Spice simulation and analysis of the amp's audio performance

My own experience has been that pickups can give out a bit more than they are usually quoted at.  I think that peak voltages are very important in guitar work.  The starting transients give you a lot of character.  It depends on a lot of factors of course.  Pickup format, loading (as I've bleated on about before), string choice, spacing from strings, picking style, using exclusive Bulgarian NOS varnish wire insulation to name just a few.  We're told that they are low output devices but I'm used to vinyl pickups and seismic transducers giving 5uV, that's what I think of as low!  I always aimed at the idea that you should be able to cope with 1V peaks without any problem in your circuitry as I can see them being generated and even surpassed all the time on a scope.  I think circuitry being able to cope with a couple of volts is enough to make you safe, after all, there is always - the guitar volume control  if things start to sound too rough.  You have to bear in mind that as things pass down the line that level will change of course, but we know the gain of most of these stages now so we can cope.

I'm thinking the same thing about the Input Buiffer stage gain Namklak, sort of.  I think you seem to be thinking towards a setup where you envisage having no clipping at all in either of the SS stages before the valves and then tweak the valve stages to give more than they do if it is necessary.  Now I understand what you are up to here but I think that may not be the best way.  It's much much easier to design and work with the opamp stages than the valves.  As you know opamp technology is much more reliable than valve technology to predict accurately at the design stage.  The gain of valve stages alone is very unpredictable and most good valve designers are working from their experience rather than from design equations.  That's why I'm not that much of a valve expert.  That said, you can learn a lot by comparison with other circuits and there is nothing new about the H&K valve designs, they are pretty much standard blocks.  I certainly want to rely most on the valve distortion but I still think a little bit of SS distortion can be useful if held in check, especially that even order single sided "bending".  What I'd like to do is to adjust things so that I get clipping occuring in the SS at a known signal level, rather than the fog of multiple stage clipping we have at the moment.  It must be mild soft clipping unlike some of the hard flat top stuff that we are seeing and there must be a lot of clean signal room beneath it.  I may not use that sound often but I would like to keep a bit of it there just in case, and just as you said, at the same time get the best expanded use out of the control ranges for what I most often do.  Does that make any sense?

I reckon the first stage needs to be trimmed so that it can maybe clip a little at the extremes for each channel with high level signals and guitar full up.  Use a lighter signal guitar or roll back a touch and it should clean up completely in this stage.  Any clipping should definitely be due to limiting diodes and not the opamp and (for me) with a network designed to ease us into assymetric clipping so it could better be described as I did, bending.  You could do away with the assymetric aspect if you like of course, we just need to alter the diode network to a couple of back to back zeners to act as a last resort and prevent the opamp internals hitting the rails.  I would also leave in the zeners across the input as they do the job of protecting the input from damage by accidental high voltages (it happens!).  They also limit the input to levels below that anomaly with the Clean channel where an extreme signal (about 4V) can defeat the feedback diodes due to it lifting the opamp -ve input pin to a silly level.  This needs to be checked and adjusted for each channel of course, we don't want to slip into the state where we make all of the channels sound exactly the same but louder.

The first stage only has two distinct switchable voicings, CL/Cr and Le/Ul.  The first is pretty flat if shifted too much to the right, and the second has a step in it's lower end which removes more of the bass.  I want to adjust these first.  Changing to a 4k7 feedback sounds fine if you don't touch anything else but what I want to do here is to lower the upper breakpoint for both voicings without losing the Clean sparkle just so it is less extreme, and lower the lower breakpoint too to bring in a bit more lower middle and even a touch of bass.  These changes will themselves alter the gain, probably drop it, without altering the feedback.  The feedback cap needs to go up anyway to bring the upper breakpoint down to a less extreme level.  I now have a feedback diode setup I like as well which I may have shown you which does very mild compression of the lower half of the signal in a way in which I can set when it starts and how aggressively it bites.

I'm proposing that we leave the basic simple shape of the Cl/Cr channels the same as original except for adjusting the top end cutoff from 49kHz to within 20kHz.  This would have to apply to the Le/Ul channels too in this stage.  That isn't extreme but it will narrow the door a bit.  Of course it could lose some of the Clean sparkle but I don't think so, and we could even consider coming down further if we want a more throaty middle.  This is done by adjusting C11 as you know and would mean changing it from 330pF to around 1n5-2n7, 5-7x as big which is a fair shift with 10k.

The reason I suggest leaving the 10k is that I have other shaping changes to suggest.  The R3, R8, C1 network shapes the bottom and top ends and I wanted to beef out the lower mid a bit as I said.  I have played around with these values to see what I could produce, keeping the C1 cap the same is not on but fortunately I wanted to go down not up.  I tweaked and tweaked and came up with the following curves.  Pick up plot "24_Input_AllChannels_Modded_Responses.jpg" from Plots on Dropbox.  It is the original and modded responses of the two selections for the Input Buffer alone.  I should point out that I got a better Spice model for the correct opamp so these readings may very well be slightly different to those I posted earlier but it will be miniscule.  Here are the stats on the 4 setups.

Green: Original component values throughout for Clean/Crunch.  High end breakpoint is at 49.4kHz, low end at 97.1Hz.  I like the sound of this channel in Clean but it's a bit thin and sharp for me on Crunch.

Blue:  Modded component values for Clean/Crunch.  Gain stays virtually the same with a 0.1dB drop.  Lowered upper frequency breakpoint to 6.44kHz.  Lower frequency breakpoint stays close at 94.5Hz, (low E string is 82Hz).  This is mild modding of what most people think of as the best aspect of the GM36 with testing whether reducing the bandwidth for the Crunch helps.

Red:  Original component values throughout for Lead/Ultra.  High end breakpoint is at 41.0kHz, low end at 874.3Hz and with a slight step in the lower section.  This starts to sound thin to my ears and then becomes coarse when it's overdriven.

Turquoise:  Modded component values for Lead/Ultra.  Gain lowered by 3.4dB.  Lowered upper frequency breakpoint to 7.4kHz.  Adjusted lower frequency breakpoint to 100.6Hz by reducing step in response.  I'm anticipating having a different top end breakpoint in the Gain stage as it is better to stagger the breakpoints a little.

Now these have both had some modded components but I've more or less managed to stick to the rules I set myself, no physically bigger caps, mainly resistor changes.  The only components I've changed to get the plots you see are C11=2n7 to adjust the upper limit and R3=120R, C1=150nF to adjust the lower curve.  The R3, R8, C1 network is only in circuit for the Lead/Ultra channels while the C11 value works for all channels.  I reckon that's a pretty remarkable change for so few components.

It should be born in mind that this represents all channels before any form of limiting.  We need to consider what we want in a feedback diode network to give us the sound we are looking for.  For you Namklak that may just be a couple of series 11V zeners back to back for overload security against opamp internal clipping, for me I will use a simple diode setup to keep a bit of milder assymetric bending and milder overall high level limiting on both sides.  I have two back to back 11V zeners as you would.  In parallel with one of them I have connected a 3.9V zener in the same orientation, a 1N4148 diode in opposite orientation and a 1k resistor all in series.  This begins to limit on the negative side around 5V just as at present but the 1k resistor is great to round off the peak over a couple of volts rather than just chop it off.  I've tried a lot of other more complex apporoaches but this simple on works as good as any.

Oh, and I have the full tonestack and its opamp buffer modelled too.  Its curves are looking really good from stock and it's surprising it's so ineffectual on this amp.  I wonder if the earlier voicing just takes all of the balls out of it before it can even do anything to the signal?

bordonbert wrote:



I reckon the first stage needs to be trimmed so that it can maybe clip a little at the extremes for each channel with high level signals and guitar full up.  Use a lighter signal guitar or roll back a touch and it should clean up completely in this stage.  Any clipping should definitely be due to limiting diodes and not the opamp and (for me) with a network designed to ease us into assymetric clipping so it could better be described as I did, bending.  You could do away with the assymetric aspect if you like of course, we just need to alter the diode network to a couple of back to back zeners to act as a last resort and prevent the opamp internals hitting the rails.  I would also leave in the zeners across the input as they do the job of protecting the input from damage by accidental high voltages (it happens!).  They also limit the input to levels below that anomaly with the Clean channel where an extreme signal (about 4V) can defeat the feedback diodes due to it lifting the opamp -ve input pin to a silly level.  This needs to be checked and adjusted for each channel of course, we don't want to slip into the state where we make all of the channels sound exactly the same but louder.

The first stage only has two distinct switchable voicings, CL/Cr and Le/Ul.  The first is pretty flat if shifted too much to the right, and the second has a step in it's lower end which removes more of the bass.  I want to adjust these first.  Changing to a 4k7 feedback sounds fine if you don't touch anything else but what I want to do here is to lower the upper breakpoint for both voicings without losing the Clean sparkle just so it is less extreme, and lower the lower breakpoint too to bring in a bit more lower middle and even a touch of bass.  These changes will themselves alter the gain, probably drop it, without altering the feedback.  The feedback cap needs to go up anyway to bring the upper breakpoint down to a less extreme level.  I now have a feedback diode setup I like as well which I may have shown you which does very mild compression of the lower half of the signal in a way in which I can set when it starts and how aggressively it bites.

I'm proposing that we leave the basic simple shape of the Cl/Cr channels the same as original except for adjusting the top end cutoff from 49kHz to within 20kHz.  This would have to apply to the Le/Ul channels too in this stage.  That isn't extreme but it will narrow the door a bit.  Of course it could lose some of the Clean sparkle but I don't think so, and we could even consider coming down further if we want a more throaty middle.  This is done by adjusting C11 as you know and would mean changing it from 330pF to around 1n5-2n7, 5-7x as big which is a fair shift with 10k.

The reason I suggest leaving the 10k is that I have other shaping changes to suggest.  The R3, R8, C1 network shapes the bottom and top ends and I wanted to beef out the lower mid a bit as I said.  I have played around with these values to see what I could produce, keeping the C1 cap the same is not on but fortunately I wanted to go down not up.  I tweaked and tweaked and came up with the following curves.  Pick up plot "24_Input_AllChannels_Modded_Responses.jpg" from Plots on Dropbox.  It is the original and modded responses of the two selections for the Input Buffer alone.  I should point out that I got a better Spice model for the correct opamp so these readings may very well be slightly different to those I posted earlier but it will be miniscule.  Here are the stats on the 4 setups.

Green: Original component values throughout for Clean/Crunch.  High end breakpoint is at 49.4kHz, low end at 97.1Hz.  I like the sound of this channel in Clean but it's a bit thin and sharp for me on Crunch.

Blue:  Modded component values for Clean/Crunch.  Gain stays virtually the same with a 0.1dB drop.  Lowered upper frequency breakpoint to 6.44kHz.  Lower frequency breakpoint stays close at 94.5Hz, (low E string is 82Hz).  This is mild modding of what most people think of as the best aspect of the GM36 with testing whether reducing the bandwidth for the Crunch helps.

Red:  Original component values throughout for Lead/Ultra.  High end breakpoint is at 41.0kHz, low end at 874.3Hz and with a slight step in the lower section.  This starts to sound thin to my ears and then becomes coarse when it's overdriven.

Turquoise:  Modded component values for Lead/Ultra.  Gain lowered by 3.4dB.  Lowered upper frequency breakpoint to 7.4kHz.  Adjusted lower frequency breakpoint to 100.6Hz by reducing step in response.  I'm anticipating having a different top end breakpoint in the Gain stage as it is better to stagger the breakpoints a little.

being impressed...very good approach !

Thanks Ignantios. Namklak and I are trying to take a solid reliable line to this process. I have the software tool to set things up and look into them, but he has a good technical knowledge of areas I don't and instincts to keep me on the right track. Too many modders just want to get into the guts and start changing things basically at random to see what happens. That way will never give you the best results. You should always start from a position where you have a good understanding of what is happening inside each stage as it is from stock and what in the circuitry makes that so. Then you know roughly what to change and why.

Just to make one thing clear for others who come across this and maybe heven't got an idea of the GM36 block diagram, the first Input Buffer stage has only two settings. There is one shared by the Clean and Crunch channels and one shared by the Lead and Ultra channels. When you select a channel, one or the other is put into action and the results are passed to the Gain stage where there are four different settings. The Gain stage is where the difference between the Clean and Crunch channels is made, and of course the Lead and Ultra. That's why we must be careful not to rush into too many changes in any stage which may basically end up making channels completely the same or too wildly different.

I'll post later on where I think I'm going with the Gain stage.

In the meantime, I'm getting smug and clever with LTSpice.  I have the tone control section and the valve sections up and running.  I'm not too convinced that the valve models in Spice are too accurate in all aspects but they give us an idea, and I don't want to play with the valve stages too much anyway unless I have to.

My first impressions are that the tonestack controls are actually fine.  I wasn't expecting that.  They show good curves with absolutely normal ranges and are almost identical to a Marshall setup.  I can post plots for these if anyone wants them but to get a picture you need to have so many of them, each one across its whole range with the other controls in various states of play, and it is a shed load of plots just to show that these are pretty normal.  I think these are really just being throttled by the weird bandwidth issues the earlier stages have.  If your bass is so far down you can hardly hear it already, cutting it any more with tone controls seems to be a "non-aural event".  And the lower mid is still recessed in comparison to the upper mid which makes the Middle control more of a lowish Treble through no fault of its own.  I'm not looking to change these at least in a first pass.

The valves also come up pretty straight forward but they would not be as accurately modelled under some circumstances.  From what I understand, they should be fine for frequency response plots but not so good at simulating the fine nature of limiting conditions.  That seems to be what I'm seeing, clipping is shown as very sharp and level in most cases.  I have the whole valve section set up as one with all valves in line, rather than as two separate stages.  Here are a few observations.  Remember, as everywhere else up to now, the signal voltages mentioned are all peak voltages representing the true level that the first clipping occurs.  DC voltages will be variable when you use real valves so all of the levels will change including signal levels for clipping.

The first stage V1A is used in all channels with two gain settings via a cathode network change for Clean/Crunch and Lead/Ultra.  It sits nicely between the rails at 263V and clips at inputs of ClCr 4.1V/LU 1.9V.  It shows a good "bending" of the sine wave signal over its output range.  Just below clipping the FFT shows a very evenly dropping line of harmonics with 2nd and 3rd at ClCr: -37dB/-58dB and LU: -20dB/-33dB so a sweet valve type distortion coming in.  The frequency response at the anode and with the next stage connected shows the now familiar hump at 7kHz over the lowest level at 200Hz at 5dB.  Low frequency response of the stage alone is flat way way down but the input cap limits this to 12Hz.

The second stage V1B is not used for the Clean channel and there are no differences between the other 3.  At first I thought that it's difficult to know what they were doing here!  Look at the cathode resistor, it's 33k.  That cuts the cathode current down so low that it puts the DC level on the anode at 338V.  That is not much headroom for the Crunch channel and would normally give very hard clipping at low levels on the others.  But, that also drops the gain of the stage down to about 3x so clipping doesn't actually become a problem.  At signal inputs of 20V there is still no flat top clipping only severe bending which is of a nice nature.  The figures show 2nd and 3rd harmonics at CrLU: -16dB/-33dB again good valve stuff.  This stage also has a cap across its load resistor which restricts the upper frequency limit.  It is absolutely flat with cutoffs at 47Hz and 1.7kHz.  At last some upper level restriction.

The third stage has a selection network to pick up the output of either V1A for Clean or V1B for CrLU.  To look at the stage in isolation I split the output of V1A from it and injected the signal there but I left the V1B output connected as that is how the valve would perform in real life.  I took a similar approach for the CrLU channels.  The anode sits at 184V so it should be fairly well balanced top and bottom.  For Clean it shows signs of balanced limiting at both ends at about 22V input.  The bottom shows a softer limit than the top (but that could just be the modelling breaking down).  Frequency response is 14Hz to 38.5Khz and there is the usual 3dB hump at the usual 11.1kHz, what did we expect?  For Crunch it limits at 3.5V input with 3dB bandwidth from 11Hz to 5.6kHz very flat.  Finally Lead/Ultra adds in a cathode network which not only changes gain but also shifts the DC level to 240V making the clipping assymetric, that's clever.  The onset of clipping is now 3.4V and bandwidth is 128Hz to 6kHz.

When V2B is added onto V1A there is no obvious difference.  I have a strong suspicion that this is reaching the limit of Spice modelling of valves.  There should be a neat little mechanism which kicks in where the cathode follower's grid steals current through the driving stage's anode resistor.  This produces lots of lovely even order harmonics which really define valve tone.  This does not happen at all in the simulation, the output from the cathode follower is exactly the same signal as on its grid.

So overall, looking at the output of the valve stages with all four valve stages in place and appropriate switching for channels:

(EDIT:  I'm editing this to correct errors I made setting the conditions up the first time around.  Apologies for any confusion.)

1) Clean.  Balanced limiting at 3.3V input.

2) Crunch.  Balanced limiting at 300mV.

3) Lead and Ultra.  Top limits at 110mV and bottom 130mV.

There is easily visible general assymetry of all signals with Lead/Ultra most obvious showing the top peaks at 83V and the bottom at -104V.

To me these results show that the valve stages are very sensitive in non-Clean channels considering that we have two solid state gain stages before this.  At first sight it seems almost like a complete independent valve amp which expects guitar level signals, but with that couple of SS stages before it the signal level could get pumped up to silly levels!  And the way they seem obsessed with voicing the channels in so many places is very complex, maybe overly complex.  Perhaps a simplification of some of these parameters would yield better results for us?  I think some of these results need to be carefully thought about before we go off and pull this to bits.  It may not be a bad approach that H&K have taken if you consider we can't see everything that they can about the design.  On the other hand there may be a much better one for classic stuff.

So there we have it, the valve stages.  As always your comments and thoughts are welcome, there is bound to be stuff I haven't thought of which I would like to know and I hope you would like to tell me.  And, though I have tried to not post a bevvy of plots this time, if anyone would like to see them just let me know what you want and I'll post them up.  I would also add that the, errrm, "information for the GM36" to allow you to follow what I am talking about is now elsewhere in the public domain and you should be able to find it through Google if you really want to compare it with my diatribe.

bordonbert wrote:
1) Clean.  Balanced limiting at 3.3V input.

2) Crunch.  Balanced limiting at 300mV.

3) Lead and Ultra.  Top limits at 110mV and bottom 130mV.

Are these values input "guitar" voltages?

I use the Ultra to get a Soldano tone, so aggressive tube gain isn't bothering me that much... But from seeing Jet City amp schematics, he uses a extra tube stage to get there I think.

But I've tried to get the Lead to be a jcm800, and it sounds uh too dirty.

So our CD came out today - all the guitar in the right channel is me and the GM. Note on the rhythm track of Going Home I was going for the 800 tone with the Lead channel, but is sounds more dirty than saturated... Kind of bummed me. The studio we did this in had great iso rooms and great recording rooms and a great engineer, but almost no musician gear. They had an AC30 and that was it for decent amps, so I used the GM. I was hoping they had a real Marshall I could record with...

http://mybluesky.co/store

Again, good work BB.

Thanks,
-Bob

AT LAST, the release!!! Wow, congratulations Namklak, you've done it and that's really something. I've tried to buy the FLAC but they say Paypal isn't working at the moment for some reason. As soon as it does you have a sale.

Yes, the figures I quoted there are the signal at the input to the valve stages that is the onset of any clipping within those stages, just as though it was the output direct from the Gain stage. It's effectively what we need to have coming from the Gain block to drive the valves to limit. We now know what we need as output from the Input Buffer and Gain stage combined to get into valve distortion exclusively.

I know I have been obsessive in pushing so much info over the last few days that it will cause mental overload but I had to get it out of the way. I wanted to have as much as was helpful there as a reference for the real stuff, the decision making! I've stopped simulating for the moment at the Tonestack and its output buffer stage now that I have good working models up to that point, even though there are a couple of stages after this where tonal changes are made once again. There is the Resonance/Presence section but that actually works so I don't see too much to come out of that, the Master Volume stage which is flat, then a final buffer stage after the Master Volume around U2A on the mainboard which feeds into the Power Amp. And of course a couple of factors in the Power Amp itself. I'll put together a couple of little synopses ( ) which might prove useful, like where the tone/gain for each channel is altered.

We're almost at the decision making time where the fun bit starts. For my tastes you were on the money when you said drop the Gain, there is far too much on tap for the needs of players like us, though we must keep in mind the fact that you may be playing with lower output guitars at some point and there must be some in reserve to cope with that.

Oh, and thanks for the tip you passed on, I tried it and I'm now a believer. I'm absolutely converted to the Noise Gate on its most sensitive setting, it makes a significant difference at all levels. I don't know how I can have missed that approach in exploring the amp. I guess I just don't think of it as a very noisy beast, I have low standards in that respect. I owe you a beer back for that one mate.

Correction.  Paypal was just teasing me.  It's working now and I am the proud possessor of a copy of My Blue Sky's latest release on the Crosse and Blackwell label.  I'm genuinely pleased for you mate.  I hope it generates a bit of interest in the band and makes you a little bit of money, it deserves it.  (And keep on supporting FLAC, it's the dog's danglies of formats!)

You go on the wall of my room in the "I know a rock star" section alongside the other notables.     Rock And Roll Hall Of Fame?  They don't make it onto my list.    (At least not in those shoes.    )

Thank you BB!  And uh what label?  Dog's Danglies???  To be frank, we just did this to prove to promoters we "do originals" and then they'll hire us to open for national acts.  I mean, yea, it's cool and all, but our gig is still Allman Brothers tribute.  And none of us are quitting are day jobs...

Back to reality...

Almost nobody who posts on this forum sets the Gain on the Lead or Ultra channels much above 3 - so why don't we just reduce the gain on those channel by mucking with the network connected to Q26?  Doing that won't have any effect on the Clean/Crunch channels and we don't have to do any down-stream gain mucking.  Changing that network plus zener tweaks, and maybe...?

Hahaha!  Some things don't travel do they?  It's an old music joke from over here.  "This next number is our latest release on the Crosse & Blackwell label."  Crosse & Blackwell make beans and soup like Heinz!

And the "dog's danglies", never come across that one?  The "mutt's nuts", the "pooches plums", all variations on the "dog's bollocks" or the "bee's knees", you know, something top of the range.  Ever seen a waddling British Bulldog from behind?  (It's a good thing, at least over here.)

Yes you are right, if it is just gain that you are trying to tweak and aren't too worried about the nature of the overdrive you get.  For my own amp, I want to see where all of the bass and lower middle is going as that is where I honestly think the amp is weakest.  The gain aspect is irritating but it can at least be corrected with the bottom third of the gain control awkward though that is, that nasty upper dominance can't.  Even though I improved the effect of my tone controls with the speaker and valve changes they still aren't right and, as I posted, I think that is just the result of a cascade of "adjustments" that H&K have made putting in lower cutoffs or steps at around the 1kHz mark.  I'd like to find a couple of easy ways to claw back some of that and then remove the bass with the tone control - only when I actually want to.  The tone controls were a revelation ro model, they are actually very good, very standard, and have the expected range of control.  Except they don't have any effect on levels when they start so low down below the mark that taking them down another few dB makes no sonic difference!

As to reducing gain, it would seem to be easy to just do it in one place but that then changes the way the two stages interact.  If you reduce it at the Input there is no control whereby you can adjust it back so some of the interaction between Buffer and Gain stages is lost.  In the Gain stage you have the ability to adjust that higher, both its input signal level and the gain of the stage itself.  I would have thought that that might be a better place to consider.  Remember those two diodes in the feedback of the Gain stage you spotted?  It turns out that, along with their series resistor, they do perform a really neat rounded soft limiting action in the Gain stage for some high signal levels.  I'm not sure whether it may be better to even reduce gain gradually over both stages, it shouldn't be hard.   That way the Clean/Crunch from the Input Buffer drives the Gain stage the way H&K have set it and we can choose how hard we want the Gain stage clipping to be pushed with its control which first of all drops the level further on its input.  Dropping gain in the Input Buffer means less Gain overdrive on tap, doing it inside the Gain stage or maybe even just after means more overdrive created where it's actually quite mild in character.  Though I would have to say the argument that the valves can produce enough themselves would have merit too.

Like I said earlier, I'm going to put together a table with each channel and all of the places where its response is tailored.  Then we will be able to see a bit more of the picture. At the moment I'm so frazzled with adjustments and settings everywhere I can't really see them separately.  Shouldn't take too long to get it sorted.  I have to say they've either done a brilliant job to manage it all, or they've completely missed the plot and made it about 10 times as complicated as it should have been.



EDIT:  I have run into a situation where I have used up all of my file attachment allowance.  I'm removing the direct downloads for the plots and putting them in a zip file on my Dropbox site.  I'll make sure to put a link to this every few posts and keep it up to date so they stay in sight.  We don't have them on this page yet so here they are:

GM36 Analysis Plots

Well, this thread has ground to a probable halt for now.  I will say that upon discovering the yet-another-ts circuit around the Master, I decided last night to not dime the master, but back it off to noon.  As I was doing that, I found that I did indeed remove some "distortion" - probably due to no longer activating the ts zener diodes around U2B.  Did last night's gig that way and just listened to the recording - hmmm, not bad.
So I'm thinking a very simple mod to get more el84 saturation is to reduce R52 to get more signal into the tube output stage - note this is after the last ss stage.  This will make the amp louder, but I use an external power soak, so that's okay with me...

Edit - I just noticed there is feedback in the output stage -it's introduced in at U2A when either Q31 (Clean) or Q2 (Ultra) is enabled. Interesting, maybe I don't want to change R52....

My advice to you is to "grip it and rip it!"  R220/R221...

220/221, whatever it takes...

Sorry guys, I actually went quiet because I thought no-one was really interested.  I'm still here fiddling in the background.  

I wouldn't think you'll get much joy out of R52 Namklak, it's only dropping the signal to about 0.7x.  I wonder if it may be there as an aid to stability too.  You pointed out that feedback from the speakers is introduced for some selections but around yet another solid state gain stage.  Your control theory experience should tell you that additional stages inside the loop mean further out on the edge as far as stability is concerned.  It may not be a problem at all but then again....  That said, your stage around U2A looks promising.  You could try increasing R102 to give that stage more gain though that may affect the feedback stability as well.  The 4k7//47pF is ridiculous if I am right as it gives a bandwidth around 720MHz.  What do they think they are designing for?  If you do change the resistor you could easily afford to increase the 47pF to give a much lower bandwidth to the stage which would help with closed loop stability when that speaker feedback is applied.  Where is the loop's dominant pole defined?

One thing I am sure you could do is to drop the resistor across the Master Volume control.  That would take down the gain of that stage so it doesn't clip as easily as you have said it might.  That would give you a larger range of control on your Master Volume again.  

That was good thinking on your part on the Master Volume.  We do know that the amp has one strength, gain to spare.  The idea is to allow you to overdrive everything to the max to offer extreme ultra distortion for the out and out metallics, (don't really know why some of them don't just buy a square wave generator), but to be able to get a wide range of overdrive conditions with the various controls backed off.  The main problem we have is that our classic rock range of that broad sweep is so restrictive it's embarassing!

One question I would ask you other classic rock guys is, are the EL84 valves capable of giving the type of sound we are looking for.  EL34s certainly can, (as I am currently experiencing), and KT66/77/88 jobs do too but I have to admit this is my first EL84 amp as far as I know.  The reason I ask is that I would like to know whether hitting those output valves harder by modding would actually get a significantly different sound which we can't approach with careful tweaking of the existing setup.

And as a final comment, am I alone in hating the sloppy lazy way that H&K number their schematics.  Top->Bottom, Left->Right or through the circuit flow should be the way all through development.  There will always be production edits to accomodate but if you make an edit these can be assigned much higher numbers above the original schema so they stand out.  You can't find anything by a component reference in their bloody diagrams.  Not like the Germans to be so sloppy!  R220/R221?  What where why how who?  

"R220/R221" - I believe he is referring to a famous quote by Michael Keaton from the movie Mr. Mom.  Back when Michael Keaton was funny....  And also when Teri Garr was uh almost hot.  I had a crush on her anyway - a cute very funny chick...  For her, see also Young Frankenstein.
https://www.youtube.com/watch?v=iX3kxAA2L4Q

Aha! One of those "cross pond" moments I guess. Terri Garr, ALMOST hot? Grrrrooowwrrrrr! I'll say. Figure to die for and that sad puppy dog look she could put on was a real plus for her. Young Frankenstein I know well, Mr Mom eluded me somehow even though I do like early Michael Keaton.

But I always thought you guys lived on that inefficient 110V stuff?

Thought I would add these to the rest of the data just for the record.  Here is the isolated Resonance/Presence section simulated.  It isn't all good looking.

Resonance is exactly what I would have expected.  Presence is - errrm - NOT!  It shows a sharply rising response with no top cut tail off after about halfway up to about 3kHz.  That could be something to do with the hard top end so many of us experience.

EDIT: In fairness I should point out that the response does peak at 7kHz and drop away but it is still at the same signal level as low frequencies when it is at 33kHz.