Express Output Waveform Shots

[LeftyStrat]

I'm certainly no expert, but I did experiment with putting that third stage in other circuits. At the time I thought it was hard clipping, but in other circuits it just tends to increase compression and sustain when not driven hard.

Since it is squashing one half of the waveform, and the waveform is AC-coupled to the next stage, this modifies the zero-crossings, making them wider on the squashed part of the waveform, and narrower on the unchanged side, so any later hard clipped stages would appear like pulse width modulation as this stage's level changed.

I've even thought about ways to simulate this effect in a stomp box. An envelope follower tracking the input level could be used to inject a DC offset before hitting the hard clipping stages.

Remove the asymmetrical clipping, you lose the changing duty cycle and a large portion of the magic.

Center bias that third stage and add a bypass cap and you've essentially neutered the amp.

[katopan]

Thanks for the comments Lefty. I appreciate you dropping by.

Since it is squashing one half of the waveform, and the waveform is AC-coupled to the next stage, this modifies the zero-crossings, making them wider on the squashed part of the waveform, and narrower on the unchanged side, so any later hard clipped stages would appear like pulse width modulation as this stage's level changed.

I totally agree with this statement, and this is how it works in any other amp. In this case the squashed part of the waveform, top of the 3rd stage plate, becomes narrower. It's backwards! So something else must be happening here. Please check it out in the waveforms I've attached and let me know what you think.

Remove the asymmetrical clipping, you lose the changing duty cycle and a large portion of the magic.

Center bias that third stage and add a bypass cap and you've essentially neutered the amp.

Once again any other amp.... I mentioned at the start of this thread (albeit that was a long time ago) that I tried this on my low powered 6P1P version. Cathode resistor changed to 1K5 and 22uF bypass cap added. It made the duty cycle become more assymetric, not less. I'm going to do the same test on my Express just so I've seen this (and I'll post it here) in the full size amp.

[katopan]

I'll hopefully get a chance to disconnect the feedback and do more tests tonight. What's the preference on how this should be done? Just disconnect the 100K? Ground the end of it? Leave the presence pot and non-inverting side coupling cap as is, or move that side coupling cap to ground like you normally would on an amp without feedback? Will all those options make any difference anyway?

[LeftyStrat]

I appreciate your efforts. Like you, I want to understand the 'why' of these circuits.

[katopan]

Interestingly, if you look at the CRO trigger point (which is the point in time of zero crossing of the speaker output, positive edge) of the 3rd stage plate, it's sitting with duty cycle shift instead of at the equal duty cycle point, and yet the grid going in is symmetrical. This is before 3rd stage cutoff. In a continuous system if you go up and them back to the same point on the input, even if the stage is non-linear the ouput will also return to its starting point. I don't get how the 3rd stage could be changing its duty cycle input to output when there's no cap or anything to shift bias. And I got poo-poo'd earlier for suggesting that the coupling cap into the PI was charging up. I now agree with that anyway because the PI input will pull up the PI cathode upon grid clipping rather than draw grid current.

[katopan]

Got some testing done with the NFB disconnected. Wasn't sure how to set it up for what everyone would like to see, so hopefully it shows you something. This is with the NFB wire taken off the output jack and grounded (effectively putting the 100K in parallel with the presence pot). Only trouble is that the presence pot is still in the tail of the PI and the non-inverting side coupling cap was still connected to the normal point. The effect of this is that the non-inverting side grid still received a signal, although it was a bastardised version of local tail feedback, not global feedback from the speaker output. You can see on that side grid that the signal is different and smaller. You can also see that the clean gain has gone up on the speaker output waveforms. But other than that, not a lot has changed. The response is very similar. The 'interesting' bit where the duty cycle changes and mixed mode distortion happens is all the same.

Any other requests or ideas?

[katopan]

And the PI.

I didn't capture the 3rd stage as it was (as would be expected) exactly the same as with the NFB connected. It's not inside the loop and therefore is just dependant on the upstream circuitry and test input signal.

[dave g]

I don't get how the 3rd stage could be changing its duty cycle input to output when there's no cap or anything to shift bias.

The duty cycle shift results from the fact that when it clips only one side of the input wave, it changes the average value of the wave. Take the integral of the clipped sine wave, and you'll find that the area under the peaks no longer cancels out the area above the valleys - in effect, the distortion introduces an extra DC component that wasn't there to begin with. The decoupling cap blocks this DC, which shifts the whole wave up or down (depending on which side is clipped) relative to the x-axis. That's where the duty cycle shift comes from.

[ampgeek]

Very cool effort here Craig. Thank you very much for taking the time to do all of this!

Quick question: How would I determine the amplitude range for each trace?

For example: Looking at the PI Pin 7 with no NFB series I see a "10V" designation at the bottom of each trace. Is that the full Y scale span value or half of the span? Or other?

I also see a "-600 mv" designation at the bottom of each trace. What does that indicate?

Thanks again,
Dave O.

[katopan]

Dave g - Thanks for the input, and I fully agree with what you've said. If you clip the top of a waveform and pass it through a coupling cap, it moves upward and creates a duty cycle shift such that the top is wider and the bottom is narrower. But as I said to Leftystrat, in this amp it is backwards. Check it out specifically in the waveforms. The 3rd stage clips the top of its plate waveform because of compression leading up to cutoff. But it's the top that is then fed through the PI and power amp as the narrower side. It's backwards! At this point this is still a mystery to me.

ampgeek - In the example you've quoted the 10V at the lefthand side is the scale of the major divisions of the y axis. So for example the EL34 speaker output waveform first clips at 2.2 divisions = 22V. The 1ms is the scale of the x axis, corresponding to 1 millisecond per major division. The -600mV at the extreme right is the trigger point level and not relevant to any scale.

[LeftyStrat]

I'm having trouble finding the plot with the duty cycle reversed. Any chance you could overlay the third stage plot and the output. I'm trying to make sense of the duty cycle being reverse from what is expected.

[katopan]

Sorry this is just a quick post. But all the waves are in phase with each other with the trgger always off the speaker output. So check out the 3rd phase plate against the PI inverting side pin 6 plate. 3rd phase plate positive side compressed and clipped, and so should be the wider side by the above standard explaination. The 3rd stage plate positive side becomes the PI pin 6 plate negative side (inverting side of the PI), which should be wider but is actually narrower. The negative side of the PI pin 6 plate becomes the V4 plate positive side, which corresponds to the speaker output positive side. These also should be wider but are actually narrower.

I'd really like you guys to see this because the duty cycle shift is so often explained by the 3rd stage cutoff clipping. Not only is it actually backwards to this explaination, but the duty cycle moves toward being equal again as the 3rd stage peaks go into cutoff. It is quite peculiar.

[Tillydog]

Neat work.

.. check out the 3rd phase plate against the PI inverting side pin 6 plate. 3rd phase plate positive side compressed and clipped, and so should be the wider side by the above standard explaination. The 3rd stage plate positive side becomes the PI pin 6 plate negative side (inverting side of the PI), which should be wider but is actually narrower.

The duty cycle shift that you see on PI pin 6 is coming from the PI output being clipped due to driving the EL34 grids positive (see the no nfb case) - it comes in at volume ~2, but the 3rd stage doesn't start clipping until volume ~6 / 6.5.

Maybe the shift on the 3rd stage helps pulls the duty cycle back towards 1:1 at higher volume settings?

What does the amp actually sound like once the 3rd stage has started clipping? (I imagine it will be quite harsh.)

Have you measured how much your raw B+ is sagging vs power?

[jazbo8]

Have you measured how much your raw B+ is sagging vs power?

First, thanks to Craig for posting the CRO shots wihout NFB. Anyway, as far as I can see (which is perhaps not much), after the initial drop, the B+ pretty much stayed constant after Volume 2, and at that point the NFB was pretty much broken as well. After Volume 6.5, everything turned to square wave, so the "sweet spot" is between Volume 2 and Volume 6.5, where we can see the duty cycle shifting.

As mentioned earlier, there are several components responsible for this - in no particular order - the relative small size of the tail resistor, the small grid resistor of stage 3 and the screen resistor. I think we can rule out the NFB and B+ supply...

[ampgeek]

Thank you Craig!
Dave O.