2019 QQ Test Lab Report - SURROUND MASTER 2

QuadraphonicQuad

Help Support QuadraphonicQuad:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
I am quite impressed with my Surrond Master V2 for stereo and QS processing. It's so good I can be quite satisfied doing a whole listening session without even playing one discrete surround recording. This allows me to concentrate on the music itself. However, I am very curious about shadow vector systems. Besides "shadow vector" being a really cool name, the concept as I understand it has credibility. Is there anyone currently developing with shadow vector technology, or is it just a ghost from the past?
 
One question... could Shadow Vector Soundfield Mapping also be applied to QS?
Looking at the patent (and I really need to dig deeper), I really do not like the use of gain riding, we use more of a "demixing" process. I think pumping may be difficult to overcome. In addition I chose tri band deliberately and I do not think more bands will improve things ......but all power to them I hope I am wrong!!
 
I would like some clarification based on what I’ve read in some other threads:

It seems that the recommended set up for the SMv2 is to set the master volume control as high as possible without “red-lining;” and then set the individual channel volume controls to achieve the desired sound output. (I.e. to match to volume levels of other inputs.) Yes?

If the above is true, what are the noticeable improvements in performance compared to, for example, setting the master volume level at 12 o’clock?
I think that is so that you get the highest level into the ADCs. Thus it gets to use the most number of bits than with a lower input i.e. has a higher bit 'depth' - so get better overall 'accuracy', also helps with the SM2 DSP surround decode.

OK, so I set the master volume control (input) to the highest level and used the channel volume controls to regulate the output into the audio processor. So what audible improvements - when converting stereo sources to Involve 4.0 - should I be hearing vs. setting all volume controls at 12 o'clock? Detail? Channel separation? Ambience? etc? With my setup, nothing really jumps out when sampling several demo tracks that I use with the SMv2.
 
OK, so I set the master volume control (input) to the highest level and used the channel volume controls to regulate the output into the audio processor. So what audible improvements - when converting stereo sources to Involve 4.0 - should I be hearing vs. setting all volume controls at 12 o'clock? Detail? Channel separation? Ambience? etc? With my setup, nothing really jumps out when sampling several demo tracks that I use with the SMv2.
A small reduction of distortion, better dynamic range but you probably wont hear it. Basic idea is to wind up the input till just before the clipping light comes up
 
OK, so I set the master volume control (input) to the highest level and used the channel volume controls to regulate the output into the audio processor. So what audible improvements - when converting stereo sources to Involve 4.0 - should I be hearing vs. setting all volume controls at 12 o'clock? Detail? Channel separation? Ambience? etc? With my setup, nothing really jumps out when sampling several demo tracks that I use with the SMv2.
We do this as it is best engineering practice, regardless of whether the ADC is sampling audio, instrumentation data, video etc., so you get the most change over the maximum number of bits so more 'accurate'. If you have the most number of bits being used you can capture smaller changes in the waveform, so for example if your signal level only change the lowest 6-bits you will only get 64 different level captured in the signal, if you get 12-bits being used, you get 4096 levels captured. So what you get is more 'detail' whether you can hear that is another matter.
 
Variomartix d0es not use VGA' to attenuate or boost selected output. It's more of a de mixer with adjustable parameters
From what I understand is the result is the same. VGA controlled signals are mixed together, the VGA isn't used to pump up the volume of the dominant channel. I also think that if your going multi-band via DSP then the more frequency bands the better. For analogue processing I think that single band is the best.
 
From what I understand is the result is the same. VGA controlled signals are mixed together, the VGA isn't used to pump up the volume of the dominant channel. I also think that if your going multi-band via DSP then the more frequency bands the better. For analogue processing I think that single band is the best.
This is a difficult discussion, it harks back to 2 years of thought crimes I did back in 2008, so I might be a bit fuzzy and my brain is frazzled!

Looking at most SQ systems and indeed the patent on Shadow Vector (all power to it - keep it up guys!!) The VGA's are applied at the outputs, this directly means you need to be very careful how the VGA's are controlled to prevent apparent pumping and image shift. This is true for example if your "logic" implies that the signal is to be "steered" to the rear left then typically you either boost the rear left or attenuate one or some of the others. One of the difficulties of that is if you boost one signal you really must attenuate the others in some manner otherwise you automatically get a "pumping" and a simultaneous image shift.

An advantage of the variomatrix as in QS is that the steering "constants" that are actually variables with the equations automatically do this see saw balancing act of maintaining a constant level output even though a boost is happening on one channel (or a cut).

In regard to multi band, I differ on that to (with the greatest of respect- who knows I could be wrong!!). I spent a great deal of time looking at music under spectrum analyzers and characteristic as far as possible what bands the fundamental and its subsequent harmonics were contained in. To prevent image smear of an instrument - meaning parts of the instrument/ voice being spread between speakers I wanted to keep as many of the harmonics in the same band group. This is to avoid the situation where a dominant primary harmonic of say a midrange instrument shifting some of the upper harmonics of a bass instrument - for example. This causes smear.

Now here is the dilemma, logic says the more bands the better individual control of the steering, but in doing so you enhance the smear. I eventually decided music was best characterised into boom, squawk and twinkle!! You know bass drums, voice/ guitars/ piano's, transient stuff like triangles etc. Many of our users have reviewed how the SM tends to move voices and instruments as a whole without smear- well this is one of the tricks!!!

I have said many times we have resisted a numbers chase but we have always moderated our persuit of big separation numbers in preference to clarity and precision of image without smear or pump as was characterised by many of the earlier systems.
 
This is a difficult discussion, it harks back to 2 years of thought crimes I did back in 2008, so I might be a bit fuzzy and my brain is frazzled!

Looking at most SQ systems and indeed the patent on Shadow Vector (all power to it - keep it up guys!!) The VGA's are applied at the outputs, this directly means you need to be very careful how the VGA's are controlled to prevent apparent pumping and image shift. This is true for example if your "logic" implies that the signal is to be "steered" to the rear left then typically you either boost the rear left or attenuate one or some of the others. One of the difficulties of that is if you boost one signal you really must attenuate the others in some manner otherwise you automatically get a "pumping" and a simultaneous image shift.

An advantage of the variomatrix as in QS is that the steering "constants" that are actually variables with the equations automatically do this see saw balancing act of maintaining a constant level output even though a boost is happening on one channel (or a cut).

In regard to multi band, I differ on that to (with the greatest of respect- who knows I could be wrong!!). I spent a great deal of time looking at music under spectrum analyzers and characteristic as far as possible what bands the fundamental and its subsequent harmonics were contained in. To prevent image smear of an instrument - meaning parts of the instrument/ voice being spread between speakers I wanted to keep as many of the harmonics in the same band group. This is to avoid the situation where a dominant primary harmonic of say a midrange instrument shifting some of the upper harmonics of a bass instrument - for example. This causes smear.

Now here is the dilemma, logic says the more bands the better individual control of the steering, but in doing so you enhance the smear. I eventually decided music was best characterised into boom, squawk and twinkle!! You know bass drums, voice/ guitars/ piano's, transient stuff like triangles etc. Many of our users have reviewed how the SM tends to move voices and instruments as a whole without smear- well this is one of the tricks!!!

I have said many times we have resisted a numbers chase but we have always moderated our persuit of big separation numbers in preference to clarity and precision of image without smear or pump as was characterised by many of the earlier systems.
Without comparing a multi-band decoder to a tri-band decoder (of equal type) we can only speculate. Before doing a repair on my QSD-1 (the mid-band wasn't functioning quite correctly) there was very audible smearing of the image, that's one reason that I prefer single band processing, I've yet to have anyone chime in that they agree with me though. The Photolum single band decoder sounds amazing to me though, much cleaner than the QSD-1. The Involve evaluation board also sounds better than the QSD-1.
This about the use of VGA's from Lynn Olson
"Oh yes, your question. Hmm, not sure. Been ages since I’ve done the math on the VGA section. The goal is for the overall decoder to have exactly the same net gain as a static, unmoving SQ decoder, and for each dynamic channel (vector) to follow a panpot law as it moves around in various directions. The patent could have been a little clearer on that point."
To me that means that the net result would be the same as with "Vario-Matrix", no pumping.
Just as you question the use of VGA's in the Shadow Vector Lynn similarity questions the use of VGA's by the DES team. I don't think that either approach uses VGA's simply to bump up the volume which was the cause of pumping in the old logic decoders.
 
The limitation of single band decoders is they have great issues separating simultaneous continuous tones, even multiband ones struggle if the frequencies are close (even ours). Fortunately music is almost never of this nature. No one publishes this test result but we did years ago, we had a QSD1 on loan with faulty output buffers but it still was interesting as a comparison. See attached. I also attached a few more conventional tests.
 

Attachments

  • Surround Master Involve Proper.pdf
    612 KB · Views: 237
  • QS tests Feb 2013 R3.pdf
    368.6 KB · Views: 235
  • DSP implementation test results V2-converted.pdf
    151.7 KB · Views: 166
Hey guys.... I know what a VCA is, I know that an OTA or FET can function as such. What the heck is a VGA? It just means a computer video output to me.
 
Chucky, It's too bad that the tests weren't done with a properly functioning and adjusted QSD-1. However I know that the Involve produces higher separation values in any case. As you said music does not consist of simultaneous continuous tones, which is in part is why I prefer single band decoding. By my logic If a channel or track is supposed to be in one location but due to tri-band processing it's split by frequency between several output channels, the sound image will be smeared. I've heard this happen with the QSD-1. Also I reason that if more frequency bands are employed any shifting or smearing should be more gradual, thus less noticeable. I often think of attempts to create stereo from mono by feeding the bass to one speaker and the treble to another, not the best method.
 
Back
Top