Sixty Cycle Blues

[dude]

Thanks for the replies, guys.

Good call on the input resistor, curtisa. I'll change that.

Yeah, the 5-band EQ will probably be a little weird. I'm too enamored with the idea of a 5-band EQ in front and back to stray from it this time. I toyed with the idea of throwing some op-amps in and having proper second-order, buffered, active filters in there, but I decided to keep this one "all tube", because, you know, "all tube" is the most important thing when it comes to guitar tone. Just look at Tom Scholz. (/sarcasm)

Here's the schematic:

The simulation looks mostly good. With the pots all set to 10% (halfway on log pots, theoretically anyway), the response of the filter portion without the follower is very flat, and adjusting the individual band levels produces pretty okay results. The cutoff between bands is definitely what you'd expect for first-order filters (read: really not sharp), but it should work okay for my main intended use: limit the input frequencies to a reasonably narrow band, and then keep the output frequencies mostly flat, but with an upper-mid peak. Could I do this with static filters? Absolutely. Do I want to be able to fiddle endlessly with 10 filter knobs to shape this tone exactly? Yes. Yes I do.

Now, the problem arises when I put the follower in front of it. On its own, the follower's output has a nearly perfectly flat response with a subtle low-end roll-off. When I hook up my EQ circuit (which also has a flat response) as shown above, however, the response of the whole system is not flat. It looks roughly the same as the response of the 5 band eq on its own (same low and high cutoff), but there is a gradual decrease from low to high of something like 3 or 6 dB.

The 150k cathode resistor should be too big to be a factor, so my only thought is that the cathode resistance or capacitance must be at play here. I'm not sure exactly how either of those are modeled on a tube.

Any ideas?

Also, Ansil, you always bring in the coolest filter ideas. It's late and I'm having trouble following your descriptions above, but I'd love to chat about them later.

EDIT: V2 is ~300v

Edited August 15, 2013 by dude

[curtisa]

The output impedance of your cathode follower is probably not low enough to drive so many stages in parallel cleanly. Try reducing the number of filter stages, or increasing the R values of your filter stages by a factor of 10 (and corresponding 10x reduction in associated C values to maintain the same frequency cutoff) and see if the response improves.

Your pots are marked as "10" - I assume that means postion at 10%, not resistance = 10. What pot values are you using?

[dude]

Yes, the "10" is 10%. The Pots are 100k. If I increase the R values of my filters I'll probably have to go with 1M pots instead.

I can also reduce the output impedance by splitting the 150k cathode resistor into something large on top and something small on bottom, and then bypassing the larger, upper resistor with a good-sized capacitor. Trouble there is that if I do a 150k resistor on top and a 1k resistor on bottom, I need about 4u for the bypass, which is a little silly. But, I just ran a simulation and it works alright.

I'll try increasing the filter R values tonight. I expect that will have better results. Plus, using smaller C values will make getting the right parts a bit more convenient. I'd probably be able to use ceramic caps all around. The reason I didn't try that before is because I have a bunch of 100k log pots sitting around, but oh well, I can order some 1M.

[ansil]

here is the bass resonance control i cobbled together using a plexi schematic i found on the web. i do not own this schematic. this is just one of the ideas i was talking about. also you can do this on one stage or more or all of them. something to think about for future referance

[curtisa]

Yes, the "10" is 10%. The Pots are 100k. If I increase the R values of my filters I'll probably have to go with 1M pots instead.

My suggestion of 10x factor was just arbitrary (makes the math easy). 3-4x factor would still offer an improvement, and you can use 250K pots for the band controls.

I can also reduce the output impedance by splitting the 150k cathode resistor into something large on top and something small on bottom, and then bypassing the larger, upper resistor with a good-sized capacitor. Trouble there is that if I do a 150k resistor on top and a 1k resistor on bottom, I need about 4u for the bypass, which is a little silly. But, I just ran a simulation and it works alright.

Splitting the cathode resistor has the side effect of lowering the signal entering into the EQ, defeating any amplification you put in front of the cathode follower, but the output impedance becomes roughly equal to the two cathode R halves in parallel.

Reducing the cathode resistor overall does not have as much of an effect on the output impedance either as the cathode resistor is in parallel with the tubes' cathode output impedance. You need to reduce it a lot to get any real effect. Reducing it too much introduces other problems (cooking the plate, need big wattage cathode R etc)

Be careful when you do your simulations that you are using a model of your "target" tube instead of the default triode model - default SPICE models are notorious for presenting impossible real-world results.

here is the bass resonance control i cobbled together using a plexi schematic i found on the web. i do not own this schematic. this is just one of the ideas i was talking about. also you can do this on one stage or more or all of them. something to think about for future referance

Bogner do a similar thing in the Ecstacy with the "structure" control (or at least, it's a switched element - just a fixed series-RC that can be added in parallel to the cathode resistor of that stage). Any reason you're not just using the pot to directly control the resistance? Seems to me you could achieve the same thing without the LDR at all.

Edited August 15, 2013 by curtisa

[ansil]

then you would need more than one pot or a dual gang triple ganged to do more stages. this way i can get multiple stages and even reverse the middle and boost the outside ones. all sorts of things. but yes you can do it with a pot.

[dude]

Okay, well I'll need to mess around with some values and see if I can get something that works well. Though honestly what I have now looks alright. I'd rather have a flatter response though.

Ansil, I like that circuit a lot. When I was reading up on vactrols I started thinking of ways to use them to shape the amp strictly from electrical control. I had a dream once of building a modeling amp that simply rearranged the circuit inside to model different famous amps. Hook a microcontroller up to a bunch of conveniently-placed vactrols, and get instant switching from Fender cleans to Marshall Crunch to Vox Bark. Of course you might need a motorized door on the back of your cabinet too

EDIT:

and yes, I'm using the built-in 12ax7 model, however good that may be. I used to have a link to a page that had SPICE models for a LOT of tubes (even the obscure TV scanner pentodes and such I've got lying around). I'll have to see if I can find it.

Edited August 16, 2013 by dude

[ansil]

with a little tweaking all those circuits you mentioned are really similar.

[ansil]

if you change your eq and make it switchable. ie not your eq you designed but your standard marhsall fender style as they are all close enough for rock and roll. you can switch from normal marshall style placement to right after first stage for a fender style. it does wonders for the tone and with a few relays or ldrs or what have you its quite easy to acomplish. have the master volume after the tone stack and you get the same basic marshall fender layout.

[dude]

Throw in an adjustment on the negative feedback and you've got yourself a pretty cool amp. Might have to try that next. Something lower-wattage this time. I've got some power dual triodes sitting around that I REALLY want to use in a push-pull power section.

[dude]

Alright, I've knocked together a more final schematic here. Looks like I forgot to calculate the bias resistances and such. Well, that can be taken care of later:

I did a factor-of-ten modification to the RC values in the EQ, and it made a huge difference. The boost and EQ now have a very good response, and the controls seem to behave as they should. I get a (normalized) +5dB at max, -5dB at min, and 0dB at center (log). We'll see how it acts in practice.

The long-tail phase splitter took me a lot of fiddling and simulation, but it seems fairly good now. The signals are very close to balanced, and I should be able to have a lot of clean headroom if I want it. I'm using EL34s for the power tubes, and I am cathode biasing. I may add a switch to change bias resistors so I can swap for 6L6 conveniently later.

No negative feedback for now. On my other project amp I have a switch to turn feedback on and off, and I really like the sound without, so I'm going to start there and add it in if I decide I need it. Overdrive will work better with feedback.

I picked the values for the bias and grid #2 resistors based on an EL34 datasheet. I'm going to look a bit more into this, though, because I'm not convinced these are correct. And because I want to understand everything. (time to read the power amp section of the Audio Cyclopedia again )

I dropped a gain stage, as the music I've been playing lately has been more in the "crunch" range and less in the "lead" range of overdrive.

Lastly, I'm moving the first EQ to before the preamp volume (like on a fender or mesa) and putting the main gain stages (2 pentodes) after. This will put the two EQs firmly before and after the gain section. I don't THINK there will be any problems putting the preamp gain immediately following the EQ with no buffer, but I could be wrong.

Any thoughts on the design are, as always, deeply appreciated. I'm about ready to order my components (the only thing I haven't got yet) and get building!

EDIT: I forgot to draw in a 1M series resistor before each band's 1M pot on the EQ. There should be one on each band.

Edited August 20, 2013 by dude

[curtisa]

Grid resistor on upper portion of LTP can be reduced to 220k - 470k. The input impedance in an LTP is bootstrapped by virtue of the reference point of the grid resistors being returned to the cathodes of the tubes rather than ground, and the apparent input impedance becomes much larger than the resistor by itself.

Not sure about the massive mismatch between the two grid resistors of the LTP. The "normal" practice is for these resistors to be the same value, otherwise the two halves of the LTP will inherently have mismatched outputs. Any mismatch in signal output from the two plates is then usually compensated for by making one plate resistor lightly smaller than the other (typically 82K vs 100K). The split plate resistor on the top half of the LTP is possibly only adding complexity to the compensation you're trying to achieve. 2u in parallel with 75K will also nuke a lot of highs.

Even if you choose to match everything (plate and grid resistors) the net difference in output is around 1dB, good enough for rock 'n roll. Still, in true rock 'n roll fashion maybe this amp is destined to break some rules.

Consider making the gain and master pots 250-500K

Edited August 20, 2013 by curtisa

[dude]

Grid resistor on upper portion of LTP can be reduced to 220k - 470k. The input impedance in an LTP is bootstrapped by virtue of the reference point of the grid resistors being returned to the cathodes of the tubes rather than ground, and the apparent input impedance becomes much larger than the resistor by itself.

I'm not sure which resistor you are referring to here.

Not sure about the massive mismatch between the two grid resistors of the LTP. The "normal" practice is for these resistors to be the same value, otherwise the two halves of the LTP will inherently have mismatched outputs. Any mismatch in signal output from the two plates is then usually compensated for by making one plate resistor lightly smaller than the other (typically 82K vs 100K). The split plate resistor on the top half of the LTP is possibly only adding complexity to the compensation you're trying to achieve. 2u in parallel with 75K will also nuke a lot of highs.

Even if you choose to match everything (plate and grid resistors) the net difference in output is around 1dB, good enough for rock 'n roll. Still, in true rock 'n roll fashion maybe this amp is destined to break some rules.

I started with the more traditional circuit with a 1M on each grid and a 100k/78k on the respective plates, but I ran into a few problems as I tweaked it. The first was a high frequency roll-off (bad on the positive triode, REALLY bad on the negative triode). The second was a pretty unequal output amplitude (I think I was seeing more like a 3dB difference - about double the amplitude).

I know the circuit I started with is pretty tried-and-true (especially since you knew the dB difference off the top of your head), so there was probably something flawed with my simulation (might have been the transience in the capacitors. I only simulated the first 20ms most of the time), but I'm intrigued enough by this setup that I think I'm going to try it. If it is horrible, I'll just copy the standard Fender/Marshall/every-tube-amp-made-in-the-60s power amp and be done with it.

In the sims I ran, this LTP circuit has a gain of only about 2-3 (which is fine), but it is pretty closely matched accross the full audio spectrum. It also can provide a pretty huge voltage swing on the output (~60v pk-pk each), so I should have a decent bit of headroom, I think, though I don't have a lot of frame of reference here.

Consider making the gain and master pots 250-500K

I assume this is because at 100k they are horribly matched with the output resistance of the EQ in front of them, and I will suffer a huge gain loss as a result? Yes, I will increase them. Is there any reason not to make the jump all the way up to 1M? Then I could use only one value of pot across the board.

[curtisa]

I'm not sure which resistor you are referring to here.

Sorry. The 1M resistor between point "A" and the junction of the 470R/10K/10K resistors

I started with the more traditional circuit with a 1M on each grid and a 100k/78k on the respective plates, but I ran into a few problems as I tweaked it. The first was a high frequency roll-off (bad on the positive triode, REALLY bad on the negative triode). The second was a pretty unequal output amplitude (I think I was seeing more like a 3dB difference - about double the amplitude).

High frequency roll-off doesn't sound right. A bog standard Long Tail Pair should be pretty flat well out beyond the normal audio band without any additional caps involved. If anything your 2u cap in parallel with the 78K resistor will make it worse. I'd be looking at your simulation software for strange results.

3dB difference is a factor of 1.4x in voltage circuits like your LTP. A factor of 2 would be 6dB.

On reflection I'm not sure if the mismatched grid resistors will have that much of an effect on the output swings if the second (unused) grid isn't driven. It does look a bit odd though, and if you do decide to implement negative feedback into this grid later on it will have to be increased back up to match the grid resistor on the first half of the LTP. My preference would be to just make the two grid resistors equal and be done with it.

I know the circuit I started with is pretty tried-and-true (especially since you knew the dB difference off the top of your head), so there was probably something flawed with my simulation (might have been the transience in the capacitors. I only simulated the first 20ms most of the time), but I'm intrigued enough by this setup that I think I'm going to try it. If it is horrible, I'll just copy the standard Fender/Marshall/every-tube-amp-made-in-the-60s power amp and be done with it.

In the sims I ran, this LTP circuit has a gain of only about 2-3 (which is fine), but it is pretty closely matched accross the full audio spectrum. It also can provide a pretty huge voltage swing on the output (~60v pk-pk each), so I should have a decent bit of headroom, I think, though I don't have a lot of frame of reference here.

A LTP should be good for an open-circuit gain of 24-28x or so. Gains of 2-3 sounds like there's something funky going on in your simulation.

Does your triode model also simulate heater warmup time? Plotting only the first 20mS may not be presenting you with true results if the simulator is trying to model the behaviour of the heaters as well (this may account for your strange HF droop as well). Can you force the simulator to only start plotting results after 30 seconds?

I assume this is because at 100k they are horribly matched with the output resistance of the EQ in front of them, and I will suffer a huge gain loss as a result? Yes, I will increase them. Is there any reason not to make the jump all the way up to 1M? Then I could use only one value of pot across the board.

1M will work OK there too if it makes it easier to to keep them all the same value.

[curtisa]

FWIW this is what I get when I sim a LTP using "standard" values. The green plot is from the lefthand plate, the blue from the right. The two halves are within 0.6dB of each other and flat from 18Hz (1dB down) out past 100KHz, with a gain of around 28dB.

If I add the following stages' coupling caps (100n) and grid resistors (220k) I lose about 0.5dB of gain and increase the LF cutoff point to 23Hz. Something is definitely amiss in your sims if you're getting results like you have been.

Edited August 22, 2013 by curtisa

[dude]

Aha!

I was missing the cap to keep the grid of the second triode at AC ground. That's pretty important.

Thanks for running that. I think I'll go with the standard, and upon further consideration I AM going to add negative feedback. I may make it variable with a pot, though. I've got a switch on my old project amp, and the ability to change the amount of feedback gives you tons of variability in tone.

[curtisa]

Ah. Yes, I missed that too in your schematic. The cap will also interact with the grid resistor to determine the LF behaviour of the LTP too, and is also another good reason to replace that 10k grid resistor in your schematic with something at least 10x bigger and the same value as the other side.

I think you need a cap in series with point "A" too (see my sim), otherwise the DC biasing gets all screwed up by the preceeding stage(s), and ruins the operation of the LTP. 22n-100n will work fine here.

[dude]

The tone stack preceding the master volume (right before "A") should block DC from the stage before (its the same circuit as the big array on the left).

The 10k grid resistor was there based on sims I ran without the grounding cap. In fact, all of the weird mods I made to the phase splitter were essentially because I forgot to add that cap. So, yeah, I'll just use a traditional LTP and be done with it. I'm already doing enough weird things with this amp as it is.

[curtisa]

The tone stack preceding the master volume (right before "A") should block DC from the stage before (its the same circuit as the big array on the left).

The grid at "A" is being referenced to 0V by the master volume pot preceding it. With the LTP grid resistors returned to the tail ("bootstrapped") rather than directly to 0V, the grids are actually sitting higher than ground, maybe 25 volts DC or so. The cap needs to be installed to prevent the DC biasing being upset by the volume pot trying to pull the grid back down to ground.

[dude]

Oh oh oh. Right. My mistake. Good catch. That would have been an unpleasant mistake. Popping fuses left and right.

[curtisa]

Well, maybe not popped fuses, but it would certainly sound ugly.

[dude]

Man I'm off today. You're right. It would just mess up the biasing on the phase inverter (when you turn the MV all the way down).

[curtisa]

So, is it time to start warming up the soldering iron?

[dude]

Nearly! Need to calculate my bias resistors for the two 6au6 stages and then order my components. I think I'll punch some holes in my chassis this weekend though.

EDIT:

I was toying with the idea of putting my master volume AFTER the phase splitter (use a dual gang pot), but I'm not sure how well that would work with negative feedback. I guess if the feedback isn't as intense as it would be for an op-amp, then adjusting the open-loop gain should still affect the closed-loop gain reasonably well. Does that sound reasonable?

Edited August 24, 2013 by dude

[curtisa]

Yup. Google "PPIMV" or "post phase inverter master volume". The only drawback I can see is that any active controls you have in the negative feedback loop (eg, presence, resonance or variable negative feedback controls) will become less effective as the PPMIV is decreased towards zero.