192 & 2" Transfer Test Ideas aka: Return Of The PT Jedi

[waterboy]

DC wrote:
"The first question of the day is:
What is a comparison format we can all agree on?"

FWIW, I think it should be a 16bit format that can be distributed and played back with PT entirely out of the loop, for those of us to whom PT is our only or primary means of listening to 24bit audo.

I don't claim to have heard it myself, but enough credible witnesses have made the asertion that running audio through the system even at unity gain "changes" the sound that this should be an elemant of error removed from the equation.

Many thanks for commiting the time and resources to conduct this test, and to Digi for allowing us to discuss it frankly on this forum.

[F Umminger]

BushPig-

If understand this test correctly, you are comparing a straight analog signal to a signal that has done a round trip through the analog-to-digital and digital-to-analog converters. That is not a valid test of bit-transparency for the AES/EBU i/o; what you are actually testing is the quality of the converters, which are not even made by us. Or am I missing something? You might also be hearing clock jitter; you don't mention what the source of the clock is.

Even if you were attempting to measure the quality of the converters (which is certainly a valid thing to test for, and this would be a decent test for it if done blindly), you had better be sure the converters are trimmed precisely to unity gain. A small gain difference will poison any A/B listening test.

In any case, if you want to test bit-transparency of the AES/EBU i/o there is no reason at all to introduce any of the thorny problems of subjective listening tests. This is an absolutely 100% objective question of whether bits going in are the same as bits going out. You should be able to just loop the AES/EBU output of Pro Tools back into the input, load up a file, and record it through that chain. Then compensate for the delay and null the two files.

I just did this by putting the Signal Generator plugin (set to full scale white noise) on a track that was sent to a mono bus. This bus fed into two aux tracks, one of which went to AES/EBU out, and one of which passed through a TimeAdjuster and then to the main output. Another aux track received AES/EBU input and sent it to the main output. All faders were set at unity and the TimeAdjuster was set to invert. By setting the TimeAdjuster to have a 17 sample delay I got complete bit-for-bit cancellation at the output as verified by a bitscope.

You can't really do this test with the analog converters because the non-integer sample delays introduced by the A to D to A process will always interfere with the cancellation. You would need a version of TimeAdjuster that could do fractional-sample delays, and it would have to be of very high quality or you would be seeing it's flaws rather than the flaws of the converters.

[johnnyv]

Dave...

First off, thanks for undertaking this test on everyone's behalf. Secondly, may I put forth that expecting listeners to listen on CD's is extremely impractical for critical A/B comparisons. First off, to do this would require 2 identical CD players. Assuming that someone has 2 identical players, next we'd be assuming that, if listening through the D/A's on each respective CD player, that both D/A's could be counted on to be 100% accurate to each other (an assumption we have to make on any format). If you're hoping people will have 2 identical CD players and should use the digital outputs, then you're expecting someone to have a D/A convertor with 2 separate inputs. In most cases, it will be highly unlikely that people can accommodate this.

May I suggest that the playback files be played back on as many DAW's as we can find. Obviously, PT and Radar would be the places to start. Both of these formats have the ability to lock the soundfiles together time-wise and hopefully we can make the assumption that the D/A convertors from channel to channel are pretty accurately similar. Perhaps some of us can fly the CD tracks onto tracks of a 3348 for comparison?

I still question the sentiment that NOTHING can be learned by listening back on one of the formats in question. Instead, I'm hoping that some will listen on PT and voice their opinions, some on Radar, some on 3348, etc and that collectively from all these opinions based on listening over different systems, some generalizations could be made.

However, for comparison's sake, the CD idea seems to me to the most impractical.

I will look forward to the files. You mention mic pres, etc. May I ask what you are recording...and where and when (if you care to share the info?). Good luck...

J

[Rail Jon Rogut]

Because the RADAR does not allow you to trim the inputs for calibration - I would suggest you do the following before any transfer:

Send a 1kHz tone into the RADAR, set it's software to -18dB ref. - take the digital out of the RADAR into the 192 and create 24 aux tracks in PT. Assign the aux inputs to 1-24 and place PT in cal mode.

See what levels you're getting into PT - they should be -18.0 dB ... but if they're not, write down what level is coming into each aux input - use that level when aligning the A/D's on the 192.

Similarly check output at the console back from the RADAR and the 192 to ensure they're consistent.... I wouldn't use the console meters (they're not going to be reliable on an 8078)... you need a high resolution digital meter.... differences of .1dB would invite skepticism.

I would probably feed the analog output of the RADAR to the analog inputs of the 192 (after it's aligned as above) and check where it reads in cal mode - then feed the outputs of the 192 back into it's inputs and align the D/A's the same level.

When are you guys doing this test -- if I'm not busy in the studio -- I could come as an independent 3rd party to oversee the technical aspects for you.

I am working down the road from where MM is anyhow.

BTW burning the mixes down to an audio CD wouldn't be my choice -- that would allow for error correction, etc to be introduced ... the best bet would be to record the final mix to a digital file via the dB converters - not a DAT machine -- why not find another product other than PT if you're not happy with recording your 2 mix digitally into that. How about Sonic - then you could just run over to Dave Collins' place with the files.

Lastly - often times when playing the same 2" back there may be alignment differences when playing back - tape travel comes into play... and not knowing how good these 2" Studers are... you may want to also do a test where you record to both the RADAR and the 192 simultaneously.

Rail

[Bushpig]

Fred,

Hi mate. Thanks for the reply. No I'm not comparing analogue to digital. I'm not sure if you know the Apogee unit I use, but I realise I should've worded my post a little better. Basically you can select the source of the D/A Input, and along with all the external input sources you can also select the A/D's digital output as a listening source to avoid having to have your 2 track recorder on input all day to hear your mix throught the conversion process. I pick this as a simple test, as you can then feed whatever signal type of your choice to the A/D for litening and merely switch quickly and easily between the direct digital feed produced by the A/D, and the very same signal looped through ProTools.

If you check the other thread "192 Bass-Lite", a chap called Blairl referred me to an article, written by a guy who's name I can't remember, which talks at length and in good layman terms about the effect of jitter on throughput monitoring but not necessarily on the playback of a recorded digital signal. Is this type of thing your area of expertise?? He points out that it is therefore possible that the numbers coming back out of a given system may very well be "readable" and identical to what went in, but that the ultimate result of jitter can still affect the sound and only ears can ever tell us that. Is he right??

Once again, thanks for the reply.

Steve Bush

[Bushpig]

Dave Carlock,

I agree with Rail. Correct calibration will be all critical I feel.

Also, as far as the results go, I don't know how or if you plan to label which mix is which once you send them out, but may I suggest you tell us which one is the 2" direct pass and then let us pick which of the two digital formats resembles the original the closest?? This is my prime concern. I just want my digital recorder to faithfully reproduce the original no questions asked, and I think the test you're proposing is a good one and mimics a proper real world session similar to what I do a lot of. I would just rather asnswer the question "does washing powder A sound like the original, or does washing powder B sound closer" as opposed to trying make everyone "guess" which is which format.

Just my 2 thruppence ha'penny!!

Good luck with it. Is anyone from Digi gonna be there??

Cheers

Steve Bush

[DaveCarlock]

OK, so as JohnnyV suggested, it would be better to toggle between two playback sources as opposed to skipping between two indexed CD tracks.

Let's open up the floor for this question:

Rail suggested saving the dB Technologies output as a file format, not to tape or CDR. Anyone know what file formats the dB converters can create? Any dB Technologies experts out there?

Let's continue to work this all the way through to the end listener--the toughest part of the test itself. If we save the tests as files, do we need to convert down to 16 bit?

I think this idea is going to lead us back to people listening through their PT systems. My concern is that a lot of PT critics will harrumph the results on principle if people are listening through PT. Let's avoid an argument and find a different solution. My goal is to do this one time in such a way that no one gets their panties in a bunch.

Doing this on CDR will allow most everyone to hear what happened, but Rail is right. The method of listening would be indexing between mixes--not as elegant as two CDR sources being toggled. Some people can hear stop & start sources quite well but I've discovered that some people even with very good ears can't. It's just a different listening skill.

Give me more input!!

DC

[DaveCarlock]

<<I agree with Rail. Correct calibration will be all critical I feel.>>

Definitely--any info like this about the RADAR is greatly appreciated.

<<...may I suggest you tell us which one is the 2" direct pass and then let us pick which of the two digital formats resembles the original the closest??>>

Exactly what I planned to do. There's been enough guess work batted around on this subject already. At this stage, we need answers.

<<I think the test you're proposing is a good one and mimics a proper real world session similar to what I do a lot of.>>

Thanks a lot, but really 99% of this is Mixerman's test. He pretty much nailed the procedure first time out. He's a smart guy and he has a good sense of humor, it's just some people don't get it.

Really, I'm just collecting the best ideas from everyone here, sorting through them and attempting to steer this topic away from the political slugfest it had become and toward progress.

<<Good luck with it. Is anyone from Digi gonna be there??>>

At this stage, no. I've thought about it and I think the independent nature of the test is really important. When Consumer Reports tests a product, I don't think they allow the manufacturer to be in attendance either.

As well, it's a bit of a p### off that I have to put so much time into this and have to pay for expenses and ask for favors to find the answer to a 300+ post issue that Digi should have taken a serious look at already. A large corporation like Avid/Digidesign could easily do this test themselves, they have resources and money. Who am I? "Joe HD-user". That's all. I am beginning to understand Mixerman's earlier statements about having Digi pay him to redo the test. This does take a lot of effort to do properly. But I'm deciding to stay positive and stay focused about getting answers.

Of course, I'll share the results with Digi as there's nothing to hide. There will be sharp technical minds there to make sure we're doing things properly. You know, the kind of guys who know how to land Apollo 13 with three paper clips and a half-torn Burger King crown. [img]images/icons/smile.gif[/img]

My motivation is to confirm that my HD system can deliver what marketing led me to believe it would. If Mixerman has discovered something here, I want Digi to make the changes to provide users with a sonically competitive product, as we all "believe" HD is.

And keep in mind, HD could test out beautifully...

Keep sending ideas! Digidesign Engineering--send your ideas too. Your pre-test technical input is welcomed and appreciated.

DC

[Rail Jon Rogut]

The dBTech has an AES out - you can feed that into any DAW... -- I would simply record it to a pair of tracks via the digital inputs of a separate PT system (could even be a mix system, it'll be 44.1k and clocked off the dBTech converters anyway). The only argument would be that PT can't record an AES data stream reliably... and even MM hasn't said that [img]images/icons/smile.gif[/img]

You can then post the files on the web as WAV files, where the person downloading can decide what to do with them - ie. either burn a CD, transfer them to DAT, play them back via whatever converters they would like from inside whatever DAW they like.

This, IMO, is the only option to have a valid test.

Also before posting them on the web, they have to be binhexed or zipped, so that there is error detection when they are uncompressed. That will ensure that the files are intact and there are no transmission errors introduced in the download process.

BTW, if you want to wait 'til a little later this month - I can get a room with an 8078 and a Studer 800 to do the tests... I just don't have the HD system, the Radar or the dBTech converters

You do have to realize, that unless there is an unbiased 3rd party present -- or at least people from both camps -- you're going to have a hard time convincing anyone of the results.

You did write: "There will be sharp technical minds there to make sure we're doing things properly."

I'd personally like names before I'll accept that... With all due respect, I've heard of technical problems related with the facility where the original test was performed.

Oops... guess that rules me out getting invited [img]images/icons/smile.gif[/img]

I will state though, that unless the alignment procedure is followed out precisely (and documented), and the transfer is done to both medium simultaneously - I personally would have doubts about any results posted.

I do know that the original test was done with multiple passes from the 2" to the different machines - and no exact alignment was done between the two machines to ensure that levels were absolutley equal.... These were just 2 reasons why the original test was invalid IMO.

Best regards.

Rail

[DaveCarlock]

<<The only argument would be that PT can't record an AES data stream reliably... and even MM hasn't said that>>

Actually Bushpig's going off about that right now--check the 300+ thread. [img]images/icons/shocked.gif[/img]

<<You do have to realize, that unless there is an unbiased 3rd party present -- or at least people from both camps -- you're going to have a hard time convincing anyone of the results.>>

We've already got both camps covered--Mixerman and myself. I'm pro-HD as a PT engineer and owner for 6 years and his reputation precedes him as an outspoken Digi critic.

I'm doing this test for myself but I'm opening up the floor in an attempt to make the best decisions in conducting an accepted test. So in the end, I'm not trying to convince anyone of anything, unless we find a problem. Then I'll be trying to convince Digi to get on it. If my test helps us all in someway by exonerating HD or pushing Digi toward fixing the alleged problem, great. But I'm not running for office. I don't need anyone's votes! [img]images/icons/smile.gif[/img]

And aside from all that, show me an unbiased 3rd party in our field on the subject of Pro Tools...

<<I'd personally like names of the sharp techs present--DC before I'll accept that. With all due respect, I've heard of technical problems related with the facility where the original test was performed.>>

Firstly, where was the original test performed? Email me privately if you don't want to announce it on the board.

Secondly, unless they prefer anonymity, I'll let everyone know who the techs are. We've got nothing to hide, but not everyone likes the spotlight.

Really, the focus should be on the test procedure, not the techs. They will basically be there to do the setups according to the mandates of the test so no one can accuse myself or Mixerman of rigging a test. Their work will be checked by MM & me and visa versa. If there is a problem with HD, they will lend their expertise to accurately measure what's going on there.

However, let me know a list of techs you would recommend that would be willing to donate their time. The cast of characters isn't finalized. I'm assuming MM may want a tech he trusts and I will too.

<<I will state though, that unless the alignment procedure is followed out precisely and documented...>>

I don't see a problem there

<<...and the transfer is done to both medium simultaneously - I personally would have doubts about any results posted.>>

This may pose a problem. I see where you're going with this in a perfect test, but by doing a simultaneous transfer:

Wouldn't multing the 2" induce loading problems? This also muddies the real world aspect of the test: 2" directly to the destination with a single snake. Any techs care to jump in here?

Good ideas, Rail.

DC