Gain structure: dBu, VU, dbFS

[Tom Morrow]

I've heard it said that the professional line signal level is +4dBu, and consumer line level is -10dBv. But why then does my Sound Devices MixPre (and the 302) have LEDs up to +20dBu if nobody ever goes above +4dBu? The manual for the Sound Devices 302 also talks about 0dBU being considered 0 VU, 20dbU corresponding to 0dBFS, and 8dBu corresponding to -12dBFS. The idea seems to be to put the peaks of your signal into that 0dBu to 8 dBu window, so that it goes between -20dBFS and -12dBFS.

The Sound Devices 302 mixer manual though confuses things by saying (#7 on page 21:)

"If the 302’s Output level is at 0 dB and the recorder is not to its clip level, increase the
recorder’s gain control until it is to the clip level."

Which implies that the clipping level should be 0dBu, not 20dBu. And if that is the case then why bother having the LEDs above 0dBu if they will never go on. I'm confused.

When properly configured, is the camera-clipping point 0dBu on the mixer, or is it 20dBu on the mixer?

The reason I'm asking is that I want to calibrate my MixPre to my camera, but given that it only has a 0dBu 1kHz tone, I don't see way of verifying/calibrating whether my camera clips at 20dBu which I think is the proper way to set it up.

[Daniel Epstein]

Tom,
Welcome to the wild world of digital audio vs Analog standard outputs and inputs. The +4DBU analog Line output standards are not the maximum signal the amplifiers can pass but the strength expected at 0db on VU meters. Consumer machines expect -10 as Line. A much smaller signal. Different Line outputs can result in distortion or a weak signal at the input. Sound Devices is taking liberty by trying to make a digital number make sense by matching the 0 levels up. This can mean underpowering inputs expecting +4DBU as line in.
Once everything is aligned with tone real world signals have a wider dynamic range so there are many times signals would go to +20DB over 0. The mixer is letting you know how big a signal it is sending above 0 at that point. Setting your limiter to avoid overloading the input device can be a matter of taste and experimenting. Most people would not want to reach the clip level of a digital recorder with signals but would prefer the mixer to limit the signal before the recorder clips. If you are running tone at 0 on the mixer to -20 at the recorder you want the headroom above -20 to record without issue.

[David Sholle]

I hope that it is OK to post this link, as it is from another forum, but I found this thread useful in helping to explain your questions:

Procedure for setting up SD302/DVX100 combination

It refers to setup of a digital 0 dBFS camcorder with the analog SD 302. I have found it useful for connecting the SD 302 to digital camcorders as well as portable digital audio recorders. The author uses the incorrect spelling "desiBel" instead of "deciBel", but otherwise I found it very useful.

Also, download the most recent version of the SD 302 manual from the SD website - it has some extra info in it that was not in my printed manual when I bought my 302 several years ago. For example, it showed how to create a full scale tone on the output, to match the 302's clipping level to the camcorder or recorder's clipping level. This is discussed on page 21 of the downloaded manual. Normally the tone output by the SD 302 is not at full scale, but is instead (from the manual):

"From the factory, the Tone Oscillator is set to output a 1 kHz tone at 0 dBu to the outputs (when the outputs are set to Line level). If you are interconnecting primarily with analog video cameras, you may want to modify to the tone output level to +4 dBu in the Setup Menu. In the Setup Menu, the Tone frequency and Output level can be adjusted, and tone can be defeated altogether."

[Rick Reineke]

Tom, I think you may be confusing dBu ,voltage mesurment with the dBFS scale on your camera and 'VU'. For simplicity, if you set the 302 ref. test tone to 0 . (SD factory set-up menu default) and set the camera's meters to read -20dBFS, you will probably be OK.
If you really want to deal with 0VU= +4dBu reference level you can change that too in the set-up menu, advisable if for instance if you were shooting with analog Betacam rig.
One could write a book on this.. some have... read up.

[Claire Buckley]

Hi Tom,

Following is an extract from a published article of some years back (I'm the author)...

May help on the path... but also get to understand the ballistics of metering in relation to levels... it's not always what you see on the meter is what you get...

From an article titled: "Broadcast Operating Levels"

EBU or SMPTE? VU or PPM? Dbu or DBFS? An array of standards and specifications with which the new entrant into broadcast audio confronts. Broadcasters have standardised on one or the other as a matter of history, and of good engineering practice as defined by specifying bodies.

But how many professionals understand the difference and the consequences of operating level? How does it relate to digital consoles, DBFS scales, and to the programme content?

EBU R68 as used throughout most of Europe specifies a test tone level at -18db DBFS which corresponds to a voltage output of 0.775V RMS. Most recognise this as the 0dbu reference, that is, the resulting voltage across a 600 ohm resistance in order for 1mW of power to be dissipated. A throw-back of the old days in the UK to the General Post Office (GPO) and BBC specification relating to balanced line feeds terminated into 600 ohms. What goes in, was what was supposed to come out - but it's okay leaving me!

SMPTE RP155 used throughout most of North America specifies a test tone level of -20db DBFS which corresponds to a voltage output of 1.23V RMS (+4db greater) recognised by users in North America as the 0VU reference. Therefore, it can be seen that things start to hot up very quickly at the business end of the scale if levels and standards are ignored. Allowing content to reach 0DBFS using EBU operating level corresponds to an output of 6.16V RMS (+18dbu), but using the SMPTE operating level 0DBFS corresponds to a colossal 12.28V RMS (+24dbu) - a +6db difference, in practical terms a voltage difference twice that of the EBU operating level. Consequences too for the audio console as output headroom becomes all the more important. Therefore console specifications of more the +24dbu output headroom should be available, typically +26dbu, or even better, +28dbu.

With colossal voltage levels appearing at the output possible, specifications relating to maximum (operating) output levels would generally be held at +12dbu (3.09V RMS). This corresponds to -6db DBFS EBU. This maximum output level applies to many of the European broadcasters. In North America, maximum (operating) output levels would be from +3dbu to +7dbu (1.10V RMS to 1.74V RMS). Therefore, employing good practice and the correct engineering standards whatever the operating levels will guarantee programme content is held within the dynamic constraints of the transmission path.

One further element to be considered however is the ballistics, dynamics, and integration time of the signal measuring instrument itself. The simultaneous use of the Volume Unit instrument (VU) and the Peak Programme Meter (PPM) with typical programme content will show marked differences in results. One may seem to underdrive, while the other might seem to overdrive. Therefore an appreciation of exactly what each instrument measures and how it responds mechanically should be realised.

The VU meter has a typical integration time of 300ms, simply, its ability to 'see' the signal and mechanically register its value takes about one-third of a second. For this reason the VU meter is considered to be relatively slow. However, the instrument detects averages not peaks - the Volume Unit so to speak. For signal paths terminating at an audio recorder using ferric oxide tape, the VU meter was considered by some a bonus, enabling maximum signal to tape in order to minimise the effects of tape noise and to maximise signal to noise ratios - tape saturation and print-through allowing (0VU at 320nw/m, DIN +4dbu). However, this does not help a transmission path where the output may be a radio transmitter. Often, a limiter device is employed to brick-wall the peak signals to the transmitter location.

The PPM on the other hand has a very fast integration time, typically a rise time of about 5ms, and a decay time of 1.5 to 2 seconds. As the name Peak Programme Meter implies, it detects peaks of the programme content. For transmission paths and transmitters sensitive to peak clipping, and at worst transient shutdown, it provided accurate monitoring and safety.

Measuring Average or Peak values has advantages and drawbacks, but providing good practice and recognised standards are employed consistently, and more important, understood, then all should be well. This being the case, the audio engineer is able to concentrate on the content and not whether their output is melting someone's transmitter, video tape or audio tape. This being true, then they can confidently say: "It's okay leaving me!"

==========

I believe this has also been published by Pro Tools in one their training modules... never checked, but that's what they said.

[Tom Morrow]

Thanks this stuff is starting to make as much sense as a mishmash of inconsistent systems can. I'm thinking to be consistent with the Sound Devices gain structure advocated in their manual, I should get things set up so that:

MixPre +20 dBu = 0dBFS on camera
MixPre +8 dBu = -12dBFS on camera
MixPre 0 dBu= -20dBFS on camera

Then I will aim to keep the signal in the orange LEDs between 0dBu and +8 dBu, The red LEDs above +8 shouldn't light more than very infrequently. The green LEDs below 0dBu will indicate too low of a signal.

The posted link above on setting up the 302 was helpful in that I learned the difference between the 302 (which I do not own, just lust after) and the MixPre (which I just got and am setting up). The MixPre LEDs are "WYSIWIG" in that there is no calibration as on the 302; the 302 can assign the 0dBu LED to any level, such as -10dBu in the case of matching consumer line level cameras.

But the MixPre doesn't have that kind of calibration; the 0dBu LED will always indicate an output level of 0dBu, nothing more or less. This means the 302 procedure of hotting up the 1kHz 0dBu signal until it reaches clipping around 20dBu is not going to work for me.

Rather I think I'll try to calibrate with an attenuator cable instead of sound devices menu. I have a SEScom -25db attenuating cable arriving tomorrow. I'll plug that from the 3.5mm tape out jack on the MixPre (which is fixed at 10 db lower than the XLR outputs indicated on the LEDs) to the camera 3.5mm input. This should get me close; 10db lower from the jack plus 25db more lowering from cable should be about 35db lower total, which is close to the -40dBu standard for microphone signals such as feed my 60d camera.

With Sescom 25db attenuating cable in the -10db output jack, I'm hoping:

MixPre 0 dBu LED = Camera Mic -35 dBu level= 5dBu hypothetical camera line = -20dBFS
MixPre 20 dBu LED = Camera Mic -15 dBu level= 25dBu hypothetical camera line = 0 dBFS
MixPre 8 dBu LED = Camera Mic -27 dBu level= 13 dBu hypothetical camera line = -12 dBFS

Since I don't know the limit of the camera mic input, I don't really know that the 25dBu that appears at the camera mic really will line up with 0dBFS for instance; this is just an educated guess, and if it turns out that I'm wrong I suppose I'll hack around with a resistance test box and make a custom cable that has just the right attenuation.

Anyway, once I've used my attenuator cable to line up MixPre 0dBu with camera meter -20dBFS, then I'll need to set the limiting screwpots so that limiting engages around -3 dBFS.

The way to do this may be to use the fader knobs along with an external 1k tone from a CD or NTI minirator mr2 . I will adjust the level of that external 1k signal with the MixPre fader until it just tickles the 0dBFS indicator on the camera. Then I will twist the limiter screwpots down until the camera meter shows the level dropping from 0 to about -3dBFS, in both channels.

Does this sound like a good plan for calibrating my mixpre to my 60d camera?

[Tom Morrow]

Claire, your article was very well written, but left me still confused about what voltages actually happen... 20dBu equals 21.9 Volts peak to peak. If that is our 0dBFS signal, does that mean that the mixer would actually send 21.9 Vpp to a hypthetical line level camera... is that high of a voltage something that really happens "in the wild", or is it more of a theoretical thing since we should give 12db clearance from 0dBFS as a matter of practice, making the max signal we would see in the mixer be 8dBu or 5.5Vpp (1.95 Vrms)?

I suppose the answer is probably "it depends" with a big asterisk of some kind :-)

[Steve House]

Originally Posted by Tom Morrow

Claire, your article was very well written, but left me still confused about what voltages actually happen... 20dBu equals 21.9 Volts peak to peak. If that is our 0dBFS signal, does that mean that the mixer would actually send 21.9 Vpp to a hypthetical line level camera... is that high of a voltage something that really happens "in the wild", or is it more of a theoretical thing since we should give 12db clearance from 0dBFS as a matter of practice, making the max signal we would see in the mixer be 8dBu or 5.5Vpp (1.95 Vrms)?

I suppose the answer is probably "it depends" with a big asterisk of some kind :-)

Remember the "0dBFS" is "zero db full-scale" the never-exceed level when the meter hits the right-hand peg and the digital signal becomes all ones (11111111....). You wouldn't deliver that high a voltage to the camera or else your signal would be hopelessly clipped flat-line.

The MixPRe's tone is set so that 0VU tone represents at output of 0.775v or 0dBu. Send the tone to your camera and set the camera's input level controls so the camera meter indicates -18dBFS (the EBU standard makes the numbers easy to juggle in your head). This gives you 18 dB of 'headroom' (not really valid with digital recording but the concept works) and the MixPre's limiters will catch any signal peaks that are high enough to approach clipping. Keep normal speech flickering between +2 and +4 dBVU on the MixPre meters and you'll be in the best of both worlds, a reasonably hot recording level to give a good signal/noise ration but not so hot as to risk clipping. If you look at the camera meters while recording that speech you'll see it hovering around -10dBFS. That's normal because a sine wave tone at an average level of -18 reads -18 on a peak meter but a speech envelope at an average level of -18 on reads about -10 on a peak reading meter.

[Tom Morrow]

I'm setting it up right now. Harder than I thought to get db-level accuracy, as one LED on these meters can span a range of many dbs.

I set the MixPre to enable the 1kHz 0dBV output tone, and fed that into the camera mic input with the SESCOM LN2MIC-ZMH4-MON attenuating cable. Using a voltmeter I read 5.5mV, which is -43dBu. That registers as about -26dBFS on the camera meter. I was hoping it would end up -20 dBFS, so we're within the ballpark here.

But as a temporary measure I replaced the SEScom cable with a radio shack variable inline headphone attenuator designed for headphones without a builtin level dial. I dialed that such that it hits the spot on the camera's meter that I think is -20dbFS exactly. Then to verify the test I fed a 1k sine wave at -12dBu into the MixPre XLR input (that is the max level before the mixpre input clipping LED goes red). Adjusting the Mixpre input fader so the mixpre 0dBu LED light just barely goes on, the camera meter stays at the same place that I think is -20dBFS. Yeah!

I then turned up the MixPre Fader to get the camera just up to 0dBFS with the limiter switched off. Turning the MixPre limiter back on, the camera meter dropped to -8dBFS, with the limiter threshold turned all the way to the right (least limiting). Turning the limiter threshold all the way to the left (most limiting) the camera shows -16dBFS. Switching the limiter off the camera goes back to 0dBFS.

Then I realized that the input limiter orange LED was activating, which makes sense since I was sending in a -12dBu signal level that was just below the clipping floor. So I lower the input signal to -18dBu, which is the highest signal before the orange LED starts to light. And for good measure I back off another 6db down to -24dBu input signal level.

I turn up the mixpre fader to get the camera signal back up to 0dBFS, and note that for some reason the red clipping light on the input comes on almost exactly at that point. Cool!

Now I play with the limiter threshold pot. At min limiting the camera meter reads about -2dBFS. With max limiting the camera meter reads -16dBFS. I decide to compromise and back the limiter pot up just a bit from minimum so the camera reads -6dBFS. I do the same adjustment for channel 2 as I did for channel 1, and walla, I think I'm done! Or at least sleepy enough that it's time to go to bed.

Now I just have to figure out a way to make up a more permanent version of the attenuator cable.

[Steve House]

You're really making this more complicated than it needs to be. With the MixPre at the factory defaults, the limiters kick on whenever one attempts to drive the output over +18dBu. In other words, the mixer will never send the camera a signal in excess of +18dBu. If you use your attenuating cable and adjust the camera's input levels so that the mixer's 0dBu output tone is aligned to -20dBFS on the camera, nothing you present to the mixer's inputs, no matter how hot, will ever drive the camera past -2dBFS so you have a 2dB safety margin before the camera clips. The mixer's input limiters will prevent the mixer itself from clipping. The mixer's output limiters will prevent the camera from clipping. No need to obsess over voltages and signal generators.

The MixPre meter is peak reading showing the output level in dBu. 0dBu on the meter means the mixer is sending 0dBu down the line. As a peak reading meter it will show flickers of about +8 on a speech waveform that has an average level of 0dBu. This is the same actual average level as a 0dBu sine wave. If tone lights the 0 LED, speech at the same identical average level occasionally lights the +8 LED. If you keep the speech signal around +4 with occasional flickers of +8 on the MixPre's meters, you'll be recording with peaks hitting around -12 in the camera and there is adequate safety margin above that to absorb both higher peaks that are too fast to be visible on the meters and the normal fluctuations in the level of the talent's delivery.

[Tom Morrow]

With the MixPre at the factory defaults, the limiters kick on whenever one attempts to drive the output over +18dBu. In other words, the mixer will never send the camera a signal in excess of +18dBu.

I think you are talking about the case where the limiting pots are set at minimum. The Mixpre manual says:

Thresholds independently adjustable from +6 dBu to +18 dBu, 10:1 limiting ratio, 5 mS attack time, 100 mS release time, Amber/Red LED indicates limiting/clipping, dual-mono or stereo-linked

But they don't give any procedure for adjusting the pots. It sounds like most people just leave them at the minimum (+18dBu, or -2dBFS).

[Steve House]

Originally Posted by Tom Morrow

I think you are talking about the case where the limiting pots are set at minimum. ..

But they don't give any procedure for adjusting the pots. It sounds like most people just leave them at the minimum (+18dBu, or -2dBFS).

True, that is what I meant by factory default. There's a very important red flag in your message. "+18dBu or -2dBFS." +18dBu is -2dBFS IF AND ONLY IF the mixer is aligned to the camera so that 0dBu tone reads -20dBFS on the camera meters. If you align 0dBu tone to -10dBFS on the camera meter and then send a +18dBu signal from the mixer to the camera, the camera meter will peg at full scale as the signal attempts to record at +8dBFS, a physical impossibility and the result iwould be a totally clipped flat topped waveform.

Most people do leave 'em at the factory setting since limiting the mixer output at about +18dBu, with 0dBu being set to -20dBFS, is a pretty standard practice, a good compromise between recording at the highest level prossible in order to get the best signal/noise ratio and yet being relatively safe from accidental clipping with the typical range of levels encountered in a scene.

Just FYI ... with SD's higher level mixers like my 442 you can set the limiter thresholds using the menu system and during the setting procedure the meter indicates the precise level you are setting.

[Tom Morrow]

I'm pretty obsessive, so I needed to get this totally right. First I realized I needed to know if the camera meter was correct, so I made a recording of 1k tones with me speaking the voltage. Playing that back, I found the voltage at which the 60d mic input clips (-14dBu). Then using that as 0dBFS played back test signals and looked at where the voltages ended up on the camera meter. It turns out the camera meter is perfectly accurate, 2db per segment, but the markings are off by 2db; the camera 0dBFS mark is actually 2db below clipping, and the camera -12dBFS is actually -14dBFS below clipping.

With this info I knew exactly where -20dBFS was on the camera meter, and I verified that it exactly hit the 1k 0dBu tone from the mixpre, with my jury-rigged variable-attenuator cable. I taped the dial on that cable down so it won't drift, and now I think I'm set.

I put little stickers saying "Voice" on the MixPre spanning the 0 to 8 dBu LEDs, and spanning the -20 to -12dBFS segments on the camera display, so I won't forget when I'm out in the field. I'll try to keep voice signals within that range, illuminating the 0 dBu LED most of the time, with the peaks of the voice lighting the 8 dBu LED.

I decided to back off the limters back to the factor default, minimum position. I verified that they are engaging at the right place... the last 0dBFS segment on the camera requires more than 10db more voltage to illuminate with test signals than the next to last segment, so it's clear the limiter is engaging there.

I think I'm good to go! The only thing left to make this a more polished installation would be to make a proper cable, approx -15db seems to be what I'd need. It turned out the SEScom -25db cable is about 8db too cold when I use it from the 1/8" tape out from the MixPre. I'm thinking that I could try feeding that cable from the XLR outputs which are 10db hotter, and that might be pretty much right on. But no hurry to do that since my variable attentuator jury rigged cable works fine.


Thanks

[Steve House]

-15dB attenuators built into XLR barrel connectors are readily available off the shelf from most pro audio shops.

[Tom Morrow]

FYI for anyone coming upon this thread trying to do something similar, here's someone who struggled with similar issues:

Connecting Zoom H4N to Sound Devices MixPre – Part II | Dan McComb

One thing I got out of it is that M-Audio seems to make barrel attenuators in 1/8" size which is what I need. But now that I have my variable attenuator cable dialed in and taped down, I don't see much reason to go with a fixed attenuator that would probably be off by a few db.

What would be ideal would be a 1/8" variable attenuator cable with right angle plug on both ends, ideally designed for the impedance of line level (high impedance) on one end, and mic level (low impedance) on the other. If anyone knows of such a beast please let me know.