Distorting/Not Distorting the phase inverter

[EETStudent]

Ive read around few places on phase inverter mods, and I read it mentioned a couple times about the phase inverter delivering a distorted or undistorted signal to the power amps. Can someone elaborate on this? Does a clean signal do as you would think, provide more clean output/headroom? Where as a distorted PI would generate more power amp distortion?

[tubeswell]

My understanding is that the distortion in output tubes can be increased if (all other things - such as tube load resistance, plate voltage, tube current - being equal) the voltage swing being delivered to the output tube grid is increased (i.e.; so as to clip the grid signal), which is dependent on the total gain that the stages in front can deliver, the PI being the last stage in that gain. Some PIs have unity gain, others (Paraphrase and LTP) have just plain old gain (in addition to their inverter function).

[EETStudent]

Tubeswell, thanks for you post! This gives me a chance to ask a few questions Ive wanted to ask for a while.

Ive seen the term 'voltage swing' a hundred times. Can you elaborate on what that means exactly?


As for clipping the grid signal. Ive read that different level power tubes require different AC voltage to fully saturate the tube(notably, the el84 requires about 28vac). Tubeswell, I cannot for the life of me locate an AC voltage going into the power tubes. Ive put my probes in what seems to be all possible points, but I never get anything. No more than a few mV, which is what you get by just touching the probes. So, when checking the ac voltage gain into my power tubes(el34) where do I probe????

Again as for clipping the grid signal. I really like power amp distortion. To acheive maximum distortion, should I go with an inverter OTHER THAN the split-load type? From what I have read, the SLPI actually reduces the power amp distortion, which is not what I want.

Finally, what tests can I do to determine if I am clipping ANY stage in the amp, preamp and poweramp. Are there some readings I can take, or do I need an oscilloscope at the grid for the particular stage?

Any answers would be greatly, greatly appreciated

[Paul P]

EETStudent, you should get yourself an oscilloscope. Look at your local
classifieds, you should find something reasonable for 100$ or less. Make
sure it works before you buy it. Then you'll be able to answer your own
questions. Really, I couldn't imagine playing around inside my amp without
being able to see what's going on. A good multimeter and a signal generator
of some sort are pretty much required too. If you were just building a kit
you could probably get by without but you want to know what's going on
and that takes test equipment.

As far as the signal to your power tubes, if you don't have ac volts on your
grids you won't have sound coming out the speaker either.

Paul P

[gmoon]

To acheive maximum distortion, should I go with an inverter OTHER THAN the split-load type? From what I have read, the SLPI actually reduces the power amp distortion, which is not what I want.

I'm not sure how the power amp distortion from split-load phase inverter compares to any other inverter type.

But the split-load phase inverter doesn't balance the output signals very accurately, and that asymmetry alone should result in distortion...

[BackwardsBoB]

An oscilloscope would be good.

The AC on the power tube grids comes from your guitar (or a signal generator, or an accordian or something). The signal swing required to reach saturation or cutoff depends on the transconductance of the tube and any negative feedback provided by a cathode load, along with other stuff.

You can overdrive the PI, or provide enough headroom so that the power tubes overdrive first. It's possible that your favorite power-amp overdrive sound is really PI distortion, and different PI configurations will sound different when overdriven.

It is possible to detect overdrive distortion in the preamp or PI by measuring the DC voltage on the anodes of the tubes in the presence of a constant AC signal. The clipped part of the output will result in a DC shift. This gets crazy on the power tubes, where the anode voltage swings much higher than in the quiescent state, with mucho HVAC, and won't read right in class AB anyway, and if the next stage reaches saturation, you'll probably see it on the voltmeter too. If you can insert some attenuation between the PI and the power tubes, perhaps by splitting the PI tube anode resistors and taking the output from the tap in the middle, you should be able to tell with your ears if the distortion is in the PI or the power tubes by comparison. If the amp gets quieter and the distortion doesn't change, the distortion is in the PI. A negative feedback loop will fight you on this.

A scope makes observing and optimizing overdrive distortion much easier.

[tubeswell]

Ive seen the term 'voltage swing' a hundred times. Can you elaborate on what that means exactly?


As for clipping the grid signal. Ive read that different level power tubes require different AC voltage to fully saturate the tube(notably, the el84 requires about 28vac). Tubeswell, I cannot for the life of me locate an AC voltage going into the power tubes. Ive put my probes in what seems to be all possible points, but I never get anything. No more than a few mV, which is what you get by just touching the probes. So, when checking the ac voltage gain into my power tubes(el34) where do I probe????

Again as for clipping the grid signal. I really like power amp distortion. To acheive maximum distortion, should I go with an inverter OTHER THAN the split-load type? From what I have read, the SLPI actually reduces the power amp distortion, which is not what I want.

Finally, what tests can I do to determine if I am clipping ANY stage in the amp, preamp and poweramp. Are there some readings I can take, or do I need an oscilloscope at the grid for the particular stage?

Any answers would be greatly, greatly appreciated

Hi EETStudent

As PaulP has said already, you need an AC swing on the grid to feed the signal into the tube, otherwise ya don't get any signal at the plate. (That's why you have coupling caps between stages, to stop the DC coming from the plate of the previous stage, but let the AC pass through, to the grid of the next stage).

As for the swing on the grid itself, it goes from more-negative-with-respect-to-the-cathode to less-negative-with-respect-to-the-cathode, and it only does so when a signal is going through it. So if you measured a grid pin without a signal, you wouldn't get any voltage swing. According to Jack Darr, a typical 6V6 runs clean with about 25V Peak-to-Peak on the grid (i,e, a swing of 12.5V either side from the negative bias point at which the grid is biased).

As far as testing for clipping goes, I guess if you put a scope on the output tube grid with a good sine wave signal going into the input, you would clearly see the waveform and determine accurately where the onset of clipping occurs. I don't have a scope, so I just use my good-ole ears.

Paraphrase Inverters and Long Tail Pair Inverters both have complete triodes that are working on each side of the inverter, so you get 'proper' gain out of such a PI (depending on how they are configured). If you want more gain going into the output tube grid, you can either have more driver stages before the grid, or change to higher gain tubes in the existing driver stages (if that's an option), or twiddle with the cathode resistors, plate resistors and B+ pre-amp supply for existing pre-amp stages (which will do different things to the signal from each stage as well), or stick a stomp box in front of the input.

[EETStudent]

Alright some pretty good stuff, thanks guys. Still though, I hear terms like 'voltage swing', and i just dont know what that means. Change in voltage? Increase/decrease?

BackwardsBob, a couple questions about your reply.
You said the AC signal on the grids of the power tubes comes from the input signal(guitar in my case). Well, I did a test where I stuck one lead on the grid one of my EL34s, and I put the other lead to ground and I cranked the controls and strummed violently. The most I mustered up was about 7vac. I mean, this cant be right, can it? It seems that Im generating more power amp distortion, but if 7 is right then I suppose not? It should be way higher I know for el34s. What do you think about this?

You mentioned either driving the PI or providing headroom so that the power tubes overdrive first. This doenst make any sense to me. Well maybe it kinda does, but I cant see the difference in the two. Overdriving the PI would surely result in overdriven power tubes, right? Or proving a clean PI would result in clean power tubes, right? I have feeling Im wrong about this, but explain to me why.

Lastly, you talked about checking the dc voltage on the plate of a particular stage, where the clipped part results in a dc shift. What do you mean by a dc shift? Do you mean that the difference in voltage with a present ac signal is an indicator of how 'hard' its being driven?

[Steve Conner]

Looks like a good discussion

First of all, PI distortion vs. power amp distortion. EETStudent is right in that it's hard to tell them apart. This is because the usual clipping mechanism is that the PI tries to drive the power tube grids more positive than their cathodes, which makes the grids conduct like diodes to the cathodes. The PI can't supply enough current to overcome this, so the signal gets clipped hard. So the power tubes are actually limiting their own output by clipping the signals from the PI!

Every well-designed tube amp should have enough voltage swing from the PI that it clips in this way, and not within the PI itself. If it didn't, it wouldn't have made as much power as its competitors with the same tubes.

Voltage swing is just a casual term that means the amount of AC output available before clipping. You should at least qualify it by saying "peak" or "peak-peak" or whatever.

The waveforms generated by strumming a guitar are complex enough that the AC volts reading on a DMM won't say much meaningful. To investigate clipping behaviour, you really want a scope and a signal generator.

[BackwardsBoB]

You could overdrive the PI and keep the output tubes from being overdriven. For example, my beloved Carvin X-series amps sometimes have a 25%-50%-100% power switch that switches in various resistors to drop the supply voltage to the PI. I'm not crazy about this feature, but at low voltage, the PI distorts and the output tubes reproduce this. Different PI architectures will produce different sounds when overdriven, and they can be different from what you'd get from a preamp. You could also get into this situation if you switch output tube types to something lower gain, or if you add cathode biasing without adding a bypass cap around the cathode resistor.

If you've got enough headroom, you can get a clean PI output and overdrive the output tubes, but as Steve Connor points out, the drive waveform will be distorted when the power tubes saturate.

Regarding the DC shift, if a preamp or PI tube isn't clipping, the DC voltage at the anode when amplifying a waveform shouldn't change much. If it clips, the DC voltage will shift, since the positive or negative going excursions of the signal will be clipped. If the tube is driven to saturation, the voltage will go up, and if driven to cutoff, the voltage will go down. This is really only useful if you are laboring without a scope, and it works best with a signal generator. Most folk with a signal generator have a scope.

Now since the coupling cap has to pass this signal to the next stage, we can see that it's response to these DC shifts has the potential to be another source of distortion. I find this interesting. The first overdriven tube gives the signal a flattop which will include low-frequency content due to the same thing that makes a DC shift with a signal generator. The signal passes on to the next stage, and if this stage is overdriven, the other polarity of the signal will receive a haircut. Now if the coupling cap is very large, and the input impedance of the next stage is high, the result will be a constant amplitude square wave at the output of the second tube, independant of at least short-term variations in the input amplitude. But if the cap is small, or the input impedance of the next stage is low, or both, you'll still get clipping, but some of the dynamics of the input will be preserved as the coupling cap blocks the low frequencies that represent the DC shift. A voltage can form across the cap that represents the envelope of the input waveform, and this would vary the level at which the second haircut occurs. You'd see the low-frequency voltage variation on the coupling capacitor as a change in the otherwise constant amplitude of the output of the second stage when both tubes are clipping. For example, a smaller cap might preserve pick attack better, though you wouldn't see the difference testing with a signal generator's constant sine-wave output and an oscilloscope. You'd need a tone-burst signal and a way to trigger on it, and the visual result would be hard to correlate to how the amp sounds.