SS poweramp with soft clipping on FET's.

[Medved]

Scheme:

Clipping:

Sim sound:
http://webdrive.reborn.ru/egorov/power/rock1.rar

30w on 8 ohms AB class completly symmetrical poweramp. Max input voltage nearly 2.8V, gain nearly 8-10, output impedance nearly 2 ohms. Clipping begins at 1v input (10w on 8Ohms output power). Efficiency up to 80% in clipping. Power dissapation from each mosfet up to 13w.
It's only a project- i have no time to check it... and have no bf245c...
Very simple, as all ingenious

[brett]

Hi.
Very interesting amp.
What do you need for V+ ?
I wonder whether a solid wallwart (24VAC 1 amp ??) would do it?

I assume that a J201 makes a good phase splitter (J1), and also work in J2 and J3, given the adjustable source resistance (R6).

Would cheaper MOSFETs work instead of the SK1058s ?  I use MTP3055s coz they're very rugged (60V 10A or thereabouts) and only cost $2 each (less in US$).

For a really simple amp, you could just add a stratoblaster (1 jfet) pre-amp stage and a big muff tone control.

Can't wait to hear from someone who builds it.....

[Medved]

Hi. Very interesting amp.

Thanks.

What do you need for V+ ?
I wonder whether a solid wallwart (24VAC 1 amp ??) would do it?

Any stabilized dc source would go. With preamp on jfets like FETslo100 or D90/D900 plus clean channel on jfets like fender twin you'll need only 100ma...

I assume that a J201 makes a good phase splitter (J1), and also work in J2 and J3, given the adjustable source resistance (R6).

No. BF245c or any other high cutoff jfet is a must. Vcutoff should be as high as possible. At least 4v...

Would cheaper MOSFETs work instead of the SK1058s ?  I use MTP3055s coz they're very rugged (60V 10A or thereabouts) and only cost $2 each (less in US$).

May be. You should understand, that there is no thermocompensation in this scheme. So, if a MOSFET has positive thermal coefficient it will probably blow. 2sk1058 is thermostable MOSFET. If you want to use another, you should use only thermostable one. Like russian 2P904A 

For a really simple amp, you could just add a stratoblaster (1 jfet) pre-amp stage and a big muff tone control.

Preamp isn't interesting for me. You can use what you want.

Can't wait to hear from someone who builds it.....

Don't wait- build it yourself. 

[R.G.]

It's interesting. A few questions.

1. The load appears to be completely floating; in fact, it appears to be floating at V1, not ground referredt at all. Also, V2 and V3 appear to be completely floating, no ground reference at all, and it appears to be important that they MUST float. Is that correct?
2. I see no feedback stabilization at all. I believe that would make either device matching or individual part tweaking critical to get a static operating point of 0V DC within a few millivolts across the load. I believe that it would be critical to have J2 and J3 both matched and thermally coupled to keep the DC point across the load from drifting. I suspect that M1 and M2 will have to be electrically matched for gain and internal resistance and thermally coupled as well. Thermo stable just means that the channel resistance increases with increasing temperature, not that the devices will self match.
3. I believe that R6 sets the AB bias point. Is that correct? If so, that is the place to apply thermal stabilization.
4. If it puts out 30W to a load, and dissipates 26W in the MOSFETs (13Wx2), then the power supply needs to make at least 56W. Since that's divided in two (V2 and V3) then you need two floating power supplies of 20V at 1.4A. The +V supply could indeed be small.
5. 30W into 8 ohms needs 22V peaks for sine waves. MOSFETs don't saturate as well as bipolars, so they typically have 3-5V across them when pulling significant currents. It looks like you need V2, V3 to be maybe 25V to make 30W. That means (I think...) that V+ needs to be about 30V. No biggie, but the power is probably more like 20W, which is OK.

I think that the clipping that's being seen is the clipping of the JFET diffamp. It looks very much like what JFET diffamps do.  This design is very much like an old design with tubes from a couple of scholarly references I remember from the Golden Age, about 1940's.

[Medved]

The load appears to be completely floating; in fact, it appears to be floating at V1, not ground referredt at all. Also, V2 and V3 appear to be completely floating, no ground reference at all, and it appears to be important that they MUST float. Is that correct?

Not entirely. 24v supply is grounded. +V is connected to a stable potential, definded by devider. v2 and v3 (and sources of mosfets of course $) ) are connected to this potential through r9 and r10.

I see no feedback stabilization at all. I believe that would make either device matching or individual part tweaking critical to get a static operating point of 0V DC within a few millivolts across the load. I believe that it would be critical to have J2 and J3 both matched and thermally coupled to keep the DC point across the load from drifting. I suspect that M1 and M2 will have to be electrically matched for gain and internal resistance and thermally coupled as well. Thermo stable just means that the channel resistance increases with increasing temperature, not that the devices will self match

Yes. There is no general negative feedback. Yes, mathing of tranies is required. I perfectlly know what means "thermo stable", and i've said no word about "self matching". Be attentive, please

I believe that R6 sets the AB bias point. Is that correct? If so, that is the place to apply thermal stabilization.

No. R12 sets the AB bias point. I'm not going to implement thermal compensation, sorry. I prefer to use normal MOSFET.
R6 is needed to bias the difamp for right working parameters. Of cource, it can be used for biasing mosfets, but...

If it puts out 30W to a load, and dissipates 26W in the MOSFETs (13Wx2), then the power supply needs to make at least 56W. Since that's divided in two (V2 and V3) then you need two floating power supplies of 20V at 1.4A. The +V supply could indeed be small.

No. It's AB.  At a time up to 43w is needed from each power supply, that is 20v@2.15A.

30W into 8 ohms needs 22V peaks for sine waves.

No.

Code Select Expand

P=U*U/R. Hence, U=sqrt(P*R). U=sqrt(30*8)=15.5v

I think that the clipping that's being seen is the clipping of the JFET diffamp. It looks very much like what JFET diffamps do.  This design is very much like an old design with tubes from a couple of scholarly references I remember from the Golden Age, about 1940's.

Yes, you are right. Difamp makes basic clipping, but power mosfets clip too. From difamp i have 20-21v double amplitude. Power stage begins clipping at 16-18v double amplitude.

[R.G.]

The load appears to be completely floating; in fact, it appears to be floating at V1, not ground referredt at all. Also, V2 and V3 appear to be completely floating, no ground reference at all, and it appears to be important that they MUST float. Is that correct?

Not entirely. 24v supply is grounded. +V is connected to a stable potential, definded by devider. v2 and v3 (and sources of mosfets of course $) ) are connected to this potential through r9 and r10.

I see. So the two floating supplies must float compared to each other, but are tied up in the air at 12.5V, defined by R9 and R10. They are connected to ground by an impedance of 4.79K. Presumably if you accidentally touch either side of the speaker to ground, this upsets things. Yes?

I see no feedback stabilization at all. I believe that would make either device matching or individual part tweaking critical to get a static operating point of 0V DC within a few millivolts across the load. I believe that it would be critical to have J2 and J3 both matched and thermally coupled to keep the DC point across the load from drifting. I suspect that M1 and M2 will have to be electrically matched for gain and internal resistance and thermally coupled as well. Thermo stable just means that the channel resistance increases with increasing temperature, not that the devices will self match
Yes. There is no general negative feedback. Yes, mathing of tranies is required. I perfectlly know what means "thermo stable", and i've said no word about "self matching". Be attentive, please Wink

OK, I thought matching would be needed; so to do a good job with this, one must match both the JFETs and the power MOSFETs. As to being attentive, I thought I did pretty well ferreting out that matching was necessary. Forgive me if I unnecessarily thought you were making the ultra-common beginner's mistake in thinking that MOSFETs will self match by using the negative coefficient of channel resistance. It is one that nearly every beginner makes.

I believe that R6 sets the AB bias point. Is that correct? If so, that is the place to apply thermal stabilization.

No. R12 sets the AB bias point. I'm not going to implement thermal compensation, sorry. I prefer to use normal MOSFET.
R6 is needed to bias the difamp for right working parameters. Of cource, it can be used for biasing mosfets, but...

OK, so R6 can be used for thermal stabilization, but that's not the only place. It looks like R6 and R12 would interact as well, since R12 sets the pull-down voltage on the sources of the MOSFETs compared to the JFET drains, and R6 sets the voltage on the JFET drains.

If it puts out 30W to a load, and dissipates 26W in the MOSFETs (13Wx2), then the power supply needs to make at least 56W. Since that's divided in two (V2 and V3) then you need two floating power supplies of 20V at 1.4A. The +V supply could indeed be small.

No. It's AB.  At a time up to 43w is needed from each power supply, that is 20v@2.15A.

I was just using your numbers, 30W into 8 ohms, and 13W in each MOSFET. I did wonder about that. So is the 43W peak, average, RMS, from each power supply? But I can readily believe that even more than 1.4A is needed in each of the floating power supplies.

30W into 8 ohms needs 22V peaks for sine waves.
No.
Code:
P=U*U/R. Hence, U=sqrt(P*R). U=sqrt(30*=15.5v

I calculated it as V=sqrt(P*R)=15.5, OK, but I added the additional voltage headroom that MOSFETs usually need in AB amps. This is normally 4-8V. I used 7. You could perhaps get by with less, depending on the exact MOSFET and conditions. Or were you counting on being able to use the entireR9 voltage to pull the gates of the MOSFETs over their sources? Be attentive, please 

Yes, you are right. Difamp makes basic clipping, but power mosfets clip too. From difamp i have 20-21v double amplitude. Power stage begins clipping at 16-18v double amplitude.

I can believe that. I think you're running out of enhancement voltage on the MOSFETs, which is good in this case because it makes the channel go higher resistance as they get starved for drive at higher levels.

[Medved]

I see. So the two floating supplies must float compared to each other, but are tied up in the air at 12.5V, defined by R9 and R10.

Exactly.

They are connected to ground by an impedance of 4.79K.

It doesn't matter. R8 and R9 should be much bigger then load. They can be 1k, 10k,100k,1meg- choose what you like best %)

Presumably if you accidentally touch either side of the speaker to ground, this upsets things. Yes?

It means that current through each mosfet will be 12A and power dissapation nearly 240W without input signal. Nevertheless, there will be no current through the load So, if you use normal heatsinks it should stay alive for sometime. Anyhow, if you are afraid then set 5A slowblo fuses in mosfet's drains...

OK, I thought matching would be needed; so to do a good job with this, one must match both the JFETs and the power MOSFETs. As to being attentive, I thought I did pretty well ferreting out that matching was necessary.

MOSFETs relative to JFETs have small parameters spread. From 10 2p904a i have 3 ideally matched pairs, one with small differences and the rest two are very different. So...
From mosfets we need only static similarity- they should have close Vgs on working current. The rest doesn't matter.
JFET's parameters should be as close as possible- it's a common  demand for difamp. Anyhow, you can ajust r1 or r2 to balance the zero...

Forgive me if I unnecessarily thought you were making the ultra-common beginner's mistake in thinking that MOSFETs will self match by using the negative coefficient of channel resistance. It is one that nearly every beginner makes.

Let's leave beginer's mistakes to beginners, ok? 

OK, so R6 can be used for thermal stabilization, but that's not the only place. It looks like R6 and R12 would interact as well, since R12 sets the pull-down voltage on the sources of the MOSFETs compared to the JFET drains, and R6 sets the voltage on the JFET drains.

R6 is used to set up the JFET's drain current, which determine how difamp works and clips. It's bad idea to use it for biasing mosfets.

I was just using your numbers, 30W into 8 ohms, and 13W in each MOSFET. I did wonder about that. So is the 43W peak, average, RMS, from each power supply? But I can readily believe that even more than 1.4A is needed in each of the floating power supplies.

It's AB class. When one mosfet is opened another is closed and there is no current through it. That's why it's needed only 30+13w peak but from each floating power supply.

I calculated it as V=sqrt(P*R)=15.5, OK, but I added the additional voltage headroom that MOSFETs usually need in AB amps. This is normally 4-8V. I used 7. You could perhaps get by with less, depending on the exact MOSFET and conditions. Or were you counting on being able to use the entireR9 voltage to pull the gates of the MOSFETs over their sources? Be attentive, please 

Mosfet stage is driven by 20-21v from diffamp (even more, than its supply voltage). It has gain of ~0.8. Hence, 16v output or little more then 30w. When input voltage for mosfet stage is 20v (gate), source is only 16v. vgs is +4v.

[R.G.]

They are connected to ground by an impedance of 4.79K.
It doesn't matter. R8 and R9 should be much bigger then load. They can be 1k, 10k,100k,1meg- choose what you like best %)

It doesn't matter - until a beginner grounds his speaker output by connection to something else, and then can't understand what happened. 

Presumably if you accidentally touch either side of the speaker to ground, this upsets things. Yes?
It means that current through each mosfet will be 12A and power dissapation nearly 240W without input signal. Nevertheless, there will be no current through the load So, if you use normal heatsinks it should stay alive for sometime. Anyhow, if you are afraid then set 5A slowblo fuses in mosfet's drains...

And that's kind of what I was getting at. I'm not sure that popping a fuse whenever you touch one speaker line to ground is a good idea. It's not that I'm particularly afraid of it - in fact, I spent part of my misspent youth seeing how brightly I could make fuses flash, like strobe lights. But again, it might be disconcerting to an uninformed user. I think you're going to need to use connectors other than phone jacks, or permanently connected speakers to avoid that.

That's not saying the amp cannot be made to work. But it shouldn't mystify users.

You'll have to pardon me if I'm running extra practically now - I'm doing safety certification work, so the issues of protecting the user and the circuit from misuse or easy-to-misuse things is on my mind.

MOSFETs relative to JFETs have small parameters spread. From 10 2p904a i have 3 ideally matched pairs, one with small differences and the rest two are very different. So...
From mosfets we need only static similarity- they should have close Vgs on working current. The rest doesn't matter.
JFET's parameters should be as close as possible- it's a common  demand for difamp. Anyhow, you can ajust r1 or r2 to balance the zero...

OK. Matching is needed. I thought so, and it is. It's good that it's easy for the power devices.

Let's leave beginer's mistakes to beginners, ok?  icon_mrgreen

I wish I could. 90% of the people who read here never post, and are fully and completely beginners. Don't you think us experienced guys should not let them get into water too deep for them to swim without at least a warning sign?

R6 is used to set up the JFET's drain current, which determine how difamp works and clips. It's bad idea to use it for biasing mosfets.

OK.

I was just using your numbers, 30W into 8 ohms, and 13W in each MOSFET. I did wonder about that. So is the 43W peak, average, RMS, Huh from each power supply? But I can readily believe that even more than 1.4A is needed in each of the floating power supplies.
It's AB class. When one mosfet is opened another is closed and there is no current through it. That's why it's needed only 30+13w peak but from each floating power supply.

Yes, AB class. I'm familiar with it.

I was just using your numbers, as I said. You said "Power dissapation from each mosfet up to 13w." So did you really mean 13W peak per MOSFET, not 13W average per MOSFET?

But that wasn't my point. My point was - there are two substantial and one trivial power supply to be built; the two big ones are 20V and between 1 and 2 amps each, and they both have to be non-ground referenced, as does the speaker. The power supply for this things needs at least three secondary windings on the power transformer(s), at least three main power supply capacitors, and so on.

I'm worried about all those beginners again. The idea that a power supply or speaker is not grounded on at least one side or the other will be foreign to them I'm not trying to be difficult, just pointing out what the first builder or two who hooks up this thing will be posting questions about. Beginners forget my maxim that a power amp is really a big power supply with a few inconsequential parts on it to let some of the power out to the speakers in a carefully controlled way. And us experienced guy gotta help beginners out, right?

Mosfet stage is driven by 20-21v from diffamp (even more, than its supply voltage). It has gain of ~0.8. Hence, 16v output or little more then 30w. When input voltage for mosfet stage is 20v (gate), source is only 16v. vgs is +4v.

Right. I was thinking about Vds under the same conditions. Perhaps your MOSFETs have a low threshold voltage. Power MOSFETs in general have a transconductance of about 1A/V above the threshold voltage. The ones I'm most familiar with have Vth of about 3V, so nothing happens until you get 3V on the gate-source. The 1058 may have a lower Vth. It's putting 16V/2A into the load, so you'll need a couple of volts on the gate above the threshold. If you have a low threshold MOSFET, great. But I was only going on what I saw in the schematic and things from other amplifiers. This one could well be magically delicious. 

[A.S.P.]

I want to hook up a non-transformerized DI-box for recording directly into my console
to this amp...  (example of R.G`s concerns);

however: 99% of the readers won`t build this amp from that schem as is...


(the 1% considering to do so, are aquainted to smoking components, arcing electrodes, and exploding caps...).

[Medved]

It doesn't matter - until a beginner grounds his speaker output by connection to something else, and then can't understand what happened. 

R.G., sorry, but in my opininon if a man could be so inaccurate to connect floating out to gnd he shouldn't build this amp. I've said, that it will survive the short term connection.

I think you're going to need to use connectors other than phone jacks, or permanently connected speakers to avoid that.

Yes, you are right. I'm going to use canon XLR 

You'll have to pardon me if I'm running extra practically now - I'm doing safety certification work, so the issues of protecting the user and the circuit from misuse or easy-to-misuse things is on my mind.

I understand, but floating out is a basic concept of this amp, so...

I was just using your numbers, as I said. You said "Power dissapation from each mosfet up to 13w." So did you really mean 13W peak per MOSFET, not 13W average per MOSFET?

Yes. Up to 13w peak from each mosfet. Did i write otherwise? $)
Average power dissapation on 1k sine wave will be less then 5w from each mosfet, and average output power will be nearly 20w.

But that wasn't my point. My point was - there are two substantial and one trivial power supply to be built; the two big ones are 20V and between 1 and 2 amps each, and they both have to be non-ground referenced, as does the speaker. The power supply for this things needs at least three secondary windings on the power transformer(s), at least three main power supply capacitors, and so on.

Right.

Right. I was thinking about Vds under the same conditions. Perhaps your MOSFETs have a low threshold voltage. Power MOSFETs in general have a transconductance of about 1A/V above the threshold voltage. The ones I'm most familiar with have Vth of about 3V, so nothing happens until you get 3V on the gate-source. The 1058 may have a lower Vth. It's putting 16V/2A into the load, so you'll need a couple of volts on the gate above the threshold. If you have a low threshold MOSFET, great. But I was only going on what I saw in the schematic and things from other amplifiers. This one could well be magically delicious. 

Ok. Let's look at diffamp. It has symmetric out, right? Each branch of diffamp has a working point near 13v- it is a potential of mosfet's gates. This potential is indepedent from power stage. If a mosfet has vcutoff near +5v then v1 should be 13-5=8v.
Do yo have any sim? Just use it and it'll be clear...

It isn't amp for beginners, in spite of it is simple. There are many hidden dangers. For example, if a difamp current is incorrect, then both mosfets will be opened for sometime, and current through one of it can be many times bigger then max working current of 2A.

[Medved]

Ok. Here is a scheme with fixed biasing of power mosfets and grounded output. May be you'll find it more safely

[Joe]

It's similar to circuit #2 here:
http://depalma.pair.com/Analog/analog.html

[R.G.]

R.G., sorry, but in my opininon if a man could be so inaccurate to connect floating out to gnd he shouldn't build this amp. I've said, that it will survive the short term connection.

Oh, no need to apologize. I thought that was the case, but I wanted the beginners to know that. Beginners make errors that they would NEVER do if they had any experience. A note that "this design is not for beginners" is a good way to do that.

Yes. Up to 13w peak from each mosfet. Did i write otherwise? $)
Average power dissapation on 1k sine wave will be less then 5w from each mosfet, and average output power will be nearly 20w.

Yeah - you didn't mention "peak", which I would have taken as an instantaneous value. 5-6W is much more like what I'd expect a 20W class AB amp to do - but this is a nonstandard amp, and I guessed that you knew something I didn't.

Ok. Let's look at diffamp. It has symmetric out, right?

Yes, given good matching.

Each branch of diffamp has a working point near 13v- it is a potential of mosfet's gates. This potential is indepedent from power stage. If a mosfet has vcutoff near +5v then v1 should be 13-5=8v.

Actually, the potential of the MOSFET gates isn't independent of the MOSFET's gates. The MOSFET gate-source potential is set by the difference between the two drains and the reference voltage to which the sources are tied. So if the reference voltage ...or... the drain voltage changes, this changes the Vgs. The gates can only be enhanced by the difference between the reference voltage and the maximum drain voltage on the diffamp. That may be entirely sufficient for the lateral MOSFETs you used. It is marginal for the more-available vertical MOSFETs that are easiest to find here.

Do yo have any sim? Just use it and it'll be clear...

I don't really need to. I had to live without simulation programs for so long that I do most circuit stuff in my head. I usually only need to simulate stuff when I need exact numerical answers or when the answers are very unclear. Besides, the schemo you showed was from a simulator input, so I figured you had the answers.

It isn't amp for beginners, in spite of it is simple. There are many hidden dangers. For example, if a difamp current is incorrect, then both mosfets will be opened for sometime, and current through one of it can be many times bigger then max working current of 2A.

I agree completely.

By the way - I think it's a great new/old approach, fitting up a MOSFET version of old tube circuits and finding that it produces a nice, soft sounding overload characteristic. And I know I've been a bit of a PITA by asking such pointed questions. But I thought it was important to the other readers to learn not only that this particular thing was not a beginner's circuit, but also some of the kinds of considerations that go into it. And you were a great guy about the questions. I owe you a few rounds of your favorite Foaming Mental Lubricant when we meet.   

[Medved]

It's similar to circuit #2 here:
http://depalma.pair.com/Analog/analog.html

Here, in Russia we say that any new thing is a well forgotten old one $)
Scheme with two floating power supplies is well known one.

Yeah - you didn't mention "peak", which I would have taken as an instantaneous value. 5-6W is much more like what I'd expect a 20W class AB amp to do - but this is a nonstandard amp, and I guessed that you knew something I didn't.

I've said "up to". Doesn't it mean "peak"? 

Actually, the potential of the MOSFET gates isn't independent of the MOSFET's gates. The MOSFET gate-source potential is set by the difference between the two drains and the reference voltage to which the sources are tied. So if the reference voltage ...or... the drain voltage changes, this changes the Vgs. The gates can only be enhanced by the difference between the reference voltage and the maximum drain voltage on the diffamp. That may be entirely sufficient for the lateral MOSFETs you used. It is marginal for the more-available vertical MOSFETs that are easiest to find here.

R.G., I started with ifr510 which has v0=4v and it works fine $)

I don't really need to. I had to live without simulation programs for so long that I do most circuit stuff in my head. I usually only need to simulate stuff when I need exact numerical answers or when the answers are very unclear. Besides, the schemo you showed was from a simulator input, so I figured you had the answers.

I can't explain it better, than i've written earlier... My english isn't so good...
Look at the scheme with fixed biasing. It's basically the same concept, but now v1 (as in the first scheme) is grounded and diodes make 550mv biasing potential. Do you agree, that it can work with any mosfet by changing the biasing potential? Do you agree that it realizes all declared characteristics (30w peak, 20w average, soft clip and so on)?
Now, just add to v1 and diodes potential of 12,5v. Do you agree, that nothing has changed? Now exclude diodes. Just leave potential of +13v on gates, +12.5 on sources (v1). Nothing changes? Now remove decoupling capacitors. You'll get exactly the first scheme. Do you understand it now?

By the way - I think it's a great new/old approach, fitting up a MOSFET version of old tube circuits and finding that it produces a nice, soft sounding overload characteristic. And I know I've been a bit of a PITA by asking such pointed questions.

What is "PITA"?

But I thought it was important to the other readers to learn not only that this particular thing was not a beginner's circuit, but also some of the kinds of considerations that go into it. And you were a great guy about the questions.
I owe you a few rounds of your favorite Foaming Mental Lubricant when we meet.   

"Foaming Mental Lubricant"... What is it?
I understand your sarcasm It's really rather funny

In Russian community we often show only pure ideas, as "information only for beginners" leads away from concepts and their understanding...

[Medved]

R.G., thanks for your splendid sence of humor Exiting reading and brainstorming. Let's continue?     

[Medved]

P.S.  Scheme in the post #10 is more beginner-friendly- one can ground output with no risk if there is no input signal 

[petemoore]

  PITA = Pain in the @ss
  Your favorite Foaming Mental Lubricant = Really Great Beer I suspect

[Medved]

Thanks.

[trevize]

beginner questions: 

_what power supply should be used with this power amp?
_how many watts should be rated for the resistors used in the circuit?

[Medved]

_what power supply should be used with this power amp?

it depends on output power. For 15-20W you need a power supply with three independent output DC voltages : 20v, 20v, 24v as shown on the scheme.

_how many watts should be rated for the resistors used in the circuit?

0.25W will be enough.