HDV audio

[ullanta]

You must be kidding:huh: Forget mics and preamps. Do a 384kbit mpeg yourself from a commercial cd. Do you hear any difference? There have been many studies on the subject of mpeg audio compression. Done by scientists and in very large groups of people and with ABX blind tests. We know 384kbit stereo is extremely close to 16bit quality. It works even in tests performed by audiophiles from various sources (hifi magazines etc). You can't ignore all this and tell me to check my preamps or my ears

Sorry, but what you're saying is sorta meaningless. All those studies proving that 384kbps is "extremely close" to 16-bit quality? What does that mean?

In general, these studies look for a percentage of people that are bothered by a certain level of compression; and in the end marketing decides what percentage of people they can stand to annoy. These tests are also very sensitive to the type of material being played.

Listening to most "pro" audio CDs is not really a valid test. Even though to sensitive ears the differences are still apparent, the lack of dynamic range in current popular audio makes it some of the least suitable material for testing. Film audio typically has a much, much wider dynamic range to deal with, which will stress any compression scheme far more.

As you say, the same engineers have decided that compressed video - e.g., HDV - is suitable for public consumption; why do you assume their standard for audio would be so much higher?

Go out, get some classical music, and do the same A/B tests. The difference can be pretty glaring.

 

[Emanuel]

Well, I didn't know. When I posted there the Robert's info I thought it would be something for a future update...is already there?

Does it mean with the PCM audio we have already uncompressed audio?

So, why the 384Kbits stereo MPEG1 audio stream if there is uncompressed audio capture?

 

[jlames]

OK. I just bought a JVC HD100 and was shooting in HDV. In moving the miniDV to NLE and in editing, I heard no problem areas whatsoever. However, that being said, perhaps being new to audio, I'm missing something. The guys at Coffee Sound have been great in helping me understand and work with audio so the set-up was as follows: 4-camera shoot. Main audio to new JVC HD 100 on Channel 1 w/ separate to Channel 2. 3 new XLR Countryman lavs hardwired to SD 302 and out to JVC. Interior shoot ala Monday Night Football w3 talent doing play-by-play. The audio in NLE sounds fine but, again, I'm new to audio. Am I missing something? If so, what should I be listening for? Your comments will doubtless help an awful lot of indies making the move from SD to HD and HDV in addition to me.

 

[limo991]

We can argue all day long on this. The tests usually include very sensitive material. That's the whole point. We are not talking about full mixes either. And they are blind. Much safer than your tests or my tests. If you start a test by believing that 384kbits/s is inferior and knowing which signal is compressed and which is not, that affects your judgement. 384kbits/s is so proven, that you will not find a higher bit rate on the standard. The tests prove even critical and experienced listeners cannot tell 384kbits mpeg from 16bit uncompressed audio. What is "sorta meaningless" about that?

HDV was a compromise based on an existing medium. The companies would provide a higher rate (40mbit/sec?) if they could.

I will stick to my original point, don't use in camera recording at a camera of this price if the recording is extremely important. Compressed or uncompressed makes no difference to me. My problem is conversion quality not compressed VS uncompressed. I don't believe 384 VS PCM has even a minimal importance in choosing a camera.

 

[TyFord]

In the case of the JVC HD-100 there are 4 audio streams captured. When in HD mode it captures not only the HDV audio but also the uncompressed PCM audio in stereo. Soooo, if there were to be any problems with the HDV audio, the PCM is there for the taking as well.

take care guys...

It was my understanding that the recording was either 16-bit, 48 kHz IN DV MODE, but that in HDV, the audio is 384 kbps MPEG1, layer 2, more commonly referred to as MPEG2.

Regards,

Ty Ford

 

[TyFord]

You must be kidding:huh: Forget mics and preamps. Do a 384kbit mpeg yourself from a commercial cd. Do you hear any difference? There have been many studies on the subject of mpeg audio compression. Done by scientists and in very large groups of people and with ABX blind tests. We know 384kbit stereo is extremely close to 16bit quality. It works even in tests performed by audiophiles from various sources (hifi magazines etc). You can't ignore all this and tell me to check my preamps or my ears

The problem with cameras is the low quality analog to digital comnversion. There is distortion and there is noise. They are 20-40dbs more noisy than a good external recording solution. It is something that we must expect from integrated designs that are just an addon to a video camera. The last thing anyone should worry about is mpeg compression at 384kbits.

Tell you what, You go for it. The facts are irrefutable. 384 is 1/5 the data rate and, for better or worse, I can hear the difference. 16-bit is the LOWEST you should be considering. Feature fiilm audio is regularly 24-bit, 48 khz MINIMUM, and that's just the dialog. EVERYONE knows tha jump from 16-bit to 24-bit is extremely obvious. In going to 24-bit, that ups the data rate another 50% to, what, just over 2 Mbps.

The recorder of choice for movie dialog is the Sound Devices 744T, Zaxcom Deva, and one or two others, all of which record at 24-bit at sample rates up to 192 kHz. HDV audio is like a dixie cup and a string in comparison.

You want to "get by" with your audio, fine. Just don't pee on my foot and tell me it's raining, OK?

Warm regards,

Ty Ford

 

[ullanta]

The tests prove even critical and experienced listeners cannot tell 384kbits mpeg from 16bit uncompressed audio. What is "sorta meaningless" about that?

Please share with us a reference to any such test?

Also, please don't pee on my feet, regardless of what you say about it!

 

[limo991]

Lets not get into a long discussion about 96khz and 192khz vs 48khz. We had enough of this for years in the Mastering WebBoard. For numerous reasons, 96khz does not mean much in real life recording situations. Actually the only 192khz test I'm aware of had mixed results (http://www.dcsltd.co.uk/). Put 24/192khz in a camera that is not good enough for 10bits of dynamic range or decent antialias filtering if you like. Have you heard anything from dCs recording at 16bit/44.1? I can assure you it sounds increddible. Formats are used as marketing tools. Quality rarely relates with the format specs. The first devices that went 24/192 were consumer implementations using dirty cheap noisy and distorted multichannel DAs on a single chip.

You can also play with numbers and refuse to listen. If your calculator tells you 5:1 is that bad how bad is your video compressed? Video already starts with a small fraction of human percection capabilites in all aspects (resolution, dynamic range) even when it is uncompressed. And compresses the hell out of it. You don't want to know how this category is cameras handles video signals in the analog and digital domain. There is practically no filtering. Alias is all over. MPEG audio is not HDV's problem. I would say MPEG video is not HDV's problem either! Tape data rate is HDV's problem. If we had even 50% more data rate for video it would look great.

http://www.heise.de/ct/00/06/092/default.shtml is a test on 128/256kbits mpeg audio. You can search the net for papers on the subject. ABX .edu make nice keywords. The whole format was based on studies by respected academics.

The industry does not focus on one person that hears or thinks he hears a difference. The industry produces products for the masses. Can you tell something is 384kbits mpeg or uncompressed when you hear it on the radio? You can probably easily tell if it is uncompressed video or HDV or SD or whatever. Many people can tell 128kbit recordings just by listening to the high end, some can tell 192khz from audiocd in direct comparisons but at 384kbits that is extremely hard even in direct ABX comparisons. From an engineering point of view that means it is good enough.

If you knew more on digital signal processing, and how it works you would care about mpeg audio artifacts appearing after processing and recompression, not about comparing it with uncompressed on the camera. I don't believe you know how a digital sidechain is implemented or a FIR filter or tube modeling etc. You certainly don't have Matlab running on your pc all the time or do ABX tests every single day. There are many ways mpeg audio can break since it relies on human perception. If you do extreme dynamics it can break since it hits the main concept behind the algorithm. If you code signals with complex high frequency response (a simple heavily distorted guitar will do) and process them through long DSP chains the format is on the edge. Same with cymbals and everything that is overly rich in the treble. I don't see anyone mentioning those. I just hear "I can hear the diffference". I can but not at 384khz. Everybody has different ears

I'm not endorsing mpeg audio in any way I'm saying it does the job when nothing better is available. By better I mean something outside the camera.

Keep up the good work and record audio in any format if you want a specific quality. I just don't want anyone to believe HDV audio is definetely bad. That does not relate to reality IMHO.

 

[stephenlnoe]

It was my understanding that the recording was either 16-bit, 48 kHz IN DV MODE, but that in HDV, the audio is 384 kbps MPEG1, layer 2, more commonly referred to as MPEG2.

Regards,

Ty Ford

Ty,

There is both HDV (mpeg-1 layer 2) and PCM (16bit 48kHz uncompressed) recorded to the tape automatically. It is a "feature" of JVC ProHD.

S.Noe

 

[wabbit]

As the first point, the analog components of these cameras are junk, giving you unwanted distortion and noise. Adding in poor AD only compounds the problem. Now you want to throw an unknown 4:1 compression alogrithm before your final edit?

It's not that it can't work sometimes for some situations; it's do you want to throw out some data and save other before you have decided on the final mix? Depending on the nature of what/who you are recording different data gets thrown out that you have no control over. When you go to edit this, it presents a real problem.

Additionally, considering the piss poor nature of audio in all these cameras, to throw in a generation lost on a 16/48 signal is just tempting fate. You are going to get screwed.

Ultimately it is possible to get useable dialog while compressing the source recording but it is threatening an already fragile audio chain.

So the question is, why would you? The general trend in the last 5 years has been to improve audio in camera (however slightly), HDV audio compression is a STEP BACK...PERIOD.

One more annoucement from the video end of the business that audio is insignificant. To paraphrase Ty, don't piss on us and expect us to not say anything.

Cheers

 

[wabbit]

Actually, Oleg has revealed himself in here and I seem to remember he had done some A/B tests awhile back with the Z1 in DV and HDV mode. I think he was mostly positive about it. Perhaps he will chime in.

Cheers

 

[ullanta]

The industry does not focus on one person that hears or thinks he hears a difference. The industry produces products for the masses. Can you tell something is 384kbits mpeg or uncompressed when you hear it on the radio? You can probably easily tell if it is uncompressed video or HDV or SD or whatever. Many people can tell 128kbit recordings just by listening to the high end, some can tell 192khz from audiocd in direct comparisons but at 384kbits that is extremely hard even in direct ABX comparisons. From an engineering point of view that means it is good enough.

Hey, dude, I get confused by the way you seem to jump around between statements like "tests proved that noone can tell the difference" and the one above.

Like I said, please share ANY test showing such results vs. uncompressed audio (preferably in English, Spanish, or Quechua).

 

[ullanta]

Ty,
There is both HDV (mpeg-1 layer 2) and PCM (16bit 48kHz uncompressed) recorded to the tape automatically. It is a "feature" of JVC ProHD.
S.Noe

Have you experienced this, and/or can you describe what it takes to get at it? I've heard it requires 2 capture passes...?

 

[limo991]

Hey, dude, I get confused by the way you seem to jump around between statements like "tests proved that noone can tell the difference" and the one above.

Like I said, please share ANY test showing such results vs. uncompressed audio (preferably in English, Spanish, or Quechua).

If you actually read the two statements carefully, you might get the point.
Tests have different results depending on the rate. If I have to explain it again, when the rate goes up the percentage of people that can hear the difference gets smaller. At 128 is is large. At 384 it is insignificant. Even in direct AB comparisons.

Language is not a problem. We all use babelfish.altavista.com for any language we are not familiar with. I gave you that reference because I thought it was not scientific in nature and anyone can read it. Sorry it is not in American English. Stereophile did a similar test and it is in english. Look it up. And remember to ask Google "are audio compression standards created by idiots and remove random data or is there a scientific basis based on psychoacoustics?"

I would like to thank Emanuel and stephenlnoe for being adults on this thread which is in a part of the site I will try not to visit again. Thanks

 

[Emanuel]

In the case of the JVC HD-100 there are 4 audio streams captured. When in HD mode it captures not only the HDV audio but also the uncompressed PCM audio in stereo. Soooo, if there were to be any problems with the HDV audio, the PCM is there for the taking as well.

Also synchro with the HD720p image -- right?