Phase correction circuit. Cheap reverbs sound expensive...

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

barclaycon

Well-known member
Joined
Jan 17, 2007
Messages
696
Location
London, England
In the 80's when I first started working Digital, I went through the same dilemma as everyone else wondering why it sounded edgy and fatiguing.
Now that we have good converters and time alignment it's not such a problem, but in the early days it certainly was.
Having discovered that phase distortion was the primary cause of the edginess, I came across a little module made by Meyer (the monitor people) and I bought one and built it into a box to take round with me. It overcame the problem of badly designed anti-aliasing filters in machines like the Sony 3324 and enabled me to make much better recordings.
I took the module apart some years ago and also checked the Meyer patent. I've tried this circuit on several other digital devices such as cheap reverbs and it really does make them sound more expensive(!). Sounds have more depth and for the want of a better word, more 'warmth'.
Might be anther project to add to the list for people who want to try this approach.
 
[quote author="barclaycon"]Yep. That's the one.[/quote]
Thanks for confirming.

It made me think of 'another' Meyer - in fact he's called Meier:

http://headwize.com/projects/showfile.php?file=meier4_prj.htm

Not sure it's alike, but at first sight it does seem related.
The Meier circuit seems to have collected quite some negative appreciation on various forums, dunno if it was deserved. OK, first some reading.
 
Hi Clint.
That was an interesting link to the Meir box. I think he is trying to adopt a similar approach, but going about it in a less precise way.
The reason I say that is that if you are going to balance out phase response (or do time-alignment), it has to be quite specific for the anti-aliasing filters that you are dealing with. It's not enough just to shift the phase of the signal and hope for the best.
The Meyer module had individual trimmers to correct the phase distortion.
A square wave was injected into the converter, or played back in the case of the D to A converter and the trimmers were tweaked for the best square wave on a scope.
(Ref. Mix Magazine Vol.8 no.5 Ralph Jones).
I found that some of the absolute worst converters and filters were on Pro Digital gear like Sony 1610 and Sony 3324. Yet the best ones were on consumer CD players. Somewhat ironic really!
Having had to re-master a lot of stuff from the early 80's, I discovered that most of the digital masters had been transferred through 1610 converters (with their horrible Soshin or Murata filter blocks). So I found a 1610 and recorded a reference square wave which I then used to line up my Meyer box. It was great to watch the reaction from people when I would flick the switch and suddenly all the 'roundness' came back to their archived master. I did the same with 1630, PCM F1 and a bunch of other popular units. So now I have I library of square wave responses.
It's interesting to use the circuit on cheap reverbs or delay units and hear how good they can be. It's obviously a price thing to skimp on converters and analogue design.
Aphex tried a scaled down approach to this with their SPR (Spectral Phase Response) circuit. Which is just a couple of 'all-pass filters' designed to make the low frequencies lag behind the high frequencies.
 
Interesting stuff. OK, so those two pots in the patent are the adjustment controls for that MS-8201 dual time correction filter module I understand.

But hey, why all this hassle, let's simply use a BBE Sonic Maximizer which magically optimizes everything without any adjustment :twisted:
Just like the Aphex it all seems to target the same things, but the approach of the Meyer-circuit looks good.
And just a few opamps; it's even hard to imagine how Aphex managed to do a 'lite'-version of the Meyer circuit.

Regards,

Peter
 
Well the Aphex circuit is attempting to make the sound 'warmer' by progressively making the lower frequencies later. Sometimes it can sound good, sometimes it doesn't seem like it's doing anything. But it's intended as an effect rather than a solution.
Similarly, the Sonic Maximizer was an attempt to add eq and compensate for the phase shift caused by the eq networks.
Anti-aliasing filters are very severe and in their simplest configuration cause extreme phase distortion. In the article that I mentioned when they measured a typical A to D, I think they said that the signal started to go out of phase from 1Khz onwards and went 180 degrees completely out-of-phase at 17Khz !
The Meyer circuit isn't intended as an effect but more of a neutralizing solution for the extreme time-distortion in these circuits.
When Apogee filters came along the situation improved dramatically, and then with oversampling the need for a brick-wall response so close to audibility was lessened.
 
this looks an interesting circuit to try on an SPX90. I'm not sure how I should determine the values of the components used. What opamps. What values of caps and resistors?
 
I started this thread to gauge any interest in this as a possible project, but there has been very little.
Applying this circuit to units like the SPX90 would yield improvements. I tried it using a Rev 7 and it was quite noticeable how much 'rounder' the reverb was. It also enabled 3 dB more level into the unit without clipping.
In use, one applies a square wave into the effects unit and adjusts the circuit for the best looking waveform (in terms of tilt and overshoot) after the anti-aliasing filter.
With a band-restriction of 20Khz or less there will always be some ringing - a 1Khz square will not have sharp edges (Have a look at the MIX article I mentioned).
In effect you are setting up a kind of pre-emphasis to do the opposite of what the anti-aliasing filter does to the phase response. I believe that the first Apogee filter blocks worked on this principle.
Anyway, like I said - not much interest.
I am going to spend time instead on doing a Dimension D clone.
 
You used the circuit from the patent?

How did you tune the input allpass filter? I didn't see any values specified for the c101 (or what is the cap @ allpass) in the patent schematic nor text. Do you have a working method for tuning the filters? Are the filters necessary when recording @ 88k or 96k?

Another idea which came to mind reading the thread was simply to use 2 allpass filters @ input and tune them to "BBE maximizer" frequencies which should be (from top of my head, could be wrong) 150Hz and 2,4k?

At least I checked with allpass filters @ mentioned frequencies in a plugin eq and it seemed "in the ballpark", it sounded euphonic.
 
Yes it is the circuit from the patent but it was supplied in a potted module.
I was going to 'reverse engineer' it if there was sufficient interest, but
apparently there wasn't!

The filters are tuned by two variable resistors in 'gyrator' type circuits.
The method of lining up the filters is to use a 1kHz square wave.
By monitoring the square wave on a scope after the meyer circuit and the aliasing filter, one can see the distortion of the waveform caused by the sharp rolloff. You then simply adjust the variable resistors to get the best looking squarewave. The low frequency one varies the 'tilt' of the waveform and the high frequency one varies the 'ringing' on the leading edge. No real indication at what frequencies these are - it's just visual.
There is a resulting delay to bring everything back in line, but in the scheme of things this is not important.

This circuit is to overcome the effect of anti-aliasing filters.
The higher the sampling frequency, the less likely it is that the effects of filtering will come into the audio range. Meyer decided to make the patent 'public domain' because A to D and D to A circuits moved on since the early days and so they didn't see a huge market for it.

I plan to experiment with a Lexicon Alex that I have just aquired. In most respect it sounds really good (reverb etc.) but the analogue sections just have simple LC filters with no real correction. So I will see if there is any improvement to be had.
Thanks for your continued interest.
 
I was thinking of using it pre-sound-card.

a) so you mean the "patent" circuit is not the real one? Which is the real circuit? Do you have a schematic? Or do you have to reverse-eingineer it first?

b) for soundcard use it should be simple to "tune" - i.e. with use of some freeware/shareware scope programs and a trusty 555 squarewave generator, right?

c) Is there any reason for using this circuit with 88k or 96k sample rates or was it meant for 44k (and related bandwidth/niquist freq. problems)?

Here's the schematic from the patent:
-------------------------------------------------


Trimpots are there, part values are there, except for the capacitor marked 101 in input allpass filter (or is that 100pF ????).


So: which is witch?
 
Do you have a schematic? Or do you have to reverse-eingineer it first?
I think that the circuit you have there is pretty much it, although I haven't checked it with the module.
for soundcard use it should be simple to "tune" - i.e. with use of some freeware/shareware scope programs and a trusty 555 squarewave generator, right?
Yep. Think so
Is there any reason for using this circuit with 88k or 96k sample rates or was it meant for 44k (and related bandwidth/niquist freq. problems)?
Probably not, as the nyquist frequency is that much higher for 88k, 96k. I think the original intention was to overcome problems with 44k or 32k circuits as the phase/response problems come right into the audio band.
the capacitor marked 101 in input allpass filter?
I believe the value of that cap is 10nF (0.01uf)


Martyn
 
Hey, if you could reverse and draw a for-sure schematic, I´m sure there would be lots of interested people, including me.
 
I think the circuit that tv supplied is pretty much it.
The module itself is encapsulated in epoxy and they even put lead shot in the base of it (!).
I tried chipping away the black epoxy some years ago so I've got a rough idea of what's in there, but to go further will mean destroying the module and then I will have to build a new one - hopefully from the information gathered.
It all takes time!
 
Is there any chance you could:

a) test this device's performance/usability with 88k / 96k samplerate

b) check if the _real_ circuit matches the one from the patent schematic

c) post the actual value of the allpass filter cap?


Thanks!
(I'm particularly curious what it this circuit does with higher samplerates, however I'm a little reluctant to jump ahead and build)
 
[quote author="tv"]however I'm a little reluctant to jump ahead and build[/quote]

This is a straightforward circuit to simulate, you might try that first to see if it does what you hope for.


On the other hand, while I'm quite fond of add-on gizmo's to make older stuff more relevant again, arent't we all using software convolution reverbs these days ? :wink: :cool:

Cheers,

Peter
 
Is there any chance you could:

a) test this device's performance/usability with 88k / 96k samplerate

b) check if the _real_ circuit matches the one from the patent schematic

c) post the actual value of the allpass filter cap?

I don't think this circuit would be very effective with higher sample rates for the simple reason that there may not be a problem there to solve. With 44k (and even with 48k) it's more than likely that the filtering in the analogue circuits will result in phase distortion within the audio range. It's one of the reasons for the harsh 'digital' sound that people used to talk about in the early days. I suggest you read the MIX article that I mentioned earlier in the thread. You can check the phase linearity yourself of your system with the square wave test. (A very useful means of testing audio circuits anyway).

I'm not going to melt and chip away the epoxy on my Meyer module unless there is a likelyhood that I can redeem the time spent and build another PCB to replace it. If there was a demand from more than just a couple of people then it might be worthwhile.

Like I said, the actual value of the cap in the first allpass section is 10nF.

arent't we all using software convolution reverbs these days ?

Er, well. Not exactly. In studios the favourites still tend to be the older units (Lexicon, AMS etc.) and there are quite a number of cheaper units that are really only let down by their poor analogue circuitry.
My idea was to purchase something like the Lex unit I mentioned (£50 from eBay) and modify it to sound like an expensive reverb. (The results from experimenting have been quite promising). It might also be that a processor that you've had sitting on the shelf for a long time (SPX90 ?) would be given a new lease of life.
Ah well...
 
Shit, yeah I overlooked your post on Epooxy.

Well, since you're the one with the "reference unit" I guess that qualifies you and only you to build a clone and compare :)

If nobody else does it, I'll whip together a layout one of these days.
 
[quote author="barclaycon"]
arent't we all using software convolution reverbs these days ?

Er, well. Not exactly. In studios the favourites still tend to be the older units (Lexicon, AMS etc.)[/quote]
I understand, was just more or less joking :wink:

FWIW, also older units there that won't need it... got a nice stereo spring reverb, all analog, no need for phase-addons.


and there are quite a number of cheaper units that are really only let down by their poor analogue circuitry.
I'm actually almost glad my old QuadraVerb+ has half it's display gone, otherwise this circuit would have yeilded me yet another project... :wink:


My idea was to purchase something like the Lex unit I mentioned (£50 from eBay) and modify it to sound like an expensive reverb. (The results from experimenting have been quite promising).
Your mentioning of Lex makes me think of the dig-I/O various of their models have (probably just I or O ? Forgot). That would make the comparison with and without using the cheap codecs already directly possible (assuming your DAW has nice S/PDIF I/O).

It might also be that a processor that you've had sitting on the shelf for a long time (SPX90 ?) would be given a new lease of life.
Ah well...
Making me curious again... :cool: I mean, I could control that complete QV+ by means of a MIDI-editor, no display needed...
 

Latest posts

Back
Top