Express Output Waveform Shots

[kyoungsteadt]

Interesting info, Kato. Keep up the good work. You are very close to cracking this nut.

[katopan]

I was going to be happy to let this thread die, but have since done some more testing some of you might be interested in. This one is for those who didn't believe me about the PI input grid conduction having a clamping effect on the PI input signal. As the clamping causes the PI input signal to move downward as it gets bigger, the input signal centre line compared to the PI cathodes obviously shifts downward. But the PI operation is all about the grid compared to the centre line of the PI cathodes, and the lower half of the PI input is now much wider at the cathode centre line crossover, and the upper half much narrower, matching the direction of the assymetry.

If you're thinking that it's not valid because my amp is not a full size real Express, please get your amp and a CRO and check it out. (Really, I'd love someone else to be posting this sort of test info for everyone to see with the 'proper' amp ).

[katopan]

I'm following this up from a few years ago, but I now have an Express to play with instead of just a 6P1P based amp. Surprises at every corner.

Here is a full range of speaker output waveforms into a purely resistive dummy load for both EL34s and 6V6s. Hard to call it the same amp with the different valves installed! These were taken with a CRO of much higher bandwidth than the other shots referenced earlier in this thread, and show more detail to the waveforms. If you feed it into a reactive dummy load (same electrical load as a speaker), which isn't shown, you get the expected inductive spikes added to areas of fast rate of change or slope.

Interpret as you wish.....

[Jackie Treehorn]

Hey, welcome back. Nice to see this thread again, too. I'm going to have to sit down with a cup of coffee and go through it again.

Regarding your latest traces, did you change the bias when you swapped tubes? It looks like the 6V6's are breaking up earlier, which would lead me to believe there's less bias voltage.

Also, which output transformer are you using?

Thanks!

[katopan]

Thanks Jackie. I did mean to post some basic voltage info along with the waveforms. No, I didn't adjust the bias for the 6V6s. Just swapped them over and turned the amp back on for more testing.

EL34s - 400V B+, -34.7V bias, 41mA plate current, 33W clean output power.
6V6s - 416V B+, -35V bias, 19.5mA plate current, 20W clean output power.

The 6V6s have a lower output because the load line is sitting well above the 'knee' of the Vg=0 curve and it hits the grid clipping point earlier. When I get the waveforms for the rest of the amp I'm expecting to see the power grid clipping come in around the same point for both types of valves, set by that -35V bias.

My OT is the one I wound myself, as per this thread - http://ampgarage.com/forum/viewtopic.php?t=19478 Primary reflected impedance 6.6K. Primary inductance 17.5H giving a -3dB cutoff freq of 60Hz. I've also done a frequency spectrum test and you can see the plot low end rolloff of the OT lines up with that calculated value. There is no high end rolloff within the audio band, so I didn't bother measuring the leakage inductance or winding capacitance.

More to come, especially that illusive 'output of the other side of the PI' none of the previous plots show, which I can guarantee won't be symmetrical with the first side.

[rooster]

Yeah, this is nice to revisit, Craig.

Man, I'm seeing all preamp distortion here. Is your 2nd cathode resistor 2.7K and is your cathode cap here a stock 22uf? Is your 3rd stage load resistor 150K?

Great seeing the 6V6 wave forms, BTW.

Does your PT produce 400VDC? Is it a typical with 300-0-300 VAC before rectification? Stock 1K choke? Well, whatever it produces (so long as the choke is correct) before rectification, can you run this again biasing the power tubes such that the plates drop 4-5VDC? With typical EL34 tubes this should be around 33ma. I would like to see the output waves with a lower bias point if possible.

Oh, when you say 33 watts of 'clean' power, your are referring to the amp at Volume 1? I think that's a little low actually, for a good Express. Well, IMO. Did I miss the voltage and frequency of the signal you are sending into the amp? It's too bad you show a wave that shows a clean wave and then the results of a boosted signal, with no volume change on the amp. Maybe something to duplicate a soft picking thing and then a dig-in picking thing. Eh, you know, what the Express was intended to do.

[katopan]

Hi Rooster. Up for any and all discussion about this. I'll try and answer your questions best I can.

OK, my amp is all stock component values as per the 'official' schematic in the downloads section, so yes 2K7/22uF, yes 150K load with a single 2.2nF coupling cap into the 3rd stage, 1K 'choke' resistor. I'm up front that it's not the downloads section layout, but is stable (barely, just like anyone's) and performs just like it should. My PT, although home wound, does produce 300Vac before rectification and I'm getting a B+ of anything slightly either side of 400Vdc depending on which side of the norm my mains happens to be.

I can run the test again at a different bias point probably after I'm finished testing the whole amp, if you still want by then. When you say lower do you mean cooler or warmer? Sorry, just a bit confused as if I drop (cool off) from the 41mA (70%) I'm at now to 33mA (about 55%), the plate voltage will rise rather than drop.

There's always debate about measuring output power of amps. If you drive it to just before clipping with a sinewave you load the power rail more than a 'real world' signal and will read lower than what's it's really capable of putting out. I won't go into huge detail unless you want as I've posted all my method elsewhere (Wattkins forum) but I strum with a real guitar and turn the volume to the point where the output waveform is just on the verge of clipping. In this case I had the amp set at the normal sweet spot around 5, and backed off the guitar volume to clean up to the verge of clipping, just like you do with this amp when playing it. That peak voltage is the max the amp will put out peak clean. Back to sinewave theory (even though one wasn't used for the measurement) Vrms = Vpeak/sqrt(2). Power = (Vrms)^2/R. Using this method was the first time I could get 13W clean measurement out of my 18 Watt, where with a continuous sinewave it measures about 11W clean. Phil has said a few times on 18Watt.com that that amp was known to actually put out 13W despite its name. I've also measured a number of other amps this way and got figures more like the published or generally known values. eg. Fender Deluxe Reverb 22W clean.

Voltage being sent to the amp is in the text at the top - 50mVpeak or 100mV peak-peak sinewave. Frequency can be read off the trace but it's roughly set to about 333Hz to be kinda midrange for guitar. Input signal voltage was kept at that the whole test. The amp volume knob was then changed to the values shown above each pic. I found that was a good setup to show the full response of the amp, clean, through to the start of symetrical clipping, to the assymetrical clipping sweet area where all the action happens, through to where it starts to head back to being symetrical again.

If my Express isn't doing what is typical I'm happy for anyone else to post clear detailed waveform captures to compare. But what you see is what my amp is doing. I've taken 3 years to contemplate and then build this amp, and to be honest a lot of the drive to do so was to be able to play one (I did get to play someone else's for 15 minutes once), have one and test the crap out of it to understand one. I've very little need for a big amp and my 18 Watt has served me well for the few times I have got to play on stage over the years.

Your comment about preamp distortion is interesting. Not quite, as the output waveforms above have plenty of power valve distortion (as can be seen in the clipping of the grid signals I'll post next), but all I'm going to say for now is I call what this amp does "mixed mode" distortion, and I don't know any other amp that does it. Don't worry. This range of tests is going to show everybody what they want to see. And if it doesn't I'll do any other tests anyone wants after the full set amp front to back are finished. Details of how this amp works have stayed in the bag for too long.

[katopan]

Next installment from tonight's testing. Here are the power valve waveforms. And yes, the intention for the first time all of us have ever seen out in public is to work back through the amp and show every signal node.

Note: CRO is set to DC and 0V is through the horizontal axis. This means not only the waveform shape can be seen, but DC component shift under load. The -35V bias can be seen as the centre point of the grid waveforms, and grid clipping can be seen to occur at 0V. The screen supply node voltage drop under load can be seen in the screen waveforms.

So keep in the back of your mind "mixed mode" distortion. Anyone see something interesting between V4 grid and V5 grid? What about V4 screen and V5 screen other than reversed assymetry?

*Edit* - Waveforms captured again triggered off the speaker output +ve edge. This means phase is aligned with the above output waveforms for every pic.

[katopan]

And the grids (max of 5 attachments per post allowed).

*Edit* - Waveforms reposted for reason stated above.

[jazbo8]

Craig,

Awesome stuff, thanks for taking the time to do such thorough job. I have a question on the charts - it appears that the grid and screen plots for V4 and V5 are shown in phase, which is great for comparisons, but for the plate plots, Volume 0.5 - Volume 3 are shown as out of phase, while the others are shown as in phase, was this done on purpose?

From the snap shot below, Volume 1-3's plots (in black) for V5 are shifted to the left:
[img527]https://dl.dropboxusercontent.com/u/132 ... eforms.gif[/img]

[katopan]

Aggghh, phasing! Thanks for raising this Jazbo8. I actually did intend to line up correct phase of each signal relative to the output waveforms I posted above. ie. Triggered to the positive edge of the thinner side during assymetrical clipping. Then I launched into the testing the left the CRO on positive edge triggering with no real regard for trigger level. It means that phasing doesn't line up as you have pointed out.

I'll redo the tests now with better regard and repost the attachments above for V4 & V5 waveforms. Sometimes it's hard to get the small waveform at 0.5 volume to trigger consistently, so that one might be hard to line up. Actually I just had a thought - I might trigger off the output waveform. My preference is to have everything relative to the real time phasing. So if you overlay plate waveforms for example, one will rise while the other is dropping. That means they can be overlaid and it will show what each is doing referenced to the same time trigger.

[jazbo8]

That means they can be overlaid and it will show what each is doing referenced to the same time trigger.

Yes, I agree if you purposely "flip" one side, then it makes comparing the two sides easier when overlaying the plots, again thanks for taking the time to do this.

[katopan]

As noted as edits in the previous posts, I retested the power valve waveforms and attached updated versions of the files. They are phase correct with the output waveforms posted earlier as I set the trigger off the hidden Channel 2 input connected to the speaker output +ve edge. And now it's not too late. Might have time for the phase inverter.

[katopan]

Here are the PI plates both DC coupled and AC coupled zoomed in to see more waveform detail. Also included the PI power node as a useful reference to see the PI headroom vs. power grid clipping point.

*Edit* - Oops, forgot to add the AC coupled ones. They're attached now.

Oh and by the way, the hiss going quiet and then coming back after a second or so when you hit a chord and then mute. That's the response time of the PI supply node. I quickly moved the volume control with my constant test signal. The screen power node sags and recovers very quickly, but the PI power node has that between 1-2 second lag in its response. Obviously the hiss going quiet then building up again is the change in gain of the whole preamp and PI with the power rail from B+3 on the schematic down the rest of the line. But this lag in dynamic response and the movement of the PI power node is a key aspect to this amp's response.

[katopan]

PI shared cathode connection and grids.

*Note* - As usual measurement of grid signals loads down the voltage at that point a lot. So magnitude of the grid signals is a lot less than it will be without the CRO probe attached. Keep that in mind. General signal shape and behaviour is still relevant.