Interstage curious

CurtZHP said:

So it would work in my case to have the plate supply going to the PP plates via the center of the LS-50 primary, provided both tubes are drawing the same current?

This is where the cathode balancing pot comes in, yes?

Click to expand...

Yes.

CurtZHP said:

....aside from my own misbegotten visions of smoke billowing from a very expensive transformer, apparently!

So it would work in my case to have the plate supply going to the PP plates via the center of the LS-50 primary, provided both tubes are drawing the same current?

This is where the cathode balancing pot comes in, yes?

Looks like the whole mess stems from my simply missing the point on balanced vs. unbalanced current on the primary.  I assumed there could be NO DC current whatsoever across that winding.

Click to expand...

Think about this:  do prototyping with a decade resistor or wirewound pot as a rheostat in line with your output stage cathodes, having some smaller resistance (560R?) in line so you don't accidentally go to zero R.  Do it with a mA meter in line for monitoring while looking at performance points like overload symmetry and headroom.  You can change that fixed R if need be once you've done the first rough in.

Do this with just the output stage, feeding test signal directly into the LS-21 primary.  Get the output stage sorted first, first rule of tube amp building.  Figure out gain and headroom with that to figure out what you need in front of it. 

Then figure out what's ideally needed to drive the LS-21.  Then figure out what's needed to drive that.....work backwards from high to low. 

read this, it's fast and entertaining:

https://ia800904.us.archive.org/28/items/High-fidelityTechniques/Langham1950High-fidelityTechniques_text.pdf

Look at those Collins I mentioned above, they are in the tech docs.    You won't find these ideas commonly on the internet, it's mostly busy reinventing the wheel with new component types.  You need to study historical circuits which AREN'T much dissected or cloned by most tube heads. 

Thanks so much for that.  Skimmed through it quickly, and I can't wait to sit down and actually read it more carefully.

Never thought of working backwards through the circuit.  Will give that a go.

Question....
While looking over numerous schematic online for PP cathode balancing circuits, I've seen some where each cathode has its own resistor and the "bottom" of each resistor goes to one end or the other of a pot with the wiper going to ground via another fixed resistor.  Others have each cathode connected to opposite ends of the pot, again with the wiper in series with another resistor in series with ground.  Is there a preferred method?  Are there pros and cons for each?

Voltage divider or none.  Limits the range, or not. 

Had some time to slap this together....

Looking at some schematics for variable mu compressor output stages, this seemed like it kept showing up. 
Ran the number I had so far through some calculators to arrive at somewhere between 1.5K and 2K for cathode resistor values.  This is where I get confused...
A few have suggested that, because of the push-pull aspect, I need to halve the cathode resistance for each cathode.  Others have said to treat them as separate except for the balancing act.  I happened to have a fairly hefty 1K pot on hand that feels like a wirewound, so this is where I landed.

The other question was whether or not I needed a bypass cap.  I would assume I do, but still trying to determine a value.

Pretty much every other detailed circuit I could find had this stage being fed from another tube stage as opposed to an IS transformer, so they also tended to include a couple grid resistors, usually in the neighborhood of 1MEG.  One RCA schematic I found, which actually had a transformer in front of all this, had those 1MEG resistors sharing ground with the center tap of the IS secondary.  With the LS21, would they even be necessary?

Or this.....

Two 1K5's or a single 750.  Downside is you have no way to balance a pair, if it matters, unless you put inline trimmers in that you can lock down.  That would seem really overkill. 

Your first drawing is fine.  Really, a 250 or 500 balance pot is going to be a better range.  The wire wound I was referring to was for the 1K5 position you have shown, so you can determine the ideal value for that position on the bench rather than guessing or trying to calculate. 

The Collins 6R is essentially this same output circuit and loading.  400R balance pot, 620R cathode resistor, 20mfd bypass cap across the 620.  20mfd was what they could get in a hermetically sealed cap package in 1947, so it was 20 or 40 from parallel connection of a pair, more space, etc. 

If you used a 250 pot and a 620 resistor you land right at 745 if it's centered. 

You can use a bypass cap as shown, or not.  Listen to it.  It's 6dB of gain to have one, and the loading is different.    You could alternatively use a pair of bypasses from either end of the pot to ground.  Historically that cap is almost always a 25mfd because that's what they could get in terms of size and cost.  You can make it 470mfd if you want.  Look at measurements and see what you like.  The bigger, the lower the cutoff, the more linear the response  and phase, with diminishing returns. 

LS-21 may not need load resistors to ground, I suspect not.  See what response looks like.  RCA transformers tended to have improved response with loading resistors, so that's specific to the parts used. 

Thanks.

I'll see what other components I can scrounge and give this a whirl this weekend.  I've seen a lot of designs that have fixed resistors in place.  Would it be safe to assume that, once I get this dialed in, I could replace the pot with fixed resistors of the appropriate values?  Or will it not stay put?

By the way, that book you provided the link to....
I got about halfway through it so far.  Very amusing as well as informative!

the balance pot would stay.  the rheostat would be replaced with a fixed resistor. 

Managed to scare up a 500R trimmer to go with the 1K pot and extra tube socket.  Put together a little test rig to try this over the weekend.

Wish me luck and have a happy Thanksgiving!

Good luck and happy thanksgiving!

So far, so good!

Taking your advice (and that of our old friend Mr. Langham...), I started at the end.  I built the push-pull stage and put it between the output transformer and the IS transformer.  Did my best to balance the stage using a pot/trimmer combination for the cathode network.  When I got things dialed in, and they seemed like they were within the tube's tolerances, I fed signal directly into the IS primary.  Got a nice clean signal out, just 16dB bigger, if my math was right.

Spent the better part of Friday night tweaking and testing.  The tube seemed awfully hot, which concerned me, so I used a temperature probe to check it.  It hovers around just under 140F, which according to the interwebs isn't too hot.  But that was mostly guys talking about guitar amps.  Never could find a definite answer on that.  I tried to do some calculations on plate dissipation, but my numbers seemed a bit low.  I was getting 234VDC plate-to-cathode.  My plate current calculations (a bit suspect...) gave me 6.9mA(?).  That gives a plate dissipation of 1.6W, which I assume is OK.

Got to feeling over-confident and decided to hook up the rest of the circuit.  That's where things went sideways.  Nothing worked.  Craziest waveforms I ever saw, if I saw any.  Pretty much wasted Saturday night.  Sunday afternoon, I set about dismantling the rest of it and put together just the second stage of the first tube.  Fed signal directly to the grid.  Got a louder signal out, but it seemed a little wobbly, like it was picking up junk from somewhere.  Tried a few things to no avail and figured it was my long clip leads acting like antennas.

Took a break and then decided to take a stab at the first half of the first tube stage.  Had to back way off on gain by changing a few things, since I ended up with way more than necessary.  Once that was figured out, both halves of that tube got along well with the rest of the circuit.  Didn't see as much junk on the output.  (Eliminated some clip leads along the way...)

With another burst of confidence, I went ahead and threw in the input transformer.  Still passed a pretty clean signal, and handled a lot higher input voltages than when I first started this project.  Above around 100mV on the input (-20dBV?) it begins to distort with the gain pot at about 2/3 of the way around.  By my calculations, the amount of gain the whole gadget provides before it fuzzes out is about 63dB.

For now, I need to update my schematic to reflect all the changes I made and clean up the wiring a bit more.  Hopefully, that improves things a little more.  Then I'll listen to some actual audio through it.

Cool

Updated schematic attached....

Turns out my gain calculation was off.  I only get about 50dB of clean gain out of this before it gets ugly.  I suppose that's still respectable.  That's with 100mV signal going in.  I tried a few things to squeeze a little more out of it, but everything I tried so far just made it worse.

Poking around with the scope probe, it seems that the second tube stage, right before the IS transformer, is where things start to grunge out.  I figured I was overdriving it with the first stage, so I reduced the gain on that one and tried to make it up on the second.  All that seemed to do was make it noisier.

Maybe I'm expecting too much of it?

I think the gain question is backwards, it should be "what input level will it take with gain pot fully open?" and "how does that relate to an actual microphone signal?".    I don't know what 100mV relates to in real life.  Is that RMS or Peak?    At what stated impedance, and then loaded by an inductance expecting a 200 ohm source.  I see for example that 100mV peak to peak at 50 ohms equates to -16dBm......none of those directly relatable standards but far beyond an input level one would expect to use with a preamp like this.    This thing is for dynamics or ribbons, only condensers with fairly quiet sources. 

Nitpicky, but R3 isn't needed.  If a load is needed, it's surely not that.  It's not hurting anything though.  LS-10's are well behaved generally, without need for a load resistance into a triode. 

R7 and R8 are extremely low values.  I would expect at least 22K if not much higher.  Yes, lowering those values raises B+ at plate, but there's a limit.  R4/5 surely land at a different point with extremely varying plate R's. 

VR1 can be 100K, and if any (realistic) lows suffer it'd still be better with 100K for C1 to be 0.22 instead.  A smaller pot value gains you better treble at lower settings. 

C2 value presents an opportunity to tune bass bump resonance against the LS-21 inductance. 

EmRR said:

I think the gain question is backwards, it should be "what input level will it take with gain pot fully open?"

Click to expand...

Good question.  I double checked.  It'll cleanly take 20mV with the throttle wide open.  At 25mV, the signal starts to show signs of getting crispy.

EmRR said:

...and "how does that relate to an actual microphone signal?".    I don't know what 100mV relates to in real life.  Is that RMS or Peak?    At what stated impedance, and then loaded by an inductance expecting a 200 ohm source.  I see for example that 100mV peak to peak at 50 ohms equates to -16dBm......none of those directly relatable standards but far beyond an input level one would expect to use with a preamp like this.    This thing is for dynamics or ribbons, only condensers with fairly quiet sources.

Click to expand...

I'm using a tone generator that displays its output as either mV or dBV.  According to the settings, the 100mV signal translates to -20dBV, for what that's worth.  Also, for what it's worth, 20mV is about -34dBV, according to the settings on the generator.  As to the question of how that translates to actual mic signals, I guess it depends on the mic.  A ribbon is certainly not getting anywhere near 100mV.  A modern condenser, maybe 50mV?  I guess I wanted to make sure I had some wiggle room.

EmRR said:

Nitpicky, but R3 isn't needed.  If a load is needed, it's surely not that.  It's not hurting anything though.  LS-10's are well behaved generally, without need for a load resistance into a triode.

Click to expand...

I had suspected that earlier.  At some point, I forgot to put it in, and then realized I hadn't.  Performance didn't seem to suffer or improve either way, so you confirmed that.

EmRR said:

R7 and R8 are extremely low values.  I would expect at least 22K if not much higher.  Yes, lowering those values raises B+ at plate, but there's a limit.  R4/5 surely land at a different point with extremely varying plate R's. 

Click to expand...

That's sort of where I landed trying to keep this thing from distorting at higher gain settings.  It seems like anything above 10K was causing massive distortion if the pot was turned above about 1/3 of the way.  But then, that was based on feeding it a 100mV signal (which is likely higher than any mic will ever give it).

EmRR said:

VR1 can be 100K, and if any (realistic) lows suffer it'd still be better with 100K for C1 to be 0.22 instead.  A smaller pot value gains you better treble at lower settings. 

Click to expand...

I thought of something smaller, but the only smaller log pot I have on hand is a 10K.

It looks like I have the basic circuit figured out.  Just need to finalize some component values on those first two stages.

Consider you have 24dB of gain in the input transformer.  You will rarely use a condenser without a pad in front of it.  Many times not a dynamic either. 

You could knock 6dBish off by wiring the LS-10 secondaries in parallel.  You could put that on a series/parallel switch but you'd have to be VERY careful about layout and wiring to not incur treble losses from wire length or position.

Plate resistor values so low are increasing distortion in of themselves.  Quick glance at a couple of classic 6SN7 stages, I see 68K-100K. 

For a reality check on pot level, my Gates SA-20 (70dB total, with an LS-10 input, triode, pot) frequently lives in the 1/5 to 1/2 zone. 

EmRR said:

You could knock 6dBish off by wiring the LS-10 secondaries in parallel.  You could put that on a series/parallel switch but you'd have to be VERY careful about layout and wiring to not incur treble losses from wire length or position.

Click to expand...

I actually did that on my last build with an LS-10!  Was also planning to do it here.  I guess that would solve most of the issue in most cases.

EmRR said:

For a reality check on pot level, my Gates SA-20 (70dB total, with an LS-10 input, triode, pot) frequently lives in the 1/5 to 1/2 zone.

Click to expand...

Come to think of it, on that other preamp of mine, I rarely turned it up more than half way with a condenser mic.

I guess I'm overthinking this.

Back to the bench.

What you said earlier about gain made me realize I've been approaching it all wrong from the get-go.
I was thinking, "How strong of a signal can I feed it without it distorting?"
Instead, I should be thinking, "How weak of a signal can I feed it that it can amplify cleanly?"

I should be dropping the output on the generator as low as it will go and looking at that.

Yep.  Even wide open and input level low enough, the output stage headroom defines the limit.