Time for a new great power amplifier, practical and simple yet advanced, production ready and fully protected. I believe in non feedback design, which in my opinion gives a lot of advantages matching my design requirements:
1) Naturally balanced inputs
2) Ability to drive real loads with high currents
3) Unconditional stable no matter load
4) Best sound possible, not best measurements
5) Very high reliability
6) Very low noise
7) Well behaved when overloaded
The schematic looks complex, but if you take out the onboard power supply, DC servo and the full protection circuit it's actually very simple, just a jfet differential input, push-pull folded cascode driver and a two stage emitter follower output stage.
The power supply is onboard, just connect a toroidal transformer with the right voltages. The amplifier is also fully protected against short circuit, high temperatures and DC on output, with a mosfet based relay.
Some of the design decisions:
1) Single power rails. It would be better having the driver part running on higher voltages than the outputs, but I would like to be able to use simple standard toroidal transformers. I have still tried to get as high output swing as possible, resulting in output voltage swinging to 4V-5V under the power rails when loaded.
2) jfet input. I believe in shortest path, and the only way to have that "single stage" high current folded cascode driver with high input impedance is if using a jfet differential input, and I have chosen the LSK389 dual part to avoid matching parts, and for the very low noise.
3) Two stage output. Again the shortest path principle and for maximum output swing, when using the high current folded cascode driver and the newest and best bipolar transistors it can be done with great results.
4) DC Servo. Without global feedback then the DC drift is depending on the whole amplifer, not just the input pair, so a DC servo is unavoidable. I have then chosen a design using low power and controlling the jfet input constant current balance in order not to influence the AC performance of the amplifier.
5) Mosfet output relay. The only way to fully protect expensive speakers is to be able to disconnect them. I hate mechanical relays, but now it's time to use mosfet relays, the parts are now good and inexpensive, and the special photovoltaic optoisolator make it simple.
6) Bandwidth limiting, I don't want a short wave transmitter so I believe in limiting bandwith to 300 Khz. I also don't trust what's it's being feed on the input so I like to have a DC blocking capacitor there.
Some measurements on a breadboard prototype with just one pair of output transistors and into 4 ohms load:
1) THD <0.5% at full power and 10 Khz, quickly going down to <0.1% at lower power levels.
2) 900 Khz !! power bandwidth without the bandwidth limiting capacitors.
3) Output impedance 0.45 ohm
Practical Implementation:
The schematic is designed for three versions, with output power up to 60W, 110W and 190W, using one, two or three sets out output transistors. I'm doing the 110W version first, if successful then the others will follow. The PCB is designed for directly mount on a heatsink, the 110W version to fit in a 2U case.
I'm a surface mount guy and some parts nowadays are only available in SMT, so the PCB is mostly surface mount, I have tried not to use the smallest parts, limiting myself to 0805 and sot23. And except for the LSK389, all other parts are stocking at either Digkey or Mouser.
I'm sending off gerbers for prototype PCB's now. When I'm done optimizing the design and if there are interest then I'm planning to start a small production and selling them a good prices....
Attached is full schematics and PCB silk screen print. All commercial rights are reserved, but feel free to use part or all for noncommercial purposes....
Comments and Feedback very welcome, but remember my design requirements....
1) Naturally balanced inputs
2) Ability to drive real loads with high currents
3) Unconditional stable no matter load
4) Best sound possible, not best measurements
5) Very high reliability
6) Very low noise
7) Well behaved when overloaded
The schematic looks complex, but if you take out the onboard power supply, DC servo and the full protection circuit it's actually very simple, just a jfet differential input, push-pull folded cascode driver and a two stage emitter follower output stage.
The power supply is onboard, just connect a toroidal transformer with the right voltages. The amplifier is also fully protected against short circuit, high temperatures and DC on output, with a mosfet based relay.
Some of the design decisions:
1) Single power rails. It would be better having the driver part running on higher voltages than the outputs, but I would like to be able to use simple standard toroidal transformers. I have still tried to get as high output swing as possible, resulting in output voltage swinging to 4V-5V under the power rails when loaded.
2) jfet input. I believe in shortest path, and the only way to have that "single stage" high current folded cascode driver with high input impedance is if using a jfet differential input, and I have chosen the LSK389 dual part to avoid matching parts, and for the very low noise.
3) Two stage output. Again the shortest path principle and for maximum output swing, when using the high current folded cascode driver and the newest and best bipolar transistors it can be done with great results.
4) DC Servo. Without global feedback then the DC drift is depending on the whole amplifer, not just the input pair, so a DC servo is unavoidable. I have then chosen a design using low power and controlling the jfet input constant current balance in order not to influence the AC performance of the amplifier.
5) Mosfet output relay. The only way to fully protect expensive speakers is to be able to disconnect them. I hate mechanical relays, but now it's time to use mosfet relays, the parts are now good and inexpensive, and the special photovoltaic optoisolator make it simple.
6) Bandwidth limiting, I don't want a short wave transmitter so I believe in limiting bandwith to 300 Khz. I also don't trust what's it's being feed on the input so I like to have a DC blocking capacitor there.
Some measurements on a breadboard prototype with just one pair of output transistors and into 4 ohms load:
1) THD <0.5% at full power and 10 Khz, quickly going down to <0.1% at lower power levels.
2) 900 Khz !! power bandwidth without the bandwidth limiting capacitors.
3) Output impedance 0.45 ohm
Practical Implementation:
The schematic is designed for three versions, with output power up to 60W, 110W and 190W, using one, two or three sets out output transistors. I'm doing the 110W version first, if successful then the others will follow. The PCB is designed for directly mount on a heatsink, the 110W version to fit in a 2U case.
I'm a surface mount guy and some parts nowadays are only available in SMT, so the PCB is mostly surface mount, I have tried not to use the smallest parts, limiting myself to 0805 and sot23. And except for the LSK389, all other parts are stocking at either Digkey or Mouser.
I'm sending off gerbers for prototype PCB's now. When I'm done optimizing the design and if there are interest then I'm planning to start a small production and selling them a good prices....
Attached is full schematics and PCB silk screen print. All commercial rights are reserved, but feel free to use part or all for noncommercial purposes....
Comments and Feedback very welcome, but remember my design requirements....
Not bad, I just don't like ZERO global feedback... As I don't like either TONES of fedback. 0.5% is good for what you did, just not good enough for my taste. Same with the output impedance - too high 5-10%. You still have local feedback (input stage R37) so... why not some global?
Maybe it was better just a little global negative feedback instead of killing the bandwidth for nothing with the cap.
I like the PS, the servo, thermal protection.
Not crazy about MOS-FETs "on" resistance series with the speaker - especially with no feedback. I would use a relay but that's me.
Ah, and I would use a separate suply for the OpAmps, now are too "close" to the current sources, even with the 1uF caps on their rails...
Maybe it was better just a little global negative feedback instead of killing the bandwidth for nothing with the cap.
I like the PS, the servo, thermal protection.
Not crazy about MOS-FETs "on" resistance series with the speaker - especially with no feedback. I would use a relay but that's me.
Ah, and I would use a separate suply for the OpAmps, now are too "close" to the current sources, even with the 1uF caps on their rails...
Last edited:
As soon as you start with global feedback you:
1) Loose the natural differential inputs
2) Loose the unconditional stability
3) Loose the well behaved when overload
And I don't consider R37 local feedback, it's just part of the differential voltage to current converter. But that might be a matter of opinion, I don't consider it feedback until you actually feed something back from an output to an input, in this case that would have been from the jfet drain to source....
In real world relays will have higher impedance than the two mosfets, and more potential for nonlinearity, especially when they get old....
The Opamp draw 7 uA with 221K on the output and the Comparators draw 9 uA each and only toggle when something bad happens, can't see how that in any way can affect anything....
Soren
1) Loose the natural differential inputs
2) Loose the unconditional stability
3) Loose the well behaved when overload
And I don't consider R37 local feedback, it's just part of the differential voltage to current converter. But that might be a matter of opinion, I don't consider it feedback until you actually feed something back from an output to an input, in this case that would have been from the jfet drain to source....
In real world relays will have higher impedance than the two mosfets, and more potential for nonlinearity, especially when they get old....
The Opamp draw 7 uA with 221K on the output and the Comparators draw 9 uA each and only toggle when something bad happens, can't see how that in any way can affect anything....
Soren
1) Differential inputs are not something that important for final quality of sound. Same results can be achieved with SE feeds.
2) Loose the stability? Only if you drive pure capacitors you might start to look into phase margin. Especialy with moderate global open loop gain/negative feedback.
3) Just say no to overloading.
I am more concerned with low damping factor and what that does to bass control. With non-linear distortions and what that does to the timbre (harmonics), but... that's me, no problem.
R37 is local feedback, the curent in the source is deducted from the other side (that has a "natural" inverted phase) diminishing the output. That looks like (local) negative feedback to me. And seems to me that R52 and R53 should have reversed the values.
Don't take it the wrong way, you have asked for comments 🙂
2) Loose the stability? Only if you drive pure capacitors you might start to look into phase margin. Especialy with moderate global open loop gain/negative feedback.
3) Just say no to overloading.
I am more concerned with low damping factor and what that does to bass control. With non-linear distortions and what that does to the timbre (harmonics), but... that's me, no problem.
R37 is local feedback, the curent in the source is deducted from the other side (that has a "natural" inverted phase) diminishing the output. That looks like (local) negative feedback to me. And seems to me that R52 and R53 should have reversed the values.
Don't take it the wrong way, you have asked for comments 🙂
Last edited:
1) Remember my design requirements, one of them are differential inputs, which I consider important.
2) As soon as you apply global feedback you need to control gain/phase, usually by slowing down the amplifier.... And there are real worlds load which are pretty demanding....
3) In real world that might not be an option 🙂
I should said that I also have an eye to pro applications, where those things above do matter....
R52 and R53 is correct, R52 is sourced an additional nominal 800 uA from the DC servo, giving 8.8 mA though R52.
Comments are very welcome, but it IS a non feedback design, as I don't see any need for global feedback, the pure THD is below what you can hear, and without feedback to get a lot of more important positive effects.... Although it would be easy to make it into a more regular amplifier with high open loops gain and global feedback, but why, there are plenty of those around, I want to make one that sounds great....
See also Pass's paper: http://www.passdiy.com/pdf/articles/distortion_feedback.pdf
Low damping factor is overrated, people forgot the impedance in the cables, bass control don't have anything to do with the damping factor, which owners of tube amps will attest to....
Anyway, the 110W version will end up a little lower, probably at 0.3 ohm as it have two sets of output transistors, the 190W version even lower.
Soren
2) As soon as you apply global feedback you need to control gain/phase, usually by slowing down the amplifier.... And there are real worlds load which are pretty demanding....
3) In real world that might not be an option 🙂
I should said that I also have an eye to pro applications, where those things above do matter....
R52 and R53 is correct, R52 is sourced an additional nominal 800 uA from the DC servo, giving 8.8 mA though R52.
Comments are very welcome, but it IS a non feedback design, as I don't see any need for global feedback, the pure THD is below what you can hear, and without feedback to get a lot of more important positive effects.... Although it would be easy to make it into a more regular amplifier with high open loops gain and global feedback, but why, there are plenty of those around, I want to make one that sounds great....
See also Pass's paper: http://www.passdiy.com/pdf/articles/distortion_feedback.pdf
Low damping factor is overrated, people forgot the impedance in the cables, bass control don't have anything to do with the damping factor, which owners of tube amps will attest to....
Anyway, the 110W version will end up a little lower, probably at 0.3 ohm as it have two sets of output transistors, the 190W version even lower.
Soren
I know how to achieve these three things with 40dB-60dB of global feedback. If you don't, you may be missing something 😉
Anyway, your approach to avoid global feedback instead of learning to use it properly is nice. Lazyness is very powerful.
Eva and soNic_real_one, on this occasion I think you misunderstand the point of the project from soekris...
There are many, many amps on this forum with global negative feedback. (You might design them yourself, and you might be happy with them, and that is good for you). However the tube guys have found that 'open loop' amps can sometimes perform very well, and this project from soekris is a way to explore 'open loop' in solid state, with the benefit of reasonably high output power.
I think it should be encouraged!
Last edited:
But how many achieve the three goals mentioned a few posts earlier in a documented way? Even teenagers and children can build an amplifier based on any global feedback schematic and post it and tell how good it sounds.
I don't care about what tube guys find out, they rarely use scientific methods so their findings are just not repeatable. Year 2010 state of the art electronics have absolutely nothing to do with them.
On the other hand, how many of these tube guys can make a tube amplifier with true voltage source behaviour (less than 0.1 ohm output impedance across the whole audio band and less than 0.1dB tonal unbalance, which would usually require strong of global feedback?)
If you can't get the result that you like, you can always persuade yourself and others into thinking that whatever else result you got is very good 😉
Last edited:
Nice one 😉 That is the first time i have ever seen power mosfets used instead of an output relay! I'm planning on doing this myself so it's nice to see someone else with the same idea 🙂
There happen to be a lot of very good devices for this purpose around now, i have seen 3.5 (maximum) milliohm on 75V devices - bye bye output relay 😀
E2A:- You need to drive the gate - source to about 15V, i don't see any PSU to deliver that or any voltage to the Mosfets? Only an opto-isolator which will need a supply to drive the Mosfets.. Unless i'm mistaken of course.
There happen to be a lot of very good devices for this purpose around now, i have seen 3.5 (maximum) milliohm on 75V devices - bye bye output relay 😀
E2A:- You need to drive the gate - source to about 15V, i don't see any PSU to deliver that or any voltage to the Mosfets? Only an opto-isolator which will need a supply to drive the Mosfets.. Unless i'm mistaken of course.
Last edited:
APV1121 is a photovoltaic gate driver. It transfers energy from the LED to the photodiode, a few microamperes are available to turn on the gate without any power supply, and turn off is active and powered with on-state gate charge. It can turn on and off the "relay" mosfet in a few miliseconds, like a relay.
http://pewa.panasonic.com/pcsd/product/pmos/pdf/mosfet.pdf
Get used to new funny parts. We are no longer in 1980s 😉
http://pewa.panasonic.com/pcsd/product/pmos/pdf/mosfet.pdf
Get used to new funny parts. We are no longer in 1980s 😉
Precise link here to the Panasonic APV1121:
http://panasonic-denko.co.jp/ac/e_download/control/relay/photomos/catalog/semi_eng_apv.pdf
It can only deliver few uA's, so to ensure quick turn-on (mosfets have significant gate capacitance) I drive it with a short peak of 28 mA during turn-on, and then steady 4 mA constant current. And as Eva says, it have built-in circuit for quick turn-off so nothing special need to be done there.
Soren
http://panasonic-denko.co.jp/ac/e_download/control/relay/photomos/catalog/semi_eng_apv.pdf
It can only deliver few uA's, so to ensure quick turn-on (mosfets have significant gate capacitance) I drive it with a short peak of 28 mA during turn-on, and then steady 4 mA constant current. And as Eva says, it have built-in circuit for quick turn-off so nothing special need to be done there.
Soren
This is true. Hi-end audio is not about accurate reproduction, Eva, you should know that by now.
Freq response ripples in the amp, bumps in the speaker response, off-axis dips, >2% distortion, gross output impedance effects, hey, if you like the sound, go for it!
Don't bother us with accuracy and hifidelity. 😉
jd
I recently heard a system powered by an Ayre amplifier. The overall performance of the system was very good and very entertaining. I understand that the Ayre amplifier runs 'open loop'. I genuinely believe that open loop amplifier performance should be investigated further. Damn, I can feel another project coming... better check my bank account...
I recently heard a system powered by a Sony amplifier. The overall performance of the system was very good and very entertaining. I understand that the Sony amplifier runs 'closed loop'. Good that closed loop amplifier performance as been so extensively investigated. Saves my bank account except for a few good books...😉
jd
There s people who need a high THD ratio, so they
can have a few more trebles , and yet, claim that their
system is "true hifi" since it has no tone corrector...
As pointed by Janneman, high end audio market is no more
about hifi but about hype and fashion..
can have a few more trebles , and yet, claim that their
system is "true hifi" since it has no tone corrector...
As pointed by Janneman, high end audio market is no more
about hifi but about hype and fashion..
Yes, I can clearly hear an amplifier with numbers substantially worse than these, say THD not well below 1% and output impedance not well below 1 ohm, and in my experience most people can, it's easy. The result is quite random, may be pleasing or ugly depending on speaker impedance and freq. response.
- Status
- Not open for further replies.