DIY Phase Meter?

[shrike]

This is my first post to this board, so please be kind if I'm violating customs here.

My experience level: I've done mods, built very simple projects, but am not at all intimidated by electronics provided I've got good resources to work from. I also like knowing /why/ things work, not just how to put them together.

What I'd like to build is a Phase Correlation Meter to operate off the line-level outs from my mixer. I'd prefer it to use an analog-type needle meter, but LED would be fine.

I've been exhaustively searching the web for ANYTHING to point me in the right direction, schematic wise, but so far have found nothing I can use. I've also searched this board and found tantalizing clues but no practical info on this kind of thing. Does anyone have a schematic they could post or email to me? Insight or Advice?

Thanks in advance,
AQL

 

[bcarso]

Welcome to Prodigy-Pro.

It sounds like what you want to do is an analog multiplication on the two signals followed by an averaging operation, that is essentially a lowpass filter. I would use a center-zero analog meter for the display as you will get correlated and anti-correlated outputs with positive and negative d.c. content. However, this will give you something valid only for fixed levels.

You might find it useful to have the instrument be multi-band and select what frequency regions are appropriate. You also might want to have a normalizing function applied, or maybe a compressor, so that the levels wouldn't interact as much with the reading.

Maybe there is material somewhere on the operation of "vector voltmeters", such as used to be manufactured by HP and others.

EDIT: This guy does it with discrete fourier transform techniques and is obviously major overkill: http://www.powertekuk.com/gp102.pdf, but at least is not just an RF instrument like most of the VV's.

Maybe you should describe your application a bit more.

 

[shrike]

Thanks!

What I'd like to have is a box that I can connect to the output of my mixer (before the main amplifier) that'll tell me the phase correlation between L and R channels. It should be operating between 20Hz and 20KHz, as that's the interesting freq range for audio.

If the physical constraints of a phase meter limit it to operating at a very narrow Frequency range (like one only), or at fixed levels, then I'll need to get more information on how phase meters are implemented in existing applications.

In that vein, I've seen meters like this on SSL boards and similar high-end mixers, and it's very useful when mixing. In the implementation I've seen, corellation meters are center-zero, as you suggested.

I hope that's useful detail -

 

[CJ]

My small brain tells me that if both L and R signals were somewhat similar, then a simple ac voltmeter between the hots would do the trick. That would probably require that the L and R grounds be common. Is that normal in most mixers?

 

[SSLtech]

There's two ways of doing it most commonly employed by analog console companies. At home I have a small collection of 'phase meters' from Sony (formerly MCI) that are labeled in such a way that they claim to be able to display the impossible: i.e. one end is +180°, the other end is -180°. (think about that one for a second!)

The most popular way to make things level independant is to amplify, clip, then compare the ratio of TIME that each channel is in the same pole compared to at opposing poles.

The resulting signal should be anywhere between 100% same pole, 100% opposing pole and anywhere in between. You can use two LM391(x) LED drivers for each pole, with green for same pole, red for opposing pole, or you can use a center-zero meter. As I mentioned, I have some very nice center zero SIFAMs for the purpose. I keep meaning to get round to it, but I also have the DK audio phase scope and phase meter, so there's no real impetus...

MCI and Sony did it wrong. Level shifts between the two channels of a 100% in-phase 1kHz sine wave signal will cause the meter do swing one way or the other... Look at how Neve did it, then look at how SSL did it. Two approaches, using mainly CMOS and some simple front-end buffering.

Keith

 

[SSLtech]

[quote author="CJ"]My small brain tells me that if both L and R signals were somewhat similar, then a simple ac voltmeter between the hots would do the trick. That would probably require that the L and R grounds be common. Is that normal in most mixers?[/quote]
Too signal-level dependant. A 100% phase-flipped signal would generate different readings as the signal amplitude changed.

Keith

 

[shrike]

[quote author="SSLtech"]they claim to be able to display the impossible: i.e. one end is +180°, the other end is -180°. (think about that one for a second!)[/quote]
I caught your reference to this in another thread - That's a real trick!

The most popular way to make things level independant is to amplify, clip, then compare the ratio of TIME that each channel is in the same pole compared to at opposing poles.

This is where the real work is done, but I have no idea how to put something like this together. I know I said I like to know how it works, but I'm not nearly facile enough with schematics to draw something like this up. . .

The resulting signal should be anywhere between 100% same pole, 100% opposing pole and anywhere in between. You can use two LM391(x) LED drivers for each pole, with green for same pole, red for opposing pole, or you can use a center-zero meter. [snip] MCI and Sony did it wrong. Level shifts between the two channels of a 100% in-phase 1kHz sine wave signal will cause the meter do swing one way or the other...

This just describes how to display the output from the earlier section, no?

Look at how Neve did it, then look at how SSL did it. Two approaches, using mainly CMOS and some simple front-end buffering.

I couldn't tell from your post: The Neve/SSL method is different from the MCI Sony method; was your amplify->clip->compare description the Neve/SSL version? By "look at how" do you mean the implementation (I've seen both boards in action - that's what I'm looking to duplicate) or the schematic (never seen them, would like to!)

 

[nrgrecording]

phasemeter_compplacement
phasemeter_pcb

but.. warning.. untested :wink: (should work, its from an old german remix mag i only did a new pcb drawing)

 

[bcarso]

[quote author="SSLtech"] ...
The most popular way to make things level independant is to amplify, clip, then compare the ratio of TIME that each channel is in the same pole compared to at opposing poles. ...

Keith[/quote]

Ah yes amplify like crazy and clip---sometimes I forget that method. You would want to have some gating on it though, I would think, so as to not work too hard at low levels.

Once you are in the clipped signal regime probably a simple CMOS exclusive-OR gate could be used as the correlator function.

 

[bcarso]

[quote author="nrgrecording"]phasemeter_compplacement.pdf
phasemeter_pcb.pdf
but.. warning.. untested :wink: (should work, its from an old german remix mag i only did a new pcb drawing)[/quote]

Heh. Saw this after the last post: there's the 4030 exclusive OR gate :green:

 

[nrgrecording]

What are the odds! :green:

 

[SSLtech]

Frank, do you have the schematic? -I think it might be an easy adapt for my center-zero Sifams...

Keith

 

[nacho459]

Yeah, since I worked on a Neve VR a while back I have wanted a phase meter too, and I have some open space in my meter bridge to do it. I don't understand why people didn't build phase meters into their "Big Knobs" and "Central Stations".

 

[alk509]

How about buying an old dual trace scope and hooking it up so that it Lissajous the left and right channels?

Peace,
Al.

 

[PRR]

> I'd like to build.. a Phase Correlation Meter

I don't know what that is. And from Keith's comments, not everybody agrees what it is.

What you want is a 2-D display. A CRT, not a meter. A Vectorscope, except those are most-all aimed at video use, not audio.

Take any 1/4-decent oscilloscope. Even an EICO 427: the stinky gas-sweep does not matter, we won't use it. Feed one channel to the X (normal) input, the other channel to the Y (horizontal) inputs. Trim to about equal gain. Feed the same signal to both inputs, you get a straight line slanted 45 degrees. Put a big rock to the side of the scope and tip it over so the line is vertical. (Actually, I just slant my eyes.)

Now a one-channel vocal potted dead center will give the vertical line. A track panned all the way to one side will give a diagonal line. If you actually have phase-shift between signals, you get ovals. Panpot stereo gives various strong slanted lines with fuzz between. 2-mike stereo gives a lot of ovals, mostly around the vertical line. You learn a lot by looking.

Instead of a big rock or slant-eyes, you can devise a matrix. Actually it is the generic M-S matrix to turn L+R to Sum and Difference.

(You can also open the scope, loosen a clamp, grab the CRT right by the High Voltage terminals, and rotate it. But that fouls up normal use.)

Instead of clipping to CMOS, consider log-amp. It suppresses the amplitude more than the relative phase.

 

[DrFrankencopter]

Probably easier said than done, but if you take an MS matrix of the signal and then rectify the M and S sections then divide the S by the M, then you'd have a measure of the phase correlation (maybe).

You'd have alot of problems with divide by near zero at low signal levels though, and may need to put in some kind of check to ensure that at low signal levels the output is gated.

How do you divide analog signals anyways....there must be a way, analog computers were able to do it, no?

Just a thought....

Cheers,

Kris

 

[chriss]

You also might want to have a normalizing function applied, or maybe a compressor, so that the levels wouldn't interact as much with the reading.

That's what Neumann did in their U479. Here is the datasheet:
http://www.klangdesign.de/pdf/U479.pdf
Too sad they don't have the schematics. Maybe silent:arts has it?

What I'd like to have is a box that I can connect to the output of my mixer (before the main amplifier)

You should think about this. The correlation-meter should be connected to the Master-Tape-Out (in parallel to the tape-machine) that you control with the master fader. The amplifier should be connected to a different output ('Control-Room' or 'CR' or so), that controls the monitor volume after the Master-Fader without influencing the level that goes to tape.

How about buying an old dual trace scope and hooking it up so that it Lissajous the left and right channels?

That could do fine but very many of these beasts do very unpleasing high-freq-noise, that would make mixing a no-fun job.

regards

Chris

 

[shrike]

chriss
Thanks for the heads up regarding the placement of this toy in the signal chain. Right now (but not for much longer) I've got to do the mixing in the box, so my main outs are just monitors. But I will make sure that the signal flow from the board to the tape is unimpeded when I upgrade. . .

nacho459
Exactly! That's why I'm going to (try to) build my own control station to compensate for whatever my board lacks. . . The of-the-shelf solutions are either missing one thing or another.

nrgrecording
Jackpot! from those I can derive a schematic, and thanks to Keith's description of how they work, I can interpret what's doing what. Strangely enough, I was looking at LM391x chips, and saw a LM393 (voltage comparator) and almost asked: "could this do the trick?"

Everyone else - thanks for your input and comments!

 

[SSLtech]

[quote author="chriss"]

How about buying an old dual trace scope and hooking it up so that it Lissajous the left and right channels?

That could do fine but very many of these beasts do very unpleasing high-freq-noise, that would make mixing a no-fun job.[/quote]
... ... ...-really? -I very much doubt it.

Are you certain that you're not getting confused with regular, 'scanning' type CRT's? -I know that with a 'Television' type of CRT, the leakage from the scanning waveform sent to the deflection coils can cause what you're describing, but think about it, with X=L, Y=R (or vice-versa) the 'deflection' signal is the audio signal... there's no fixed HF scanning. Likewise for M/S displays.

As I mentioned earlier, I have a DK audio MSD-500 series meter. It displays L+R vertically, and L-R horizontally. It also employs some serious AGC (I don;t mean limiting, I mean 'upward' gain control as well as 'downward' so that the signal doesn't usually appear as a little dot at the middle of the display. If you put a sine wave at 0VU on both channels, you get a full-height vertical line. If you pull the fader down 6dB, it stays full-height. Id doesn't overload or clip, but it stays full height. -If you put a 500Hz tone on the left and you put a 1mS-delayed 500Hz tone on the right, you get a circle. The circle stays the same diameter as you fade it down to about -20VU, and doesn't clip (turn into a square) until about +20-something VU... That's over 40dB of AGC.

We also have a Woehler CRT display in one of our machine rooms, which produces a L=vertical/r=horizontal display. One of our engineers (an important and senior one) says that the "slant-eyed" way of looking at it is "right" to him, and he can't read the DK Audio M/S display (which to me is the "right" way of doing it: M/S or L+R/L-R) makes hom feel like his head is screwed on sideways... Well the Woehler doesn't have the same AGC and so a -6dB reduction in level makes the area covered by the trace shrink by 75%... I don't like it, since dynamic program is often a lot further below peak than that!

Anyhow, I'm rambling... A correlation meter done the right way is good. The MCI/Sony ones read badly. Even the one on our 9098i is a little suspect.

Keith

 

[shrike]

[quote author="SSLtech"] A correlation meter done the right way is good. The MCI/Sony ones read badly. Even the one on our 9098i is a little suspect.[/quote]

I'm still a little unsure as to the difference between how MCI/Sony implement their PCMs and how Neve/SSL do. . . Could you explain it again for the dense I've read your first reply about 50 times and still can't figure out where you're talking about MCI/Sony and when you're talking about Neve/SSL. . .

-AQL