Wireless audio - noisey!

[Michael Clark]

I filmed a wedding Saturday. I have three cameras, 1 Marantz pmd660 recorder connected to the church's sound system, one Zoom H2 recorder near the pastor's podium, and one Sennheiser 500 G3 wireless system. It was my first time using the wireless. I put the wireless system in auto scan mode, and it picked up a frequency right away (if I'm using the wrong terms please forgive me). The great thing about this system is that it has level monitors on both the transmitter and receiver. Plus Peak and RF lights which were a huge plus. I came home, not having heard the audio yet, feeling really good about how it was probably going to sound. Unfortunately when I put it on the computer, it came out sounding like this (http://clarkvideoproductions.com/bad_audio.mp3)

The church sound guy messed up my mp3 recorder plugged into their sound board - it only ended up recording the music tracks from a CD player he had plugged in. It didn't record any of the church sound board's (pastors mic, special singing, etc.). So that back-up audio source is shot. I know there isn't a magic bullet when it comes to fixing bad audio, but is there anything that can take some of the hiss and/or waviness out of it? I have Soundbooth CS5 but have not used it much. ANY help would be appreciated. Also, does anyone have any insight as to what may have happened? The sound guy (who apparently knew as little as I did!) promised they were on different channels and would not interfere with each other. I don't know WHAT happened. I've got decent audio from the Zoom H2, and also decent audio from a lapel wired olympus recorder I put on the groom, so I've got backups, I just would like to know a little more if possible to prevent this in the future. Thanks!

[Don Bloom]

Michael, while I'm not an audio expert I can guess that at least of the following happened.
1. The "soundguy" as you stated knew nothing about wht he was doing and might have done something to bang your audio withou realizing it.

2. There were crossed freqs somewhere although many of the churchs I have worked in over the last 27 years are still running VHF OR a UHF freq that is so far from mine we're not on the same planet

3. you simply got unlucky and the moon and stars were in alighnment and BAM! You got it in the shorts.

Having said that though, please allow me to ask, how long you've had your wireless, how often you use it, how familar you are with it, did you do a sound test prior to the event and did you wear headphones.
I'm not trying to be mean or a jerk but it sounds like you haven't had the unit too long or at least haven't used it a lot, didn't do a sound check with the wireless prior and didn't monitor the wireless. If you had I'm sure you would have heard the noise in the headphones or some combination of the above.
All I'm saying here is there are things that go wrong for no appearant reason but one of the biggest reasons people have trouble on the job is because before the job they don't check or know the gear they're using well enough to have 100% confidence in it.
Again I'm not dissin or flaming but looking at all the reasons you had challenges on this job and glad you had enough backup to get thru it.

[Michael Clark]

No offense taken! This was my first job with the Sennheiser, although I've used older UHF systems in the past. I did sound checks on it multiple times leading up to the event, using the same Canon T2i that I plugged it into at the event (same exact setup). Unfortunately the T2i only has one jack and I couldn't monitor it live from the cam. I checked the receiver's levels multiple times throughout. But you're right, I should have tested it probably AT the venue while their sound system/wireless was on, and played it back before the ceremony began. I did test it afterwards (today) and it seems to be acting normal again. Ah well I'll have to chalk this up to a learning experience perhaps. I think in the future I may plug the wireless receiver into the Marantz or Zoom instead of using their sound board. That way I can monitor the audio live. Thanks!

And again, if anybody knows anything that may improve the quality of what I've already got, that would be appreciated as well.

[Chris Harding]

Hi Michael

The bottom line, of course, is to pre-test the wireless audio and monitor with headphones so you get a "What I'm hearing is what I will get" .... I try to hook up the wireless system and also get the priest/officiant to stand close to the groom. My mate had a case where the officiant's headset/ mic was perfect BUT he then used a handheld mic so the guests could hear the vows and that totally screwed with the cams wireless mic but only when they were close!!! I know you can't monitor what's actually going thru the T2i but at least monitor at the receiver!!!

While Don is around here...a BIG thank you regarding the 3 mics into 2 XLR situation!!! Yes, Don, I built a switch box so I could switch between the 2nd transmitter on the readings lectern and my on-cam Rode mic ....the priest couldn't get the Church lectern PA mic to work so he began fiddling with switches and accidentally turned my transmitter from "audio" to "standby" ..I had communication but audio was now muted so I just flicked my switch box and let the Rode take over...not the greatest audio of course (I was about 15' away!!!) but certainly a lifesaver.

Don't worry Michael, the next one will be good!!! Just avoid any mixer feeds...most operators have no idea what they are doing and you are at their complete mercy!!!

Chris

[Johannes Soetandi]

Originally Posted by Michael Clark

Unfortunately the T2i only has one jack and I couldn't monitor it live from the cam.

My comment is non related to the wireless audio.. But I've heard with T2i you can monitor the audio using earphones (via the A/V output). You just need to buy some cable converter plug and install Magic Lantern. You can also adjust the volume input and monitor the dB level on the T2i monitor.

When I heard all these I'd been so tempted to switch my 7D to T2i!

[Jim Snow]

Originally Posted by Michael Clark

The church sound guy messed up

Unfortunately, that describes too many church "sound guys". Most of these are volunteers with little to no training or knowledge. I have adopted the policy of trusting no "sound guy". The worst offenders are many DJ's who are actually entertainers, not sound guys. The next worst offenders are church "sound guys."

I have developed a self defense strategy that assumes the sound guy is an idiot. If it turns out that he isn't, that's a nice surprise. If I'm not familiar with the sound setup and mixer, I try to become familiar with it when possible, I talk to the 'sound guy' beforehand and find out what kind of mixer he uses so I can be prepared beforehand. It's a subject for another thread but you don't want to hook up to the main output bus from the board which essentially is what is being fed to the speakers. Every time the "sound guy" tweaks the volume, he is tweaking yours too. You don't want that. You want to be able to adjust the volume only when you need to. Some of these "sound guys" do what I call 'feel good tweaking'. They turn the volume up whenever they get jazzed up and start to 'feel good' - - - and surprise! you're clipping all of a sudden before you even notice it sometimes.

You don't have to be rude but you still need to do what you need to do to unsure that YOUR video is properly done. Whoever you are doing it for expects that, not a debriefing on why the bad audio was someone else's fault.

[Oren Arieli]

Since everyone has advised you on what to do next time, I'll be the voice of the dreaded 'fix it in post'. If you want to give a shot to some narrow-band spectral noise reduction, I made a tutorial that might help. This technique works best when the noises are isolated within a narrow frequency band, and not buried within the approximately 100hz-8000Hz range of human voices. Good luck. YouTube - ‪Advanced Noise Reduction in Adobe Soundbooth‬‏

[Greg Miller]

What I particularly notice is gain pumping. This was undoubtedly recorded with AGC at some point in the chain... presumably at the final recorder.

I hear two specific problems from the AGC. Listen to the distortion at the beginning of the word, "but." That's because the AGC had pulled the gain up prior to that word (presumably there was a long pause in the speaking, although you've made the sample too short to hear the pause). I would guess that the beginning of many of the minister's phrases (when they follow long pauses) are clipped like this.

The second problem from the AGC is that the background noise -- from the room, from the minister's breathing, and from the wireless interference -- is pulled up and made much louder than it really was. AGC will always exacerbate any problems with noise floor. Listen to all the shuffling, minister's breathing, minister's lip smacks, etc... all that garbage should be, at a guess, perhaps 10dB lower. The AGC is pulling up the level of all the noise.

So my advice for future events is: make sure all AGC in the recording chain is turned off. It will make your track sound a lot cleaner in general (including reducing any noise from the wireless system).

Unfortunately, without knowing the slope and attack time of the AGC, it will be very difficult to accurately correct for this in post. If I were in your shoes, I would try to play around with a bit of downward expansion, to see if I could get a more natural sounding track. (You could adjust gains manually, but that would be painfully tedious.)

After getting the noise level back down where it should be, then I'd try a bit of judicious NR to deal with the RF garbage.

[Greg Miller]

Here's a sample of your file, but with the gain corrected "by ear" to undo the effects of the AGC in the recording.

You'll note that the background noise (wireless noise, crowd noise, breathing noises) is much less objectionable than in your original file.

The sample was much too short to even attempt setting up any downward expansion, to make this repair automatically. I suspect it would be not trivial to do that, but if you could do that successfully, it would be less time-consuming than repairing the entire file's gain by hand.

You'll also note that the beginning of the word "but" is still clipped. In fact you can see the clipping if you look at the waveform. (But note that the level of the flat-topped waves is about -4dB FS. So either the clipping occured ahead of the recorder, or the file you posted online has had its gain changed relative to the original file from the recorder.)

[Michael Clark]

Thanks to everyone for the great advice! You are absolutely correct on the AGC. I will look out for that next time. I tried finding some information online about how to perform a downward expansion but am not having much luck. Is it the same or similar as a gate? I looked under the effects tab but didn't see either except "amen gate - aggressive". This actually made it worse. Yours on the other hand sounded great! Any further information on how to do a downward expansion would be appreciated. Thanks again to everybody!

[Greg Miller]

Michael,

Downward expansion is similar to a gate, but much less drastic.

Both processes involve setting a threshold level.

Audio above that level is essentially not changed. With a gate, when the input audio is below the threshold, the gain is reduced pretty drastically, essentially (or nearly) cutting off all audio below the threshold level. You might use a noise gate at a fairly low threshold (perhaps -50dB or lower) to get rid of some low level hum or random noise. Using a gate on your present track would be much too abrupt.

Downward expansion is similar, but much gentler. Audio below the threshold level is reduced somewhat, but not drastically eliminated. The amount of reduction is determined by setting an expansion ratio.

Let's assume the threshold is set at -30dB, and the expansion ratio is set at 2:1.

If the input file is -30dB (or anything louder than that), the output file will be exactly the same level.

If the input is -40dB, that's obviously 10dB below the threshold. Since the expansion ratio is set at 2:1, the output level will become 20dB below the threshold (10dB x 2), for a final level of -50dB. At that point you'd have a 10dB reduction in the level of the unwanted audio.

If the input is -50dB, that's 20dB below the threshold. With a 2:1 expansion ratio, the output level will become 40dB (20dB x 2) below the threshold, for a final level of -70dB. At that point you have a 20dB reduction in the level of the unwanted audio.

Even at that, expansion can sound too drastic, if it makes the background noise pump up and down audibly. So you can also set a lower limit... for example, you can set the maximum amount of gain reduction to 10dB, 20dB, or whatever.

So a noise gate essentially "slams shut" and kills everything below threshold; downward expansion just gradually turns down the gain as the input falls below the threshold.

What I did manually was give you about 10dB to 12dB of noise reduction... very gentle. I did it manually. I listened to the file to see where the AGC started to raise the gain. I also noted where the gain seemed to stop rising. I marked those two points, and gradually lowered the gain over that time period. Then I kept the gain down until the beginning of the next desired audio.

In other words, if the minister was talking up to 1.00 seconds, and I heard the gain rise from 1.00 to 2.00 seconds, I made a ramp down: I left the gain unchanged at 1.00, but gradually reduced it until it was -10dB at 2.00 seconds. Then I kept the gain down until the beginning of the next desired word.

You might or might not be able to do this successfully using a downward expander. There is a huge amount of background noise, and the expander might accidentally interpret that as "desired" audio. No way to tell except to try.

As I said, the sample you uploaded is much too brief to be able to play with expansion and see if you can make it work for this particular file.

And note that there is also a lot of distortion in your file. That's because the gain was set much too high at some point in the chain. Actually the AGC saves you, because as soon as it hears the beginning of a word, it drops the gain, which prevents the rest of the phrase from being distorted. Even if you reduce the background level, and filter out some of the HF interference, you won't be able to get rid of that distortion. So... do you think it's possible to salvage the file to your (or your client's) satisfaction?


Michael, please clarify one thing, just to satisfy my curiosity: was the file that you uploaded shipped directly from the file as it was recorded? Or did you process it, change the level, or alter it in any way before you uploaded that sample?

[Michael Clark]

Greg - the information you provided is very valuable in providing me with some great insight - thank you. I can tell you how that audio link I provided came to be. The wireless receiver plugged straight into a DSLR (with AGC on). I then imported the .mov file into Premiere CS5, and imported a segment of the audio to mp3 (256bps).

So is a downward expansion performed manually, or is there a filter or effect I can apply to the entire track? Sorry to seem dense!

[Greg Miller]

Michael,

As I said in post #9 and #11 (above), in this sample I changed the gain manually to try to restore what I thought were more correct levels.

As I said in post #9 and #11 (above), the sample you posted is too short to set up any automatic expansion process in the software.

(Basically, I'd want to listen to a large number of samples to hear how the expansion is working, to be sure I have the various parameters set successfully. I can't make that assessment on a sample that is only ten seconds long.)

[Michael Clark]

Please excuse me. I was in a head-on collision (not kidding) on Wednesday night and so since then I've been on muscle relaxers and pain killers. But I have changed out the audio for a longer clip:

http://clarkvideoproductions.com/bad_audio.mp3

[Greg Miller]

My gosh, Michael, I'm truly sorry to hear that. I hope you weren't hurt badly. I'm sure the medications are relaxing more than must your muscles.

I'll take a look at your longer file tomorrow, and see if I have any luck using downward expansion in software.