stevesherrick
Well-known member
One more thing, which is probably already in the spec, but should mention just out of principle, is 48v phantom power.
Steve
Steve
Audio & FPS
- Thread starter boothba
- Start date
mike the beginner
Well-known member
Good points Steve. I asked a similar question recently but no response. Maybe most people getting a red camera will be using a decent quality mixer with a good pre amp and controls already within the mixer so they are only looking at the audio as a sort of backup in case things go wrong.
I do hope that the audio is top notch in any case particularly the XLRs. Some people want these connections on the back others think its fine at the sides? If red is to be trully great there should be no obvious weak areas.
Michael
glenn chan
Well-known member
What about keeping it simple, and offering 4 tracks 16-bit 48khz?
96khz would be nicer, but most of the time you are just recording dialogue (and a little bit of ambience). 48khz is probably good enough, depending on how well it's implemented (freq. response, aliasing, etc.). If you're doing indie work, you may not distribute with 96khz quality and the recording quality may not be that great to begin with.
The downside of 96khz is added cost, post complications, lower datarate for video.
2- I would have good limiters higher on the list. The ones with soft knees (like on Sonosax mixers, good compressors) sound a lot better than ones without. In indie and especially ENG situations, good quality limiters would be useful (for unexpected changes in volume).
3- For 24-bit 96khz quality and lots of tracks, it wouldn't cost too much to get a laptop with Metacorder (or other multi-track recorder).
4- In many recording situations, there's so much background noise that you don't need to record that much dynamic range.
Just throwing some ideas out there.
96khz would be nicer, but most of the time you are just recording dialogue (and a little bit of ambience). 48khz is probably good enough, depending on how well it's implemented (freq. response, aliasing, etc.). If you're doing indie work, you may not distribute with 96khz quality and the recording quality may not be that great to begin with.
The downside of 96khz is added cost, post complications, lower datarate for video.
2- I would have good limiters higher on the list. The ones with soft knees (like on Sonosax mixers, good compressors) sound a lot better than ones without. In indie and especially ENG situations, good quality limiters would be useful (for unexpected changes in volume).
3- For 24-bit 96khz quality and lots of tracks, it wouldn't cost too much to get a laptop with Metacorder (or other multi-track recorder).
4- In many recording situations, there's so much background noise that you don't need to record that much dynamic range.
Just throwing some ideas out there.
stevesherrick
Well-known member
Glenn, I can see where you are coming from on the 48K issue. Doing post audio, I have to say I really do enjoy working with 24bit 96K audio. It just sounds better to me when I apply a lot of processing to tracks. And you can work at these sample rates in some of the NLEs. It would be interesting to see if they could do some kind of audio conversion in RedCine for offline purposes. That way you could offline with the lower res audio and then if it gets sent to a post audio house for mixing it could be conformed with the 24bit 96K audio.
And I agree with you about the limiters. Having quality limiters would be a major plus.
As for Metacorder, there are some people using it with great success, but it is expensive when you consider you still need a computer to run it, a good audio interface, power (battery or AC). I love my SD 744, and think it's the perfect mid-budget recorder. DEVAs are great but expensive.
One thing I forgot to mention as a necessity for this camera's audio function is accurate metering. That is an important consideration whether you are recording separately or not. If you are recording audio, you need to have accurate meters or things can really get screwed up.
And keep in mind that with its ability to record 4 tracks, you could be recording with say a boom and a lav and record ambience tracks on the other channels. Or you may have no dialogue for a given shot (let's say a beautiful establishing shot overlooking a small rural town) and you have the opportunity to get a nice, rich soundtrack. With the higher quality bit depth and sample rate you'll get a better track. Again, this is me thinking outside the realm of a movie setup and thinking about those who might use this camera's audio functionality to its fullest. I'd be recording that ambience to my 744, maybe even at 192K, but that's not going to be everyone's situation, so I'm trying to keep that in mind when thinking about the design.
Steve
And I agree with you about the limiters. Having quality limiters would be a major plus.
As for Metacorder, there are some people using it with great success, but it is expensive when you consider you still need a computer to run it, a good audio interface, power (battery or AC). I love my SD 744, and think it's the perfect mid-budget recorder. DEVAs are great but expensive.
One thing I forgot to mention as a necessity for this camera's audio function is accurate metering. That is an important consideration whether you are recording separately or not. If you are recording audio, you need to have accurate meters or things can really get screwed up.
And keep in mind that with its ability to record 4 tracks, you could be recording with say a boom and a lav and record ambience tracks on the other channels. Or you may have no dialogue for a given shot (let's say a beautiful establishing shot overlooking a small rural town) and you have the opportunity to get a nice, rich soundtrack. With the higher quality bit depth and sample rate you'll get a better track. Again, this is me thinking outside the realm of a movie setup and thinking about those who might use this camera's audio functionality to its fullest. I'd be recording that ambience to my 744, maybe even at 192K, but that's not going to be everyone's situation, so I'm trying to keep that in mind when thinking about the design.
Steve
10s
Veteran
24 bit 96khz is the new professional std, 16 bit 48k is good but no longer the std. CDs are 44.1 16 bit, but many of the current recordings are at least 88.2, 96k, 192khz, all 24 bit depth. Red should be 24 bit, 96k, 4 channel. Recording direct to the camera could be better than the dual system if proper design up front is worked out with all the required tools. I've conducted many listening tests with all sorts of participants and many can detect qualitative differences up to 24 bit 96khz. At this sample rate one can detect a sound that is somehow just better. If Red does not include this spec then a work around with better recording gear utilizing quality preamps and limiters will be the norm until the designers get the message. I'm amazed at how difficult it is for this message to get through.
> 24 bit 96khz is the new professional std.
I would not venture to disagree with that, but I'm just wondering... are there any cameras made today with audio signal paths clean enough to justify more than 16 bits? Which ones are those? What is their actual measured SNR over a 20 Hz-20 kHz bandwidth? And is that with a mic-level input...?
I know the RED team are miracle workers but with all the high-speed digital activity inside the RED chassis, I believe it will be a miracle squared if they can also manage integrated audio with noise below the 16-bit level.
As far as I know, all the pro 24/96 audio recording gear is dedicated to that task alone, with no built-in video hardware to generate noise.
I would not venture to disagree with that, but I'm just wondering... are there any cameras made today with audio signal paths clean enough to justify more than 16 bits? Which ones are those? What is their actual measured SNR over a 20 Hz-20 kHz bandwidth? And is that with a mic-level input...?
I know the RED team are miracle workers but with all the high-speed digital activity inside the RED chassis, I believe it will be a miracle squared if they can also manage integrated audio with noise below the 16-bit level.
As far as I know, all the pro 24/96 audio recording gear is dedicated to that task alone, with no built-in video hardware to generate noise.
10s
Veteran
You're correct, there aren't mfgs making cameras with these audio specs that I'm aware of, but that's the point, they aren't making RED video specs/price point either. Given the RED modular design, that can easily accommodate add-on components, i.e, preamps, and the high end philosophy RED is using in it's design approach, it's ironic for RED not to pursue high end audio. The S/N spec is in the AD converter & preamps, once the signal is digitized it's just numbers so there should not signal degration in the path once your past the A/D conversion. The components are readily available OTS components requiring moderate adaptation. I get the feeling that the picture will look great but the sound will equal Creative's sand blaster, go figure.
filmmaker1977
Carbonite Member
well, i know they are or the most part is 16 bit/48khz.. i mean, those cheap USB devices for PC audio but they are effectively noise free.. so, it won't be difficult to get it from the red chassis if well done..jbeale said:
Greg Voevodsky
Active member
yes. At least have 2 channels at 24 bit 96khz. 4 channel at the lower bit rates.
filmmaker1977
Carbonite Member
greg, i was talking about those USB audio devices.. not about the red audio but yes i think 24 bit/96khz will be what we'd expect..
Certainly it is desirable. I just wanted to point out that achieving analog audio noise levels below 16 bit quantization inside of a video camera is a HARD job, and as far as I know* this has not yet been done on any camera at any price. (I mean mic-level analog audio inputs; as you say, it's not an issue on the far side of an A/D converter.)
I'm not saying the Red team can't do it or doesn't desire to do it. Just pointing out this is not just a case of replacing a 16-bit A/D chip with a 24-bit one. To actually use those extra bits (beyond just marketing
you have to keep the noise below microvolt levels, meaning a total system layout designed to protect the analog signals from digital interference, and layout changes and board re-spins when the first designs come up short. With the various other world records Red is going for, on what seems to be an extremely rapid development schedule, 16 bit audio doesn't seem so out of place to me from the design point of view.
*I don't have direct experience with the latest high-end gear
I'm not saying the Red team can't do it or doesn't desire to do it. Just pointing out this is not just a case of replacing a 16-bit A/D chip with a 24-bit one. To actually use those extra bits (beyond just marketing
you have to keep the noise below microvolt levels, meaning a total system layout designed to protect the analog signals from digital interference, and layout changes and board re-spins when the first designs come up short. With the various other world records Red is going for, on what seems to be an extremely rapid development schedule, 16 bit audio doesn't seem so out of place to me from the design point of view.*I don't have direct experience with the latest high-end gear
Last edited:
Graeme_Nattress
Red Team
I'd have thought that the bigger issue 30 or 40 or 50 db background noise, perhaps a TV studio will have less than 50db, but there's still background noise unless you're in a superb anechoic chamber, Above 100db it starts to get painful. I'd have thought that the 96db dynamic range that 16bit audio delivers would be more an adequate for capture. Sure, process in 24 or 32bit to avoid rounding error etc., just as we process video in 32t bit float on our GPUs, even though the source is often 8bit (now it will be 12bit he he he) but surely > 16bit is overkill for real world audio capture?
Graeme
Graeme
Steen Dongo
New member
Re: 16 or 24 bit audio ?
Re: 16 or 24 bit audio ?
Actually no !
More often than not the real world of audio locationrecording will not allow you to get even near to using all 16 bits. Doing dialogrecording on a 16bit format will typically leave some 75 percent of your recording mapped on the first 8bit leaving the last 8 bit for temperamental outbursts. When you enter audiopost you end up ganging many tracks of 8bit quantization distortion..... and it sounds just terrible. On feature film recording limiters are not especially well seen. Avoid them if you can - use them if really have to. I consider 20 - 24 bit ideal. Even the cheapest Sony VTRs use 20bit AD converters and purchasing micpreamp circuitry incl. phantomsupply matched to really good converters is not an unsurmountable problem today. (Well maybe due to space constraints
To sum it up: I'd rather have AD converters giving me 24 bit of dynamic range while accepting 48kHz samplefrequency instead of 96kHz if that's the compromise to reduce storage space. Nyquist leave the 24kHz well outside of myhearing capabillity anyways
respectfully
Steen (I used to work as a Tonmeister but was degraded to directing) Dongo
PICTUREWISE COMMUNICATION'
DENMARK
Re: 16 or 24 bit audio ?
Graeme_Nattress said:
Actually no !
More often than not the real world of audio locationrecording will not allow you to get even near to using all 16 bits. Doing dialogrecording on a 16bit format will typically leave some 75 percent of your recording mapped on the first 8bit leaving the last 8 bit for temperamental outbursts. When you enter audiopost you end up ganging many tracks of 8bit quantization distortion..... and it sounds just terrible. On feature film recording limiters are not especially well seen. Avoid them if you can - use them if really have to. I consider 20 - 24 bit ideal. Even the cheapest Sony VTRs use 20bit AD converters and purchasing micpreamp circuitry incl. phantomsupply matched to really good converters is not an unsurmountable problem today. (Well maybe due to space constraints
To sum it up: I'd rather have AD converters giving me 24 bit of dynamic range while accepting 48kHz samplefrequency instead of 96kHz if that's the compromise to reduce storage space. Nyquist leave the 24kHz well outside of myhearing capabillity anyways
respectfully
Steen (I used to work as a Tonmeister but was degraded to directing) Dongo
PICTUREWISE COMMUNICATION'
DENMARK
Muttondraw
Well-known member
Steen Dongo said:
I don't think this is practical, top quality 24 bit audio converters will at best give you 20 useable bits in the real world and that is ignoring all the noise from the associated analogue circuitry.
I don't have a problem with 16 bits as long as the circuitry is well engineered. One thing that is really important though is to minimise jitter. There is no way to recover from jitter introduced at AD conversion.
Martin
> ...75 percent of your recording mapped on the first 8bit...
Wow! That implies 48 dB headroom for temperamental outbursts, which seems like a lot... I wonder if it is possible to find better-modulated talent?
As far as ambient acoustic background noise, I have never even come close to shooting in a quiet enough environment to push 96 dB dynamic range and I don't know where you actually find a space that quiet. Maybe in a well-isolated sound booth, but obviously there you're recording with something else besides a camera.
Wow! That implies 48 dB headroom for temperamental outbursts, which seems like a lot... I wonder if it is possible to find better-modulated talent?
As far as ambient acoustic background noise, I have never even come close to shooting in a quiet enough environment to push 96 dB dynamic range and I don't know where you actually find a space that quiet. Maybe in a well-isolated sound booth, but obviously there you're recording with something else besides a camera.
Steen Dongo
New member
jbeale said:
As a soundrecordist you have to take what comes your way from both actors and directors. If a dialog line calls for whispering the first 5 words and shouting the last...? Believe me you'll appreciate the headroom
Unless you've got the stomach to insist on taking the scene again ? Perhaps it's a bit of a european thing ? We hate rerecording dialog and try to avoid it if possibleThe advantage of 24bits is not just increased dynamic range but the absence of quantization errors, which as I stated above sounds horrible, especially as you start adding up tracks. I know for shure that higher bitdepth is one of the most common reasons for recording sound separately from the camera. (Add to that the option of having more tracks and not beeing tied up to the camera, of course
So if I should record sound onboard the RED I would very certainly prefer 24bit 48 kHz:
As far as jitter is concerned ? Well it's definately hard to fight bad design or faulty implementations in the first place, but it seems to me that audio in general is something like 5 -10 years ahead of video when it comes to the digital realm. A lot of experiments have been conducted and practices been established. It's like recording music: Nobody likes to record anything in 16 bit, but when all the 24bit tracks are summed it's actually quite ok to deliver proberly dithered 16bit 44.1kHz on a CD.
Anyways... it's just my humble opinion your milage may certainly wary
Respectfully
Steen
PICTUREWISE COMMUNICATION
DENMARK
glenn chan
Well-known member
Nyquist leave the 24kHz well outside of myhearing capabillity anyways
I think that's a misinterpretation/misapplication of Nyquist. To avoid aliasing, practical systems have to implement low pass filtering to get rid of high frequencies. The amplitude response for high frequencies is significantly reduced, so performance won't go up all the way to 24khz with perfect amplitude response.
If you don't try to anti-alias, then the system can reproduce 24,000hz cleanly. However, it will have problems with frequencies slightly above 24,000hz- these frequencies will cause a change in the recording. They will cause low frequencies to appear in the recording.
The real world gets a lot more complicated since certain underlying assumptions and conditions for the Nyquist theorem can't be realized in practice. For more detail than you probably want, the Wikipedia article on the Nyquist-Shannon Theorem has some good information.
http://en.wikipedia.org/wiki/Nyquist-Shannon_sampling_theorem
For something easier to understand, just look at how images are resized in Photoshop. When you make an image smaller, Photoshop gives you some different options in how the image is filtered. The nearest neighbour scheme tends to give weirdness (sort of like aliasing) on "high frequency" components like lines and fine repeating detail. The bilinear and bicubic methods don't do this nearly as much.
However, if you make an image 2X and then 0.5X and then 2X and then 0.5X, nearest neighbour gives the best (perfect) results. The bilinear and bicubic methods do not. In sampling theory terms, this is related to "phase shift error". Basically, bilinear filtering results in fine detail getting blurred in different directions. When you want to perform the opposite operation, you can't do this- you don't know which direction the components originally came from.
I hope that is a clear explanation of some of the issues when it comes to sampling. What you also need to take into account is human perception and real-world performance of various systems. There are some things we can't notice and some things we tend not to notice. This allow certain schemes to work very well even if they do not perfectly (mathematically) reproduce a signal (i.e. mp3 compression).
Also, most people listen to music on low-quality devices that can't reproduce high frequencies well. This tends to mask mp3 compression artifacts. Also, workflow and economic limitations make a difference. Not everyone is shooting 70mm or 3-D, even if it is better quality.
2- I would argue that 4 tracks of *well-implemented* 16-bit 48khz analog inputs is a good balance. It would give excellent quality and has excellent workflow. Relevant issues would be:
-how good are the preamps
-how good are the limiters. To my (untrained) ears, the really good limiters (i.e. sonosax mixers, RNC compressor) are pretty transparent for around 15dB (of compression). Very acceptable up to 25dB.
-how easy is the audio interface. Can the sound person monitor off the camera easily. Can he/she easily adjust things like gain on each channel.
-will more audio affect video quality/bitrate.
-is the audio oversampled to reduce aliasing? how good is the anti-alias filter?
Done well, 4 tracks of 16bit 48khz can be pretty good, cheap, and fast.
Where higher quality or more tracks are desired, just go double system. Some double system solutions such as a laptop with Metacorder is very cost-effective, due to economies of scale. Workflow-wise, double system is inferior but the Red team may come up with something smart that works.
What if there was integration between metadata systems (audio) and Redcine- when Redcine ingests the EDL, it can also output an OMF containing synced audio?
Also, when Redcine is outputting the offline cutting copies, it would be nice if it can automatically sync the video to the audio recordings (much like the dailies process with film). One of the problems here is that audio sync is often manually checked when creating film dailies- keykode does not work that well.
I think that's a misinterpretation/misapplication of Nyquist. To avoid aliasing, practical systems have to implement low pass filtering to get rid of high frequencies. The amplitude response for high frequencies is significantly reduced, so performance won't go up all the way to 24khz with perfect amplitude response.
If you don't try to anti-alias, then the system can reproduce 24,000hz cleanly. However, it will have problems with frequencies slightly above 24,000hz- these frequencies will cause a change in the recording. They will cause low frequencies to appear in the recording.
The real world gets a lot more complicated since certain underlying assumptions and conditions for the Nyquist theorem can't be realized in practice. For more detail than you probably want, the Wikipedia article on the Nyquist-Shannon Theorem has some good information.
http://en.wikipedia.org/wiki/Nyquist-Shannon_sampling_theorem
For something easier to understand, just look at how images are resized in Photoshop. When you make an image smaller, Photoshop gives you some different options in how the image is filtered. The nearest neighbour scheme tends to give weirdness (sort of like aliasing) on "high frequency" components like lines and fine repeating detail. The bilinear and bicubic methods don't do this nearly as much.
However, if you make an image 2X and then 0.5X and then 2X and then 0.5X, nearest neighbour gives the best (perfect) results. The bilinear and bicubic methods do not. In sampling theory terms, this is related to "phase shift error". Basically, bilinear filtering results in fine detail getting blurred in different directions. When you want to perform the opposite operation, you can't do this- you don't know which direction the components originally came from.
I hope that is a clear explanation of some of the issues when it comes to sampling. What you also need to take into account is human perception and real-world performance of various systems. There are some things we can't notice and some things we tend not to notice. This allow certain schemes to work very well even if they do not perfectly (mathematically) reproduce a signal (i.e. mp3 compression).
Also, most people listen to music on low-quality devices that can't reproduce high frequencies well. This tends to mask mp3 compression artifacts. Also, workflow and economic limitations make a difference. Not everyone is shooting 70mm or 3-D, even if it is better quality.
2- I would argue that 4 tracks of *well-implemented* 16-bit 48khz analog inputs is a good balance. It would give excellent quality and has excellent workflow. Relevant issues would be:
-how good are the preamps
-how good are the limiters. To my (untrained) ears, the really good limiters (i.e. sonosax mixers, RNC compressor) are pretty transparent for around 15dB (of compression). Very acceptable up to 25dB.
-how easy is the audio interface. Can the sound person monitor off the camera easily. Can he/she easily adjust things like gain on each channel.
-will more audio affect video quality/bitrate.
-is the audio oversampled to reduce aliasing? how good is the anti-alias filter?
Done well, 4 tracks of 16bit 48khz can be pretty good, cheap, and fast.
Where higher quality or more tracks are desired, just go double system. Some double system solutions such as a laptop with Metacorder is very cost-effective, due to economies of scale. Workflow-wise, double system is inferior but the Red team may come up with something smart that works.
What if there was integration between metadata systems (audio) and Redcine- when Redcine ingests the EDL, it can also output an OMF containing synced audio?
Also, when Redcine is outputting the offline cutting copies, it would be nice if it can automatically sync the video to the audio recordings (much like the dailies process with film). One of the problems here is that audio sync is often manually checked when creating film dailies- keykode does not work that well.
Graeme_Nattress
Red Team
"The advantage of 24bits is not just increased dynamic range but the absence of quantization errors," They're the same thing. SNR in a digital sampling system is basically the ratio of quantization error to max signal, which for 16bit is 96db. If the audio you're recording has a dynamic range of < 96db, then you gain no benefit from > 16bit recording. 24bit implies a 144db SNR which cannot be met by your audio amp that connects to your speakers, and given the general noise floor in real world environments, would never, in practical use, be used. 24bit A-to-D's generally struggle to go beyond 20bit performance anyway.
96khz can be useful though, as you can use milder AA filters and hence get less ringing in the audio band.
96khz can be useful though, as you can use milder AA filters and hence get less ringing in the audio band.
visceralpsyche
Well-known member
Given that specs for Red already dictate 24bit/48kHz I'm not worried at all. If it can be pushed to 24bit/96kHz then all the better, but current specs are "good enough" for a cinema camera. And yes, audio is my main forté in cinema! If you want more professional sound then you'll be using an offboard recorder anyway (plus Red has said there will be word clock or better included as well).
filmmaker1977
Carbonite Member
though this is the matter.. to handle or to try 24 bit/96khz on board not from outside..