right channel gain will be +0.96dB. Both gains will fade down to unity (0dB).

When we zoom out CUE 2 - ADJUSTMENT (http://www.centre.ws/DVinfo/ClickFix-11.gif) to set up the gain correction this time, we find that we cannot use a fade time of 0.1 seconds this time, because the next click (Cue 3) occurs before that... in fact in about 0.039 seconds. So this time, we'll set up our fadetime to be 0.035 seconds long.

When we zoom back in on Cue 2 after repairs, CUE 2 - FIXED (http://www.centre.ws/DVinfo/ClickFix-12.gif) we see similar results to those at Cue 1. The overall curve is pretty well aligned, so again our gain values were correct. But, again, there is some high frequency roughness in the curve, indicating that the click started to occur just a bit before the large voltage transition.

Let's move on to Cue 3. CUE 3 RAW (http://www.centre.ws/DVinfo/ClickFix-13.gif) Again, it looks as if there are just three spurious samples, between the parts of the waveform that we plan to re-align.

Here CUE 3 RAW - SNIPPED (http://www.centre.ws/DVinfo/ClickFix-14.gif) is Cue 3 after removing the spurious samples. The first left sample = -4592640. The second left sample = -4061440. The ratio of the second sample, in relation to the first sample, is 0.884. The log of that ratio is -0.053. Multiply that log by 20, and you find the second sample is -1.07dB lower than the first sample.

The first right sample = -4250880. The second right sample = -3771136. The ratio of the second sample, in relation to the first sample, is 0.887. The log of that ratio is -0.052. Multiply that log by 20, and you find that the second sample is -1.04dB lower than the first sample.

Incidentally, here's a good point to stop and visually check our calculations. Let's look at the right channel. Using the vertical scale at the right side of the display, you see that the first sample is -5.91dB and the second sample is -6.96dB. To compare levels expressed in dB, just subtract... but be careful of the signs! (-6.96dB) - (-5.91dB) = -1.04dB. Visually we can see that the second right channel sample is -1.04dB lower than the first right channel sample. That confirms our calculations above.

OK, let's set up a gain adjustment. Initial left channel gain is +1.07dB. Initial right channel gain is +1.04dB. Again, both gains fade down to unity (0dB). We can set a duration of 0.1 seconds. CUE 3 - ADJUSTMENT (http://www.centre.ws/DVinfo/ClickFix-15.gif)

Here CUE 3 - FIXED (http://www.centre.ws/DVinfo/ClickFix-16.gif) is the waveform after the adjustment. Again, alignment is good; gain is correct. Again, there's some high frequency roughness because the click actually started a bit before the big voltage transition.

Again, when I play the file, starting before Cue 1, and ending before Cue 4, the clicks are fixed!

OK, I am going to continue with this process, but I won't write out all the gruesome details here for the rest of the clicks. I'll just compile a list for the sake of reference.


CUE# L #1 L #2 Ratio Log dB R #1 R #2 Ratio Log dB
4 -5022976 -4331264 0.862 -0.064 -1.29 -4666880 -4084992 0.875 -0.058 -1.16
5 -4453888 -3833856 0.861 -0.065 -1.30 -4188672 -3568640 0.852 -0.070 -1.39
6 -4382208 -3905536 0.891 -0.500 -1.00 -4184576 -3788032 0.905 -0.043 -0.86
7 -4158208 -3494912 0.840 -0.075 -1.51 -3959040 -3337728 0.843 -0.074 -1.48

If you want to see them, here are the images, after the spurious samples are removed, but before the gain is adjusted.
CUE 4 RAW - SNIPPED (http://www.centre.ws/DVinfo/ClickFix-17.gif)
CUE 5 RAW - SNIPPED (http://www.centre.ws/DVinfo/ClickFix-18.gif)
CUE 6 RAW - SNIPPED (http://www.centre.ws/DVinfo/ClickFix-19.gif)
CUE 7 RAW - SNIPPED (http://www.centre.ws/DVinfo/ClickFix-20.gif)

Cue #8 CUE 8 (http://www.centre.ws/DVinfo/ClickFix-21.gif) is different! I hear a glitch, but I don't see the same abrupt voltage jump. So we can't fix this one by the technique we've used so far. We can see some high-frequency disturbance in the waveform, and there are certainly ways to clean this up, but we will intentionally pass over this one because it doesn't help us with the pattern that we're trying to analyze.

OK, let's play the repaired file from Cue 1 to Cue 8. AUDIO FILE - CUES 1-7 REPAIRED (http://www.centre.ws/DVinfo/R09_0006-snippet-demo-fix-part1.wav) (I've uploaded just the first 37 seconds of the original.) Listen immediately after the pair of cues #5 and #6. It sounds to me like the overall audio level drops there, and then slowly comes back up. Assuming that the planetarium's audio track doesn't have this issue, I suspect that there's some sort of AGC or peak limiting taking place in your recorder. The fact that the audio drop seems to coincide with the pair of clicks leads me to wonder whether (1.) you had the AGC or limiter turned on, and (2.) whether the clicks are a manifestation of a badly designed AGC or limiter in the recorder.

(To be continued... hopefully later this weekend)

OMG...

WOW!!!!

Man, Greg, I know I have said it before... You ROCK!!!!!!!!

I apologize for not replying sooner. I greatly appreciate all of the amazing and professional help you have given me.

After analyzing some clicks in those two files, I've concluded that we can use the "snippet" file for this demonstration. That will confine us to a smaller file and make it easier to refer to times within the file; also, the sample values will be more manageable. So we are going to work with "R09_0006.snippet.wav" in the following exercise.

Ahh! That's great! I was wondering if that export would be an exact copy of the original WAV. I have taken down the full length file for now... Let me know if you need me to upload it again. :)

(To make things easy for myself, I will use my online PC, which has CoolEdit Pro v.2.1. I'm going to ask you to flip back and forth between this text and a number of screen captures. I'm sorry if this will be inconvenient. It would be simpler if I could just project my computer on a screen and explain this in person; but that's not going to happen. When we get to the end, I hope you'll feel this was worthwhile.)

No problem at all!!! I totally appreciate you doing all of this! I am learning a ton of new stuff! Much appreciated. :)

...<snip>...Here CUE LIST (http://www.centre.ws/DVinfo/ClickFix-01.gif) is a list of the times of the clicks I isolated. Let's take a look at the first of those clicks. CUE 1 RAW (http://www.centre.ws/DVinfo/ClickFix-02.gif) ...<snip>... Now let's zoom in a bit, both timewise (the horizontal scale) and amplitude-wise (the vertical scale). CUE 1 RAW - ZOOMED IN (http://www.centre.ws/DVinfo/ClickFix-03.gif) ...<snip>... Before we repair this click, we need to make an assumption. We will assume that the samples before the click are the correct values, and that the samples after the click are the wrong value. (Either way, the five samples in the middle -- the ones with the reverse slope to their waveform -- are obviously incorrect values.)


Man, that's interesting! I never imagined that my audio could be analyzed to this level of detail. Your observations are very interesting and revealing.

Having made that assumption, we need to choose between two options. If we eventually need to sync this audio to some video, then we should keep every single sample in the file. In that case, we will have to correct the five samples with the "reverse" slope, as well as the samples after the click. But, if we do not need to sync this later, it will be half as much work if we just discard those five bad samples; then we will need to repair only the samples after the click. Since this is an "audio only" exercise, we're going to choose the second option. ...<snip>... OK, let's play the repaired file from Cue 1 to Cue 8. AUDIO FILE - CUES 1-7 REPAIRED (http://www.centre.ws/DVinfo/R09_0006-snippet-demo-fix-part1.wav) (I've uploaded just the first 37 seconds of the original.) Listen immediately after the pair of cues #5 and #6. It sounds to me like the overall audio level drops there, and then slowly comes back up. Assuming that the planetarium's audio track doesn't have this issue, I suspect that there's some sort of AGC or peak limiting taking place in your recorder. The fact that the audio drop seems to coincide with the pair of clicks leads me to wonder whether (1.) you had the AGC or limiter turned on, and (2.) whether the clicks are a manifestation of a badly designed AGC or limiter in the recorder.

Sheesh, you are a genius!!!!

I am betting that I had both AGC and low cut switched on... That part of the audio you mention after cues #5 and #6 definitely do dip... I can almost say for certain now that I at least had AGC turned on.

I can't believe that you were able to fix those clicks! I am looking forward to hearing your end result, that is if you have further time to work on this. I completely understand if you are busy... Definitely let me know if you want me to get you something off of your Amazon wishlist!!!!

Speaking of Amazon, I purchased those books you recommended. I am looking forward to reading through them.

Sorry again that I did not reply sooner. I can't tell you how much I appreciate all of your help. :)

Have an excellent day!!!!

Cheers,
Micky