Express Output Waveform Shots

[katopan]

Hi everyone. Firstly I should introduce myself. I've been hanging out over at the Wattkins and 18Watt amp forums regularly for the last few years. Reading through the threads here I recognise a number of members from over there. I've built a few amps including an 18W, 6V6 harp amp, submini push-pull, and recently done a lot of experimenting with some of the readily available Russian valves. Have known about Amp Garage for some time but until now was reading missing out on the attachments and member only areas. Been impressed with various clips of the Express and although I don't have a use for an amp of that size, a low power version really appeals. I remember when the AX84 4-4-0 appeared and have seen a number of other Wreck style amps floating around the net using lower power output stages. I've done a lot of reading here with much interest on theories on the 'clean to mean' and how the Express is so magic for what is a fairly straight forward amp. The question of what makes amps do what they do has been asked many times here, but definitive answers are not forthcoming.

I'm currently playing around with a low power Express based amp using Russian valves in this thread on Wattkins (login required) http://www.wattkins.com/node/17851 Built a version with 6N2Ps (instead of 12AX7s) and 6P1P power valves, then rebuilt it to accommodate 6F3P (equiv of the 6BM8) as PI/PA. I followed the AX84 Gothik ring layout approach instead of following the Express layout, but don't have any stability issues. Done some extensive testing on the Express preamp & PI and have looked into the effects of B+ sag, screen supply sag, screen compression, 3rd stage operation, dynamic shift of the PI operating point with 3rd stage output clamping by charging up of the 0.1uF PI coupling cap with PI grid clipping current, dynamic shift from discharge of the power grid coupling caps, etc. I'll be more than happy to share oscillosope captures and all the other info I have in this forum when I get all that organised over on Wattkins. Found some interesting stuff that was not expected and isn't covered by any discussion here that I could find.

So my question is does anyone have oscilloscope shots of an Express output at various levels of clean & overdrive? I've looked at every detail of the pics on John Polstra's website, but these seem to be the only waveform pics around. It'd be great to see some with a scale and I'm looking for specific detail which will help me match the PI output into the smaller power valves I'm using. At the moment I've getting great tone both clean and overdriven but clean to mean is changing volume more than I thought it would. Thanks, Craig.

[katopan]

Maybe I could ask another question... I've read threads by RJ and others with a lot of interest in using lower power power valves. I particularly share that passion. I remember one thread where RJ used split plate resistors on the PI, and 40% signal was fed to an EL84 output stage. Lots of threads mention using 6AQ5s and even a few look at the 6P1P and 6F3P that I'm trying out. But the threads all kinda fizzle out. Also without dropping the PI output somehow the drive level into the power stage when PI and then 3rd stage clipping occurs is all wrong.

Any advise on PI split plate ratios that suit these smaller valves like the 6P1P and 6F3P? I'm at the point using the 6F3Ps where the PI sensitivity and larger output swing is too different to a 12AX7 and making me think about going back to the 6N2P PI and 6P1P output stage. Then at least the PI operation will be back to what it is in a real Express and I can play with the split plate ratio to suit the 6P1P grid clipping threshold.

[guitarmike2107]

Hi Craig

Try this

http://home.polstra.com/amps/wreck1/scope/

[rooster]

The pics of the Express clipping are nice to see, not seen those before, thanks for posting this..

Unfortunately, as he gets to the last series, he does the 100k side of the PI/Output tube and not the 82K/Output tube. Had he done this the magic of the Express would be revealed. Can you contact this person and have him finish the job?

[martin manning]

So, if this guy did that what magic would we see? How would the pictures be different from the ones shown?

[rooster]

OK, you asked. You can perform this yourself if you have a dual bias meter around. Basically plug both tubes into the meter and amp, turn the amp up to 2 o'clock and beat the shite out of your guitar. You will see the power tube closest to the PT go ballistic while the other one will behave pretty typically.

The one side he tested was the typical one, the one being fed with the 100K resistor. If he ran his test on the 82K side, this would be worthy of a pic.

Do you know this person? If yes, tell him to run this again because he missed the real pictorial example of why an Express sounds like it does. He sort of sees some of this in the pics he does show, BTW, when he looks at the OT (speaker output), but well, he is not able to truly duplicate the spikes of a HB pickup with that test signal. Still, going the extra mile and scoping the 82K side, I think you would be able to see the contrast between the two sides and therefore get a much better idea of what is going on with the PI circuit and why the amp sounds like it does.

BTW, this is why, if you run your Express full out, you should swap your power tubes after each session - one tube is working much harder than the other.

[katopan]

Thanks, I really appreciate the replies. I've kinda jumped into this forum with a very specific interest. But I'm chasing this down and will be happy to share everything I find with you all.

The home.polstra.com pics are the ones I've grilled over. Rooster is spot on in that we just needed to see both sides of the PI output. Reason is because unlike other amps, they are different - more below. For anyone interested here is the AX84 thread where John Polstra discusses the waveforms posted on his site. http://www.ax84.com/bbs/dm.php?thread=328710 You'll note in that thread that he sold the amp and therefore couldn't do any more testing. Some interesting discussion but they don't quite hit the mark with a conclusion.

Biggest issue is I'm not testing a full size Express. If I had access to one on my bench for 3 hrs I'd document the hell out of it and post a full description of static waveforms and how things move with the dynamics. But I can work with a real pre and PI at Express voltages, just with different power valves. Throw in a heap of reading here and elsewhere over the years and some theory and hopefully it'll come together.

Ok, so a couple of things I've found so far which I haven't found discussed anywhere....

The common assumption is the assymetry in the output waveform is from and when the 3rd stage clipping kicks in. Any other amp with an overdriven (hot or cold) preamp stage moves the bias across the following coupling cap to make the output assymetric, and changes the duty cycle (adds 2nd harmonics, etc). But with for example the top of the waveform clipped, it will balance by moving upwards after the coupling cap making the clipped side wider and the non-clipped side narrower. In the Express it's the other way around where the clipped side (top side going into the PI) get narrower. If you measure what's going on across the coupling cap 3rd stage to PI you find that PI grid clipping is drawing a heap of current and charging up that 0.1uF coupling cap. This pushes the 'centre' of the signal going into the PI down and actually dynamically changes the operating point of the PI. This coupling cap and the PI input grid clipping at this extreme level of signal is acting like a simple clamping circuit and the centre of the waveform going into the PI moves a long way down causing the non-clipped side to become wider instead of the clipped side. Just as some proof too, if you change the 3rd stage cathode resistor from 10k to 1.5k or even warm biased at 1k (which gives too much gain on that stage and isn't useful for playing through, only testing) the 3rd stage output becomes completely balanced but you get more assymetry in the PI output, not less. This is because there's more signal now on the top which with the clamping effect pushes things down more feeding into the PI. Oh, and in the AX84 thread there's comment about how rounded the top of the 3rd stage waveform is for a cold stage cutting off, esp compared to Marshall, etc. That's because it's the PI grid clipping cutting off the top of the signal, and it comes in before the 3rd stage actually cuts off. It goes flat once the 3rd stage cutoff kicks in a bit later.

The feature where if you play a power chord and then silence the strings, the hiss has disappeared and then fades back up. A few people have said you get about 40V sag after the 1k screen supply resistor. I've simulated the full amp in Spice (I know, sims have limitations) and it says you get 15V screen supply sag from zero output to just on clipping, and a further 40V from just clean to full out saturation. You can replace the PI supply dropping resistor to set the PI and preamp supply voltages 55V lower, and test the difference with no signal. The hiss level isn't that much lower than it is at normal supply voltage. This feature isn't caused by power supply sag. I haven't finished investigating this one yet but it is likely that the PI temporarily staying in cutoff while the 0.1uF coupling cap discharges back to normal, or something like this somewhere else, is actually blocking the signal (noise, hiss) from the previous stages for a brief moment after being hit with a power chord.

Lastly if you disconnect the 1.5k power grid resistors from the PI output and connect them to ground, you can see what the PI is doing without the power grid clipping. A side note is you also don't get the power supply sag as the power valves don't have signal, but we can change supply dropping resistors to emulate that, and the result is just a bit of a smaller output swing anyway. We all know the output gets assymetrical (as described above) but part of the secret is that it's not just the duty cycle that changes but the height is also different between the two halves. The output of the half of the PI which drives power valve V4 has a wider top which means V4 which is running more than 50% duty cycle, V5 positive input is the narrower one or less than 50% duty cycle. This translates to the bottom of the speaker output waveform being the wider side. But as the assymetry kicks in the PI output feeding V4 not only get wider but also drops in height with a shift in bias voltage of the 22nF power grid coupling cap. Correspondingly the V5 side gets narrower but also increases in height. If the levels are set right by the PI loading and the PI power supply sag compared to the power valve grid clipping threshold, you find the V4 side goes into power grid clipping but then actually backs back out of it with more signal, where the V5 side just drives harder and harder into power grid clipping. The result at full bore is that the bottom half of the output waveform is actually dominated by PI cutoff clipping and the top half is power grid clipping. Mixed mode distortion! As Rooster said V5 is the one nearer the PT. It's getting driven so hard both with plate current and screen current. Meanwhile V4 runs cool because it's actually not being overdriven. That's specifically what I'd be looking for and trying to show with oscilloscope shots. I've seen it with my scope on the preamp & PI setup I have in my test amp.

Anyway, sorry for the long post but there's so much to explore here. This is all just a starting point. Hope some of you are interested!

[surfsup]

I've simulated the full amp in Spice

Interesting thread. Has anyone uploaded a tw spice file?

[katopan]

More than happy to clean up my PSpice (Orcad) Express file and accompanying valve model library and post it. Here in this thread or somewhere else?

I've done a fair bit of amp modelling in PSpice and compared to my findings of the real thing. Using the model files from the Duncan Amps site and worked out how to make them work as library objects. They model the bias and characteristic curves very well. The distortion or overdriven responses are rougher and although things like grid current, screen current, etc are included in the model, the overdriven responses have to be taken with a grain of knowledge of the reality. Nevertheless a valuable tool.

[vibratoking]

Very interesting. I would love to have your Orcad files and libs. I have a full verysion and have wanted to import the tube libs, but I haven't gotten to it quite yet. I would appreciate it.

[martin manning]

The asymmetry in the output waveform is a result of the LTP PI being driven hard, and this design will exhibit that behavior even when driven with a pure sine wave. When the third stage output (PI input) is clipped on the top (due to the cold bias), that part of the signal ends up coming through on the top of the output wave, and so it is actually on the narrower side. Note in figure 4 of the linked scope traces http://home.polstra.com/amps/wreck1/scope/ the 3rd stage output is showing some rounding of the tops, but the output wave is already quite asymmetric, and even more so than shown in figure 5 where the 3rd stage is clipped hard.

Rooster, this is what your bias meter is showing you, a higher average current on the 82k-side power tube, and this is the result of the PI's asymmetry.

This is a very typical LTP PI and power stage, and so none of this will be unique to a TW. I think the unique feature here is that the power amp is driven very hard beginning at very low volume settings. The two stages in front of the PI have no attenuation at all, and together make about 50 dB of gain. No wonder these are loud!

This means that a PPIMV, even a good one like the KF Type-2/Lar-Mar, will change the character of the amp. You maintain the asymmetry and other PI distortion, but lose the power tube distortion when it's turned down. An output attenuator would be the only way to keep it operating as intended, hence the TW Air Brake, I guess.

[Jackie Treehorn]

No, the pi is pretty symmetrical. You can see both sides looking at the output from the output transformer. Any imbalance in the pi would be apparent in it's linear range, too, since we're talking about a voltage difference/ac imbalance. It clips when it hits the bias voltage on the positive swing, so if the output tubes are both biased with equal negative voltages the clipping will be symmetrical.

When the bottom of the waveform gets wider as the amp is driven harder, it's because of crossover distortion/bias shift as the bias shifts closer to class b and the waveform shifts down around the zero crossing.

There's no significant output tube clipping in the express, you can see in the traces the pi is clipped whenever the output is. The good distortion is in the pi, so a post pi master could work really well and eliminate the crossover distortion as a bonus.

[martin manning]

Crossover distortion is about the middle of the waveform, not the width of the upper vs lower halves. A LTP will create that kind of asymmetry all by itself, in isolation, when driven hard with an undistorted sine wave.

[rooster]

OK, Martin, when you say this is 'very typical' in a LTP PI, um, no, it's not. It's the tail in this particular one really, @10K that's responsible for most of it. Change it to 22K and it goes away. And Fender went away from it. However, this said, if you take a PLexi and put it through it's paces, it still will not behave as radically as the Express. And yet it has the same LTP PI.

So finally you are left with the gain circuit before the PI as another difference. Cathode follower vs. the typical triode. But still, it is the amount of gain, the KF tuning thing, that offers up the sustain of the note within the preamp - the preamp 'bloom', if you will - that finishes with the BLOOM in the output tubes, caused as they try to deal with this crazy lop-sided event. Incredible really.

It's science, sure. But let's consider that it's science that is unique and quite organic in it's delivery. And let's not forget that Mr. Fischer DID figure this out and DID offer it to the world. And yes, the scope will show all of this, although a scientist might miss it. As fun as a scope pic is, reading the results is a complex and difficult art. Really, just like a mamogram is a simple pic, they say only one Dr. in a thousand can actually use it to the patient's benefit. Hm.

P.S. I also need to point out the 1K choke in the power supply as a partner to this BLOOM event. The Plexi circuit used a resistive value of 160 ohms, definitely taking away a lot of compression - which reduces the BLOOM factor as well.

[katopan]

Appreciate the discussion about this. You're right Martin that any LTP PI driven hard enough will have this behaviour. I changed the 3rd stage cathode resistor to make it warm biased and with a very symmetrical sine wave on its output even before it cutoff I got even more assymetry out of the PI. I suppose I'm just trying to point out that charging up of the 0.1uF coupling cap actually shifts it's operating point to go from balanced to unbalanced as the PI starts to grid clip the previous 3rd stage. Before clipping the PI is balanced with no assymetry. It actually shifts in its operating point.

Jackie, the symmetrical clipping you see on the positive side of the PI outputs is from the power grids clipping. If you disconnect the power grid resistors you get to see what the PI is really doing all by itself, and it is different on each side. Any bias shift in the power stage would apply itself equally to both sides, and you don't really get any with fixed bias anyway.

Rooster, the Plexi has no hope of driving the PI as hard because of the loss of the tone stack in between the cathode follower and the PI input.

I should get a chance to clean up and post the PSpice stuff later today.