Express Output Waveform Shots

Express, Liverpool, Rocket, Dirty Little Monster, etc.

Moderators: pompeiisneaks, Colossal

Post Reply
katopan
Posts: 594
Joined: Mon Aug 22, 2011 10:10 pm
Location: Melb, Australia
Contact:

Re: Express Output Waveform Shots

Post by katopan »

Thanks Jaz. I appreciate your interest in this all the way through.

Finely tuned sports car, yep. I'm a much better engineer than I am a guitar player, although music has always been such a big thing in my life. But it is a privilege to own such an amp. I'm not worthy, I'm not worthy..... :lol:
katopan
Posts: 594
Joined: Mon Aug 22, 2011 10:10 pm
Location: Melb, Australia
Contact:

Re: Express Output Waveform Shots

Post by katopan »

This is probably the wrong thread to move into this, but what the hell. It's time to talk about 'that' 1K 25W screen supply chain resistor.

A page back I posted B+ and screen supply voltages at different volume levels with my test signal going in. At zero we have 8Vdc across the 1K screen supply resistor. At 1 where it's still clean there's 17Vdc across it. At the sweet spot where the maximum duty cycle shift and mixed mode distortion occur we have 66Vdc across it, and fully cranked it's 74Vdc. That equates to 0.064W at volume 0, 0.3W at 1, 4.3W at 6.5 and 5.5W at 10.

http://www.mouser.com/ds/2/351/XC-600041-198125.pdf

The above datasheet is for the part number as linked in the Express BOM in the Trainwreck files area. It gives temperature rise and ambient derating curves for these resistors. With a maximum dissipation of 5.5W obviously we don't want to install a 5W resistor. But what would happen if we put in a 10W resistor?

A 10W resistor running at 5.5W has about 130degC temperature rise from the chart. That's 150degC if the ambient temperature is 20degC, but I can guarantee with a heatsource like that inside your chassis it's going to be hotter in there than 20degC. Solder melts between 183-190degC. A quick search tells me UL sets a limit of 105degC on PCBs. FR4 is rated to 130degC but survives much better below 95degC. I don't know about the material used in a TW perfboard. But in short it would certainly burn the board, is possibly getting close to melting the solder (resistor leads conduct heat and will be hotter than the ceramic body), and will heat nearby components like the main filter caps (radiant heating will be higher due to higher resistor surface temperature, although general temperature rise will be the same due to the dissipation of 5.5W of heat).

Lets run through the same with the recommended 25W resistor. It's running at a max of 22% of rating, which gives a temperature rise of around 75degC. So that's 95degC if the ambient is 20degC, or in reality probably higher inside the chassis. But it's got a way to go yet to hit the UL limit, and even further for any board rating or further still for the solder melting point. The board will survive much longer. Nearby components will also be better off with much less radiated heat. It will still feel too hot to touch (60degC is the max industry exposure for surfaces which can be touched). Still seems like it 'gets really hot' even though it's only working to 22% of its rating.

In my amp I made up a small aluminium bracket and screwed down the resistor to heatsink it to the aluminium chassis. I used 10W rating and the chassis surface directly above the resistor only gets mildly warm fully cranked for an extended period (definitely tested this out while obtaining all these waveforms). Much better for the resistor and much better for the rest of the components with the heat having a path to get outside the chassis rather than dissipating inside it.

So was the 25W resistor what Ken had in his box at the time, or was it a smart decision that considered temperature rise and the impact on longer term amp reliability and life?
User avatar
rooster
Posts: 1616
Joined: Wed Apr 02, 2008 8:50 am
Location: Pacific NW

Re: Express Output Waveform Shots

Post by rooster »

Craig, whew, great effort here for sure. Thank you.

Thanks presented, I have to say, the increased tail resistor results really stun me. I have done this with other circuits and found that increasing the tail quickly reduces the perceived overall distortion/overload of the amp. To hear/see that this does not change things noticeably seems impossible. Ha, not doubting your documentation but wondering what in the Express circuit would make this behave this way. Perhaps my results with other amps are not relevant here. Hm, I will give this a lot of thought in the future. Initially I have to think that it's the gain from the preamp that may be influencing the results.

Oh well, thanks for all the work!
Most people stall out when fixing a mistake that they've made. Why?
katopan
Posts: 594
Joined: Mon Aug 22, 2011 10:10 pm
Location: Melb, Australia
Contact:

Re: Express Output Waveform Shots

Post by katopan »

Yes it's all about the preamp gain and shear magnitude of the signal going into the PI. Once it's grid clipping on the PI input it's just so far beyond things like tail resistor making any difference. The behavior remains the same despite circuit values being changed.
cxx
Posts: 60
Joined: Mon Nov 05, 2007 4:17 am

Re: Express Output Waveform Shots

Post by cxx »

There are a couple things I'm curious about. Looking at the PI output waveforms when clipping shows a ramp at the top that seems to be a result of the variation of the cathode voltage. The el34 grid waveforms don't have this. Is this the result of the grid clamping or is it an anomaly of measuring at the PI plates? Looks like it takes place between 1 and 2. Looks the same with and without NFB.

The v4,5 plate waveforms and output are pretty much the same with v5 inverted. From the output, looks like about 21 volts clean at volume 1 or 27 watts. Maybe there is more between 1 and 2 but it doesn't look like it given the next waveform clips about at the peak level. The waveform at 3 isn't square which would be about 55 watts but more of a trapazoid which would be closer to square than sinusoid which would be about 50 watts. The funny thing is that as the wave becomes more square at higher volumes (the sides are more vertical) the lower part develops a curved ramp instead of a flat bottom. This would reduce the output power. This seems to happen most about 6.5 on the volume. Not sure you could hear this effect though. I can't see a reason for this in the inputs to v4,5.

So where does this curved part of the waveform come from? Is it on every amp that clips?
katopan
Posts: 594
Joined: Mon Aug 22, 2011 10:10 pm
Location: Melb, Australia
Contact:

Re: Express Output Waveform Shots

Post by katopan »

Sorry cxx, busy few days and I've been meaning to get back to this.
cxx wrote:Looking at the PI output waveforms when clipping shows a ramp at the top that seems to be a result of the variation of the cathode voltage. The el34 grid waveforms don't have this. Is this the result of the grid clamping or is it an anomaly of measuring at the PI plates?
The power grids are held at 0V when clipping. Grid current is flowing (although not enough to drive a +ve voltage into AB2 operation) and charging up the PI to power grid coupling caps. With the downstream end of the coupling cap clamped or held at the power clipping threshold of 0V, the cap charging up is shown by the PI plate waveform top ramping upward. This is standard behaviour of a LTP PI driving a power stage to clipping. It shows up the same in my 18 Watt and will in any other push-pull amp.
cxx wrote:The v4,5 plate waveforms and output are pretty much the same with v5 inverted.
This is always the case with a push-pull output stage. When one side reaches cutoff, the signal voltage is coupled through standard transformer action from the side still conducting to appear on the cutoff side.
cxx wrote:The waveform at 3 isn't square which would be about 55 watts but more of a trapazoid which would be closer to square than sinusoid which would be about 50 watts. The funny thing is that as the wave becomes more square at higher volumes (the sides are more vertical) the lower part develops a curved ramp instead of a flat bottom. This would reduce the output power. This seems to happen most about 6.5 on the volume. Not sure you could hear this effect though. I can't see a reason for this in the inputs to v4,5.

So where does this curved part of the waveform come from? Is it on every amp that clips?
No, it is totally unique to this amp. Yes you hear it - it is a key component of the tone and also reduces the clipped output power to be closer to the clean output power compared to a full headroom squarewave. This is the mixed mode distortion I've been on about. PI clipping sneaks underneath the power grid clipping threshold and becomes dominant on only the lower side of the output signal. The upper side remains power grid clipping dominant. The slope is kinda reverse of what happens as described above with power grid clipping. With the PI clipping having moved under the power grid clipping threshold, the upstream side of the PI to power grid coupling cap is held at a constant voltage (actually it drifts up slightly as the charging current through the cap backs off, and therefore the voltage drop across the PI plate resistor backs off) and the downstream side charges or discharges depending on how you think of it through the power grid leak resistor to the bias supply. When the signal goes up to PI clipping the signal cap is lifted (no change in voltage across it), and then because the PI output is flat from PI clipping the cap charges up so its downstream voltage heads toward the bias voltage (like any RC filter - won't pass DC). That shows in the output as a curve heading back toward 0V, but of course before it gets anywhere near steady state the signal flips the other way for the next half of the cycle.
cxx
Posts: 60
Joined: Mon Nov 05, 2007 4:17 am

Re: Express Output Waveform Shots

Post by cxx »

I've been reading over this thread again and, as much as I hate bring it up as I appreciate all the information provided, it seems as though the characteristics that you attribute to the unique sound of the express are not attributed to the unique characteristics of the design.

Seems like from volume 0-6.5 the third stage does not really clip or provide much coloration to the waveform, just gain. The PI and PA are pretty generic and the PI is somewhat insensitive to parameter changes. Most of the characteristics are attributed to the clipping of the PI/power stage with its negative feedback loop.

The things that seem unique to me are the higher than optimal impedance of the output transformer, the removal of low frequencies (<530 Hz) of the filter between 2nd and 3rd stage and the cold bias of the third stage. The 1k vs choke seems important, but pretty generic.

It seems that I can get the same effect out of a low gain amp by fronting it with a boost, regular into amps that filter the low end or a treble boost for those that don't.

Did I miss something?
katopan
Posts: 594
Joined: Mon Aug 22, 2011 10:10 pm
Location: Melb, Australia
Contact:

Re: Express Output Waveform Shots

Post by katopan »

You raise an interesting point cxx. My thoughts are:

Higher than optimal impedance of the PT creates more screen current draw under signal. This works with the 1K screen supply resistor to get the right amount of sag for the PI headroom to dynamically move the right amount clean to mean. Those two unique things directly impact the characteristics the waveform posts I've shown.

You're right that the 3rd stage does not clip or contribute colouration in the volume range of interest. I now believe that the 3rd stage bias is all about gain control rather than for its clipping characteristic. It's often been called the cold clipping stage but doesn't function like that at all (like the 39k stage in a Soldano does) until things are very heavily overdriven, and then it actually works to even up the duty cycle which is counter intuitive.

The 18.2K power rail to the PI is also unique and has a direct impact on the dynamic response of the PI headroom change as the amp goes in and out of overdrive. This has a massive contribution to the feel when playing the amp. Get rid of this and the static waveforms will look the same, but the feel of the amp dynamics won't be there at all.

None of this waveform analysis has looked at frequency shaping. Two comments are often made when talking about Express tone shaping. The TMB tonestack is very bass heavy. Some people who find that out after modelling on Duncan's Tonestack Calc mod the component values to be more like a Liverpool, and think it's good that the boomy bass is gone. Next there's the 530Hz low cut filter between the 2nd and 3rd stage. But it's only first order and has a sloped rolloff. I've read that the bass heavy tonestack is good in the front end for full cleans, and the low cut before the 3rd stage helps with keeping things nice when everything 3rd stage and after goes into overdrive.

But the tonestack and low cut filter are both present the whole time, clean to mean. People have described them like they separately contribute clean and mean. The Express TMB tonestack is bass heavy. But if you adjust the response curve with the low cut filter (because they are in series) you'll find it doesn't remove all frequencies below 530Hz, it just adjusts them with a rolloff slope. With the Express values it has the effect of making the TMB tonestack more balanced again, if not just a little bright.

Good discussion points.
ve3vxo
Posts: 5
Joined: Fri Jul 18, 2014 4:05 pm
Location: Canada

Re: Express Output Waveform Shots

Post by ve3vxo »

This is my first post here. Many thanks to all the contributors and Katopan for all the hard work building, testing and documenting. I've been reading everything I can find about TWE amps but this thread is a goldmine!

So back in 2011 before Katopan went whole hog and built a full blown clone he was interested in a low power version. Myself and a friend are about to embark on this journey. I found the AX84 4-4-0 project and thought it looked great. Since this little amp has triode output it obviously will always be somewhat different than a full power TWE but I am interested in trying to retain as much of the character of the original amp as possible at household volume levels.

If understood the analysis correctly the power supply sag due to the screen circuit plays a major role in what is being called the sweet spot of the amp where a bias shift allows asymmetry in the PI to drive one of the power grids into conduction and the other not so much. This is all great for the undertsanding of the TWE but is there ever going to be a return to the low watter discussion? Since triodes have no screen circuit, I'm wondering is there a way by tweaking the resistors in the power supply chain of the 4-4-0 to mimic the screen induced sag of the actual TWE so that the same interaction between the PI, PP stage and NFB will result to get the same kind of classic trainwreck feel in a smaller amp? I am assuming that Katopan was originally interested in exactly this, or has it become all academic at this point?

Joe
katopan
Posts: 594
Joined: Mon Aug 22, 2011 10:10 pm
Location: Melb, Australia
Contact:

Re: Express Output Waveform Shots

Post by katopan »

Thanks Joe, and you're most welcome.

You're exactly right in that the 4-4-0 can't do it with the triode output stage. Not only that but in my opinion even beam tetrodes don't cut it fully. If you want the same response you need proper pentodes.

So no, this hasn't become all academic for me. I've got a couple of other projects that have been holding me up from a low power Express but that is still the aim. I developed the 'Thrifty Wreck' on the Wattkins forum as my first steps forward that. It's nominal 10W (measured 7W clean) from a pair of Russian 6P1Ps, with 6N2Ps (equiv characteristic to 12AX7 but different heater connections) in the preamp and PI. With what I know now from all this testing I owe it to that forum to optimise the circuit values to match the response as close as I can get it. But ultimately the 6P1P is a beam tetrode and doesn't have the same screen current characteristic as a pentode. For the 10W sort of power level using 6K6-GTs would be better, but the power chain and PI split load resistors would need tweaking to suit that valve.

My real interest now is in lower power than that. 10W has proved itself to me to be too loud when playing alone but not quite enough to take on to stage. My ideal stage power is around 20W. Thanks to the multiple taps on my PT, I can have a 300V B+ at the flick of a switch on my Express build which gives a response very close to what you get at 400V B+ but takes the edge off with output down from 33W to 18W measured clean with EL34s.

Next step is a 2W or 4W version depending on how things go, with VVR. The testing I've posted here has given me ideas on how to get the same response from a single ended version (more complex than just feeding the Express preamp into a SE power stage, like many other designs I've seen). If that doesn't work then it'll be a 4W push-pull design. Either of these will be based on the 6F4P/6DX8/ECL84 still with 6N2Ps for the preamp. All the waveforms I published here of the full size one give a basis of comparison to get the small one working properly.

So that's where I'm up to in the Express journey!
ve3vxo
Posts: 5
Joined: Fri Jul 18, 2014 4:05 pm
Location: Canada

Re: Express Output Waveform Shots

Post by ve3vxo »

Thanks for the warm welcome Craig

I'm glad to hear you are still working toward something in the 2W range. Are you sure a similar action can't be had with a triode output? My understanding of these tube circuits is not that great yet but if I understand anything it seems that as the signal level rises the way the PS voltage sags and the sequence of it between stages is what creates the magic. Correct me if I am wrong here. It seems to me that the combination of the screen current and the plate current would cause the sag to happen in a non linear way which wouldn't be the same in a triode but what if we used a non linear device in the current path feeding the B+ supply? With a mosfet in there and clever control of the gate we should be able to create any kind of transfer characteristic no? This would offer the ability to tweak things a lot and try different ideas. If a small pentode doesn't end up behaving like an EL34, then what?

Joe
katopan
Posts: 594
Joined: Mon Aug 22, 2011 10:10 pm
Location: Melb, Australia
Contact:

Re: Express Output Waveform Shots

Post by katopan »

You could probably get triodes to work somewhat in the ballpark. Yes the plate current draw in push-pull can give you some sag, and with the right resistor values that might be able to give enough PI power node sag for the same effect.

But I've come to a conclusion that I'm a screen compression junkie. I love the way it sounds and I love the way it feels when you play. And I can see in what I've posted the way it compresses the cleans before you hit clipping. You won't get that with a triode. You won't even get it to the same degree and character with a beam pentode. Smaller pentodes don't necessarily sound like an EL34, but they do behave the same in this respect from the testing I've done.
ve3vxo
Posts: 5
Joined: Fri Jul 18, 2014 4:05 pm
Location: Canada

Re: Express Output Waveform Shots

Post by ve3vxo »

I did some searching around on the 18watt and watkins forums you had mentioned earlier on. I found a design called a microwreck on the PP section of the forum that sounds exactly like what you intend to do except the fellow used miniature military spec tubes. I think it worked out so well he left off on the forum, never published the design and started selling them on the hall amplification website as the Lil Devil. Have you seen that one?

http://www.ppwatt.com/node/16013

I have begun ordering parts to build something. So far I have only got an OTPP5 from musical power supplies and some turret board but I made spreadsheets from the BOM of the 4-4-0 and have selected part numbers for everything to order. I'm glad I found this thread before I finished ordering stuff and started building a 4-4-0.

You had said you are busy with other things but when you get time I would like to build along with you. Tubestore is close to me and has the 6N2P for $9.95 and the 6DX8 for only $6.95! I can have all the parts very quickly it seems. Can you tell 'm excited?

Joe
katopan
Posts: 594
Joined: Mon Aug 22, 2011 10:10 pm
Location: Melb, Australia
Contact:

Re: Express Output Waveform Shots

Post by katopan »

I'm very familiar with Dana's Lil Devil. You'll find the development thread on that amp right here. Killer sound from the clips and what people say who bought one. In fact that amp was the inspiration for me building a submini version of the 18 Watt TMB Superlite, which I wrote a build thread for on 18Watt.com many years ago. That was all before $2 Russian valves. Submini's were at that stage half the price of new production 12A?7s (for me in Australia anyway) so were a great option for a 2W amp. The submini 18 Watt Superlite which I named the "2 Pot Screamer" still blows people away when I show it to someone new. But we're in a new era where $12.50 a valve isn't cheap anymore.

Dude, nothing against Tubestore but you can easily get 6N2Ps for $2 and 6F4Ps for $1.50.

Check out http://www.wattkins.com/node/18753 for what a pair in push-pull can sound like, and http://www.wattkins.com/node/18361 for single ended. There will be a series of TMC threads of different flavours over on Wattkins when I start building, but it will be a few months away yet.
ve3vxo
Posts: 5
Joined: Fri Jul 18, 2014 4:05 pm
Location: Canada

Re: Express Output Waveform Shots

Post by ve3vxo »

Well there's no denying the tone of a SE amp and the TMC-1 sounds very sweet! But I already have a 5w SE amp and I'm looking for the dynamics the TWE is famous for but that I can crank in my basement without my ears bleeding.

Tubestore is local but even with saving on shipping I'd still be paying more than the prices you're talking. Where do you order your Russian tubes online then? I'm in Canada.
Post Reply