Thank you all very much for your encouraging, friendly comments.
Bill Mueller wrote on Sat, 18 October 2008 16:46 |
Do you have any audio files from these experiments?
|
Hi Bill, Yes I have saved all of the files and all of the original sessions.
However the big line stages test was made in a Logic session with the Pro Tools engine and hardware per AES in and out to and from the Lavry converters. The current versions of Logic and Pro Tools are working together ok but I have the feeling that I should better convert the test into a Pro Tools session because the Logic software doesn't record any more files. As in that experiment no mix engine was involved and all you need to adjust is volume here and there for those units that have no volume control but have fixed level changes like API Eqs it's no problem to play it back with the Pro Tools software. As long as all components during the recording process are absolutely identical it schould be fine to do so. I have never touched the trim pots on my Lavry converters since I have started the test.
Quote: |
Your descriptions of the leftover files make perfect sense to my personal experiences and are exactly what I was asking for and in a small way, expecting.
|
I did some more 0-tests yesterday and the picture is complete when you listen to the differences between the original file and the run through one unit A/B and then listen to the 0-test. I know the files in and out and still it takes some consideration to interpret the leftover.
It is much easier to hear the differences of the DAAD- A/B tests performed with the Lavry converters than with the Pro Tools 192. It helps to not use any volume controller after the Lavry DA but plug it directly into an active monitor with input control as my Avocet was one of the candidates.
However the leftover experiment only worked sample accurate with the PT DAAD.
It is a complex matter of constant changes over time in frequency, amplitude and speed of change but there is a distinctive character that each line stage shows.
The peak level of the leftover is just one static aspect and those peaks can happen at 15 Hhz or 50 Hz. Though it is fun to make it a sport who is closest to 0.
We have a new winner in this ( nonscientific) fun competition:
The Malotki 1:1 transformer with -43 dB is now second.
New No 1 is a Haufe 1:1 transformer with -47,5 dB. I took this tranny from an old Neumann console.
Isn't it funny that an old transformer is so much closer to 0 than the best discrete audio unit? The GML 8200 had a leftover of 33,9 dB.
Some classical units : API 550b, BA 1084, Helios 1976 channel buffer, V72s
The peak level of the leftover of the API and the 1084 are pretty close to each other but the sound of the two is significantly different.
Both have a constant peak at 15 KHz.
Again that is not only caused by what the unit adds but also what it leaves out and possibly phase anomalies.
The API shows a broad leftover that falls from 15 KHz down to 400 Hz. The loudest almost biting audible part of it is in the upper midrange. The amplitude changes are fast and to my understanding that correlates with the typical API- sound as I know it.
Leftover peak -19,5 dB
The BA 1084 shows the same peak at 15 KHz but it isn't as broad ( only down to the upper midrange) and sounds more like treble. However the most sicnificant difference is the loud boomy bass which makes a big part of the level of the leftover. It peaks at about 50 - 200 Hz.
Leftover peak -18 dB
The Helios leftover sounds completely different.
I bet that David and Dan will like what I have found.
The peak level is at -31,8 dB, the same as or even a hair better than the Cranesong Avocet (31,5 dB) close to the GML.
There is no audible midrange leftover but low frequencies at 50 Hz and very high treble , pretty narrow at 15 Khz.
Again that correlates with the typical sound of my console. It has a neutral midrange but adds some low end and silky highs.
Maybe I should add that this is after we have completely rewired all channel in and outputs and inserts of the console which was one bitter consequence of my line-stages test.
The cables that we have removed had the highest capacity of all cables Manfred had ever measured.: 360 pf per meter. The new one measures about 40pf per meter.
That alone made a huge huge difference affecting transparency, dynamic and dry bass. The old highly capacitive PVC cable sounded like a mushy compressor in comparison.
( Manfred suggested to sell it as : " Vintage Warmer Cable")
My beloved original V72s and V74a had the highest leftover:
V72s -15dB
V74a -15.5dB
Typically the V72s leftover had the most energy in the high treble.
Obviously those nonlinear distortion are something our brains like.
( Except for the classical engineers in the test of coarse)
Here is a quote of Paul Wolff about audio design philosophy that I have taken from another thread:
"There are several directions that designs take:
1 Make it as clean and pure as possible because that is their goal.
2 Make it as clean and pure as possible because they aren't at a level to design for a tone.
3 Just make it and see where the sound ends up, could be good, could be bad.
4 Make it have a certain tone because that is the goal.
5 Make crap because you know that someone will only want to spend $10 on a mixer.
The clean direction works provided that you have something that is attractive to the brain going in.
The #3 may work and may not work, depending on what was going in.
The #4 adds that "thing" that the brain likes, so it makes just about everything sound "better".
The #5 makes everything sound like shit, but the owner really doesn't know or care...
Personally, I put very little emphasis on the specs, My stuff was a theory in my head and it came to be. When I listen, I may pick the cheaper mic or the tape with print through (old days) if I liked it better.
There are definitely things that turn on the brain. If you can harness those things, add to it a song that you like and record it in a way that doesn't destroy it, you have something that can make people happy and you money at the same time. Who knows, you might even get laid...
__________________
Paul Wolff "
End of quote.
Transformer and transormerless designs:
The GML, the Avocet and the Helios have in common that they are discrete transformer-less designs and they show pretty similar results.
The API, the 1084 and the V72, V74 ( both tube) are discrete transformer designs.
I wanted to find out whether that alone predicts part of the results.
I combined the two transformers in the test with my Helios channel.
Both of the transformers had proved to be overall the most neutral audio devices in the test.
The combination of the trannys with the Helios stage however raised the level of the leftover of the Helios by
1,2 dB Haufe and 3,6 dB.Malotki.
Or in other words the Helios stage raised the level of the leftover of the Haufe tranny by 16.9dB and the leftover of the Malotki by 14,8 dB.
Haufe single -47,5 dB
Helios -31,8dB
Haufe/Helios -30,6 dB
Malotki single -43 dB
Malotki/Helios -28,2 dB
Different combinations lead to different level changes. That is probably due to nonlinear impedance changes over the frequency range.
However the results are still far away from the API/1084/V72s.
The combination transformer-discrete stage alone does not determine the result.
It's a matter of the individual design and matching of all components.
Quote: |
Now my question is; are the summing differences a matter of amplitude or phase?
|
I can only guess and I believe it's a complex mix of everything.
One example:
Over the time Manfred had built several line stages and summing stages. He copied a typical discrete class A Siemens output stage just to show how it would sound without a transformer and he built a symmetrical version , IC based versions, we experimented with negative feedback with different designs .....
From these experiments I learned that the amount of negative feedback can have a significant effect on the amplitude or at least on what we hear or interpret as dynamic. Most designs added a sort of crack ( nonlinear distortion) in the upper midrange when negative feedback was higher. That distortion can sound like a dynamic treble resonance that adds the same upper midrange or treble to percussive instruments or even to all instruments. Suddenly the attacks of a tambourine , Hi Hat and snare all shift to the same range. But the sound of the mix is more percussive, Snare have more snap and there is an amount of that effect, that gives the impression of more transparency and more dynamic. On the other hand a solo violin or a distorted solo guitar often sounded extremely ugly through those stages to my ears and to the classical engineer involved in the test.
Alone that one artifact caused by negative feedback might be one reason for the impression of or even a measurable higher amplitude that a line stage or summing stage can add to a mix.
Transformers can milden this effect when they compress fast transients and add harmonic distortion in the low frequencies. On the other hand transformers can even enhance that effect when the impedance of the next stage is too high.
The Malotki transformers gets a very nice, silky treble resonance and attack when the following impedance is about 20 KOhm. Most of the listeners found that effect pleasing.
Yesterday I added a little 0-test experiment with negative feedback to illustrate the effect:
I sent the signal through an GML mic pre, added a symmetrical line pad of -15dB with very short ( almost no) cables and set the GMl on the lowest Gain 15dB.
Then I made a second run with another additional pad of -20dB and set the GML gain to 35dB in order to reduce the units negative feedback
and made the 0-Test.
The leftover of the first file was -32,6 dB
The leftover of the second file -33,9 dB
But there was a big sound difference between the two leftover files.
Manfred said that the amount of change in this case - 1,3 dB
( although that means something as it was the same unit with just another gain setting)
was not as important as the content.
The leftover of the second file sounded much milder in the upper midrange, more even and less dynamic. In other words the frequency response of the second rerecorded file with the raised gain and lowered negative feedback soloed was closer to the original, it had a fuller, warmer midrange and more fundamentals.
And the first rerecorded file sounded more jumpy and dynamic and harder.
That's the reason why I like to use a pad of -20dB in front of some mic pres for distorted guitar sounds. It sounds healthier with lowered negative feedback.
All my tests with the line stages included one stage at a time only.
That would be the signal path of a passive summing box like the folcrom.
But a console or an active summing box has line stages in each channel, a console has many of them in each signal path. There each signal gets it's individual treatment with a mix of all of the effects discussed above and then meets the others each with it's own enhanced or diminished new set of harmonics in the mix bus.
In the analogue world this mix of harmonics can happen far above our hearing range and then the inter-modulation ( my private little theory) probably affects the audible frequency response in a dynamic way. Pure, very interesting chaos.
My tests just scratched the surface of that complex matter but I hope it showed that there
must be a significant and measurable change of sound whenever you send individual signals through a mixing board as there is already a significant measurable change even if only one mono mix is recorded through one analogue stage.
But most important!
The 0-test only worked with two files recorded with two passes trough the same converters.
The original files and one DA/AD run through converters showed that there is no way of getting those two nulled.
Midrange quality converters always take away from the sound, detail, room information, frequency response.
My vintage Helios console is much closer to the original sound than one run through
the Digi 192 converters.
And the closest to the original were two transformers, one from the mid 60s and one from the 80s.
Quote: |
One note. You have the groundwork for a scientific paper here. I don't know if you took the extreme precautions and documentation necessary for such, but if so (and especially if you have the files) it would sound to me like you could get a paper published.
|
Thank you Bill,
I have all of the files and all of the sessions but only a few written notes because first it wasn't intended to make it such a big and long experiment.
I can remember every detail and as I have never changed the setup it is pretty easy to recall. But I didn't take the necessary precautions and documentary to make it a scientific work.
Best regards,
Peter