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

[mediatechnology]

Putting the final touches on the "Quadrature Summing Filter" board.



I eliminated the relay which forwarded I and Q as "stereo" to the final output.
That would almost never be used except for people wanting to do sin cos detection and draw flower petals on the oscilloscope.
I provided the separate I and Q outputs on a header.
The board is 3.7" x 2.75".

[mediatechnology]

This is the current schematic for the "Quadrature Summing FIlter" layout.

I decided to eliminate the ability to route the separate I and Q components to the final outputs to save a relay and board space.
This mode would not be used in most applications.
I and Q now come up on a header where they can be used in a sin/cos output configuration for instrumentation applications.
J1 optionally routes the Left input to both allpass chains for a "one in two out" mode.



The board is 3.7" x 2.75"

[mediatechnology]

Quadrature Summing Filter boards are on order and should be available in about 7-10 days.
I'll be building one to check the layout as soon as they arrive.
The BOM for parts excluding PC board looks to be about $101 US.

[mediatechnology]

While I'm waiting on Quadrature Summing Filter boards to be delivered I fired up the Protoboard to make some sample files.

Some time ago I downloaded a problematic synth sample from Gearspace that was posted in this thread: https://gearspace.com/board/so-much-gea ... oting.html

I hadn't had the opportunity to process that file until now.

The first section is the uber-wide original version which on the vectorscope looks like a big round ball of steel wool.
The second section is conventional (L+R)/2 mono.
The third is I+Q/1.4 mono.

For identical in-polarity inputs the gain is structured so that (L+R)/2 is unity gain and (I+Q)/1.4 is also unity gain.
When the inputs are out-of-polarity the conventional mono output has no output.
The (I+Q)/1.4 output - when the inputs are out-of-polarity is also at unity gain.
Thus on the "I+Q" output in-polarity and out-of-polarity fold down to mono with equal weight.

The first section is the uber-wide original version which on the vectorscope looks like a big round ball of steel wool.
The second section is conventional (L+R)/2 mono.
The third is (I+Q)/1.4 mono.

Quadrature Summing Filter Synth Demo wav file: https://proaudiodesignforum.com/content ... h_Demo.wav


Quadrature Summing Filter Synth Demo

I also tried a fully-uncorrelated noise file: https://proaudiodesignforum.com/content ... e_Demo.wav

The first section is stereo noise.
The second section is conventional (L+R)/2 mono.
The third is (I+Q)/1.4 mono.


Quadrature Summing Filter Uncorrelated Noise Demo

Somehow, as if by "magic" - but actually predictable - the RMS power of the I+Q sum is 3 dB greater than L+R.
The powers of the "stereo" and I+Q segments are statistically identical.

[mediatechnology]

More sound files.

Davie Allan and The Arrows "Shape of Things to Come" originally recorded in LCR. Sequence is Stereo, L+R then I+Q.

https://proaudiodesignforum.com/content ... s_Demo.wav



Once again I+Q has 3 dB more signal power than L+R.

[mediatechnology]

I should be receiving boards Tuesday.

[mediatechnology]

The layout for the Quadrature Summing Filter checks out.


Quadrature Summming Filter PC Board Assembled


Quadrature Summing Filter Frequency Response Red/Blue L+R Green/Pink L-R


Quadrature Summing Phase Response Left Is 0° Reference Right "Q" Output Displayed

[emrr]

Looks great Wayne!

[mediatechnology]

Crest Factor changes using the Quadrature Summing Filter

I've been listening to a lot of different music "mono-ized" by the Quadrature Summing Filter and found that both the power and crest factor are affected depending upon what is in the L-R Side channel.

Two things I've found prior to this post:

1) Unless the source is mono or nearly-so to begin with the I+Q version always sounds more like the stereo version and better than the mono.
2) The I+Q version retains the signal power of the stereo version.

Some of previous tests were with stereo L-C-R hard-panned music from the pre-panpot era.
These elements showed a 0 to +3 dB increase in signal power because L-R has significant signal power and it gets folded back in.

As I was monitoring the 'scope in X-Y and comparing the peak levels of this particular song - Grace Potter and the Nocturnals "Runaway" - I saw huge changes in peak, but not average levels.
This is a fairly typical modern, wide, slightly clipped with a lot of transients in the Side channel.

The order in the following is stereo, L+R and I+Q.



The song was sampled over the entire length with the K-Meter's peak and RMS hold functions engaged.
You can see the RMS levels are about the same with mono just slightly less and I+Q almost exactly equal to stereo.
The peak levels however are greatly exaggerated in the I+Q version with the crest factor going from about 4 dB in stereo and L+R to over 9.5 dB with I+Q.

This suggests that the Side channel has very high crest factor and contains transient information which is lost completely when mono'd and also that Side has very little signal power.

[mediatechnology]

This suggests that the Side channel has very high crest factor and contains transient information which is lost completely when mono'd and also that Side has very little signal power.

I decided to test this theory by feeding a mono version of the song into the inputs.
As it turns out the exaggeration of peaks still occurs.

What I'm apparently observing is the "unwrapping" of clipped peaks in the original recording due to the all-pass filters' phase rotation.
I've seen this effect before in the EEQ.

The increase in signal power is real, the increase in peak level is also real but not from, in the case of clipped source material, lost information.