M670 compressor

[creal]

So - while I have no doubt that these things can improve performance on it's predecessor, it remains very close to it!

Along with the new features and the simplification of the psu design and build, it could well represent a real step forward in buildability and stability. And an incremental improvement in performance, tube and other matching issues allowing.

I would surely see it as a MkII version of the mighty and controversial PM660 (actual might may vary)

Fine, I'll keep this new proto to myself then, since it will always be just "MKII". A variable load DOA sidechain with threshold, which further simplifies the PSU and improves performance and safety, needs one less transformer, has an easy to use tube matching rig incorporated to it. But like you said "remains very close to the original". Since apparently all this design is as simple as "adding a cap" looks the folks here can take over. As a bonus I no longer have to shift through datasheets, calculate bias points and loads or measure a thing.

If only somebody had told me earlier it was this easy!

Very sad to read this news, personally i'll like to try this proto. with your agreement, of course...
Like everybody in this place, i know your works and the different mods you've made before.
Always, pertinent observations, advises and high grade builds.
Don't leave this new varimu as proto, i'm sure many peoples will find this new project interesting and useful in their setup after try it.

Peace,
cyril

 

[telefunk]

About the plate configuration of the PM670: Did you have that DC offset/distortion with different output trafos? That asymmetry seems to be partly the result of certain vari-mu tube/trafo combination. 

 

[lolo-m]

Sorry Kingston but I have a few questions about your design :

First of all I really like the suppression of the pot in the varimu stage for a classic T pad.

About Fart generation : did you modify the plate arrangement to a "a la PRR varimu" before or after removing the 21 position switch in the varimu stage ?

About the cap in the varimu signal path : I'm a bit sceptic especially for ultra fast attack times. A classic well balanced pushpull configuration just kill the frequencies generated by the control voltage as noises are killed by a symetric input. The negative cathode bias exist in a lot more complex way in the Fairchild. Thump isn't killed by the transitorised cathode bias. Thump is killed by well balanced tubes or super slow attack times... A varimu push-pull with a cap in the path ... Well, I don't know...

And don't be stupid, if the specs you posted are real, just remove the schemo, go and talk to Eva Manley and sell it. Their varimu has got a "tube mastering grade noise floor" of -85 dB. Yours claim -106 or so  ...

 

[Kingston]

About the plate configuration of the PM670: Did you have that DC offset/distortion with different output trafos? That asymmetry seems to be partly the result of certain vari-mu tube/trafo combination.

The first question, both actually already answered directly to you, twice now.

http://www.groupdiy.com/index.php?topic=28274.msg653850#msg653850
http://www.groupdiy.com/index.php?topic=51478.msg654282#msg654282

I find it strange you specifically mention wanting to discuss the topic, yet consistently ignore the responses.

Second question with the transformer vs. tube imbalance related to this topic,

http://www.groupdiy.com/index.php?topic=51478.msg654249#msg654249
with a generalised version here, especially by PRR http://www.groupdiy.com/index.php?topic=24961.0

Am I talking to just one person or are there several of you under that "telefunk" name? I've seen this in another forum and it's confusing as hell and makes you question basic intelligence. You mention a "tech". Please let him take over from now on.

 

[emrr]

the very first comments about crossover distortion:

http://www.groupdiy.com/index.php?topic=29056.msg372105#msg372105

 

[Kingston]

What is the software you are using for testing?
Thanks!
Best,
Bruno2000

http://audio.rightmark.org/products/rmaa.shtml

About Fart generation : did you modify the plate arrangement to a "a la PRR varimu" before or after removing the 21 position switch in the varimu stage ?

I've tried both. It makes no difference. Neither does connecting CV to input transformer secondary center tap if there is one. I felt this combination was most convenient with input loading from Lundahl datasheet.

But it does make a difference in sidechain. A 10-50k dual input pot (or a rotary switch) is very sensitive to wiring length and will easily oscillate, hence the grid stoppers which eliminate this completely.

About the cap in the varimu signal path : I'm a bit sceptic especially for ultra fast attack times. A classic well balanced pushpull configuration just kill the frequencies generated by the control voltage as noises are killed by a symetric input. The negative cathode bias exist in a lot more complex way in the Fairchild. Thump isn't killed by the transitorised cathode bias. Thump is killed by well balanced tubes or super slow attack times... A varimu push-pull with a cap in the path ... Well, I don't know...

There's no thump with mismatched tubes. Two triode sections somewhat even this out. Then let the tube current imbalance into the transformer secondary, the lundahl and a cap seems to handle this well. The cap doesn't have an audible effect on the 1uF "attack" setting. I didn't actually think to measure this one. I will find out the hard way with the much more powerful (ie. faster) DOA sidechain which I've yet to measure in real world.

And don't be stupid, if the specs you posted are real, just remove the schemo, go and talk to Eva Manley and sell it. Their varimu has got a "tube mastering grade noise floor" of -85 dB. Yours claim -106 or so  ...

-106dBFS. With the AD/DA I'm using that results to -93dBu. Still pretty good.

As for commercial implementation, the DOA sidechain might have a bright future, which is one of the reasons I'm not posting it here outright. People seem quite content with the existing projects, judging by many of the responses here. The audience for very high performance but cheap mastering compressor is very likely elsewhere.

 

[emrr]

I may have missed some of these points being addressed.  What observations have you about:

use of B+ feed R's at T2/T4 without 4.7 cap, so current balance through transformer in play?  Still takes 99% (guess) of current out of transformer.

match loaded T1 secondary versus more lightly loaded?

use versus non-use of stereo link diode iso?

addition of B+ branch iso caps per channel?

do you find FF of much real use?  I tried it while modifying DaveP's 26C build, and it was extremely unwieldy in action with that unit. 

curious if you hacked enough to try AC heaters, and had any observations.  I realize that thought totally undoes your -12 scheme in the PSU. 

 

[Kingston]

For Gus,

Left channel: 10k load
Right channel: 620ohm load

Use of 10uF cap is a negligible performance improvement. Right channel then has -0.80dB at 10hz.

Gain drops 4dB at 620ohms (compensated in the above images). Some would think that a problem, but there's actually plenty to spare. And no change in THD%+N.

 

[Kingston]

use of B+ feed R's at T2/T4 without 4.7 cap, so current balance through transformer in play?  Still takes 99% (guess) of current out of transformer.

Left channel: 4uf7 OT cap
Right channel: no cap

That bump in the frequency response is an artifact of the strangely dynamic cross-over distortion at low frequencies. Which looks like this at worst:

(yellow waveform with cap, blue with no cap)

This behaviour is consistent between all transformers and tubes (whether matched or not) I've tried which I've specified at length in previous posts. Looks and measures exactly the same with B+ at OT center tap with no possibility for the capacitor of course.

match loaded T1 secondary versus more lightly loaded?

I don't have a good enough square wave generator to reliably capture differences in transient response. When using 10-30k load recommendation from lundahl datasheet there is no difference in frequency response. My FFT measurement caps out at 40khz. Beyond that, who cares.

In short, Lundahl makes crazy good transformers.

use versus non-use of stereo link diode iso?

If there's no diode, the time constant capacitance and resistance is in parallel between channels. I didn't bother with this in my PRR176 variant and the longer time constant can be easily heard in linked mode. In the M670 they stay the same length whether linked or not. I seem to recall that the diode should be picked for the least forward voltage drop, but I didn't actually compare types.

addition of B+ branch iso caps per channel?

The B+ regulator is decent, but not directly from the gods and if there is any possibility of interaction between channels this will take care of it. I didn't actually measure any, so file under force of habit.

do you find FF of much real use?  I tried it while modifying DaveP's 26C build, and it was extremely unwieldy in action with that unit. 

It's unwieldy if wired from before t-pad because you can blast very loud volume into sidechain and have the vari-mu stage at low volume at the same time. There are indeed interesting "inverse" compression effects, stomp box style. But used in moderation it changes the very transparent or even invisible feedback compression into a more aggressive attack. Somewhat reminiscent of 1176 which is strange because that one is feedback design throughout.

curious if you hacked enough to try AC heaters, and had any observations.  I realize that thought totally undoes your -12 scheme in the PSU.

For almost a decade now I've shied away from AC heaters. Few bad experiences from lazy wiring mistakes and I've done DC ever since. And it's nice to have "free" DC heaters from the shared bias supply.

 

[telefunk]

Am I talking to just one person or are there several of you under that "telefunk" name?

No, just me. Sorry about the last one, wrote it in total hurry, left it halfway and should not have send it  :-[. You are absolutely right, you answered this question already several times.

But like I said, this is absolutely not my field and i'm just trying to learn here. Still you don't have to act that rude. If we would be talking my professional field, you surely would not follow me two sentences.

Whatever.

 

[emrr]

use of B+ feed R's at T2/T4 without 4.7 cap, so current balance through transformer in play?  Still takes 99% (guess) of current out of transformer.

Left channel: 4uf7 OT cap
Right channel: no cap

That bump in the frequency response is an artifact of the strangely dynamic cross-over distortion at low frequencies. This behaviour is consistent between all transformers and tubes (whether matched or not) I've tried which I've specified at length in previous posts. Looks and measures exactly the same with B+ at OT center tap with no possibility for the capacitor of course.

Yeah, that looks better.  Addition of cap is also viewed as plate to plate AC coupled feedback, as opposed to DC coupled feedback.  Perhaps there's a feedback scheme across the original T2/T4 plans that would provide a parallel answer, and simple drop-in upgrade to the original.

Here's another one I maybe missed:  Obvious improvement with cap at T4 too?  Matters much there?  I do recall noticing the output amp distortion point in the original design was a limiting factor.  Maybe better with R feed, but maybe C addition makes little difference there?  If so also an easier retrofit.  Apologies if mentioned and braincells fail to recall. 

match loaded T1 secondary versus more lightly loaded?

I don't have a good enough square wave generator to reliably capture differences in transient response. When using 10-30k load recommendation from lundahl datasheet there is no difference in frequency response.

I'm less worried about transient response than source loading.  Sounds like you have indirectly seen no real change in time constant behavior either.  If true, I'd be inclined to raise those values just to lighten the loading.

use versus non-use of stereo link diode iso?

If there's no diode, the time constant capacitance and resistance is in parallel between channels. I didn't bother with this in my PRR176 variant and the longer time constant can be easily heard in linked mode. In the M670 they stay the same length whether linked or not. I seem to recall that the diode should be picked for the least forward voltage drop, but I didn't actually compare types.

Right, as I'd said in pimp/mod.  Someone else in the other thread claimed there was no difference in the PM670 whether isolated or not.  I was dubious.  I used schottky diodes in some early NBC AGC's I linked for stereo, and they have worked fine with no change in constants. 

do you find FF of much real use?  I tried it while modifying DaveP's 26C build, and it was extremely unwieldy in action with that unit. 

It's unwieldy if wired from before t-pad because you can blast very loud volume into sidechain and have the vari-mu stage at low volume at the same time. There are indeed interesting "inverse" compression effects, stomp box style. But used in moderation it changes the very transparent or even invisible feedback compression into a more aggressive attack. Somewhat reminiscent of 1176 which is strange because that one is feedback design throughout.

It was almost entirely inverse in that one design, so I ignored it there.  It would have required special stepped attenuator scaling, by the feel of it.

 

[Kingston]

Addition of cap is also viewed as plate to plate AC coupled feedback, as opposed to DC coupled feedback.

In a way yes, and tied to triode current mismatch (if any!). I think this is the main mechanism for reduction of cross-over distortion, but the math is unknown to me.

Here's another one I maybe missed:  Obvious improvement with cap at T4 too?  Matters much there?  I do recall noticing the output amp distortion point in the original design was a limiting factor.  Maybe better with R feed, but maybe C addition makes little difference there?

Sorry, yeah you asked this already.

I didn't detect the strange cross-over behaviour at T4, so this combo is really a safety precaution with healthy paranoia. It's been a while since I measured sidechain in isolation and I don't have the data saved, but THD+N was way below 1% and since headroom is quite a bit above the vari-mu stage, clipping is a non-issue. In the end, it's just a really fancy battery charger.

I'm less worried about transient response than source loading.  Sounds like you have indirectly seen no real change in time constant behavior either.  If true, I'd be inclined to raise those values just to lighten the loading.

Nice one. Ear-marked for more testing.

It [feed-forward] was almost entirely inverse in that one design, so I ignored it there.  It would have required special stepped attenuator scaling, by the feel of it.

Another somewhat unstable but interesting combo is balanced input tied directly into output, a feedback/feedforward "mixer" of sorts. It's less unpredictable than plain-feedforward, but completely tied to the output impedance of the unit driving this, and the input impedance of the unit receiving the compressor. Sacrilege really (Y-cable connected wrong!), but works strangely well.

Still you don't have to act that rude. If we would be talking my professional field, you surely would not follow me two sentences.

While electronics isn't my profession, just a passionate hobby, you try to keep cool if I show up at your gig consistently shouting "Derivative! That's not even your poem!" and "My daughter plays solea better and she is four!".  ;D

 

[emrr]

Crossover distortion, looking back at bluebird's comments, seems to be a them of that tube type. 

I'd investigate whether SC amp headroom suffers from R feed versus L feed, if no other significant differences noted.  If differences noted, juggle the apples and oranges towards the better evil.    I think I posted an estimate of original SC amp distortion point, but it wasn't terribly precise.  Amazing that the original designers couldn't be bothered to post SC amp gain figures to help with troubleshooting, I think I hooked up the damn boards to deliver that info before anyone else did, and I still don't have a finished build. 

Maybe for original builds there are simple add-on mods for T2 and T4 that will reduce GR amp crossover distortion and improve side chain headroom/distortion.  Those will require more work, and really should go in the pimp/mod thread if they come to light. 

 

[telefunk]

While electronics isn't my profession, just a passionate hobby, you try to keep cool if I show up at your gig consistently shouting "Derivative! That's not even your poem!" and "My daughter plays solea better and she is four!".  ;D

You're welcome to try me! I've seen worse and never lose my cool. I don't doubt your daughters abilities to play soleá but better is subjective in my field, although Soleá compás needs to be in right measure, otherwise it'll sound like Miles's Scetches of Spain (Soleá but not really, beautiful anyway). ;D peace, and respect.

EDIT: maybe PM670 was like Scetches of Spain...beautiful in it's own way but didn't really deserve that number 670 in the end)

 

[alexc]

Fine, I'll keep this new proto to myself then, since it will always be just "MKII".

If only somebody had told me earlier it was this easy!

What you plan to do is entirely your affair - I was merely commenting on what you posted, which was a schematic.

Good luck with whatever you do or don;t do.

CHeers

 

[Kingston]

I'd investigate whether SC amp headroom suffers from R feed versus L feed, if no other significant differences noted.

This one I've investigated in detail, just not in this particular project. Let's assume B+ at transformer center tap delivers highest possible headroom, measuring 140VDC directly at the plates, both sides of class A/B device. Now configure a sensible value R feed at the same B+ voltage, let's say a common 5-10k per plate. Plate voltage at both sides drops 30-40V to maybe 100VDC. Headroom doesn't drop quite as much since the plates are now working against the resistance. Need to raise B+ roughly the amount of that voltage drop to deliver equal THD figure.

This seems to work across projects, which is partly why the PSU of this project also allows experimentation.

[edit]

oh and if we are being pedantic, the shape of the grid-voltage/plate-current curve should be a bit different between R feed and L feed since plate voltage is moving a whole lot more against the resistance. In practise I haven't heard this difference while compressing. Not quite sure how to measure or trace this difference and it hasn't really bothered me.

 

[emrr]

Yeah, on paper L feed is better, but here T4 is under rated to the point that it may make little difference in this case.  It may be happier without this much DC through it and distortion may improve more greatly from this than corresponding loss from R feed, even with same B+. 

 

[3nity]

I can be a beta tester if desired man!
I was looking for a tube compressor anyways!

It all looks like its the right project right now!

 

[radiance]

Go Kingston! ...let's do this point to point...

 

[bruno2000]

What is the software you are using for testing?
Thanks!
Best,
Bruno2000

http://audio.rightmark.org/products/rmaa.shtml

What "soundcard" are you using?
I agree, point to point.
Thanks!
Best,
Bruno2000