Express Output Waveform Shots

[jazbo8]

IME the output valves(tubes) clip first due to grid conduction (not cut-off). At this point, the PI still has headroom left.

The wording I used was lifted from Merlin's book, I think "cut-off" just meant "clipped" so either cut-off or saturation would qualify, with the end result of the tubes running out of gain.

So, IME, the NFB loop does collapse / spiral out of control, or however you want to describe it, but this has consequences and produces a discontinuity in the response of the amp - a tiny change in input produces a very marked change in output. The NFB certainly doesn't just stop or 'disappear' gracefully.

I was speculating that the NFB in the TWE drops out fairly early on (comparing to the other amps), so the "sweet spot" that most players use is no longer dependent on it. Otherwise, the abrupt transistion from clean to overdriven of the NFB would be quite noticible, which is not the characteristic typically associated with C2M. I hope some of you agree with this line of thinking...

Again my experience, but my Express does a lovely clean.

I was exaggerating "TWE does no clean" bit to make a point, it certainly has a lovely clean tone. But relative to other amps, perhaps it still has a bit more dirt or you have to dial the volume way down on the guitar or play very softly.

[cxx]

That could be the case if the PI had the grunt to drive the power grids into AB2. But they hit a ceiling and can't go any further. That effectively breaks the feedback loop before the PI itself clips, and the feedback return signal can't get any bigger.

Also remember in all this that from a feedback/control theory point of view, the open loop gain PI input to speaker output is relatively low, and the amount of feedback is relatively small. So the amount of correction provided by the feedback is only partial.

So you are saying that you run into the positive grid scenario prior to any clipping by the PI and the PI can't provide the current.

Seems like the PI has a gain of about 28, the power tube of about 11, the transformer about 1/28 to the 8 ohm tap. The voltage divider back to the PI of about .05 for a net of .55 . Could be I'm missing something.

You're correct, except the screen power chain resistor in place of the choke makes all the difference between the two amps. It creates the sag of the PI power node that allows one side of the PI clipping to sneak in under the power grid clipping threshold. In a Marshall it has the same PI input grid clipping creating asymmetry (although not as much as an Express) but the choke screen power chain keeps voltages higher and both sides remain power grid clipping dominant when the amp is heavily overdriven.

What does this mean "has the same PI input grid clipping creating asymmetry "? Are you still talking about the power tube grids or something else?

Are you saying the entire power rail sags except for the bias supply, and the relationship changes between these voltages helps avoid power tube grid clipping?


Seems like my understanding of power tube clipping may have been wrong. The more I look the more it seems like it's all about the grid going positive and clamping. I haven't found an example where it hits the rail. All the data sheets where push pull examples are given show the bias to be just negative enough to accommodate full clean power. Anyone have an example where you run out of plate voltage first? You see this in preamp triodes all the time.

[jazbo8]

For those that do not have Merlin's book, here is excerpt from Aiken Amplification on NFB:

"Negative feedback makes the amp sound "tighter", particularly in the low end, where the speaker resonant hump has the most effect on amplifier output. This is better suited for pristine clean playing or a tight distorted tone, while a non-negative feedback amp has a "looser" feel, better suited to a bluesy, dynamic style of playing. The other disadvantage of a negative feedback amplifier is that the transition from clean to distorted is much more abrupt, because the negative feedback tends to keep the amp distortion to a minimum until the output stage clips, at which point there is no "excess gain" available to keep the feedback loop operating properly. At this point, the feedback loop is broken, and the amp transitions to the full non-feedback forward gain, which means that the clipping occurs very abruptly. The non-negative feedback amp transitions much more smoothly into distortion, making it better for players who like to use their volume control to change from a clean to a distorted tone." hmmm... sound like he is describing C2M.

And on power tube grid conduction:
"The standard AC-coupled phase inverter or single-ended driver stages used in nearly all guitar amplifiers will not allow grid current flow, so they are class A1/AB1/B1 amplifiers."

Elliot Sound also has a good overview of the output stage (especially section 9), it's well worth a read as a refresher.

[katopan]

*Edit* - Bias voltages added to below as requested.

I've got some discussion to get back to on this. Sorry I haven't had much of a chance in the last week. But for now here's some more test results.

B+ voltage sag and HT current

This test was done in the same manner as the waveform captures, with a 50mVpeak (or 100mV peak-peak) 333Hz sinewave applied to the amp input. So the values will be directly related to the waveforms I have posted at the different Volume knob settings. Tone and presence all at 5 out of 10. One multimeter on the B+ voltage and a second on the HT line measuring current (I have a B+ fuse, so I pulled the fuse out and clipped the meter across the fuse holder). Then I did the test again with one meter on the B+ voltage and one on the screen supply node voltage.

Volume - B+ voltage - HT current - screen supply node voltage - bias voltage
0.0 - 411Vdc - 85mA - 403Vdc - -36.6Vdc
1.0 - 396Vdc - 140mA - 379Vdc - -36.6Vdc
2.0 - 381Vdc - 216mA - 323Vdc - -38.6Vdc
3.0 - 376Vdc - 247mA - 312Vdc - -38.9Vdc
4.0 - 374Vdc - 260mA - 309Vdc - -38.9Vdc
5.0 - 373Vdc - 265mA - 308Vdc - -38.9Vdc
6.0 - 372Vdc - 269mA - 307Vdc - -39.0Vdc
6.5 - 372Vdc - 275mA - 306Vdc - -39.0Vdc
7.0 - 372Vdc - 279mA - 303Vdc - -39.1Vdc
8.0 - 372Vdc - 281mA - 299Vdc - -39.3Vdc
9.0 - 372Vdc - 281mA - 298Vdc - -39.3Vdc
10.0 - 372Vdc - 280mA - 298Vdc - -39.3Vdc

Keep in mind that the bias voltage under signal is a bit misleading. The source resistance of the bias supply isn't ideal and the imbalance in power grid signals means they don't really cancel once everything's clipping. The power grid signal movement and magnitude feeds into the bias supply resistance and impacts the bias reading.

Clean to mean guitar signal

So I simply put the CRO across the signal coming from the guitar while I played both full guitar volume down to where it cleans up at my typical amp settings. Amp volume is around noon (5 out of 10) but either side adjusting for pickups. Les Paul with humbuckers is just a smidge under 5 on the amp. Strat style guitar is 5.5-6 on the amp.

So it's difficult to give a measurement figure for such a harmonically complex and dynamic signal, so please take these numbers as a guide. They are just ballpark comparison figures to compare to the 50mVpeak I used for all the waveform testing.

With the LP up full I can get pretty big peak values if I strum hard, but normal sort of playing with strumming chords, bigger peaks are around and sometimes over 500mVpeak but general in between peaks are something around 200-300mVpeak. Solo notes are 100-200mVpeak and die away from that pretty quick.

Strat style with volume up full and strumming chords gives bigger peaks above 100mVpeak but in between peaks are around 70mVpeak. Solo notes are around 40mVpeak and die away from that.

My CRO goes down to 50mV/div minimum so measurements of turning down for clean are pretty rough. But with the LP at around 3 on the guitar volume gives a good clean on the amp with no real volume drop, and it shows up as around 15mVpeak. My Strat is a bit funny in that the volume pot measures about 7% resistance set half way, so it does clean up earlier than what you'd normally get with a 10% or 15% volume pot. In fact it cleans up a bit quick, which is opposite to the normal problem where it cleans up too close to the bottom end of the pot travel. But it's clean around 6.5 on the guitar volume and on the CRO that is showing as peaks of under 10mVpeak. Remember I turn up the amp volume a bit for this guitar to compensate.

In the perspective of my 50mVpeak test signal that I used for the waveform captures, a real guitar is a smaller signal for clean. In the test waveforms it's clean just before clipping at a volume setting of 1 and by 2 is well into symmetrical power grid clipping. So my measurements of a real guitar rolled back off for clean supports the normal use amp volume setting of much higher than 1. To put that into some perspective my amp volume pot reads about 30K to ground at 1 out of 10, 70K at 2 and 187K at 5.

[jazbo8]

Thanks again for the measurements. So the guitar signal for the cleans is way down - hope I get the following right... You needed 15mV to get the cleans with the amp volume set at 5-ish, assuming a typical log taper pot, that's equivalent to 50mV input signal with the amp volume set just a tad over 3, so by that point there is already significant sag in the B+ and screen grid voltage, and the NFB loop is broken, etc.

Also found this quote from GK on the MV thread:
"In a non-master Express the power supply sag puts more of a limiter on the fully distorted volume bringing that down more in relation to the cleans making the net volume more consistent and usable."

So the order of importance for the C2M effect seems to be: B+ (well by default since everything depends on it) -> Screen/PI Voltage -> PI clipping characteristics.

[katopan]

Last but not least here is the test of increasing the PI tail resistance. I disconnected the 10K tail resistor and put a 33K in series for a total of 43K tail resistance. The 5K presence pot is still connected and feedback connected as normal.

PI gain is reduced a bit as expected, but the behaviour is again all the same. I couldn't include the PI grids in these waveforms. With the much bigger tail the loading effect of attaching the CRO is even worse and totally stuffs up the output waveform, so it wasn't worth including. But the output, power grids, PI plates and PI cathodes show everything anyway.

[katopan]

And here's the PI waveforms for the bigger PI tail resistance.

[cxx]

For those that do not have Merlin's book, here is excerpt from Aiken Amplification on NFB:

"Negative feedback makes the amp sound "tighter", particularly in the low end, where the speaker resonant hump has the most effect on amplifier output. This is better suited for pristine clean playing or a tight distorted tone, while a non-negative feedback amp has a "looser" feel, better suited to a bluesy, dynamic style of playing. The other disadvantage of a negative feedback amplifier is that the transition from clean to distorted is much more abrupt, because the negative feedback tends to keep the amp distortion to a minimum until the output stage clips, at which point there is no "excess gain" available to keep the feedback loop operating properly. At this point, the feedback loop is broken, and the amp transitions to the full non-feedback forward gain, which means that the clipping occurs very abruptly. The non-negative feedback amp transitions much more smoothly into distortion, making it better for players who like to use their volume control to change from a clean to a distorted tone." hmmm... sound like he is describing C2M.

I had read this description. I know it is probably heresy, but seems like running out of gain may have been attributed to the wrong stage in this quote. The PI is the one that is the differential amp providing the corrected signal to the power tubes. The NFB signal results in a lower input to the power tubes. The PI is the stage that runs out of gain when the NFB fails to correct the power tube input signal. You can see it as it approaches the limit. The PI gets waveforms get pointy as it tries to correct the flatness of the output. When the PI starts clipping it can't make the pointy waveforms any longer.


Katopan, Thanks again for all your work. The number of traces in this thread is amazing.


I was interested in your sag results but was wondering about the bias supply. A constant bias voltage would shift the bias significantly as the B+ and screens sagged. I would think that the bias supply was relatively unaffected by the sag, but I'm not sure .

You have investigated many avenues and I see that it's a work in progress, but I'm not sure if any conclusions have been drawn relating to the special nature of the express. There has been some speculation but it would be interesting to put down what is known or postulated.

[jazbo8]

I had read this description. I know it is probably heresy, but seems like running out of gain may have been attributed to the wrong stage in this quote. The PI is the one that is the differential amp providing the corrected signal to the power tubes. The NFB signal results in a lower input to the power tubes. The PI is the stage that runs out of gain when the NFB fails to correct the power tube input signal. You can see it as it approaches the limit. The PI gets waveforms get pointy as it tries to correct the flatness of the output. When the PI starts clipping it can't make the pointy waveforms any longer.

You lost me there... Please correct me if I am wrong, I thought GNFB occurs in a closed loop and the gain A is the total of the PI and the PA, the NFB factor is a fraction of the full output, once clipping occurs within the loop, whether it is in the PA or the PI, there can be no excess gain left for feedback correction.

[cxx]

I had read this description. I know it is probably heresy, but seems like running out of gain may have been attributed to the wrong stage in this quote. The PI is the one that is the differential amp providing the corrected signal to the power tubes. The NFB signal results in a lower input to the power tubes. The PI is the stage that runs out of gain when the NFB fails to correct the power tube input signal. You can see it as it approaches the limit. The PI gets waveforms get pointy as it tries to correct the flatness of the output. When the PI starts clipping it can't make the pointy waveforms any longer.

You lost me there... Please correct me if I am wrong, I thought GNFB occurs in a closed loop and the gain A is the total of the PI and the PA, the NFB factor is a fraction of the full output, once clipping occurs within the loop, whether it is in the PA or the PI, there can be no excess gain left for feedback correction.

I'm saying that as long as there is gain left in the PI it can produce the signal necessary, which includes the negative of the distortion in the PA. I suppose that they both fall off once that limit is reached and I am thinking of them as sequential stages.

[EtherealWidow]

Katopan, thank you so much for posting those figures! Are all of those measurements VDC? It's really interesting to see the B+ drop levels off once the power tubes start clipping pretty hard at about "3" on the volume. It's crazy how this amp has so much sag with a seemingly stiff looking power supply on the schematic.

[katopan]

No worries EtherealWidow. Yes the B+ and screen power node voltages are Vdc (I edited the post to make it clear).

This amp overdrives really hard and draws a heap of current while it's doing it. The PT is my own winding, but has the same core lamination size and stack height as a 'proper' one, and the winding window is full. But as Sagitt and I have discussed in our winding transformer threads, if you design a suitable PT for an Express using the standard sort of current density and flux density, you just have to go up a core size to fit it all. I don't know the specs (turns and wire dia) of the 'proper' PT, but anyone who knows how to size transformers can quickly work out that it is undersized compared to conventional design. It still works and doesn't get too hot, but sags as you can see. Even if the 'proper' one has less turns with a thicker wire, or more turns with a thinner wire, it can't be that different because the copper still has to fit in that undersized winding window.

So it looks like I've got one more test to go back and do - I'll measure the bias supply with the same test signal at all volume settings and add to the post above with the other voltages and HT current. I am expecting that the bias will also sag (become less negative) and prevent the bias cooling off as the B+ sags. The B+ sag is PT sag as well as first filter cap ripple, and as the peak voltage out of the PT reduces, so must the bias supply. But it'll be checked and posted here.

After that I'm calling all the testing done unless any of you think I've missed something and want me to test out something else. Please let me know and I'll see if I can do it.

[katopan]

Hopefully we're reaching an understanding about the feedback. Once the power grids clip, the output can't go any bigger. The feedback signal is clipped by the power grid clipping, and so any further increase in the PI signal has a larger increase of signal into the power grids. But power grid clipping stays at the same level and you get signal clamping. The PI can't drive the power grids into AB2, so the output still goes no further. But then our PI starts to grid clip, clamping the 3rd stage positive output swing and pushes the rest of the PI input waveform downward creating the PI asymmetry. This asymmetry creates movement in the power grid signal on the downstream side of the coupling caps, and that in conjunction with the extreme PI power node sag (from the screen power node sag) leads to the lower side sneaking in under the power grid clipping threshold to become PI clipping dominant on the lower wider side and remaining power grid clipping dominant on the upper narrower side (what I call mixed mode distortion).

So yes the feedback helps with clean to mean. It does transition quickly compared to an amp without NFB. If you've played a Marshall 18 Watt clone of whatever variety you'll know there's a wide sweep of clean to overdriven with their beautiful screen compression and no NFB. The quicker transition of the Express gives it the ability to cover so much ground clean to really mean. It's touchier on the guitar volume pot. The preamp gain is such that you can drive harder into power distortion than a lot of other amps, but the asymmetry and mixed mode nature of the distortion means it still sounds good at very high levels of distortion. Amps where the output is purely very heavy symmetrical power grid clipping sound fizzy and are basically square wave generators at that point, so they don't get to the same levels of distortion while keeping remotely good tone. Other amps take the power distortion only as far as having it still sound good, and then start to mix in preamp distortion. But this arrangement doesn't clean up as well because preamps don't behave the same way, there's no screen compression and no real power supply sag as they go into clipping.

Conclusions are scattered throughout the testing and posts. The above covers a lot of what I would consider as a summary. I'll try and think how to present it better, maybe dot point, so we can have a full list of findings from all of this. It's been as interesting working out what isn't contributing as much as what is.

[katopan]

I've measured the bias voltage at the different volume knob settings and added them to the post above.

Unless anyone's got more suggestions I'm calling the testing done. I'm glad some of you enjoyed it. Take the time to digest and if anyone has questions or their own observations from what I've posted or their own measurements, I'm more than happy to keep the discussion going.

Still planning a post sometime next week in which I'll try and summarise the observations all in one place.

[jazbo8]

Craig - great summary of the TWE's operation - it's the combination of design features that create its unique sound, but it's also a demanding amp to play - the player's touch and working the guitar's volume knob would take years of practice and not to mention pure skill - it's like a finely tuned sports car, best left to professionals

Thanks again for all your hard-work, I think it is the most in-depth documentation of a guitar amplifier I have ever seen! Valuable information indeed.

Cheers,
Jaz