Need tips on using interstage transformer - phase inverter

[Phil_S]

Gibson Minuteman has an interstage TX... I have one on the shelf. This is very intriguing.

Yes, I own one of these, GA20-RVT c. 1965. It's been in a closet for several years. Interesting amp. Stock configuration, the tone is brittle. I did a few "improvements." It's an OK amp. It's got great reverb and tremolo. It uses puny i/s transformer. IIRC, the turns ratio is about 2.4:1. Mine blew the transformer (not sure why) and I thought it was made of unobtanium until I found an eBay auction where a guy was selling the whole transformer set for dirt cheap. So, I got lucky. A Hammond replacement did not work well. MM makes a pricey clone of it.

This Gibson uses half of a 12AU7 to drive the i/s transformer. There is a 6L6 amp in the 1965 line up that uses the same transformer also driven by a 12AU7. This suggests to me that current is more important than voltage [gain/swing] in driving the transformer. I think that is a hint for the design work. I suggest looking at the GA95-RVT Apollo or the EA300-RVT schematic.

[stephenl]

I adapted this from a HiFi design. My favorite build by far. I also have PowerScaling on this.

Interstage Driven KT66 PA.jpg

[Roe]

The Fender ps400 is a class AB2 output stage is that also what you are shooting for?

yes, I would prefer ab2 over ab1.

my reasons for trying an interstage transformer are:
(1) it differs from all my (16) other amps
(2) it gives better balance, and
(3) allows class ab2, while
(4) only requiring one triode or pentode

[Stevem]

In that builds schematic in post 270 , those cathode test points tp1 and tp2 come off of 1K resistors?

[dorrisant]

About 12 years ago there was a build posted somewhere that used a small choke in place of the cathode resistor in a PI circuit to provide more drive current somewhat like a class AB2 stage.

That would be more about providing a constant current source for the PI to improve its balance. That can be done with a MOSFET too, I believe.

I thought there was something that could be done with LEDs that could do the same.

[Colossal]

About 12 years ago there was a build posted somewhere that used a small choke in place of the cathode resistor in a PI circuit to provide more drive current somewhat like a class AB2 stage.

That would be more about providing a constant current source for the PI to improve its balance. That can be done with a MOSFET too, I believe.

I thought there was something that could be done with LEDs that could do the same.

Or a LM317 regulator!

[dorrisant]

Or a LM317 regulator!

That sounds adjustable. Do tell.

[Colossal]

Or a LM317 regulator!

That sounds adjustable. Do tell.

I think it was on DIY Audio...a hi-fi design by Bruce Herran. A push-pull pair biased with an LM317.

[Stevem]

I have not had time to find ( I will on the weekend )the build I posted about with the choke used a the PI cathode resistor, but I do now recall it was a Hammond 157G choke and the PI tube was a 6SN7.

[R.G.]

my reasons for trying an interstage transformer are:
(1) it differs from all my (16) other amps
(2) it gives better balance, and
(3) allows class ab2, while
(4) only requiring one triode or pentode

I'm solidly in favor of (1) - doing something unusual just because it pleases you to try it out has always been a good thing.

As far as (2), you can't necessarily say the balance is always better unless you get a truly high quality transformer. The necessity for special provisions inside transformers that carry wide bandwidth signals, on top of the price and weight, account for them being driven out of the main path of audio practice. You're going to need a good (ie. expensive) one. A somewhat subtle issue with transformers is phase shift and stability. Guitar amps already are limited in how much feedback they can use in the power amp, just because feedback from the speaker outputs has to have traversed the OT. A transformer introduces two time constants all by itself, and the magic number for oscillation is three or more. Using a second transformer in the power amp signal path can force you to use little or no feedback to get rid of motorboating, and it may still oscillate because of the high frequency rolloffs. Not that it can't be done, but there is some trickiness involved.

As noted, there are ways to fix the balance issues of a normal long tailed PI. The simple model of PI, with a single tube as a diffamp, does lose balance as it's driven hard. Part of this is the transfer function of a tube as a transconductance, and part of it is the way that long tailed pairs are tailed with resistors, not current sources. A current source in the shared cathodes forces the tubes to trade the same current back and forth, irrespective of the degree of drive. There are other refinements on the diffamp that could be tossed in as well.

For (3), running AB2 means you're trying to force the grids positive on the output tubes in the face of the transition from no conduction to grid conduction. Transformers take the approach that they trade high voltage, low current drive on the primary for low(er) voltage, higher current drive in the secondary. It helps. But you can get much, much lower impedance drive for the output tube cathodes more simply and cheaply by using MOSFET followers to drive the output tube grids. That will provide a much lower impedance and much less wobbling right at the crossover into grid conduction than a transformer ever could, and without the phase funnies, weight, and quality issues. There is some tinkering to be done with this, but it does work.

The nice thing about using FET followers is that they live naturally in the high voltage world, and don't require either heaters or additional tube spaces. They also make a cathode current source for the PI simple to accomplish.

This is, as the military folks say, a target-rich environment.

[Roe]

(1) it differs from all my (16) other amps
(2) it gives better balance, and
(3) allows class ab2

As far as (2), you can't necessarily say the balance is always better unless you get a truly high quality transformer. The necessity for special provisions inside transformers that carry wide bandwidth signals, on top of the price and weight, account for them being driven out of the main path of audio practice. You're going to need a good (ie. expensive) one. A somewhat subtle issue with transformers is phase shift and stability. Guitar amps already are limited in how much feedback they can use in the power amp, just because feedback from the speaker outputs has to have traversed the OT. A transformer introduces two time constants all by itself, and the magic number for oscillation is three or more. Using a second transformer in the power amp signal path can force you to use little or no feedback to get rid of motorboating, and it may still oscillate because of the high frequency rolloffs. Not that it can't be done, but there is some trickiness involved.

As noted, there are ways to fix the balance issues of a normal long tailed PI. The simple model of PI, with a single tube as a diffamp, does lose balance as it's driven hard. Part of this is the transfer function of a tube as a transconductance, and part of it is the way that long tailed pairs are tailed with resistors, not current sources. A current source in the shared cathodes forces the tubes to trade the same current back and forth, irrespective of the degree of drive. There are other refinements on the diffamp that could be tossed in as well.

For (3), running AB2 means you're trying to force the grids positive on the output tubes in the face of the transition from no conduction to grid conduction. Transformers take the approach that they trade high voltage, low current drive on the primary for low(er) voltage, higher current drive in the secondary. It helps. But you can get much, much lower impedance drive for the output tube cathodes more simply and cheaply by using MOSFET followers to drive the output tube grids. That will provide a much lower impedance and much less wobbling right at the crossover into grid conduction than a transformer ever could, and without the phase funnies, weight, and quality issues. There is some tinkering to be done with this, but it does work.

The nice thing about using FET followers is that they live naturally in the high voltage world, and don't require either heaters or additional tube spaces. They also make a cathode current source for the PI simple to accomplish.

Yes, (3) I'd be happy to try FET followers if I know how to implement it. The SVT/SSS driver is too complicated for most amps. And (2) using current sources instead of resistors sounds promising. Would this help: https://ampgarage.com/forum/download/file.php?id=19492?

I any case the interstage transformer has disadvantages: it makes negative feedback difficult to apply and a post phase inverter (larmar) master volume won't work

[Stevem]

My feeling on this whole thing is that from what I have heard there's a point where a greater more accurate balance makes the guitar sound too sterial even for clean tones.

Some minor phase canceling works for me guys what can I say?

[stephenl]

I noticed a mistake in my interstage drawing above - I ended up with KT66's + 8K primary OT. I originally had KT88's with a 4.2K Primary which sounded a little too clean for me. The interstage transformer that Heyboer made for me is bigger than most 25w OT's.

I had a pic of a choke tail PI amp stashed away:

bhl-15-values.gif

Martin Manning posted a nice source follower design in this forum that you can search for. I haven't gotten around to trying it yet - but plan to.

65DR_MOSFET_Driver.png

[martin manning]

...and there is a +/-160V bipolar supply posted in the Dumble section that runs from the standard 50-60V bias tap if you don’t want to add another small transformer. Several people have used that for SSS builds.
https://ampgarage.com/forum/viewtopic.php?f=4&t=29882

[R.G.]

My feeling on this whole thing is that from what I have heard there's a point where a greater more accurate balance makes the guitar sound too sterial even for clean tones.

Some minor phase canceling works for me guys what can I say?

That may well be. Music is and should be a personal and emotional experience. Technical accuracy may not be ideal. However, once you know how to make something very accurate, you also know how to make it inaccurate, and you generally know how to make it inaccurate in the way and amount that you'd like.

The long tailed pair PI is a special case of a differential amplifier. The general idea is that the current for both sides is set by the resistors in the cathode circuit and is kind-of a fixed amount. The "long tailed" part of the description comes from the idea that if the voltage across a resistor doesn't change much, neither does the current. So the common resistor is made "long" - a big resistor with lots of voltage across it, and relatively little voltage change. Using a resistor means that there is still some current change, especially on big signals. The current change on big signals affects both halves of the diffamp. It's a kind of common mode gain change. This effect is used with transistor diffamps for out-and-out gain change in multiplier circuits.

For a single-side-driven differential amplifier, you can think of the setup as a common-cathode, ordinary tube gain stage at the input, with the second stage being a common-grid amplifier driven by the cathode of the first stage. If the transconductance of the tubes is high and the impedance at the cathodes is high, you get nearly the same gain from the common cathode input stage and the common-grid second stage. However, tubes being tubes, the transconductance is never high enough to make this work perfectly, and the gain of the second, non-input stage is a bit lower.

You can make up for the lower gain by diddling the plate resistor values. That's why many long tailed pair PIs have an 82K plate resistor on the input and highest gain side, and a 100K on the plate of the second half. It helps equalize the gains, and restore some balance, if a bit crudely. And this tells you one way to make the sides of the plates NOT balanced: un equalize the plate resistors. This works even if you have theoretically perfect tubes or high gain solid state devices. You can even make it dirtier by making one or both plate "resistors" have non-linear things like diodes and such attached to make it really unbalanced and dirty in ways that you choose.

Putting a constant current source in the cathodes makes the balance of the LTP better because it raises the gain from the input to the cathode of the input stage more toward unity, and does not eat signal from the input side cathode, leaving more for driving the common-grid second stage. It also forces the sum of the two halves to be truly constant, so any increased current from one side is truly taken out of the other side. Again, you can mess with this by making the constant-ness of the current source on the cathodes be less constant.

I'm a techie - it makes me feel good to know how things work, and how I can tweak them.