PI input caps

[daz]

I was thinking about the Pi and the fact that in a typical marshall style PI with NFB you have a .1uf cap for the NFB to pass thru to the grid, yet on the main input signal side you have a .022 or even less. (i use a .0047uf) Note that as i often say here i am no tech, so began to wonder why this is. The PI is supposed to take 2 180 degree out of phase signal and send each to 1 side of the PA. So why would you want them to differ in frequency range? Why would it not be better to have the 2 signals close to the same? So i put a .0047uf on the NFB side to see how it sounds. Can't really try it much at this time of nite, but it seemed good like that. Maybe tomorrow i'll feel differently and toss it out. But what is the reasoning behind imbalanced frequency ranges for each side of the PA? I do realize the signal being derived from the main input side it should be the same, but after it's journey thru the PA and OT what if there are lower frequencies that get into the signal. But even if not, why a .1uf?

[Enzo]

Caps don't have a frequency, the circuits they are in has a frequency. Any frequency response of a cap is always calculated with other circuit component values.

That aside, the two caps do different jobs. The input cap passes the signal through to the amp, consider it is part of a larger circuit, and they may want to limit the range of signal that passes to the amp. The feedback circuit is there to reduce distortion and increase stability. And tone shaping. If you look at a presence control, it is a variable feedback system.

Look at your band PA system. Your power amp has a graphic EQ to make it sound good. You don't change the graphic settings from when your guys sing to when the higher frequency women sing. Your power amp is designed to sound a certain way. A smaller input cap may attenuate the lower end, but that isn't a brick wall, so whatever reaches the power amp, we want it to handle. On the other hand, this is a guitar amp, and no one said it needs to be flat.

[Chuck H]

It has to do with the IMPEDANCE of the circuit the cap is acting in. As circuit impedance lowers a higher value capacitance is needed for the same results as a higher impedance circuit. Not being really high tech (read=not going to do the formulas) all I can say is that the PI input is REALLY high impedance. Typically estimated between 2M and 5M!!! The circuit feeding the feedback loop is very low impedance. But it is raised by the series resistor. The .1uf cap you mention is part of the presence circuit. WRT the impedance of the NFB/presence circuit in a Marshall amp, the .1uf value passes treble and upper midrange frequencies. Which are bled to ground and effectively deleated from the feedback circuit. This is how the boost of these frequencies is achieved with the presence control. Now, the input to the PI being a very high impedance needs only a fraction of that capacitance value to pass frequencies well below those passed by the .1uf cap in the presence circuit. This is just the electronic facts of life. They are two different circuits with two different criteria. Even though they are somewhat linked together at the same tube the inputs for either circuit is at different sections of the triode.

If I were really cool I might spout some formulas for how to figure the pass frequencies for the cap in question on these circuits. But I've never found that to be that helpful. They are different circuits. Use your ears to choose an input cap for the PI. Anything over about 4700p probably won't make much difference. And use your ears to choose a presence cap. Values larger than .1uf will dig into lower mid frequencies, higher values will effect only extreme treble.

Not observed much is the fact that many amps (other than Marshalls) use very different NFB circuit values. Yet the presence cap of .1 remains consistent!?! When NFB loop values change there must be a corresponding change to the cap value if similar frequencies are to be passed. One of my builds uses a 22k series resistor in the NFB loop and a 1K resistor at the bottom of the PI tail (typical Marshall values would be 100k or 56k with a 4.7k or 5k) I use a .47uf cap for the presence control to pass the same frequencies as the .1uf cap does with the standard Marshall values in the NFB circuit. The reason the cap value is larger in my circuit is that I've lowered the impedance of the NFB circuit. The lower impedance requres a larger capacitance to pass the same frequencies.

[daz]

Thanks guys. A lot of that went whoooosh, but i got some of it. Chuck, i may have misread you here, but i meant the .1 at the grid of the PI, not the presence cap. I too have played with the .1uf presence cap (in both directions) but always seem to go back to .1uf like you suggested. for the record, my NFB comes off the 16 ohm tap and has a 33k resistor and 250k pot (yes, i know it's high) in series. The rest is typical marshall. I tend towards little to no NFB, but with the higher voltages i tried i was looking at maybe tweaking the NFB to be more usable and calm the brightness and hardness of the bigger preamp voltage. (read the my latest post in the other thread) Right now i feel adding enough NFB to do that kills the punch and aliveness of the tone too much.

[Chuck H]

Gotcha. I thought that you expressley said that you use a 4700p cap for the PI input and you mentioned the .1uf value pertaining to the NFB loop. And the only .1uf in the NFB loop for a Marshall is the presence cap. No matter. I think the subject was covered. FWIW I just use on line calculators to balance impedances with cap values. I CAN do it myself if I must. But I use any tools I can to make the design process easier (and more fun).

[Merlinb]

I was thinking about the Pi and the fact that in a typical marshall style PI with NFB you have a .1uf cap for the NFB to pass thru to the grid, yet on the main input signal side you have a .022 or even less. (i use a .0047uf) Note that as i often say here i am no tech, so began to wonder why this is.

The signal input cap can be anything you like. All that matters is what actually ends up on the grid, not what you might have filtered out before it got there. Once the signal has reached the grid, that is all the LTP cares about.

But you (usually) want the feedback to operate equally at all frequencies, so a big cap is used on the other grid. SO the circuit doesn't need to 'look' symmetrical as far as the caps are concerned. You can think of the input-side cap as being separate from the LTP and feedback loop.

[daz]

I think you guys have pretty much convinced me a .1uf there is the way to go and that changing that isn't going to change anything, at least not for the better. So when i fire it up today i'll listed for a few seconds while that .0047uf is still soldered in out of curiosity, but unless I can't deny an improvement which i see isn't likely to happen it will go. Thanks.

[yunger]

what does the cap in question do when there is no negative feedback?

[Chuck H]

what does the cap in question do when there is no negative feedback?

I think it reduces local NFB within the PI circuit. Increasing gain and improving LF balance.

[clyde1]

I read somewhere the lower (feedback) cap should be 10 times the capacitance of the input cap.

[Steve Conner]

The input impedance of a long-tailed PI can be extremely high. Sure there's a 1M resistor on the grid, but it's bootstrapped by the NFB signal. So even a small input capacitor can let through a lot of bass.

The input and feedback capacitors are made different values to help stability of the NFB loop at low frequencies. Each capacitor adds a pole to the loop, so you want to stagger the poles to different frequencies, to avoid building up too much phase shift at any one frequency.

If no NFB is used, neither point applies.

[pdf64]

'The input impedance of a long-tailed PI can be extremely high. Sure there's a 1M resistor on the grid, but it's bootstrapped by the NFB signal. So even a small input capacitor can let through a lot of bass...If no NFB is used, neither point applies'
Doesn't that grid leak resistor get bootstrapped by the local NFB inherent in the LTP, irrespective of whether there's additional global NFB?
Thereby increasing the effective input impedance, though not by as much as if additional / global NFB was applied.
Pete