Why is the input cap value to PI stages much lower than the FB input?

[nevetslab]

In the process of listening to a fresh restoration of my 44 yr old Traynor YBA-1A, before making any major moves in mod’s, I found myself puzzling over why the input cap value to the Phase Inverter stage is 1/5 the value to that on the Feedback side of the PI stage. I went and looked at Ampeg’s, Fenders, some Mesa amps, Orange, Vox amps and others.

All the Fender amps I looked at were at least 1/5 the value or way smaller. Some Ampegs were the same value, others 1/10, some Mesa amps were 1/5 the value, others were the same value. The Vox amps I looked at also the same value, Orange Amps were 1/2 the value. I didn’t spend a lot of time looking.

The main reason I can see is that input cap’s value (to the PI stage) sets the LF corner of the power amps’ frequency response, and you’d set it high enough so you don’t drive the output transformer into hard saturation at low frequency, while still having good low frequency response. Guitar amps will usually be set an octave higher than bass amps. If you have the budget in manufacturing for a really good output transformer with full power @ 30Hz before it folds up, you can set the LF corner lower than a cheaper OT.

The OT on the Traynor YBA-1A isn’t what I’d call huge. It’s a 50W amp, and will produce 50W into 8 ohms @ 1% THD, which is the onset of transformer saturation @ 30Hz. I haven’t fully characterized the OT yet, but the non-linearity curve begins ascending rapidly below that, and the tubes are at visible clip (EL-34's).

The -3dB LF corner is 63Hz with the stock values (22nF), while changing to 100nF dropped it to 34Hz, an octave lower. I gave a listen to it, and liked the change, giving a deeper bass response. At real high levels with my 5-string Fender Jazz, that might be another story. I’m still on the fence with this OT for use with KT-88’s and a separate screen supply for them.

[Chuck H]

It really just comes down to the NFB loop design being an all band effect for the purposes of stability. Snarky problems with ripple, ESR related cross coupling at filter nodes and even EM interference at random can cause LF oscillations if the loop is not capable down to DC. That said... I have done NFB loops that filter LF from the path without issue, BUT, good design practice dictates otherwise.

Conversely the PI coupling cap might be valued for an intentional rolloff. But make no mistake. The frequencies that must be accounted for in the power amp aren't limited to those passed by the preamp outside of those amps that operate in a perfect world. Which none do.

[potatofarmer]

Also the input impedance of the PI's grids is huge, ~2-5x Rg.

[Chuck H]

Yes, that too. I was going to mention it, but it seems from the post that nevetslab already gets that. He mentions it and offers that he calculated a large difference in pass frequencies anyhow.

It may also be an economic issue WRT amp design. Most NFB loops don't include a DC blocking cap. Since passing frequencies far below what is needed doesn't really have any drawback, adding a cap would simply add expense to the design.

[nevetslab]

I looked at that too. In the case of Fender Twin Reverbs, that input cap to the PI is like 1nF, but you have the bootstrapped grid bias resistors of 1M, so there's still very good LF response. On the Traynor, where they're using 47k's instead, bootstrapped, the impedance calculated (from measurement data) was 2X (around 115k). I'd guess the FB coupling cap's value at input of the PI is much larger (or the same in many cases) so it accounts for good LF in the loop. On amps that have no feedback, such as the Vox AC15 & AC30, Orange AD30, both values to the PI stage are the same. I didn't notice any lacking NFB having different values.

[potatofarmer]

So the better question is why is the second cap so large? I think Merlin wrote in one of his books that it's largely historical and recovery from blocking distortion can be improved by making it smaller. Just another example of the time-honored tradition of guitar amp cut & paste.

[Chuck H]

Are we talking about the global NFB or the local NFB? in the case of the local NFB bypass cap (usually .1uf on about every guitar amp that uses one) that value is supposed to be 10x the input cap. I'll guess this assumes 1M grid bias resistors and I'll assume that the 10x "rule" should change according to grid circuit impedance.

[nevetslab]

In this particular case, there is only the one NFB loop, and it is the 0.1uF cap that is very commonly used on the LTPI circuits, as used in this Traynor YBA-1A, which uses the old 59 Fender Bassman circuit, apart from the NFB value. 100k for the Traynor, 27k for the Fender.

[Chuck H]

Some of the discussion made me think it was the local NFB of the PI since the .1uf cap value was mentioned. Though there is no NFB "circuit", it's local NFB, the .1uf cap on PI circuits IS a feedback related component. The .1uf cap incorporated into the FB "loop" circuit on amps is normally part of a "presence" circuit and serves to REMOVE feedback according to it's value. Which does change with impedance and you're right to notice that the .1uf value seems to be used almost universally regardless of this. There are even amps and amp mods that stick to the .1uf value with such relatively low circuit impedance as to make the effect useless for anything but risking oscillations, adding little or nothing in the audio spectrum to the circuit function.

[pdf64]

'Traynor YBA-1A, which uses the old 59 Fender Bassman circuit, apart from the NFB value. 100k for the Traynor, 27k for the Fender'

Bare in mind that with the 5F6A, the global feedback loop is sampling the signal voltage of a 2 ohm load.
An 8 ohm load circuit, eg the YB1A, of the same power level will have twice the signal voltage.
Hence it would be appropriate to scale up the feedback resistor value, ie to 54k, to keep the feedback ratio the same.
For reasons unknown, Traynor doubled up again on the value.
Pete

[nevetslab]

I also misquoted the NFB, as there is the local feedback, being that used to increase the input impedance of the grids of both inputs of the LTPI.

One of the things I've noticed on a lot of amps is WHERE the series cap associated with the Presence control is located (NOT the input from the NFB loop). In the case of this Traynor, and other amps I've serviced, there's DC on the pot wiper, so there's wiper noise when you adjust it. On this amp I'm rebuilding, I put the cap ahead of the pot, both in series to ground, eliminating the DC noise. I hadn't thought about the relationship of the output load with regards to this seemingly universal value of 0.1uF. Very good point. I've heard many a presence control having little tonal effect, and others with considerable effect. Blinders on again.

[nevetslab]

I forgot about the 59 Fender Bassman having the four 8-ohm speakers in parallel, presenting a 2 ohm load.

I haven't yet played with the feedback value on this amp. It's on my mind to do, as part of my quest in rebuilding this Traynor YBA-1A is to increase the power output above 50W/8 ohm. Good point you brought up on the FB value for the scaled ratio, and what Traynor did, doubling it again in value. I just looked to see what they did with the YBA-3. It too is 100k, what we would expect, it being a 100W amp. Makes me think production oversight, or, simply found it didn't matter?

[Chuck H]

On this amp I'm rebuilding, I put the cap ahead of the pot, both in series to ground, eliminating the DC noise.

I'm having trouble picturing a circuit that would do this without adding a resistor. Does the entire NFB signal run through the cap then? That would eliminate NFB on the LF.

[nevetslab]

Marshall JCM 800's and 1959 Plexi's have the 0.1uF cap ahead of the presence control pot, wired as a rehostat, or at least between the top of the pot and the wiper, so there's NO DCV on the pot. I've added the circuit from one of the Fender Bassman's #5f6a circuit for clarity:

bassman_5f6a_schem.pdf

It is just the series cap & variable R to Gnd off the PI side of the FB resistor. There 's DCV on the NFB from the cathode circuit of the tail of the PI. Fender and Traynor had the cap connected to Ground, so you had DC on the wiper, which, of course makes noise when you turn the knob.

[Chuck H]

I'm still unclear about how you're keeping DC off the pot and maintaining the integrity of the FB loop without involving other components. I'm familiar with the stock circuit. I wanted to see yours.