Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Peter Weihe]

I had a conversation with the Neumann engineer Martin Schneider about the results of our 0-tests and he sent me a list that I have attached  to this post.

The list shows how much level difference between the reference signal A and the subtracted signal B causes how much level difference ( leftover) in a 0-test.

Martin sat down and wrote the list and I would like to thank him for allowing me to post it on the forum.

We agreed that the performed 0-test can prove that two signals A and B are identical if there is perfect cancellation or if not they are not identical.
That was Bill's original request. The test showed that an analog mix and a digital mix can not be identical.
As the level differences caused by various inserted circuits are frequency dependent
and in some of the units probably also due to phase shift the matter is complex.
The interpretation is tricky.

However the list shows that even with an ideal phase small level differences between signal A and B lead to big differences in the "underworld" - leftover in our 0-test.
It also shows how good some of the inserted devices like the Haufe transformers really are.

An example:
the leftover of the Haufe transformer was -43 dB in my test.
Leaving the aspect of possible phase shift aside and only looking at the level according to the list the level difference caused by the Haufe was at least not greater than 0,06 dB.

A circuit causing a leftover of -32.9 dB in the 0-test caused a level difference of 0,2 dB in real life.

Those units in the test that we love for their sound that caused leftovers of about 18, 19 dB in the "underworld" of the reversed phase-test would have a level difference of about 1dB -1.1 dB according to the list.





Again I made the test with music and of coarse not every frequency was equally represented. The level of the leftovers was always taken at the individual peak, which was dependent on the specific nonlinearity in frequency and dynamic of each circuit.
Another musical material would probably have caused different leftovers.

Regarding the relatively small differences I am amazed that we could hear them by just comparing signal A and B.
What a great tool the ear/brain connection is.

Another interesting result is that even the tiniest differences of 0.0X dB were exactly doubled in the level of the leftover ( +6dB ) when two circuits of the same model were connected in series.
That shows how consistently individual units of the same model show the same character. Ok we  used well respected classical brands only.



Best Peter

[Andy Simpson]

Peter Weihe wrote on Sun, 06 December 2009 21:46

...Regarding the relatively small differences I am amazed that we could hear them by just comparing signal A and B. What a great tool the ear/brain connection is.

Hi Peter,

I would not be quite so quick to assign greatness to the ear/brain in this case.

In my opinion, the results and examination of 'leftovers' from a 'sum/cancellation test' are extremely unreliable if not totally misleading.

For example, if we take a 16bit recording and truncate it to 8bit (without dither) and do the 'inverse sum' test we find the 'leftovers' somewhere around -48dB and below.

Andy

[ericbridenbaker]

If i might put forth some support for the idea that null tests do have their place, to a certain degree of precision.

i'd reason that in the analog domain no two circuits can be made to be "exactly" identical (but very close). in the digital domain hardware can experience both repeated and non repeated errors in the instruction stream.

very interesting how something many decibels below peak can cause things to sound quite different.

[Peter Weihe]

Andy Simpson wrote on Tue, 08 December 2009 15:24

Peter Weihe wrote on Sun, 06 December 2009 21:46 ...Regarding the relatively small differences I am amazed that we could hear them by just comparing signal A and B. What a great tool the ear/brain connection is. Hi Peter, I would not be quite so quick to assign greatness to the ear/brain in this case. In my opinion, the results and examination of 'leftovers' from a 'sum/cancellation test' are extremely unreliable if not totally misleading. Andy

Hi Andy,

thanks for your reply.

I was not referring to the sound of the leftover when I said that I was amazed that we could hear it.
That would be easy to hear for everybody because those sounds are obvious.

When I wrote:

Peter Weihe wrote on Sun, 06 December 2009 21:46

we could hear them by just comparing signal A and B "

I was talking about a simple A-B test. We compared Signal A, the reference signal and Signal B , which was the same signal but running through an extra circuit like a transformer or a line level stage.
Those differences are subtle and some circuits like the Haufe transformers caused a leftover of -43 dB in the 0-Test which was the device with the highest cancellation that we have found so far.

Nevertheless we could clearly hear those differences in a blind A-B test before we even started the 0-test by just switching from signal A to signal B.

Regarding how small the difference between those two signals must be to cause that leftover, yes I find the ear to be a great tool. Sure it can be fooled.

But one of the reasons why we started the whole procedure was, that some of us claimed that that their ears keep telling them, that digital summing sounds different from analog summing.
Well it's not possible to isolate and examine the analog summing process without the use of converters and at least an analog summing stage, so there is no abstract summing process in a vacuum.  Our 0-test that included DA and AD converters and various analog stages and examined them separately in order to see whether or not they  left the signal unchanged so two signals would perfectly cancel out, proved that an analog mix will never be identical to a digital mix.
At least with the circuits we had in our test.

So the question was: Are they identical or not. Answer No.

Andy Simpson wrote on Tue, 08 December 2009 15:24

In my opinion, the results and examination of 'leftovers' from a 'sum/cancellation test' are extremely unreliable if not totally misleading.

They can be misleading , I agree.
And that's why I wrote in my last post:

Peter Weihe wrote on Sun, 06 December 2009 21:46

As the level differences caused by various inserted circuits are frequency dependent and in some of the units probably also due to phase shift the matter is complex. The interpretation is tricky

I would like to add that it is impossible to tell what the inserted circuit exactly alters!
But a 0-test can prove whether two signals are identical in every detail.
And that's what Bills's original request was about.

However I found that certain aspects in the sound of the leftover clearly corresponded to the alterations that the inserted circuit caused as we could hear them in the simple A-B test.

So yes I still find ear -brain connection an amazing and the right equipment to choose my tools for audio.

Best Peter

[Andy Simpson]

Peter Weihe wrote on Tue, 08 December 2009 18:44

Hi Andy, thanks for your reply. I was not referring to the sound of the leftover when I said that I was amazed that we could hear it. That would be easy to hear for everybody because those sounds are obvious. When I wrote: Peter Weihe wrote on Sun, 06 December 2009 21:46 we could hear them by just comparing signal A and B " I was talking about a simple A-B test. We compared Signal A, the reference signal and Signal B , which was the same signal but running through an extra circuit like a transformer or a line level stage. Those differences are subtle and some circuits like the Haufe transformers caused a leftover of -43 dB in the 0-Test which was the device with the highest cancellation that we have found so far. Nevertheless we could clearly hear those differences in a blind A-B test before we even started the 0-test by just switching from signal A to signal B....

Hi Peter,

My point regarding the example of the (nasty) 8bit truncation is that the -48dB distortion products of an 8bit truncation are very far from subtle.

However, as the tabulated input of the Neumann engineer describes, simply changing the gain of one of a pair of identical inverse polarity files will alter the gain of the summed 'leftovers' in apparently drastic fashion.

Not only will very audible distortion (ie. 8bit truncation) lead to 'apparently low' leftover (-48dB) in the ideal cancellation test, but in the case of identical undistorted files, where linear gain differs by only 1dB, the leftover will be around -22dB (files that will sound essentially identical in A/B).

In other words, if the gain is not absolutely equal in the inverse-sum test, looking at the level of the leftovers will not be very useful.

Of course, I would expect the distortion introduced by the devices under test to be repeatably audible but I would not expect much correlation between leftover level and audibility of distortion, since according to the test methodology, an 8bit truncation should be less audible than the other distortion introduced in the test.

That the Haufe transformer achieved the lowest leftovers result most likely indicates that the gain was most closely matched in this case, rather than saying very much about the distortion introduced.

Andy

[johnR]

Quote:

That the Haufe transformer achieved the lowest leftovers result most likely indicates that the gain was most closely matched in this case, rather than saying very much about the distortion introduced.

On a wide band music signal as used in the test, part of the leftover will presumably be caused by deviations in frequency response, and not just non-linear distortion. I don't think that makes the results any less useful however.

[adoucette]

If the budget is too low to have me mix on a proper board (often the case), I mix ITB and give them the option to do final tweaks and a external summing through a large format console. I'll take the finished mix to another studio and run everything into the board, set the faders to 0, pan FX L and R and then make any final tweeks necessary before printing the mixes back into protools. I'm happy with the results, not the same as mixing entirely on the board but certainly alot of the boards character rubs off onto the tracks. I have a few favorites, the most favorite being an 80 input SSL4000G series

[compasspnt]

Andre, how would you charactrise the differences with the "Zero Sum" method through the desk track-for-track v. the fully ITB method?

[adoucette]

compasspnt wrote on Sun, 03 October 2010 08:54

Andre, how would you charactrise the differences with the "Zero Sum" method through the desk track-for-track v. the fully ITB method?

"Magic"

No just kidding. Well not really, it really does sound magical, but I can specify a bit more.

The biggest difference for me is the clarity of the stereo image, more control, smoothness and tightness in the low-end, and a much smoother, shinier and more natural top-end. You can literally hear the analog desk wiping away the digital "pixelation" from the top end of the digital mix. Naturally, when you bring in each track fader to fader you benefit from the components like the transformers to help color the sound in a way that you would come to expect from mixing on a desk. I also take advantage of the VCA groups which colors the sound in an even more pleasing way. This method also allows you to get a chance to make any final revisions in a different light/perspective using a machine that is nothing short of brilliant. Lastly, I am a huge fan of the 4000 series buss compressor so I like to just touch the needle with that on the way out. And if the mix is still a little dark I'll patch in a pair of EQ's to brighten it just a tad. The SSL studio has a nice pair of Lang  PEQ-2's

IMO Its a relatively in-expensive way to bring some of the benefits of a large-format console to your mixes that has a price tag that most groups will OK. Also, I like bringing business to the studios still repping the large consoles, I would like to see them around for many many many many many years to come. I actually may be doing this with some mixes soon, if that happens I will gladly upload the pre-mastered mixes with the ITB mix and the "Zero-sum" approach for people to A/B.