MV, Powerscaling in 50W, 100W Amps

[nickb]

Recently the EVH5150III combo schematic came to my attention. On page 5 you can see another take at power scaling (the screen type so to speak). Nothing new as concept except that the "high voltage on pot" problem is avoided and the circuit is driven by opamps.
As you know because of the poor regular dual pot tracking and matching the most simple designs suffer from non linearity which especially at low levels results in crossover distortion and other artifacts.
In this the POWER pot is a dual gang one with the Master which is made maybe to prevent getting that nasty crossover distortion at low levels when the PI is hit with stronger signal. Something similar is going on in the YJM100. No variable feedback like in YJM100 though.
I'm wondering how linear this design would be provided there's some feedback going on from the current sensing resistor (BIAS CURR). Looks like a project for the weekend.
Maybe someone who actually used the amp would be able to tell us how it works.

[ATTACH]41485[/ATTACH]

The cathode current would need to be over 500mA for the bias current trip to kick in. Therefore I guess its purpose is a current limit. I don't think you would see 500mA other than under fault conditions.

The dual ganged pot means that as you reduce the master, you also lower the screens so can still get breakup at the new lower volume.

I see the bias is adjusted by monitoring the screen voltage via U301A to reduce crossover effects.

Here how I did it on a 94 Twin. I may have tweaked a few values from what is shown.
Edit: I have updated this diagram from the original post to reflect the expected component values and make it a little clearer.

[GainFreak]

The cathode current would need to be over 500mA for the bias current trip to kick in. Therefore I guess its purpose is a current limit. I don't think you would see 500mA other than under fault conditions.

I don't know how those 500mA were calculated but it doesn't make sense to me and it's way over what is normally seen for two tubes. What I see is a 50x multiplier at U301B which takes the 60mV (max screen voltage) from the current sensing resistor which results in ~3V fed back to the Power pot.

Your setup looks more complicated and is that a dual ganged pot? Also have you measured the linearity and how both voltage are tracking each other which is very important at low levels?

[nickb]

I don't know how those 500mA were calculated but it doesn't make sense to me and it's way over what is normally seen for two tubes. What I see is a 50x multiplier at U301B which takes the 60mV (max screen voltage) from the current sensing resistor which results in ~3V fed back to the Power pot.

The -ve input of U301-B sits at +0.5V. Therefore the +ve input needs to be greater than this 0.5v to get the diode to turn on and the screen voltage to be reduced. The cathode sensing resistor is 1 ohm ==> 500mA. The x50 gain is just part of the loop gain of a simple proportional servo. It has to be high enough to keep the error small but not so high as to be unstable.

Your setup looks more complicated and is that a dual ganged pot? Also have you measured the linearity and how both voltage are tracking each other which is very important at low levels?

The reason I showed you what I had done is that it's a great deal simpler. Many less components, no opamps, no additional power supplies - The dual pot is key to the simplicity. It doesn't need to be 'linear'. You just need enough current flowing at idle to avoid crossover distortion. It works very well.

[GainFreak]

OK, I see now how you derived the 500mA value but that's only possible if a short occurs or something.
Power supply for an opamp is not a problem. Those small DIP DC to DC converters will do the job. There are other ways as well.
By 'linear" I meant that the bias voltage must change with the same percentage as the screen voltage otherwise the tubes will be biased hotter or colder (because of the dual pot tolerances) and that's a problem especially at those so much desired low levels.

[nickb]

OK, I see now how you derived the 500mA value but that's only possible if a short occurs or something.

I did say that above "I don't think you would see 500mA other than under fault conditions."

Power supply for an opamp is not a problem. Those small DIP DC to DC converters will do the job. There are other ways as well.
By 'linear" I meant that the bias voltage must change with the same percentage as the screen voltage otherwise the tubes will be biased hotter or colder (because of the dual pot tolerances) and that's a problem especially at those so much desired low levels.

By that definition the scheme I showed you is 'linear'. You need to understand that so long as the idle current is high enough to prevent excessive crossover distortion and low enough to stop the tubes for over dissipation then that is plenty good enough. The way I did it was set the top resistor in the bias divider for the full power condition and then set the bottom one for the low power case. That method is a bit iterative and could be improved. The pot tolerances is probably not a big issue - near enough is good enough. This made the current approximately constant - think it varied by something like 10mA over the range*. Of course, if you want exact, then linear is the wrong answer you need something like (Vs)^(2/3).

*Edit: I was talking from errant memory and it was a while back. In fact the current was varied from about 5mA to 35mA

[GainFreak]

The pot tolerances is probably not a big issue - near enough is good enough.

What I noticed in my previous experiments is at voltages 150V and below even a tenth of a Volt bias change would shift the operating point considerably that's why I'm insisting on "linearity" which in this case means both voltages to follow each other closely. At higher voltages of course this is not such a problem.

Of course, if you want exact, then linear is the wrong answer you need something like (Vs)^(2/3).

The response of the pot is not so important and it has nothing to do with the problem I described above.

[nickb]

What I noticed in my previous experiments is at voltages 150V and below even a tenth of a Volt bias change would shift the operating point considerably that's why I'm insisting on "linearity" which in this case means both voltages to follow each other closely. At higher voltages of course this is not such a problem.

There was something wrong with your measurement. That just doesn't happen.

The response of the pot is not so important and it has nothing to do with the problem I described above.

I have no idea where you got that from. At no point did I mention the pot taper. I completely understood you and clearly the reverse is not not true. Go back and look over it again - maybe it will fall into place.

[GainFreak]

There was something wrong with your measurement. That just doesn't happen.

Well, the change may not be that considerable but it's there and you can clearly see on the scope how the wave form changes.

[nickb]

Well, the change may not be that considerable but it's there and you can clearly see on the scope how the wave form changes.

Can you post some scope captures showing the phenomenon along with the DC and load conditions?

[GainFreak]

Maybe next weekend.
I'm making a PCB to test the EVH setup so hopefully I'll have some data on that one as well.

[KevinOConnor]

Hi Guys

TUT4 and TUT6 shows hundreds of circuits that can be used to Power Scale an amp, and the bulk of these use small noncritical pots with little voltage across them. SSH describes the methods as text only, but in both TUT4 and SSH is a chart that shows the performance difference between Full PS and PS-TT (Power Scaling all relevant points versus only two-thirds of them). In both methods, total audio power reduction is the same, but with Full-PS the dissipation in the tubes is reduced so tube life extension is better with Full-PS than with PS-TT.

It is a common misconception by those who have not read the books or talked to me about this, that the operating point changes as Power is dialled down - it does not. In my amps, I use Full-PS, but in amps with just SV-TT the transfer function of the tube remains constant just the same, so the sound stays the same provided the drive into the tube is compensated, hence the need for the Drive Compensation control. For automatic drive compensation, ganged controls can be used, or tracking signal attenuation can be used. As far as the transfer function, all that matters is that the proportion of grid and screen voltage stays the same - and preferably for plate voltage to track, too.

There are many errors that designers make when they "interpret" the Power Scaling schemes, and most of it easily avoided and some of it is quite unfortunate. For example, in the YJ and Slash models from Marshall, the basic character of the amp is changed by the specific interpretation of the circuit and parts choices made. This is a PS-TT approach but poorly implemented, apart from the fact that power reduction is achieved. The drive for the mosfet is not good nor is the choice of mosfet, and it is these things that change the sound of the amp. A tech i know bypassed the EPA and had a Marshall tone again, then reconnected EPA and that family tone was gone. The tech asked me what was going on there, I told him, he made some mods, and now his customer has the only Slash amp that has true Marshall tone to boot with correct Power Scaling.

Have fun

[GainFreak]

For example, in the YJ and Slash models from Marshall, the basic character of the amp is changed by the specific interpretation of the circuit and parts choices made. This is a PS-TT approach but poorly implemented, apart from the fact that power reduction is achieved. The drive for the mosfet is not good nor is the choice of mosfet, and it is these things that change the sound of the amp.

I studied AFD and YJM scheamtics very carefully and tried that implementation of PS using a simple pot and one bias section only, no uCU. The only obvious difference in AFD/YJM not found in most PS designs is the variable feedback.
Also those amps have a lot of other things going on and are not pure tube tone amps so yes, some difference should be expected especially with the variable feedback.
I don't know why you say that the MOSFET drive is not good. MOSFET is setup as a cathode follower. The only obvious reason could be the MOSFET higher input capacitance. I'm not very much into theory but in a CF that shouldn't matter much. The whole setup works fine though.
How does that CF destroy the Marshall tone compared to the "usual" follower used in most designs? Does it generally destroy the tone and what if we're not after "Marshall tone" but some other tone?

Can you post some scope captures showing the phenomenon along with the DC and load conditions?

If you check page 2 of the thread someone has researched the subject already and posted a nice video clip. Because of too many commercials it's a bitch to download from that link so I uploaded it on another site:

http://www.megafileupload.com/kxZm/P...asVariable.avi

[KevinOConnor]

Hi Guys

I've had more than one tech relay the same info regarding bypassing the EPA and having a more traditional tone from the YJ and AFD. The reason is obvious if you know how the circuits work and you have experience with mosfets.Besides the fact of whatever tone it is that the amps are supposed to provide, the fact that bypassing the EPA changes the tone tells you it is poorly implemented even without expert knowledge.

The bulk of mosfets are optimised for switching NOT linear applications. For linear, you have to select a much larger die than for switching and although on the surface the specs for the mosfet used are excessive, there are better devices to use here. And, it is not how the mosfet drives the screens that is at fault, rather how the gate of the mosfet is driven that matters. High-z gate drive destroys dynamics of an audio circuit even when the mosfet is just part of a straight voltage regulator for that circuit. Experience with actual circuits bears this out and I doubt a sim will show it.

In the AFD and YJ, the computer controls bias and monitors the tubes. They did this simply because the processor was already there for the DSP effects - a case of "why not" which is usually for the wrong reason and with a result that is of no benefit. Bias in most amps is entirely stable with simple bias-set pots. Marshall has an ongoing problem with bias circuits... If only they had stolen a good one.

CFs in Marshalls has been discussed at length in other threads. Personally I don't like how the threads branch here. On other forums the threads are linear and easy to follow; here with the branching you have the same Q asked multiple times in different branches of what is supposed to be one thread. Maybe it is just my simple monkey brain that has a problem with that?

Have fun

[g1]

Personally I don't like how the threads branch here. On other forums the threads are linear and easy to follow; here with the branching you have the same Q asked multiple times in different branches of what is supposed to be one thread.

In the upper right area of this page, there should be a button called 'display'. Can you check that it is set to 'linear' ?

[nickb]

I've had more than one tech relay the same info regarding bypassing the EPA and having a more traditional tone from the YJ and AFD. The reason is obvious if you know how the circuits work and you have experience with mosfets.

I believe this is true as in these particular designs have a very low output impedance meaning and effects due to power supply sag are lost. But that is just my own opinion. Phrased as though from a voice of authority, it does seems rather odd to say things like "obvious to those in know" rather actually coming right out and giving us the benefit of your vast wealth of knowledge when the purpose of the discussion is to inform rather than obfuscate.

Besides the fact of whatever tone it is that the amps are supposed to provide, the fact that bypassing the EPA changes the tone tells you it is poorly implemented even without expert knowledge.

You cannot deduce the the circuit is poorly designed simply because it changes the tone. Shall we all immediately dispense with our tone stacks then? After all, they do change the tone? Of course I suppose you meant to say "changes the tone in a bad way", but that is subjective. On the contrary, it seems that considerable thought has gone into it and no doubt was subject to many engineering and audio reviews prior to be released to production. I have do doubt it accomplished precisely what the marketing, engineering and I suppose YJM himself wanted. Some others may not like it but that it no more than their opinion.

The bulk of mosfets are optimised for switching NOT linear applications. For linear, you have to select a much larger die than for switching and although on the surface the specs for the mosfet used are excessive, there are better devices to use here.

A big problem with using switching MOSFET in a linear application is one of thermal stability. Swithers have very high transconductance which pushes the zero temperature coefficient point up to a higher drain current. They are thermally unstable when operated below that current. The thing is this only applies to operation when used in the constant current mode or close to it. In this case the current is determined by the load and not the MOSFET so it's not a concern. What irks me is there seems to be a hint that switching FET are sonically inferior without any supporting evidence. I'm not disagreeing they there may be better choices, although I'd be hard pressed to find one, but rather than keep us hanging again why don't you just come out and say what and why?

And, it is not how the mosfet drives the screens that is at fault, rather how the gate of the mosfet is driven that matters. High-z gate drive destroys dynamics of an audio circuit even when the mosfet is just part of a straight voltage regulator for that circuit.

There's no doubt how the gate is driven is crucial, after all it IS the control terminal of the device, but I doubt Hi-Z within reason has got anything to do with it. But again, do give the whole story instead of just a shard. I'm always prepared to listen and learn.

Experience with actual circuits bears this out and I doubt a sim will show it.

This seems to be an appeal to authority. I can say with absolute certainty that billions of $ and even lives are dependant on sims that produce accurate results. Of course a properly done one will show the problem if only you would tell us what it is supposed to be so we can go looking for it.

In the AFD and YJ, the computer controls bias and monitors the tubes. They did this simply because the processor was already there for the DSP effects - a case of "why not" which is usually for the wrong reason and with a result that is of no benefit. Bias in most amps is entirely stable with simple bias-set pots. Marshall has an ongoing problem with bias circuits... If only they had stolen a good one.

The reason they chose to do it this way is because it allows them a degree of control and flexibility that would be almost impossible to achieve any other way. They didn't do it 'just because the processor was there" but because it was a better way of doing and, dare I say, might be the best way to do it.