The Studer "90°" Dome Filter Stereo to Mono Quadrature Summer

[mediatechnology]

I ran a comparison of the Quadrature 90 Degree Filter versus Stereo and conventional L+R Mono.

The waveforms tell a large part of the story.

The first cut is conventional stereo, the second is L+R and the third the "quadrature" sum. (Right-click to enlarge)


Supertramp Breakfast In America Stereo, L+R Mono, I+Q Mono

The levels were adjusted for 0 dB in 0 dB out using tone in all three cases.
The L+R level is attenuated by 6 dB; the I+Q attenuated by 3 dB so that they're all the same.
The level differences shown above are how they appear and sum naturally.
The I+Q sum has greater peak level and RMS power than the L+R because it has Side information folded in.

Here is the soundfile with all three cuts in series.

Supertramp "Breakfast in America," Stereo, L+R Mono, I+Q Mono: https://proaudiodesignforum.com/content ... Q_Mono.mp3

The statistics show both a power increase and a peak level increase when the "lost" Side information is recovered.

L+R Mono statistics:


I+Q Mono statistics:


The I+Q sum has about 3 dB more peak level and is about 1 dB hotter in perceived loudness.
The I+Q sum has the balance and "life" of the stereo original - the L+R mono is more restrained because it lacks all of the L-R information.

[mediatechnology]

I decided to run a fun little experiment with tone to demonstrate how elements can disappear when mono'd.

I used Cool Edit to lay down 440 Hz tone at -6 dBFS in mono.
I created a second file using 220 Hz tone, also at -6 dBFS and feeding both channels, but with the relative polarity inverted in the right channel.
I then summed them.

The resulting file was processed by the Quadrature summing filter.
The first section is the unaltered stereo file followed by the I+Q sum, the L+R sum, etc.
The L+R sum can be easily identified - it's missing the 220 Hz tone to to phase cancellation.

Disappearing "out-of-polarity" 220 Hz tone: https://proaudiodesignforum.com/content ... _IQ_LR.mp3

[FrankLacy]

Would you consider processing some song, recording the normal and quadrature mono sums, but then only publicly naming the files A and B (randomly selected) ?

[mediatechnology]

Sounds like a plan Frank.

I have a transfer of Paperback Writer I recorded today...

File A: https://proaudiodesignforum.com/content ... File_A.mp3

File B: https://proaudiodesignforum.com/content ... File_B.mp3

[jaddie]

Just found this thread.

The wide-band phase shift network show up in several other interesting applications. In the original Dolby Stereo matrix 4:2:4 surround system, the goal was to produce an LtRt track where L+R (-3dB) is Center, L-R (-3dB) is surround, there is a problem panning from Center to Surround which would null out one channel during the pan move. The problem was solved by using the same wideband phase shift networks appled to the signals creating L-R, where one (L) is shifted +90 degrees, and (R) is shifted -90 degrees, so the null you'd normally run into with that pan move was alleviated, but L+R and L-R decoding still worked. The phase shift networks were in the original Dolby Stereo surround ecoder. Third-party work-arounds used time delay rather than 90 degree networks to aleviate the null, but to the detriment of the total system.

Even before Dolby Stereo, both SQ and QS 4:2:4 Quadraphonic system encoders used 90 degree networks for rear channels for much the same reason.

[mediatechnology]

You might be interested in this bit of documentation: https://proaudiodesignforum.com/forum/p ... 427&p=4861

At Dallas Sound Lab we had a Dolby Surround decoder on loan and I recall (perhaps not correctly) pulling the board out and seeing the Sansui QS chip set in it.

[jaddie]

IIR, the QS chips may have been used to create the matrix, which was slightly different in the Dolby matrix vs stock QS.

The second version of the Dolby Cat. 150 decoder card used the Tate/Exar chip set to hype adjacent channel separation, which would have been only 3dB without steering logic. These were developed and used in quad decoders first. There were 3 chips, a detector and two multipliers. The time constants were chosen to be “syllabic”, so just a bit slow, but also not easily “falsed.” However, if you were mixing for the Cat. 150 decoder, you would sometimes have to place a low level transient in advance of the desired surround effect to open up the logic and avoid incorrect steering. The steering logic misfire and need for pre-transient was nicknamed a “Todd-ism”, with full respect and admiration for Craig Todd, who piloted the use of the Tate steering logic in the Cat. 150, and now holds the position of Chief Technology Officer and Senior VP of Dolby Labs.

Later versions of the Cat. 150 (150E) used Sanyo ProLogic chips to drive VCAs which improved many performance aspects. The entire process is DSP-based now of course.

[fredbloggstwo]

Hi Wayne,

did you ever make up a PCB for this filter?

Kind regards

Mike

[mediatechnology]

Hi Wayne,

did you ever make up a PCB for this filter?

Kind regards

Mike

Thanks for asking.
I've been waiting for someone to ask.
I'll bump it up the list - I want a 90 degree filter and I get an empty Protoboard back by making a board.
Will try to fit it in a small format board.

Curious: Will you need I and Q outputs without mono summation? That would appear useful in sidechains and level detectors.

[fredbloggstwo]

Hi again,

the version I was thinking of was the 'Recorealator' version that you posted on Mon Aug 02, 2021 at 2:58.

Cheers

Mike