Diff Pix as promised

[3daudioinc]

Here are the settings from Spectrafoo, just so you know. I was primarily interested in audible differences which is why the parameters are set so high. I only compared mixes that were easily audibly different. The charts below are spectragraph charts of only the difference signal, when compared to Mix file 24. All of the charts below were made with these exact same settings.

NOTE: These screen shots were taken before I discovered the defective plugin that was adding a -132 bump below 96 Hz. Since the screen resolution is much higher than -132, I do not know that it would effect what you see here, but I thought I should mention it anyway.




One thing I noticed about the SSL mix was that it seemed to have more low end, primarily below 80 Hz. There was also a crispy thing going on in the top end. The graph below confirmed both these observations.




The Oxford mix also revealed something very different from the SSL, yet clearly the difference signal is frequency dependent. See below.




Here's the chart on the Dangerous 2 Bus below. Notice that there is almost no difference in the 250 Hz region. Why, you might wonder? I sure do.




And the Pultec mix showed a substantial difference. The comb filtering that is so obvious below (as curved cascading stripes) is a result of slightly different sampling freqencies in the AD and DA.




The difference signal of the Trident shows a different type of sampling frequency discrepancy. If you've listened to the difference signal, you know what I mean. The vertical stripes where the level is the loudest is when the difference signal is the greatest, due to Fs differences.




And lastly, here's the Yamaha 2000 analog mix below.




What do you think of these differences?

I think the difference in the SSL and Oxford mixes are the most interesting. Why is there a difference in the frequency response? And the D2B, why is there such a difference everywhere except in the 250 Hz region? Interesting to say the least.

[ May 07, 2003: Message edited by: 3D Audio Inc. ]

[ May 09, 2003: Message edited by: 3D Audio Inc. ]

[3daudioinc]

Before you ask, the difference signals between most of the DAW samples was much less interesting and revealing, which is why I didn't bother posting it. It also would have required changing the scale by at least 60-80 dB, which I did not want to do. I didn't find any differences that were interesting enough to post.

[Fletcher]

What on earth does that stuff mean? I like the "colors"... but haven't really taken enough drugs in recent years to fully appreciate them.......

[Brian T]

The colors are supposed to flash and accompany music from the SpectraFooFighters. Kind of like the pulsing laser light boxes from the '80s.

Guess you had to be there.


BT

[3daudioinc]

Originally posted by Brian T:
The colors are supposed to flash and accompany music from the SpectraFooFighters. Kind of like the pulsing laser light boxes from the '80s.

Guess you had to be there.

I suspect Fletcher was "there."

For those unfamiliar with "the ways of Foo," allow me to explain.

Think of the sonic palette as a landscape. What you are looking at is like a 3-D topographical map with the colors representing Height. So, the dark blues are -63 and below, the greens are -50 to -33 and yellow to red represent the range above -30 up to 0. So the red represents the greatest value (or highest peak) and the blues are lesser values (or low hills).

Since this is a representation of the difference signal, it's like taking the 3D sonic landscape (topographical map) of one mix (File 24) and laying it on top of the inverted topograpy of another mix. Then the difference signal is represented in colors. So the mixes that are the "least similar" will have the most of the green, yellow and red. Those most similar will have nothing but blue, while those that are incredibly similar (say to -100 or below) would look like a black screen.

What is shows to me is that the Oxford is incredibly similar except for the lowest 3 octaves. How can that be? Is it possible for summing to alter the frequency response of the summed signals? Can you sum 22 identical signals and end up with a signal that has different frequency content than another platform's summing? Evidently so.

The reason I included the SSL plot is that it makes it easy to see why someone would think it sounds different, regardless of the way the pieces fit together. It looks like there is a loudness curve on the mix. Extra stuff below 250 and more between 2K and 16K. Where did that stuff come from? All the input signals were identical. The difference here is evidently at least partly in the conversion, since it's analog.

But the Oxford mix? What's up with that? Is there more low end? Or less low end? Either way, there is a frequency dependent difference between the summing on these two platforms that is NOT subtle. They are nearly identical above 250 Hz. What causes that? Anyone have a clue? I don't.

[Brian T]

I don't know what causes the Oxford/PT differences, but I can spot them audibly 100% of the time.

I know I should probably just let it go, but does this cause anyone besides me to question either the listening skills, technical skills or ????? about the previous tests that cited PT Mix and the Oxford as indistiguishable? Much was made of that at the time, IIRC.

Somebody screwed up. Them or us? (I know what I think.)

Lynn, BTW, can you run the differences on the #18? To my ear, that and #1 are the most pleasing of the DAWs, and in the case of #18 I believe the Peak and RMS levels all match to 1/100 of a dB and it nulls down to -117 IIRC. That would be very interesting to me as a difference picture, since that is the combination of the deepest null with the greatest sonic difference to my ears.


Regards,
Brian T

[3daudioinc]

Originally posted by Brian T:
Lynn, BTW, can you run the differences on the #18? To my ear, that and #1 are the most pleasing of the DAWs, and in the case of #18 I believe the Peak and RMS levels all match to 1/100 of a dB and it nulls down to -117 IIRC. That would be very interesting to me as a difference picture, since that is the combination of the deepest null with the greatest sonic difference to my ears.

I can do that. The scale will have to change though. You also may or may not be aware that there is some digging going on into the differences present in cut 18. The person who did the testing was not able to get a difference signal better than -117. The dealer here in town, suspecting that was incorrect and should be better, contacted me and using the same source files on a different system was able to attain a perfect null. So they are investigating what caused the differences, trying to get to the bottom of this.

So are you hearing something you like that is due to something that is wrong, but your ears thinks it's right? Would you still like 18 better if it cancelled to -132 or lower? I don't know. When I hear back from them, I'll let you know what they discovered.

[Steve Devino]

So this is saying that the mixers you have posted deviate greatest from what was put into them.

I guess its no surprize. it also does not mean anythign in terms of sound quality really. I mean the SSL had a very hyped top and bottom, yet that is the sound the world luvs (like a built in loudness control).

It also means that something tracked on an SSL will need some work to sound right on a flatter DAW and vica versa.

It also looks like to you can get the flatter DAWs to sound awesome if you track with great sounds to begin with.

[Kevin Perry]

Same with the Dangerous 2 bus. What doesn't sound better with a little notch at 250-300?

[3daudioinc]

ATTENTION

Please read my post here about something that I discovered that might make a difference in these plots. It seems there is a defect in a plugin that I used that skewed the comparison difference numbers. But since these plots have a minimum resolution of -80 or so, they may not show any change in the -132 to -144 region.

I think it was Bob Ludwig that said "Never turn your back on digital." That saying is becoming truer every day.

[ May 07, 2003: Message edited by: 3D Audio Inc. ]