Studer 1 194 533. 90 degree shifter. LTspice file

[abbey road d enfer]

At a time, there were discussions about using this type of circuit in side-chain of limiters. The idea was that producing two versions of the signal shifted by 90°, one could apply the trigonometric equation sin²x+cos²x=1 to extract the signal's amplitude with an attack time that would perfectly match the frequency. In theory, the attack time would be exactly a 1/4 cycle whatever the frequency.
AFAIK no commercial unit resulted of it.

 

[gyraf]

In digital, it's called a Hilbert detector iirc

An added benefit is the possibility of summed/unitary L/R/M/S sidechain without loosing information

..my G24 works this way, works surprisingly well..

/Jakob E.

 

[abbey road d enfer]

Imy G24 works this way, works surprisingly well..

Does it mean you can dispense with an attack control (for limiting)?

 

[gyraf]

..most probably, yes - although I don't really want to go that way (The G24 is not primarily aimed at limiting, but at compressing/levelling). It behaves very civilized even at quite fast attack settings at low-frequency-dominant material.

/Jakob E.

 

[13engrsapper]

It's hard to see any of the values.

I wrote: "I agree you would never want to process with this and leave it in stereo unless you wanted an effect."

I'll add that I did listen to it after processing in "stereo" and the imaging, as one would expect, was bizarre and pulled hard to one side owing to the 90° phase difference between outputs. (On my schematic I have a mode where I Q forward to the outputs. I put that there for use as a keyboard stereo DI for "monofication" downstream. It's likely a mode that would never be used.)

Having said the above pretty much everything monitored in mono sounded more like it's stereo counterpart, in terms of spectral and instrumental balance, when summed as I+Q instead of conventional L+R. As many of you know conventional Mono, by definition is "lossy" because L-R is eliminated. When summed as I+Q, L+R and L-R have equal weight.

It will destroy a stereo mix if presented in stereo.

Monitoring in mono on some mixes I+Q summation made modest improvement on others it was quite large. I can't honestly say I+Q made anything sound worse when monitored in mono.

I stumbled into a thread over at Steve Hoffman's about people monitoring early stereo recordings (mostly Beatles) in mono. I've thought about doing a listening test there. It's been awhile since I read it but I got the impression that some people felt the mono button made the unlistenable listenable.

If I were producing background music that I know was going to be played at Target on a mono distributed 25 or 70V system I could see this being in the chain. I'm assuming Studer and LAWO designed these things to live full-time in a mono AM airchain.

There are two niche apps that might benefit us here:

Jakob, in an earlier thread, looked at using the 90° filter in a single-detector sidechain to provide equal weighting of Mid and Side. At the time the sims showed they did not have equal weighting so I'm not sure what came of that. I think it might make a good addition to the GSSL sidechain.

I really want to see how this works with problematic stereo keyboards. For this reason it may some day move from Protoboard to PC board.

In any case the 90° filter is a very niche thing.

Turn it into a guitar effect.

 

[abbey road d enfer]

..most probably, yes - although I don't really want to go that way (The G24 is not primarily aimed at limiting, but at compressing/levelling). It behaves very civilized even at quite fast attack settings at low-frequency-dominant material.

/Jakob E.

Jakob, I don't want to make you divulge things you'd rather keep quiet, but:

• Why did you choose to do so? Was it started by curiosity, or had you been tipped and expected a particular quality?
• Please confirm you're using the shifters only in the side-chain?
• Have you compared that with a more conventional detector?

 

[gyraf]

- The G24 is basically a LR/MS compressor in-one: Fully passive, autotransformer step-up plus (fast) opto cells. Selling impedance to get a bit of makeup gain. Floating balanced, meaning we can't just attenuate into ground. So we attenuate between channels in stead - four opto cells in an array between channels, +to-, -to+, +to+ and -to-. This led to the realisazion that the two first of these are in charge of the signal's M component, the two last manages the S component. So it was trivial to implement a pot to dial in wether you want stereo compression (equal M and S, pot centered) or compress more on the M- or S- dimensions. Pretty usable according to test pilots from early on.

Which means that the first version of circuit had to have four sidechains, summed post ratio: L,R,M,S - or it wouldn't play nice in all cases. These four sidechains getting annoyingly complicated to get precise, I was looking into a way of summing in AC already for simplicity - and thereby achieve a single unitary sidechain signal that would describe the envelope of all L,R,M,S at the same time.

The +/-90 degree circuit, I knew, had the property of completely decorrelating incoming signals (when you monitor it on a phasescope it'll be converting any straight line into a perfect circle, that is, spreading incoming signal equally to all four dimensions). So this made me a single signal always describing the loudest part of any of the four implicit "channels" - that I could collapse to a mono sidechain envelope signal without compromising precision too much

The impact on attack timing behavior wasn't anything we aimed for - we were actually pleasantly surprised after we implemented it. At that time I did not realize that this sort of thing was being implemented for envelope detection in some (advanced) DSP systems as "Hilbert detector"

- Yes, sidechain only. Horrible for audio (unless you are taming stereo bass synths).

- The initial conventional detector was just that - conventional, and behaved as such. Our test panel found the fastest timings much more usable/agreeable in the phase-summed-version (at that time we never realized why). The nice fast-behavior later led us to implement a continuous feed-forward/feedback control that also, by it self, had very interesting impact on the "texture" of timings

/Jakob E.

 

[abbey road d enfer]

So this made me a single signal always describing the loudest part of any of the four implicit "channels" - that I could collapse to a mono sidechain envelope signal without compromising precision too much

Thanks for this extensive answer, Jakob.
My simulations (and Wayne Kirkwood's) show there is a 3dB difference between the individual channels and the I/Q outputs. It should not matter much. It's better than +6 or minus infinity...

At that time I did not realize that this sort of thing was being implemented for envelope detection in some (advanced) DSP systems as "Hilbert detector"

Audio is such a niche activity.
The only industrial application I have found for 90° shifters is in PLL's, particularly those used for synchronizing electric generators.

 

[zamproject]

So interesting conversation !

I wonder now if combining a 1/4 wave attack frequency dependant with an 1/2 wave or a full wave all pass in the audio might result in a "perfect" look ahead limiter ?
Long time since I look at some EMT schematic, IIRC there is all pass in the audio, but not sure for sidechain

Best
Zam

 

[user 41968]

I think I have that EMT schematic and will look. There's definitely an allpass in it.

WRT using this 90° filter as a guitar effect I think you'll find it to be incredibly lame.

A dynamically-controlled one would be useful. I experimented with some 4 pole and 8 pole VCA-based phasers here: A VCA-Based Phaser - Pro Audio Design Forum

My intent was to make it signal-controlled to emulate the DynaFlanger I helped bring to market for MicMix back in the late 70s. Rosalfonso Bortoni asked me about the VCA-controlled allpass a couple of weeks ago and we may see it on the THAT pedal page eventually.

 

[abbey road d enfer]

So interesting conversation !

I wonder now if combining a 1/4 wave attack frequency dependant with an 1/2 wave or a full wave all pass in the audio might result in a "perfect" look ahead limiter ?
Long time since I look at some EMT schematic, IIRC there is all pass in the audio, but not sure for sidechain

In the 266, the APF is in the audio, with a delay that matches the attack time. The peak detector is supposed to be super-fast. I tried to sim it, but it looks like the total sim time would be about 100 hours...!

 

[zamproject]

it looks like the total sim time would be about 100 hours...!

There is for sure serious analog mathematics computing in the 266

 

[richiyobs]

here tried to implement Studer shifter to sum 2 channels feedback at Sidechain input in my diode bridge comp prototype
my attempt is to have one control board for both channels
the feedback is balanced so i m summing positive to one side and negative sign to the other...
should it work?
Thanks for your attention
best
richi

 

[abbey road d enfer]

I don't understand what you're trying to do. Post a schemo.

 

[richiyobs]

sorry, i explained myself in a wrong way!
thanks for attention
best

 

[abbey road d enfer]

sorry, i explained myself in a wrong way!

You're summing two versions of the input signals that are shifted by 90°. The result is one signal that stands in the middle, 45° from each other. There's no point in doing this, because it's just a signal identical to the input signal but with a weird phase response.
The trigonometrical formula is sin²x+cos²x=1
That means you hace to square the signals before summing them. IIRC EMT did that with 1496 multipliers.

 

[richiyobs]

thank you for clarify my mistake!
what about running like this...same story?
thanks in advance for your support

 

[abbey road d enfer]

thank you for clarify my mistake!
what about running like this...same story?
thanks in advance for your support

Assuming a big approximation, you are OR'ing the peak values of the two signals.
That has nothing in common with a²+b²
|a| ∧ |b| ≠ (a²+b²)

 

[richiyobs]

not so expert and used to formulas, but i will study this concept soon... hope to find a way to implement a solution for my needs
thanks for your time anyway, i ll go back to a simple Sidechain sum for the moment
best
richi

 

[richiyobs]

https://czh-labs.com/products/audio...MInIm72c3u9gIVFrLVCh3lIwQTEAQYASABEgL6s_D_BwE
Mr. Kahn Symmetra-Peak device