The time had finally come for me to see what all the fuss was about when it comes to Class A. For all their shortcomings they must have *something* going for them to have such a loyal following. So I dredged through the junk box and stuck a few bits and pieces together to see just what would happen. The results were so pleasing that I started putting it all together in a chassis as a proper project. This thing is nominally a 50 watt per channel stereo setup with a source follower output stage and an inductor in the source lead. It runs a 28 v rail with 3.5 amps current so that's about 100w dissipation. The advantage of using an inductor is that you only need half the supply voltage because the inductor will swing the missing half rail for you. Therefore the efficiency is at best a horrible 50% rather than a woeful 25%. The downside is that suitable inductors are heavy and expensive. But I already had two of them so that was lucky.
How does it sound? Very smooth and relaxed, crank up the volume and it doesn't get stressed at all. Easy to listen to. Most enjoyable. Not as nice as AKSA's amps that I had a listen to the other week (they are just soooooo good) but the development time on mine was about 2 days if you add it up. I sit there and look at the small handful of parts that is producing all that nice sound and think the results are way out of proportion to the simplicity of it. When I get it further assembled (it is only a rats nest at the moment) I will post a picture or two. In the meantime here is the schematic. Any comments, criticisms, suggestions etc are most welcome.
GP.
P.S. I edited this post and for some reason the schematic fell off. There are other ones later in the thread though.
How does it sound? Very smooth and relaxed, crank up the volume and it doesn't get stressed at all. Easy to listen to. Most enjoyable. Not as nice as AKSA's amps that I had a listen to the other week (they are just soooooo good) but the development time on mine was about 2 days if you add it up. I sit there and look at the small handful of parts that is producing all that nice sound and think the results are way out of proportion to the simplicity of it. When I get it further assembled (it is only a rats nest at the moment) I will post a picture or two. In the meantime here is the schematic. Any comments, criticisms, suggestions etc are most welcome.
GP.
P.S. I edited this post and for some reason the schematic fell off. There are other ones later in the thread though.
Attachments
Last edited:
DC offset
Yes, the speaker does in fact have about 600mV across it. Doesn't worry me but it might make some people's hair stand on end.
It was either that or use an electrolytic cap to couple to the speaker. I *think* I chose the lesser of the two evils. The resistance of the choke is a little under 200 milliohms. It only gets about 2 watts dissipated in it and seing there is about 4-5kg of copper (about ~100 metres of 1/8 inch thick wire) it doesn't warm up much. I am eventually going to low pass filter the dc voltage across it and use it basically as a sense resistor to control the source follower bias so it's current will remain stable regardless of temperature.
Since the last posting I put a 1000uF 16v cap across the 100R source resistor of the first fet. The ac open loop gain increases quite a bit loudness-wise but with the feedback loop closed there is practically no difference gain-wise but it does make the NFB loop work that much harder so it looks like a worthwhile mod.
GP.
Yes, the speaker does in fact have about 600mV across it. Doesn't worry me but it might make some people's hair stand on end.
It was either that or use an electrolytic cap to couple to the speaker. I *think* I chose the lesser of the two evils. The resistance of the choke is a little under 200 milliohms. It only gets about 2 watts dissipated in it and seing there is about 4-5kg of copper (about ~100 metres of 1/8 inch thick wire) it doesn't warm up much. I am eventually going to low pass filter the dc voltage across it and use it basically as a sense resistor to control the source follower bias so it's current will remain stable regardless of temperature.Since the last posting I put a 1000uF 16v cap across the 100R source resistor of the first fet. The ac open loop gain increases quite a bit loudness-wise but with the feedback loop closed there is practically no difference gain-wise but it does make the NFB loop work that much harder so it looks like a worthwhile mod.
GP.
Re: DC offset
I must have been dreaming. I measured it again and it's 300mV. So the choke has about 1 watt dissipation at 3.5 amps. A voice coil with 5 ohms dc resistance would dissipate 18 milliwatts With my 10 inch test loudspeaker the cone offsets about 1/2 mm. Not worth losing sleep over, hey?
GP.
Circlotron said:
I must have been dreaming. I measured it again and it's 300mV. So the choke has about 1 watt dissipation at 3.5 amps. A voice coil with 5 ohms dc resistance would dissipate 18 milliwatts
With my 10 inch test loudspeaker the cone offsets about 1/2 mm. Not worth losing sleep over, hey?GP.
caution: backseat designer
A possible benefit to using an electrolytic coupling cap is that this would turn the output section into the type of circuit used by Pathos, in which the current through the output device is apparently held to a nearly constant value by the storage and release of energy from the resulting LC tank network. I've always been curious about this circuit, but never tried it myself.
http://www.pathosacoustics.com/inpoleng.htm
A possible benefit to using an electrolytic coupling cap is that this would turn the output section into the type of circuit used by Pathos, in which the current through the output device is apparently held to a nearly constant value by the storage and release of energy from the resulting LC tank network. I've always been curious about this circuit, but never tried it myself.
http://www.pathosacoustics.com/inpoleng.htm
problems, problems...
If there was a tank cct there, it would go to a very high impedance at only one frequency, not what you need for a flat frequency response! That company claim to have a worldwide patent but I couldn't find any reference to it in a quick look at the USPTO.
A constant current load for the source follower would indeed be a good thing. I was looking at some of Nelson Pass' amps with the current source at the top and I was thinking if it were arranged to be a negative resistance device that complemented the changing current drain of the load with changing voltage across it, then there is half a chance you could have a constant current load. That would be excellent. How it would go with a reactive loudspeaker load is another story though. Still, I think it's an interesting idea.
Cranked up the amp tonite with sinewaves and an 8 ohm dummy load and it seems it will swing down to -25 volts and then clip (that's fine), but it will only go to +15 volts before becoming decidedly nonlinear. Especially at 15 kHz upward it seems to lose a whole lot of slewing capacity above +15v (i.e.when Vds is < ~13v). But then I am using a *very* large die 500v Hexfet that has been optimised for *switching* not linear service. Once the drain-source voltage is less than about 13 volts in this case, and the drain current would be about 2 amps (perhaps not relevant) and the Vgs is about 6 volts and rising quickly because the negative feedback is starting to take action, the fet just seems to go into lazy mode. Admittedly, the gate is normally supposed to be rammed up to 15 volts or so in about 20 nS and down at the same rate as per it's intended usage. Even then this fet never was particularly fast which is why I ended up being given them. Turnoff time was in the order of 150 nS. The ones that were chosen in place of them turn off 15 amps 500v in 30 nS. So they're a bit of a slug all in all.
It's only in the last 10v of decreasing Vds that the wheels fall off with linear usage. At this voltage the nonlinear Drain-Source capacitance starts to go through the roof. Given the fact that the gate voltage is still in the linear region, perhaps this capacitance is showing up really badly. I might just try a low voltage hexfet tomorrow night and see what happens. Drat!
Oh well, at least I'm learning things and nothing has blown up. Actually this amp seems very resistant to blowups and I have accidentally shorted things inside the cct several times. Not like the CDA amp; you only have too look at that and it pops the output fets. 
It's 1:25 AM. Yoicks! Way past bedtime!
GP.
If there was a tank cct there, it would go to a very high impedance at only one frequency, not what you need for a flat frequency response! That company claim to have a worldwide patent but I couldn't find any reference to it in a quick look at the USPTO.
A constant current load for the source follower would indeed be a good thing. I was looking at some of Nelson Pass' amps with the current source at the top and I was thinking if it were arranged to be a negative resistance device that complemented the changing current drain of the load with changing voltage across it, then there is half a chance you could have a constant current load. That would be excellent. How it would go with a reactive loudspeaker load is another story though. Still, I think it's an interesting idea.
Cranked up the amp tonite with sinewaves and an 8 ohm dummy load and it seems it will swing down to -25 volts and then clip (that's fine), but it will only go to +15 volts before becoming decidedly nonlinear. Especially at 15 kHz upward it seems to lose a whole lot of slewing capacity above +15v (i.e.when Vds is < ~13v). But then I am using a *very* large die 500v Hexfet that has been optimised for *switching* not linear service. Once the drain-source voltage is less than about 13 volts in this case, and the drain current would be about 2 amps (perhaps not relevant) and the Vgs is about 6 volts and rising quickly because the negative feedback is starting to take action, the fet just seems to go into lazy mode. Admittedly, the gate is normally supposed to be rammed up to 15 volts or so in about 20 nS and down at the same rate as per it's intended usage. Even then this fet never was particularly fast which is why I ended up being given them. Turnoff time was in the order of 150 nS. The ones that were chosen in place of them turn off 15 amps 500v in 30 nS. So they're a bit of a slug all in all.
It's only in the last 10v of decreasing Vds that the wheels fall off with linear usage. At this voltage the nonlinear Drain-Source capacitance starts to go through the roof. Given the fact that the gate voltage is still in the linear region, perhaps this capacitance is showing up really badly. I might just try a low voltage hexfet tomorrow night and see what happens. Drat!
Oh well, at least I'm learning things and nothing has blown up. Actually this amp seems very resistant to blowups and I have accidentally shorted things inside the cct several times. Not like the CDA amp; you only have too look at that and it pops the output fets. It's 1:25 AM. Yoicks! Way past bedtime!GP.
Re: problems, problems...
Perhaps I used the term "tank" too loosely given that the load is in series with the C and would reduce the Q of the circuit. I presume that this, combined with the damping influence of the output follower MOSFET, is what makes it all hang together. But again I'm just speculating, having never built or tested one.Circlotron said:
I agree with Nelson, original but I wonder also how much you can "kräma ut" before your ears hurts by distortion (max useable output power)? Bandwith?
Do you have a picture of your "monster"?
Your choke, where did you get it?
300 mV DC, nothing to scared of and if you have to choose between the electrolythic cap I would have done the same thing...but I have had very big problems to sleep....0.0000001 mV is my kind of style.
Do you have a picture of your "monster"?
Your choke, where did you get it?
300 mV DC, nothing to scared of and if you have to choose between the electrolythic cap I would have done the same thing...but I have had very big problems to sleep....0.0000001 mV is my kind of style.
Hope you treble unit is not DC - coupled.
I would personaly accept a max of 50mV. But Congratulation with you home made amp..
Always nice when somebody design a amp from the scratch! This is very rare around here!
.1nV ... What is the noise level of a 50Ohm resistor!?!? ... Okay.. so you have no thermal drift or is it a snapshot of 1 ps?
Sonny
I would personaly accept a max of 50mV. But Congratulation with you home made amp..
Always nice when somebody design a amp from the scratch! This is very rare around here!
.1nV ... What is the noise level of a 50Ohm resistor!?!? ... Okay.. so you have no thermal drift or is it a snapshot of 1 ps?
Sonny
peranders said:
I'll try and post a picture on Saturday if I don't feel to lazy. the chokes were originally made as filter chokes for a 350vdc 15 amp power supply that was part of 2 UPS's. They ended up never getting used and so I hauled them home. Even took them with me when I shifted house 8 years ago.
Originally they measured 22 mH each and had a 1/4 inch wide airgap between the facing EE cores. I pulled out the spacer and tapped the two halves of the stack together untill the centre legs of the E's met. This left about 1mm gap between the outer legs. Applied 40vac and measured the current to come up with the 85mH. Just lucky actually. They will actually take 240vac @ 50Hz without saturating although they do buzz a little bit. Overkill? You bet! Now, as far as that distortion problem is concerned, talk about looking for a complex reason for the problem! I got to thinking about things and now I think it is because when the Hexfet starts to run out of puff on the positive half cycle - guess what? The output waveform starts to round off and so of course the bootstrapping starts to fall on it's face too! So now at the very moment the driver stage has sit up and work the bootstrap cct goes on a holiday.
What a fair-weather friend! What I think I will do now is make some sort of current source load for the driver instead of the 1k resistor. Maybe a current mirror using a pair of pnp's. Bootstrapping would not be particularly necessary then perhaps. But of course I will still try it in conjunction with the current source to try and get the most linear cct possible. The gain of that stage should be quite high by then I think.GP.
You too? Is it because of me? Is it a contagious? I asure you that I haven't tried to infect you by purpose. Just lucky actually.
"Originally they measured 22 mH each and had a 1/4 inch wide airgap between the facing EE cores. I pulled out the spacer and tapped the two halves of the stack together untill the centre legs of the E's met. This left about 1mm gap between the outer legs. Applied 40vac and measured the current to come up with the 85mH. Just lucky actually."
Hmmmm......I wonder what that 1/4 inch air gap was in there for?
Can you tell us Peranders?
H.H.
"Originally they measured 22 mH each and had a 1/4 inch wide airgap between the facing EE cores. I pulled out the spacer and tapped the two halves of the stack together untill the centre legs of the E's met. This left about 1mm gap between the outer legs. Applied 40vac and measured the current to come up with the 85mH. Just lucky actually."
Hmmmm......I wonder what that 1/4 inch air gap was in there for?
Can you tell us Peranders?
H.H.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.