Edit: Direct-coupled valve compressor

[Winston OBoogie]

So,
I have a valve differential amplifier that is an output stage.  It is directly coupled from another differential input stage.  
The DC level at the anode/grid nodes is not constant and moves as part of a sensing system on the input stage grids.  The change in voltage at these nodes might be 100V +/-.
My output stage must be able to drive a fixed load to full capacity at all times so I can't have appreciable changes in current.
Therefore, I have a constant current sink on the output stage cathodes.
Currently (!) the sink is a fairly simple high voltage cascoded MOSFET affair and is operating within safe limits.   For the record, sometimes the changes in voltage I'm sensing may happen fairly quickly, although we're not talking uS.   Rise/fall times of 1mS might be the fastest I'll encounter.
Anyway, I've been plowing through a few old Tektronix schematics for inspiration and I'm about convinced that I don't need anything more than what I have.
But, does anyone feel there would be benefit to another approach?   Pentode sink? choke sink?  Fixing the thing in treacle?

Thanks,

John

 

[abbey road d enfer]

I'm afraid you're on your own on this one!
Maybe adding cathode-followers in the outputs...

 

[Winston OBoogie]

I'm afraid you're on your own on this one!
Maybe adding cathode-followers in the outputs...

Hi A.R. d enfer,
Yes, maybe I am    but  thanks for taking the time to reply! 

Cathode Followers will no doubt get thrown in the mix at some point to see if they help or not.

Cheers,

John

 

[PRR]

> Fixing the thing in treacle?

It does rather sound like a 'scope stage. But your description is too foggy for me to ponder the need for treacle.

What are you worried about? Bad sound? Silicon contamination of hollow-state sound? Bad measurement? Unexpected smoke? Poor production yield? Or are you just in new land, unsure of your footing, yet unwilling to spill all your beans?

Rather than micro-managing the output to give "full capacity at all times", you could just brute-force it. If you need 1mA at 50V, run the stage at 50mA and 250V, let it wander. MORE POWER! is, sometimes, an answer.

You surely know many ways that simple negative feedback can be applied to stabilize a power stage. Op-point setting is easier when you have P-type devices, but many clever techniques exist for N-only technology.

Cascoding for HIGH-voltage has tricks, as you know full well. As a general thing, you have a voltage divider, with resistors, bypassed with caps to equalize rapid change. OTOH, 1mS change is not rapid. OTOOH, MOSFET gates have large capacitance. But you know your Rs and Cs.

And you can avoid most DC and common-mode troubles by abandoning silly-state direct-coupled fashion-driven thinking. Transformer, tube, transformer..., repeat as needed. Yeah, if fashion-crazed customers don't gag at the thought, they will gag at the cost.

 

[Winston OBoogie]

It does rather sound like a 'scope stage. But your description is too foggy for me to ponder the need for treacle.

I was being a bit cagey, sorry  :

What are you worried about? Bad sound?

Yes.  Always.

Silicon contamination of hollow-state sound?

Not necessarily

Bad measurement?

No.

Unexpected smoke?

I have plenty of sandbags.

Poor production yield?

It's just a DIY doodle.  

Or are you just in new land, unsure of your footing, yet unwilling to spill all your beans?

That's pretty much it.  A cigar, if you smoke 'em, for ya.

The idea came up on another forum for eliminating the coupling caps in an Altec comp type circuit.
I built a quick proto using a pair of 6ES8's in // that were cascoded by a 7N7 and directly coupled to another 7N7 diff amp.  The current sink under the 2nd stage was a cascoded IRF820 type MOSFET job.
It seemed to work quite well and I hadn't even paid much attention to the details so I figured I must be forgetting something, it can't be this easy?  I then started to 2nd guess the approach!

My goal was to eliminate an expensive interstage transformer and also eliminate the settling time effects of the coupling caps on the 2nd stage.  Anyway, I have since gone back to listening again to a choke loaded 1st stage and also a transformer coupled arrangement and find the results somewhat better.  I also admit to quite liking the Altec splat and splurt as it is for some things.

Not giving up but you did give me some ideas to try in the meantime.  Thanks!

Such as:

Rather than micro-managing the output to give "full capacity at all times", you could just brute-force it. If you need 1mA at 50V, run the stage at 50mA and 250V, let it wander. MORE POWER! is, sometimes, an answer.

You surely know many ways that simple negative feedback can be applied to stabilize a power stage. Op-point setting is easier when you have P-type devices, but many clever techniques exist for N-only technology.

Cascoding for HIGH-voltage has tricks, as you know full well. As a general thing, you have a voltage divider, with resistors, bypassed with caps to equalize rapid change. OTOH, 1mS change is not rapid. OTOOH, MOSFET gates have large capacitance. But you know your Rs and Cs.
And you can avoid most DC and common-mode troubles by abandoning silly-state direct-coupled fashion-driven thinking. Transformer, tube, transformer..., repeat as needed. Yeah, if fashion-crazed customers don't gag at the thought, they will gag at the cost.

 

[abbey road d enfer]

I just had an idea the other day; what about putting another differential pair in parallels with the original one, but driven with no-signal and receiving an inverted version of the control voltage (with proper level-shift of course)? Then the total current MAY be somewhat constant, reducing the DC shift...

 

[Winston OBoogie]

I just had an idea the other day; what about putting another differential pair in parallels with the original one, but driven with no-signal and receiving an inverted version of the control voltage (with proper level-shift of course)? Then the total current MAY be somewhat constant, reducing the DC shift...

Abbey Road d'enfer (I concur btw, not the literal but the idiomatic use  )

I've had only two sips of coffee and am still going through "re-entry" so I need a while for things to kick /sink in.

At the moment, I'm not seeing how to do this with my example.  Sorry.  Thanks though and I'll come back later and think about it,

Cheers,

John.
 

 

[PRR]

> eliminating the coupling caps in an Altec comp type circuit.

The caps are not a major problem.

When R-C coupled, the massive DC voltage shift is a real problem.

> My goal was to eliminate an expensive interstage transformer

Onct upon a time, that was about the cheapest coupling Xformer around. But my local radio-parts shop was already out of them in 1980 when I cleared out most of their "obsolete" iron at $0.50/lump.

But that's getting astray. Yes, I would expect a simple current source to be fine. I have built negative 300V supplies just so I could use a dumb resistor. (And BTW, with triodes the tail-source may not want to be infinite impedance; mostly depending if you count from cathodes or grids.) I'd use a simple BJT, except today MOSFETs are more available in 500V ratings. Not sure that a cascode buys you much, though it don't cost much extra.

With three plate-drops stacked up, you need a tall power supply.

The common-mode voltage shift can be cancelled (as Abbey hints). The plan below reduces CM shift from ~~75V to ~~1V. As-is, your next stage could be simple resistor-tail with quite negligible fault. There are many obvious variants.

 

[Winston OBoogie]

Thanks PRR and A.R. d' E.

Yep. I get it now.  


"I have built negative 300V supplies just so I could use a dumb resistor."

Me too.  ALA Philbrick K2's.  A few -V and a piece of sand is easier though.



"And BTW, with triodes the tail-source may not want to be infinite impedance; mostly depending if you count from cathodes or grids."

Gotcha, hadn't thought about that.

'I'd use a simple BJT, except today MOSFETs are more available in 500V ratings. Not sure that a cascode buys you much, though it don't cost much extra."

I would too except I have lots of IRF Hi-Voltage MOSFETS around.  
Cascoded: In this app., probably no benefit. I stole it from another PCB as was.  Just being lazy really.

"With three plate-drops stacked up, you need a tall power supply."

Yep.  I have it.  

"The common-mode voltage shift can be cancelled (as Abbey hints). The plan below reduces CM shift from ~~75V to ~~1V. As-is, your next stage could be simple resistor-tail with quite negligible fault. There are many obvious variants."


Thanks for taking the time to do that Paul, much appreciated.  I was being dumb and not getting what Abbey was saying/hinting at.
I will play with the thing some more when I can.

I suspect that it'll not be worth some of the effort involved for DIY'ers.  I'll treat it as an exercise.  

I have gotten pretty decent results using a cheap 'n' cheerful transformer primary as the VCA load (R C coupled to the diff. amp) which might be the way to go for easy but still "Hi-Fi".    I see some fairly inexpensive filter chokes that have sufficient L (fairly high dcr though) that I'd like to try too.  They may be too sloppily wound for use as a load?  

Thanks gentlemen,

John

 

[emrr]

Keep it up guys, fascinating reading. 

 

[Winston OBoogie]

Keep it up guys, fascinating reading. 

Hey Doug,
Not much to add at this point until I get time to fiddle again.

However, I just realized I'm probably OK to de-rail the thread a bit since I started it?

I'd like to ask, since you have a good collection of vintage comps, do you particularly like an aspect of any of the units?  Is there anything in a design that you've noticed and that seems to hint/point in any particular direction?

Another de-railer:  I think the 7N7 (and the 12volt 14N7) that I'm using are great tubes.  Of course, they're Loctal but that's OK.  They compare to the better of the 6SN7's but not as pricey. 

John

 

[tomelectro]

Neat!  What's the purpose of the cathode transistor (Q1?) on the offset canceling circuit? Is it just to anchor the cathodes near ground? EDIT: Now that I think about it some more, looks like the cathode voltage would go too negative without it, degenerating the gain reduction bias.

 

[emrr]

Keep it up guys, fascinating reading. 

Hey Doug,

I just realized I'm probably OK to de-rail the thread a bit since I started it?

I'd like to ask, since you have a good collection of vintage comps, do you particularly like an aspect of any of the units?  Is there anything in a design that you've noticed and that seems to hint/point in any particular direction?

Another de-railer:  I think the 7N7 (and the 12volt 14N7) that I'm using are great tubes.  Of course, they're Loctal but that's OK.  They compare to the better of the 6SN7's but not as pricey. 

John

It's yours to derail.  ;D

I don't seem to like any of them until I expand the time constants.  If I had to pick one thing, it probably boils down to the flavor(s) of the old iron.  I used to buy up lots of vintage iron, thinking I'd do a short run of custom limiters, and every time the iron became worth too much to use in a product build.  So I've yet to go there.  Some of the variable bridge shunt methods sound cool.  I haven't heard the more common Collins 26W, but it's predecessor the 26C is pretty amazing; same method there.  I'd think a bridge shunt worth exploring with modern iron. 

I've been marveling lately at how different any comparison of Gates SA-39's can be.  4 possible different inputs, 4 different outputs, 4+ different circuit revisions, over more than 20 years.   

Loctals seem like a great place to work, given the lack of demand. 

 

[Winston OBoogie]

Neat!  What's the purpose of the cathode transistor (Q1?) on the offset canceling circuit? Is it just to anchor the cathodes near ground? EDIT: Now that I think about it some more, looks like the cathode voltage would go too negative without it, degenerating the gain reduction bias.

As you probably know, normally, if there is a transistor beneath a pair of cathodes, it is an NPN and is configured as a current source.  


In this case however, Q1 is a PNP and is configured as 'grounded base'.  In fact, you could think of it as half of a PNP diff. amp with it's input grounded if you like.  This is part of the voltage adjusting loop.  Current enters Q1 emitter and voltage (but, being grounded base, no current) is amplified (developed across the 1K R3) and sent via the Q2 follower to Q3 and Q4 which are also grounded base voltage amps, this time being NPN.  From there back to the anodes of the valve pair.  
Quite nifty is our PRR  




Edited because I wrote PNP when I meant NPN.  Fixed. 

 

[Winston OBoogie]

I haven't heard the more common Collins 26W, but it's predecessor the 26C is pretty amazing; same method there.  I'd think a bridge shunt worth exploring with modern iron.  

Cool!  Thanks Doug.  I've never played with one of those but I'll find a schem, and fiddle with bits I have.

I've been marveling lately at how different any comparison of Gates SA-39's can be.  4 possible different inputs, 4 different outputs, 4+ different circuit revisions, over more than 20 years.    

Same thing with Universal Audio and the 175 comp output transformer.   Except the variations happened month to month or batch to batch, not over 20 years.  Same general spec throughout for the transformer but any one of 5 different winders.  Also, when UTC couldn't supply an A-18 interstage, they'd use an A-19.  Or otherwise.  

Loctals seem like a great place to work, given the lack of demand.  

That's why I like 'em.  No demand...yet!    I found the distortion was typically a decade below a 12AU7.  



Cheers, thanks again,

John

 

[emrr]

[quote author=emrr]
I haven't heard the more common Collins 26W, but it's predecessor the 26C is pretty amazing; same method there.  I'd think a bridge shunt worth exploring with modern iron.   [/quote]

[quote author=Winston O'Boogie] Cool!  Thanks Doug.  I've never played with one of those but I'll find a schem, and fiddle with bits I have.[/quote]

http://www.groupdiy.com/index.php?topic=24540.msg291495#msg291495


[quote author=emrr]
I've been marveling lately at how different any comparison of Gates SA-39's can be.  4 possible different inputs, 4 different outputs, 4+ different circuit revisions, over more than 20 years.    [/quote]

[quote author=Winston O'Boogie]Same thing with Universal Audio and the 175 comp output transformer.   Except the variations happened month to month or batch to batch, not over 20 years.  Same general spec throughout for the transformer but any one of 5 different winders.  Also, when UTC couldn't supply an A-18 interstage, they'd use an A-19.  Or otherwise.  [/quote]

BA-6A too; all legend, and not a single comparison or commentary about all the different iron used.  I've yet to see anyone but myself bring it up.  Of course, if I try to sell any BA-6A iron, the question gets asked.  I mean, really, who would know? 



[quote author=emrr]
Loctals seem like a great place to work, given the lack of demand.  
[/quote]

[quote author=Winston O'Boogie]That's why I like 'em.  No demand...yet!    I found the distortion was typically a decade below a 12AU7.  

[/quote]

I've never done comparative distortion measurements between large and small bottle tubes, but my ears always seem to tell me this.  I've gotten to the point I outright avoid pro gear with 12AX7's.  I leave the room like someone just farted; and it might have been me, out of surprise and disgust.  I realize it may be a personal problem, but I'm not feeling that way yet.  The 12AX7 part, that is.  ;D

 

[tomelectro]

Thanks for the good explanation. I just wasn't quite sure why Q1 needed to be there at all.

Same thing with Universal Audio and the 175 comp output transformer.  Except the variations happened month to month or batch to batch, not over 20 years.

That's interesting.  What other iron did they use? I've got one on the bench right now with a bad output xformer (ADC 322) that I'm planning to replace with a UTC A34.  I'm pretty sure I did the same mod to another one years ago which worked out pretty well, but I'm not sure if the A34 was really meant for push-pull with B+ CT. Anybody else try one?

Thanks,
Tom

 

[PRR]

> What's the purpose of the cathode transistor

Virtual ground.

The cathodes must have "low" resistance to ground or THD rises.

We want to know the cathode current, to compensate it.

The emitter is pretty solid ground (far below cathode impedance). Collector current is 99% of emitter current, for any current the tube is useful at, and can readily be sensed without disrupting the tube.

There are some flaws with the plan shown. I now see one which is rather embarrassing. Ah, well, that's what you get for free. It's a concept. Details are left to the profiteer.

 

[Winston OBoogie]

There are some flaws with the plan shown. I now see one which is rather embarrassing. Ah, well, that's what you get for free. It's a concept. Details are left to the profiteer.

Hopefully no profit for anyone on this one.  At least there won't be for me.  If I get a circuit that works decently (I planned on my own implementation of your idea/schema), I'll post it for any who care to build it. 
Although I think I'm still leaning towards a cheap choke pair load of the VCA, R,C'ed to a diff-amp.   
Again, I will play with this though as I'll learn something by twiddling with the details.

Thanks again. 

 

[Winston OBoogie]

What other iron did they use? I've got one on the bench right now with a bad output xformer (ADC 322) that I'm planning to replace with a UTC A34.  I'm pretty sure I did the same mod to another one years ago which worked out pretty well, but I'm not sure if the A34 was really meant for push-pull with B+ CT. Anybody else try one?

Thanks,
Tom

Tom,
I've never seen an ADC output in a 175 compressor.  I would maybe suspect that it is a replacement part done years ago but, who knows? 
The actual original part used was custom to UA and was internally numbered as PA-5946.  When the various companies wound it for them, they wound it per the UA spec and documentation.
As it happens, I ordered, 2 days ago, a first run of 12 pieces as per the docs. for a project.  We'll see how they turn out...

I don't know the particulars of the A-34.  However, if there is no DC current imbalance in the primary then it should be fine with H.T. applied to the centre tap.  The correct primary to secondary ratio is 7:1.

John