Tube comp/limiter under development

[NewYorkDave]

Larr was right. Look at how the overshoot disappears after careful balancing of the vari-mu stage:

It's fun to slow down the sweep and watch how the symmetry of the envelope changes as you adjust the balance. Matter of fact, this would be a good method for making that adjustment--except it'd damned inconvenient in a studio control room as opposed to a lab bench!

Here's the frequency response, as it stands now:

It's not stellar, but it's not ghastly either. Some tweaks may flatten it out, otherwise I'll have to resort to more-suitable iron.

I think the wiggles in the response at higher GR settings were caused by my using too short of a release time for this test. They'd likely smooth out with a longer release time. I'll try that next chance I get to putz around with the circuit.

 

[Larrchild]

Excellent Dave. Tube matching will keep that consistent thru the GR range.
So it does this properly:

 

[SSLtech]

[quote author="NewYorkDave"]

It's not stellar, but it's not ghastly either. [/quote]
It looks about the same as the Mercury 66 that I measured recently. Same ever-so-slow-rolloff down the audio band below 3kHz... comparable overall total rolloff as well... And if Dave Marquette is putting out a finished product with that amplitude curve, then that response in and of itself is plainly acceptable. -The Mercury sounds good too. -I've had one next to a real original Fairchild and the Fairchild did indeed measure better at the low end, but not by a monsterous amount...

Marvellous result with the overshoot (or -perhaps more accurately- undershoot)! :thumb: :thumb: :thumb:

I'm finishing off my own ÜBER-compressor at the moment... it's s-o-o-o-o close to being done I'm tempted to publish some teaser-pics, but I'm manfully resisting! -There's not a tube in it though...!

Keith

 

[rafafredd]

-There's not a tube in it though...!

Is it the PWM thing?

 

[analag]

Not hijacking your thread, Dave but since the development of your compressor has inspired me to tweak mine a little bit, here is my spectrum analysis. These are strictly Edcor XSM600:10K going in and XSM10K:600 coming out. Parallel 6BZ8 on each side of the differential, no matching was done.
(1) No GR
(2)-3 GR
(3)-10 GR
(4)-10 GR maximum release.
Sounds better than it looks.

analag

 

[rafafredd]

interesting fact is that you have more bass on -10dB GR than no GR, Analag... Would you post more graphs with deaper reduction? Like -20dB or even more?

 

[analag]

[quote author="rafafredd"]interesting fact is that you have more bass on -10dB GR than no GR, Analag... Would you post more graphs with deaper reduction? Like -20dB or even more?[/quote]

Sure not a problem, Fred. More than -96dB SNR too. No overshoot. Been testing with a Tek 464, 456M, 466.

analag

 

[Larrchild]

Been testing with a Tek 464, 456M, 466

Using this tricilloscopic method, did you find the balance pot changed the over/undershoot in yours?

You can see it on the control voltage, clear as day.

 

[analag]

[quote author="Larrchild"]

Been testing with a Tek 464, 456M, 466

Using this tricilloscopic method, did you find the balance pot changed the over/undershoot in yours?

You can see it on the control voltage, clear as day.[/quote]

No balance pot in the thing, man. I told you I found a trick around that. If you embrace a little sandstate ass-istance this tube phenomena can take on a whole new life, bro.
Wanna see some tricilloscope action?

analag

 

[Larrchild]

Yeah, not to mention wanting to rid myself of a balance pot. I've been messing with putting sand in the cathode here for just the same reason. I ain't sceert.

 

[analag]

[quote author="Larrchild"] I ain't sceert.[/quote] Neither am I as you already know. I could debunk a lot of the theories that have been equated about this particular (660) design by some of the "big minds" on this here forum, but to keep the peace I'll keep my mouth shut.

For real aggressive compression I use 6DJ8 or 6N1P which are not remote cutoff tubes. Very nice crush. Aggressive control over the current flowing through the tubes really help them to snap to the ebbing of the gain reduction voltage much better.

No need for a big power amp to attain satisfactory attenuation, not at all, no sir.

Yet I extracted the final bits to this puzzle from this thread of Dave's. And as we can see, his compressor is in development so I look forward to seeing where he will take it.

In threads like these, I am but a student.

analag

 

[NewYorkDave]

I could debunk a lot of the theories that have been equated about this particular (660) design by some of the "big minds" on this here forum, but to keep the peace I'll keep my mouth shut.

Don't worry about peace or politics, man. If you've got the goods, bring 'em. That's what the Drawing Board is all about.

 

[NewYorkDave]

OK, then. Moving right along...

Tube matching will keep that consistent thru the GR range.

I lucked out, since my randomly-selected tube turns out to be pretty well-matched. I can set the balance at a moderate level of GR--say, 6dB--and it'll still be in balance when slammin' with 30dB of GR.

then that response in and of itself is plainly acceptable.

Yes, in this day and age, we've been trained to expect +/-0.5dB from 20 to 20 whether we need it or not. But the threshold of perception for the average listener is 2dB in the mid-band, so few listeners are going to hear if it's down 2dB at 20Hz or 20kHz. It's a reasonable statement that an overall system response that's down 1 or 2dB at the extremes is not going to offend the human ear much, if at all.

On the other hand, the rolloffs of individual pieces of equipment in a system accumulate, so it's a worthwhile goal to aim for the flattest response we can achieve without the sacrifice of other important elements of fidelity. If you cascade three units that are each down 2dB at 20Hz, you've lost three-quarters of your acoustical power at that frequency, which is approaching perceptibility.

I looked at the response of my VCA and output sections separately to get an idea of where the rolloffs were occurring. Here's what I found:

The response of the output sections droops 1-2dB at the upper and lower extremes, and I attribute this to the output transformer (a bit undersized) and the lack of corrective feedback. The next curve shows the response of the VCA section bridged by a high impedance and the third curve shows its (slightly flatter) response with a resistive termination. You can see a rise at 20kHz that's almost a mirror-image of the output section's droop in the same region, so it's easy to understand why my composite spectrum was pretty flat at the high end.

The last curve shows the VCA circuit response with a reactive termination. Now the VCA's output is rising slightly at the top and bottom of the range; put it together with the output section and the end result looks like this:

A quick check reveals that the changes have not affected the limiting action:

 

[NewYorkDave]

More measurements today... Here's the frequency response with input and output level controls at 50% resistance--which is the worst-case as far as rolloff caused by series resistance acting against the Miller capacitance of the tubes:

It's still not bad.

I also took some THD+N measurements (but didn't grab any spectra, sorry). Max output is +24dBM at 4%. THD+N at other output levels: 1% at +21dBM, 0.7% at +20dBM, 0.5% at +18dBM, 0.35% at +14dBM. All of these numbers were measured with a 1kHz sine input and no gain reduction.

As far as THD under GR goes: 0.8% with 20dB of gain reduction (+24dBM output reduced to +4dBM).

 

[analag]

Very interesting stuff, do you have any Edcor trannie in the signal path, Dave. I have a few experiments of my own to do, but my day job is upon me....thus no time until weekend.

analag

 

[Larrchild]

Re: THD:

Dave, make sure your control line is dead-free of hum and artifacts, as they get modulated into the THD spec quite readily. The hygene of the supply used to pre-bias the detector diodes is uber-critical.

Then it's all tube matching. But on a really good day, min-THD, min-Noise floor/hum and lowest thump occur at the same balance-pot setting.:sam:

 

[Winston OBoogie]

[quote author="Larrchild"]Re: THD:

Dave, make sure your control line is dead-free of hum and artifacts, as they get modulated into the THD spec quite readily.
[/quote]

Speaking of artifacts: when using a solid-state as opposed to a valve rectifier for the control line, anyone found any benefits to using something along the lines of a Schottky or FRED bridge?

 

[NewYorkDave]

I've definitely gotta try something other than 1N4148. They're not working too well in this circuit above 5kHz or so, especially with reverse bias applied.

I don't have any fancy Schottky or whatever laying around, but I do have some 1N34A. I have to check the voltage ratings, though--they're pretty low, IIRC.

I'm powering the breadboard off a clean, regulated supply with low source impedance.

 

[Larrchild]

something along the lines of a Schottky or FRED bridge?

Definitely the FRED bridge.

(i know, one fred is installed backwards. I always do that!)

 

[dale116dot7]

I found 1N4148 diodes to be pretty leaky if the release resistance is much higher than maybe 400k or so. After looking at various solid-state rectifiers, I settled in a 6AL5 just like the original UA, Altec, etc. It has an amazingly low leakage current and a surprisingly low voltage drop. I also started out with 1N4148's in the G1176 project because I didn't have any low-leakage diodes. They didn't work very well either.

-Dale