Epic audio: feature request - digital audio only

[Stephen Pruitt]

Brian. . . of all of the people who write posts on this forum, my "respectfulness quotient" must certainly be near the top of the class. There was nothing personal in my response to your comment at all, so I am very surprised that you somehow took it that way.

And, yes, the RED's preamps are certainly inferior to what any reasonably high-end mixer would have. This is no insult. It's a camera. I understand that. What I was trying to make clear is that if we had a digital line-in, we'd be able to have perfectly pristine audio in perfect synch and solve all sorts of work-flow hassles, as well.

Further, by-passing the preamps does not completely solve the audio problem, as any analog line-in will be adding cable noise all along the route from the mixer to the camera. An AES input is the obvious solution. A lot of people will be very disappointed if it is not included in Epic.

Stephen

 

[Eddy Robinson]

But if shooting our first feature taught us one thing it is this: It is vastly more difficult to get great audio than it is to get a great image.
Stephen

As a sound guy, you have no idea how happy it makes me to hear this from a camera person. In Hollywood they can and do ADR everything, but they have a huge budget with which to do it. For indie films, ADR means losing more money out of a post-budget that is usually already less than planned before shooting. Good production sound is hard, and sometimes it takes some extra time to get it - but it is much, much cheaper than 'doing it in post'. To hear you say it pays off for all the time I've spent learning about things like focal length in order to better understand what the camera team is doing, and how to do my job without getting in their way or asking too many questions.

There's an unfortunate knowledge gap among some producers about the way the picture comes together. The net effect is that in production, they spend a lot of money on the camera (plus lighting, grip etc.) and the sound comes in between makeup and catering on the budget sheet. It's rare for me to talk to a producer and be offered a good rate immediately, usually I have to fight for cash, then fight for pay for a boom operator, then fight to get on the tech scout to learn about the location (for which I have never been paid...). The same thing often happens in post - very often the first chunk of money goes to the editor, there's another (reduced) chunk for the color correction, some money (not much) left for music, and very little money left for the sound.

I could go on and on about how badly this tends to work out in practice, but I won't. I'll just try and sum it up in a few rules of thumb:

0. Budget. On a big film, it's 5%. For an indie film, it needs to be 10%.
1. PAs rarely turn into good boom ops, any more than they do focus pullers.
2. Location scouting. We can save you so much production time this way.
3. It's usually NOT cheaper to just 'fix it in post', unless we tell you it is.
4. We hear everything behind the camera too. No, you can't filter it out.
5. Get your ADR/ wild lines/ FX on set. You will save sooo much money. Yes, you take sound on inserts. Only the sound guy knows what should be 'MOS'.
6. Dailies - anywhere you would bring the DP, take the sound person.
7. Post - consider the Murch way, keep sound involved from day 1. It's often much cheaper.
8. Very rarely will music fix a dull scene, though it can make a good one great.
9. Mixing is like compositing - it's complex and expensive.

I'll add my vote for AES on the camera, or more specifically for digital input - it's only a firmware tweak to switch betweeen AES and SP/DIF, which is a little more common on small recorders.

I understand why people want analog audio in, but it's also one more thing to to go wrong - it needs to be lined up with a test tone every time, and anyway you never, ever want a mic going directly into the camera unless it's something super-simple. You can't watch focus, framing etc. and still do a good job managing audio gain.

 

[Eddy Robinson]

I have been working on very low budget short films lately see locker13.com and it has all been dual system. I mean this is a very low budget film we feed a wireless reference usually. I have yet to work with a sound mixer that had a digital mixer.

Really? You've never had a sound guy with a Fostex PD-4, Tascam HD-P2, Sound Devices mixer, or (if you're lucky) high end Fostex, Deva or Cantar system? Almost all field recorders since DAT days have digital output. It's actually one less cable (vs 2 for stereo), and it saves a couple of minutes every day in sending down a test tone and calibrating the meters.

Most Red users are just getting up to speed on how to handle analog inputs.

Which is another reason I'd rather just hand someone a digital XLR cable. Plug it in, are you getting audio at the camera y/n, if yes then we're done. It's not that big a deal, but as a sound person I'd much rather not have to check switches and pots on the camera in addition to my own gear. With digital, I know the camera is recording exactly what I'm hearing.

If you just feed a line level from an analog mixer you don't have deal with our "inferior pre-amps" and if you think our A/D conversion sucks then you should rely on your audio guy's digital tracks. How can you insult RED in one breath and expect anyone to take your input seriously?

Generally the line level input on most prosumer+ cameras is quite good enough nowadays - even at 16 bits like on a DVX100 (you won't hear that much difference unless you're recording music). But even at line level, there are differences in what happens when you get close to the clipping theshold, which is easy to do when an actor is whispering at the beginning of a scene but then starts yelling or screaming (eg pretty much any horror film will have several such scenes). Every time an actor is substantially louder or quieter than they were in rehearsal, any uncertainty about what went onto the production track = a 5 minute production delay, or a half-hour post task.

 

[Eddy Robinson]

The only caveat here is that this needs to cross platform to Scarlet as well since that's what I (and many indie types) will be buying.

Word, on both the Traveler and Scarlet.

 

[JonathanF]

Eddy,

I think my sound guy (who just did Slumdog) would give you a big wet sloppy kiss for all that. I could not agree more. We treat sound as important as picture and while we do record to an external recorder, we send wireless to the camera so we have sync sound in dailies and all of your points are well supported by my crew.

We're doing doc work and sound is indeed a much lager part of our budget than on a feature (having done both) and your figures are just about dead on.

I'd take AES on camera in a heartbeat and I know my sound guy would wet himself too.

Good on ya mate,

J

 

[Jochen Schmidt-Hambrock]

I´ve been shooting cine style with no audio on cam.
Because there would be too much crap hanging off the camera.
Two tracks IN amounts to 4 cables or wireless devices.
Unless you want the dp to monitor the audio with headphones - and you dont want that.

I´ve been shooting interviews with just a directional mic on cam: No problem.
Interviews with just a wireless feed on cam: No problem.
Shooting 2nd unit or stock with just a stereo mic on cam for ambience: No problem.

Please dont take that option (recording uncritical audio) away.

Jochen

 

[JonathanF]

I´ve been shooting interviews with just a directional mic on cam: No problem.
Interviews with just a wireless feed on cam: No problem.
Shooting 2nd unit or stock with just a stereo mic on cam for ambience: No problem.

Please dont take that option (recording uncritical audio) away.

Jochen

I don't see how adding AES onto the board in any way takes that away from you.

J

 

[Jochen Schmidt-Hambrock]

If its adding AES to the analog inputs you are right. Could be good.
BUT:
Main question is always: Who is monitoring the recorded signal? (Dont know the english term. In german its "Hinterbandkontrolle".) And with all things digital controlling the rec level is only the trivial part of the job.
There it gets enormously more complex: Do you have a separate wordclock input on cam? I usually avoid self timing AES. Too many things can go wrong that you will not notice in a noisy environment.
I own lots of digital audio gear and hear new interesting ways to produce digital errors all the time. Connecting more than 2 pieces of digital audio without wordclock ist much too risky.
Long story short: Just having some AES inputs is not a safe way to clean audio.

IMHO good convertors are.
And I dont want to buy some AD converters and have them hang off my Red.

Jochen

 

[Tom Visser]

As reasonable as this may be- it won't always be the case.

I would suspect that it would always be the case.

I like the idea - about AES inputs, the one thing to be wary of, though, is that you then open a certain randomness factor into latency. Every (digital) camera to date has had A/D converters built in, so therefore the manufacturer knows the latency of the converter because it does not change. When the sound is printed to "tape", the camera knows how to correctly align the audio to image so that it is frame accurate. If you use 3rd party A/D converters, there is a chance that some converters could have an unusually high latency... not the end of the world, but something that should be accounted for. If AES inputs were incorporated, the camera should also have a ping / loopback test to measure the latency and then adjust the printed audio track.

It is somewhat non standard, however, and although I feel that A/D conversion does make a difference, the reality is that any competent A/D design is mostly indistinguishable from any other competent design, unless you are dealing with specific types of audio situations where "critical" recordings are being made. Recording dialogue and typical production ambient tones don't really qualify. The weakness in the majority of on-camera audio is typically in an underperforming microphone preamp, poor metering, and of course, the operator... or lack thereof, the common desire for mics to be hooked up to the camera and then "set and forget" with respect to the levels. You also deal with the reality that at least for a period of time, most all professional sound mixers would not be able to immediately work on an AES only shoot with the equipment they have. Battery powered A/D converters are not rare, but they aren't common either. Probably not an issue for those that plan ahead, but there will be instances of sound crew showing up, geared for what is standard in the industry, and then swearing to themselves and at your camera as they walk away from the set, off to go find an A/D converter at the last minute. Those that use digital recorders, such as the Sound Devices 7##, Zaxom Devas, Nagra V or VI... they will be set, but not all mixer are equipped with recorders too and many are still outfitted with analog only mixers. I think it would be a short growing pain and one that would be justifiable to go through, though.

But rather than reinvent the wheel, my final recommendation would be simply to equip the camera with good time code I/O and then a utility audio input for an IFB feed for scratch audio. Today's IFB devices are of such high calibre, if you really did in fact want to use the track, you could always hook up something like a Sound Devices MM-1 to it or the RED accessory microphone kit and do without the sound crew, as ill-advised as that may be.

 

[JonathanF]

Main question is always: Who is monitoring the recorded signal? (Dont know the english term. In german its "Hinterbandkontrolle".) And with all things digital controlling the rec level is only the trivial part of the job.
There it gets enormously more complex: Do you have a separate wordclock input on cam? I usually avoid self timing AES. Too many things can go wrong that you will not notice in a noisy environment.
I own lots of digital audio gear and hear new interesting ways to produce digital errors all the time. Connecting more than 2 pieces of digital audio without wordclock ist much too risky.
Long story short: Just having some AES inputs is not a safe way to clean audio.

The sound person should be monitoring the recorded audio at the mixer. He/she's going to run everything through a mixer and then send the AES output to the camera. No one on the camera side has to do squat. If my DoP or AC is worrying about sound the sound guy is not doing his job.

On cameras where you can't record AES I'm pretty settled on recording it all off camera and just sending audio off the sounddevices mixer to the camera wireless for reference and worryling about sync later. Been working for feature films for decades and we all understand the process. Frankly, with wireless, there are so many potential pitfalls with Rf, you can't even count on that.

Any good sound person is going to moan and complain about sending sound to the camera anyway and the ONLY time I recall not having a guy moan about it is when we sent AES fo the 900.

The preamps in the cameras pretty much all suck, so if you can use AES you just let the sound guy do all the heavy lifting with gear they like and just send the signal to the camera. This is what you do on the F900 and it works like a charm. The levels and everything else are managed off camera (as it should be).

Now I don't pretend to be a sound guy and I don't play one on TV, but I just had this discussion with our sound engineer and this is pretty much exactly what he wants for the new camera.

Ta,

J

 

[JonathanF]

I like the idea - about AES inputs, the one thing to be wary of, though, is that you then open a certain randomness factor into latency.

This has never been an issue on the F900 and I have not heard of one single issue with AES or latency from people shooting on the F900. Have you?

You also deal with the reality that at least for a period of time, most all professional sound mixers would not be able to immediately work on an AES only shoot with the equipment they have.

Why should this in any way close those of us who are using mixers that DO support AES from using it. Like I said, this IS available on the F900, and with that it seems a sensible addition to Epic. If people aren't willing to make the investment in gear that supports AES, they can always rent.

I know a 788T or a Zaxcom Deva 5.8 is on our purchase list for production as soon as we know that AES3 will be supported on the new cameras.

J

 

[Tom Visser]

I suppose it depends on where you fall in the workflow. If you're not an editor or the location mixer, you probably wouldn't have heard about the problem, if there was one, as it would have been handled without your involvement. Its not really a big "problem" per se... just an offset that might become evident. Most converters, such as those used in the typical Cooper Sound digital card, Sony DMX, Sound Devices 7##, Nagra V / VI, are fairly low latency, don't have effects DSP to slow things down, etc... so if there was an offset, it would only be by a frame or fraction of a frame, and really only evident if a TC slate was used so that one could visibly see the offset, you wouldn't be able to hear it. But as people start bringing other types of gear into the equation, like live digital mixers with DSP, converters with built in sample rate conversion, and especially computer based rigs, those that send the audio into a computer / DAW and back out via Firewire / USB through an interface - that is where you could potentially end up with audible offset. If RED were to go this route, I'm only suggesting that RED provide an automatic delay compensation function. Hook up the gear as it will be used in production - press "PING" - RED sends out an audible 12KHz test pulse or series of pulses via a cheap built in piezo transducer, it measures the return via the digital input, and then applies the appropriate offset to the recording medium. I just feel that one should incorporate such a function into a camera, to eliminate any possible issues and thus, improve the confidence and minimize the expense of using such a system.

Regarding monitoring the AES signal... that is the advantage of the digital domain. Assuming that you don't clip the signal on the mixer side, once it is sent digitally to the camera, it would be impossible for the signal to clip on the camera. The same is not true with analog domain, as the calibration could be slightly off or not done correctly and if you have headroom on the mixer, it is possible you could be overmodulating the camera... thus the need for a confidence return monitor. (or something like RFI, EMI, or other analog domain disturbances were introduced) The issues that AES inputs do introduce, however, is one must have warning indicators if the camera either looses lock, or detects the clock slipping (short losses of lock due to drifting clock rates).

Wordclock is the standard, but honestly the embedded clock in a standard AES signal should be good enough. Wordclock is better in the studio, as it is more flexible - you can daisy chain or D/A wordclock to multiple devices so that everything stays in sync. Here, since the camera would be the terminus of a relative simple digital interconnection, I don't think that you would necessarily HAVE TO use word clock and could get by well enough with the embedded signal. The audio recorder / mixer would be the time code master and the camera would slave to it.

addendum: This might actually be a good opportunity for RED to sell an AES/EBU + LTC timecode wireless accessory or support a 3rd party device, such as some sort of Lectro/Zaxcom product that would more or less slot-in into the camera, keeping a wired option / backup.

 

[Dustin Sims]

This is probably a dumb question: why do digital audio recording methods utilize XLR connections and not optical cables (at least that I'm aware of)? Is it as simple as the fact that nearly everything has XLR connections? And would you even be able to get better audio from optical cables? I'd think you wouldn't have as much interference or signal loss with long optical cables.

 

[Tom Visser]

depends on what type of optical and digital format you are talking about.

toslink or EIAJ connectors use a plastic "fibre" which is really not all that of a long range solution. the cable is sort of durable, but tight kinks or constant bundling could easily cause the cable to snag. SPDIF over toslink can handle 2 channels of audio up to 24bit/96kHz. There is also a format made originally by Alesis called ADAT that can handle up to 8 channels at 24bit/48kHz or 4 channels at 24bit/96kHz (then its called SMUX). this is more of a consumer and permanent installation format. it is popular in low / mid level studio applications because of the quantity of devices out there that support ADAT and the Apple Mac computers happen to support optical SPDIF. for consumer video, the SPDIF transceiver can usually also pass Dolby or DTS bitstream for surround applications.

There is also a glass fibre, originally developed by AT&T, I believe, that was used for a time in consumer audio. It was expensive, both the cables and transceivers, supposedly very high performance at the time, but is not currently being used.

SPDIF over coaxial cable, usually terminated with RCA or BNC connectors uses cheap transcoders and cheap cables. The cables can be rugged, although a bit stiff, because it uses 75ohm video cable. It also can contain a copy protection bitcode that was supposed to make the signal more difficult to copy. (but not really)

AES/EBU uses a twinaxial cable, a shielded twisted pair with an overall shield or 110ohm cabling. The signal itself is very similar to SPDIF. It can come in flavors that have 2 channels per cable or a single channel per cable. it generally uses a locking XLR-3 connector, although it is possible to see implementations that use coaxial cables (AJA for example) and in studios, on equipment that has a lot of AES I//O, it will often be a DB-25 multi-pin connector. There is no copy bit support and there is a professional "flag bit" present that tells the equipment to treat the signal as AES. Because the connector is locking and it will work with regular XLR cables in reality (theoretically, you should actually have 110 ohm twinaxial cable) and the fact that it is a professional transport, makes it THE standard for professional gear.

Both SPDIF and AES can be used with or without a wordclock, basically the synchronization signal on a separate coaxial cable. It used to be true that external wordclocks offered more stable performance and was a necessity where there were lots of different components hooked up together that needed to stay in synch. Modern equipment with better locking hardware and more agile in reclocking to poor sources, although hooking up equipment that is not simply daisy chained, does in fact still require a centrally distributed word clock.

 

[JonathanF]

Regarding monitoring the AES signal... that is the advantage of the digital domain. Assuming that you don't clip the signal on the mixer side, once it is sent digitally to the camera, it would be impossible for the signal to clip on the camera.

Exactly, and that's why AES is such a great feature request for the new camera. In fact, if there's ever _another_ upgrade to the R1 it would be a great addition to that board.

:J

 

[tj williams]

AES module Analog Mic in module MicLine In module XLR digital in Module....
It will be fun to be modular!
Tj

 

[Jochen Schmidt-Hambrock]

The sound person should be monitoring the recorded audio at the mixer.

Ok

He/she's going to run everything through a mixer and then send the AES output to the camera.

And this is your primary audio signal? No recording to another system?

On cameras where you can't record AES I'm pretty settled on recording it all off camera and just sending audio off the sounddevices mixer to the camera wireless for reference and worryling about sync later.

Good procedure. But why do you think that it is different with digital?

Frankly, with wireless, there are so many potential pitfalls with Rf, you can't even count on that.

Right. Same goes with wire. In analog and digital.

The preamps in the cameras pretty much all suck, so if you can use AES you just let the sound guy do all the heavy lifting with gear they like and just send the signal to the camera. This is what you do on the F900 and it works like a charm. The levels and everything else are managed off camera (as it should be).

And this is the only advantage of AES IN on camera. But - as you said it yourself - going from the mixer to the camera is not without problems. So you have to monitor the AES IN.

To repeat myself: Who does it? I would never ever put the burden of checking an AES signal on a DP.

Jochen

 

[Jochen Schmidt-Hambrock]

Wordclock is the standard, but honestly the embedded clock in a standard AES signal should be good enough.

It would be so nice if you would be right. I could have spent all those hours chasing sync errors on something better.

When you are only daisy chaining two digital audio devices without wordclock, it is ok (mostly...... it should be..... unless...).

But what if the situation on set is just a bit more complex: Hasn´t HDMI also embedded digital audio? What about a timecode synced RED? Two locked REDs?

Now who is AES sync master?

I have been in digital audio from its very beginning and in my experience, as soon as you ask that question, you have trouble. Next thing is you have people scratching their heads and checking manuals. In some studios you are not even allowed to say "wordclock" because you will get clicks in the audio.

Dont get me wrong. I am in favour of AES In on the RED. I just dont know if I´d ever use it.

What I really want on my EPIC is 4 good preamps in 24 bit and a decent limiter.

Jochen