Dears,
Since I joined this forum few months ago, I can see there is big interest in unconventional amplifiers, so I post this work of mine that was cooked and baked almost quarter of century ago, late 90’s and early 2000’s. So far I did not see someone published such schematics, so please criticize, comment and say what you think. I’m not “Great Amp builder” , In my life I designed and built only might be 6 power amps. All of them were my circuits and worked well from the beginning, most of them I still have. Also , I never used SPICE simulation, just paper calculations . Ill post circuit diagram that’s hand written.
History of this is that in late 90’s I got read a lot of Nelson Pass DIY publishing, first I built my version of Balanced Zen preamp (similar to original but with added current sources and followed by 24pole – 4 gang switch balanced attenuator) ) and then built fully differential phono MC preamp. That was permanent hook on fully differential topology and Nelson’s ideas!
Next was to finish the system with fully balanced amp. I went through Son of Zen pdf over and over again, however I was so greedy I wanted also power, ability to drive difficult speakers and not complete room heater. So I cooked something. The Zen version 6 was not published yet at the time I made complete circuit and build, you will see some similarities in the feedback arrangement. What was published was Nelsons X amplifiers, where I read reviews and measurements and got mouthwatering , diagrams of those were not obtainable and I still did not see it today.
This amp is my cooking for bad or good, nothing to do with Mr Nelson, except he gave great inspiration to think differently, so thanks very much for that!
Good things about this are:
Since I joined this forum few months ago, I can see there is big interest in unconventional amplifiers, so I post this work of mine that was cooked and baked almost quarter of century ago, late 90’s and early 2000’s. So far I did not see someone published such schematics, so please criticize, comment and say what you think. I’m not “Great Amp builder” , In my life I designed and built only might be 6 power amps. All of them were my circuits and worked well from the beginning, most of them I still have. Also , I never used SPICE simulation, just paper calculations . Ill post circuit diagram that’s hand written.
History of this is that in late 90’s I got read a lot of Nelson Pass DIY publishing, first I built my version of Balanced Zen preamp (similar to original but with added current sources and followed by 24pole – 4 gang switch balanced attenuator) ) and then built fully differential phono MC preamp. That was permanent hook on fully differential topology and Nelson’s ideas!
Next was to finish the system with fully balanced amp. I went through Son of Zen pdf over and over again, however I was so greedy I wanted also power, ability to drive difficult speakers and not complete room heater. So I cooked something. The Zen version 6 was not published yet at the time I made complete circuit and build, you will see some similarities in the feedback arrangement. What was published was Nelsons X amplifiers, where I read reviews and measurements and got mouthwatering , diagrams of those were not obtainable and I still did not see it today.
This amp is my cooking for bad or good, nothing to do with Mr Nelson, except he gave great inspiration to think differently, so thanks very much for that!
Good things about this are:
- Its absolutely stable without compensation caps
- 20+ years of use and I think only twice I opened top lid to change driver PS supply fuses, they blow after long time no use (both time when I was moving the house and amp was sitting unplugged for months)
- In those years many people auditioned this amp connected to Sonus Faber Signums, this passed looong audition test. The crowd include many audiophiles (some with levitating turntables and beer size vacuum tube SET amps) , professional musicians and sound editors, if not saying Whoooah (which many did), everybody said this sounds just all right.
This is roughly the circuit diagram of the main amp without power supply and power decoupling capacitors. Sorry this is just hand drawing. I tried to use TinaTi but it was too much challenge for my nerves today. I will describe what is done and why.
If anyone is interested this is main PCB layout, I'm mechanical engineer and always used Autocad, also for PCB drawings.
If any of skilled members would like to move my mess of hand sketches to SPICE and help with improving, I would highly appreciate!
If anyone is interested this is main PCB layout, I'm mechanical engineer and always used Autocad, also for PCB drawings.
If any of skilled members would like to move my mess of hand sketches to SPICE and help with improving, I would highly appreciate!
The whole thing is an differential amp, made of +T1 and -T1, followed by push pull mosfet pairs (5 of them per + and - output, total 40 pcs of IRFP240 and IRFP9240 for stereo).
Very little feedback (in inverting amp mode) is arranged with R3 and R2, one being 360kohm, and other 3k3, that would give gain of 109 V/V if amp would be indefinite gain.
However, I reduced a gain with R1 to match 19 V/V overall gain which is used . In years when this was conceived and built, many discussions were going on about benefits of NO Feedback at all. I knew that some will be necessary with push-pull output, so minimum is applied.
Below differential pair +T1 and -T1 are current sources 2x T2, they are driven by DC servo opamps to ensure no DC at speaker terminals and close to ground output DC
On top of differential pair are + and - T3, which form a bias network for output mosfets. Here I wanted absolute thermal stability so special PTC resistors are used. One is mounted on one N, and other on one P output mosfets. Their characteristic is that the resistance goes up at 60°C. Basically its thermo regulated heater.
When amplifier is cold, bias current of outputs is 1.86 Amp. When full warmed up it is 1.12 Amp.
When music is listened on low level, A class is going up to approx 6 watts in 8 ohm, then it goes to AB class .
Of course, when party is on, this is AB class amp.
Now I opened this amp to see what I exactly did then, and at first it was not clear for me why I did bias as it is, but I remember: first I adjusted minimum AB class bias with 2k pot without PTC resistors in place to achieve AB class regardless of heatsink temperature.
Than I added PTC's and adjusted A class bias with 5k pot. As the heat goes up (from bias or music) PTC keep output transistors at 60°C , what amp class do we listen, dont know!!!, Small levels its class A, high levels its class AB
Very little feedback (in inverting amp mode) is arranged with R3 and R2, one being 360kohm, and other 3k3, that would give gain of 109 V/V if amp would be indefinite gain.
However, I reduced a gain with R1 to match 19 V/V overall gain which is used . In years when this was conceived and built, many discussions were going on about benefits of NO Feedback at all. I knew that some will be necessary with push-pull output, so minimum is applied.
Below differential pair +T1 and -T1 are current sources 2x T2, they are driven by DC servo opamps to ensure no DC at speaker terminals and close to ground output DC
On top of differential pair are + and - T3, which form a bias network for output mosfets. Here I wanted absolute thermal stability so special PTC resistors are used. One is mounted on one N, and other on one P output mosfets. Their characteristic is that the resistance goes up at 60°C. Basically its thermo regulated heater.
When amplifier is cold, bias current of outputs is 1.86 Amp. When full warmed up it is 1.12 Amp.
When music is listened on low level, A class is going up to approx 6 watts in 8 ohm, then it goes to AB class .
Of course, when party is on, this is AB class amp.
Now I opened this amp to see what I exactly did then, and at first it was not clear for me why I did bias as it is, but I remember: first I adjusted minimum AB class bias with 2k pot without PTC resistors in place to achieve AB class regardless of heatsink temperature.
Than I added PTC's and adjusted A class bias with 5k pot. As the heat goes up (from bias or music) PTC keep output transistors at 60°C , what amp class do we listen, dont know!!!, Small levels its class A, high levels its class AB
The measurements:
When it was built, I had only analog frequency generator up to 100kHz and good Fluke true RMS multimeter. Than I measured that frequency response was up to 100kHz minus few %, that was good enough to put it in use.
Few days ago it was taken out of setup into work bench;
Measurement set up is windows PC with REW, RME Fireface UCX interface in ASIO mode, calibrated, with one reference loop in - out, and other output going throuhg amplifier back to input. For RME input I organized 4 precision resistors to bring signal level back to unity gain, without buffering.
Amp was measured with 4 ande 8 ohm load for low medium power, for high power I used 10 ohm as this is only 100W resistor I have.
When it was built, I had only analog frequency generator up to 100kHz and good Fluke true RMS multimeter. Than I measured that frequency response was up to 100kHz minus few %, that was good enough to put it in use.
Few days ago it was taken out of setup into work bench;
Measurement set up is windows PC with REW, RME Fireface UCX interface in ASIO mode, calibrated, with one reference loop in - out, and other output going throuhg amplifier back to input. For RME input I organized 4 precision resistors to bring signal level back to unity gain, without buffering.
Amp was measured with 4 ande 8 ohm load for low medium power, for high power I used 10 ohm as this is only 100W resistor I have.
This is torture test: 41 V RMS in 10 ohms:
This is obviously last useful power before clipping, 1.12% TDH, now with dominant odd order distorsion.
As I was testing this for minute or two, I felt burning smell, quickly shut off the signal and looked, my work bench started burning under load resistor, amp was just cool as ever
This is obviously last useful power before clipping, 1.12% TDH, now with dominant odd order distorsion.
As I was testing this for minute or two, I felt burning smell, quickly shut off the signal and looked, my work bench started burning under load resistor, amp was just cool as ever
Unfortunately I don't have analog oscillator any more, only square wave I can generate is through 192Khz/24bit RME interface, this is naturally not square wave due to digital frequency limits.
To make picture little bit clear, I used digital scope measurements with both inputs: RME closed loop and signal through power amp afterwards reduced by resistors network to nominal interface input level . Yellow trace is always RME closed loop input: green is imput through power amp with different loads that are mentioned
This is 1 kHz square at 8 ohm (zoom to see difference between traces) :
This is the same thing into 4 ohms:
Than 10 kHz square into 4 ohm, this ntime at 8Volts RMS output, to see slowing down with higher frequency and power:
To make picture little bit clear, I used digital scope measurements with both inputs: RME closed loop and signal through power amp afterwards reduced by resistors network to nominal interface input level . Yellow trace is always RME closed loop input: green is imput through power amp with different loads that are mentioned
This is 1 kHz square at 8 ohm (zoom to see difference between traces) :
This is the same thing into 4 ohms:
Than 10 kHz square into 4 ohm, this ntime at 8Volts RMS output, to see slowing down with higher frequency and power:
More, I tried to measure output impedance:
) , I call this 0.25 Ohm output resistance at full load, or 32 damping factor at 8Ohms....?
- With 42 Volts RMS and no load i connected 10 Ohm resistor
- Lots of sparks while connecting alligator clip!
) , I call this 0.25 Ohm output resistance at full load, or 32 damping factor at 8Ohms....?Before I need to pick up my kid from school, few pictures from 2003 and 2004, when it was assembled:
600VA transformer with 4 windings for separate l and R channel power. On top is higher voltage transformer for drivers:
This is power c stage; 40 IRF-s for current delivery, here you can see thermo regulating bias PTC resistors mounted on top of IRF-s (green bubbles):
These are power supply filters; 100,000 MicroFarad per rail per channel (totally 400,000 microfarad) for power. And inductors followed by about 15,000 MicroFarad pre filter for drivers power supply
On top of everything, main PCB board, left and right are L and R voltage gain amplifiers, in the middle is regulated PS for those and DC servo opamps.
600VA transformer with 4 windings for separate l and R channel power. On top is higher voltage transformer for drivers:
This is power c stage; 40 IRF-s for current delivery, here you can see thermo regulating bias PTC resistors mounted on top of IRF-s (green bubbles):
These are power supply filters; 100,000 MicroFarad per rail per channel (totally 400,000 microfarad) for power. And inductors followed by about 15,000 MicroFarad pre filter for drivers power supply
On top of everything, main PCB board, left and right are L and R voltage gain amplifiers, in the middle is regulated PS for those and DC servo opamps.
All in all, its quite mediocre amp THD wise, but it's much better than many. Especially simple design makes really decent HF response.
No exotic unobtainable components are used here, nevertheless its quite expensive design: just power filtering caps are some 200 Euro, followed by 80 TO3 IRF-s needed to get semi matched 5 in parallel output transistors, only half of purchase: 40 pieces are used.
After quarter of century in use, and me being equivalently older, reviewing old diagram after the lid was finally opened; here is dusty picture of today:
I will try to update this machine;
- For sure I will open it up completely and replace main transformer. Even existing one is sealed in epoxy, this 600VA still makes some buzz. Nothing is coming out of the speakers, transformer itself makes noise, tried with mains DC blocker and filter, that's not the cause.... I have one 800VA toroid that's quiet (probably because I ordered 1kVA core double vacuum-varnished and backed. That goes in now and for that I need to disassemble everything, this buzz is not really loud and not heard with quiet music, but its matter of OCD.
When I'm already into digging, some other improvements can be made, here in forum we have many amp masters, suggestions are very welcome!
These are my ideas for improvement:
Thanks for looking at this and any comments,
Drazen
No exotic unobtainable components are used here, nevertheless its quite expensive design: just power filtering caps are some 200 Euro, followed by 80 TO3 IRF-s needed to get semi matched 5 in parallel output transistors, only half of purchase: 40 pieces are used.
After quarter of century in use, and me being equivalently older
, reviewing old diagram after the lid was finally opened; here is dusty picture of today:I will try to update this machine;
- For sure I will open it up completely and replace main transformer. Even existing one is sealed in epoxy, this 600VA still makes some buzz. Nothing is coming out of the speakers, transformer itself makes noise, tried with mains DC blocker and filter, that's not the cause.... I have one 800VA toroid that's quiet (probably because I ordered 1kVA core double vacuum-varnished and backed. That goes in now and for that I need to disassemble everything, this buzz is not really loud and not heard with quiet music, but its matter of OCD.
When I'm already into digging, some other improvements can be made, here in forum we have many amp masters, suggestions are very welcome!
These are my ideas for improvement:
- Main differential pair transistors (also voltage gain transistors) are IRF620 and they have hard job to do (very high voltage swing). I wanted to use IRF610 but at the time they were not available for purchase. Now I have matched IRF610 to use. What do you say, to replace them with IRF610 or even better with some more exotic fets? Any suggestion for transistors?
- Output Mosfets I would like to keep, don't want to get another 40 pcs of something else.
- Also Bias adjustment and Current sources T2 and T3 IRF620 are probably not not worth to change to something else, or?
- I can change also drivers transformer and get higher voltages for drivers (now it is +46V and -58V) . That would bring drivers into more linear region....
- Increasing this voltage also opens questions if might be vacuum tubes could be inserted instead of fets, 2 x 6H30P tubes per channel would probably do the job, however than I need to make new main PCB from scratch...
- Feedback resistors are of high value; this can be proportionally reduced for noise purposes, but than again, there is no noise from the speakers that can be heard, so what to hack?
- And finally the feedback, at the moment this amp is using very little of it. I can short connect R1 to get much more gain, then reduce R3 to some 40- 50 kOhm to get correct overall gain. My present DC servo network would not work any more than, but it could be fixed by bringing fixed voltage to T2 and making it constant current source, and bringing outputs of opamps to R4 (instead of referencing it to - 58V power supply) . Some resistor value calculations will be needed but that is manageable. As result this would bring more NFB and likely better THD, I'm not sure if sound would be better ?? Many people tested that THD under 0.1 % is not perceived by humans...
Thanks for looking at this and any comments,
Drazen
Looks like your amp does pretty well at low powers, but at higher powers, perhaps when it switches to class AB, distortion increases. Your power supply is working very well, the mains components are pretty low. Did you experiment with the bias settings? You'd really want to operate this in class A as long as it's thermally possible.
You got it right, that was intention. to have A class amp for first watts (and home listening) and still massive output capability with cost of higher distortion. Idea was is when we party and volume is high up, everyone will be pretty drunk by then and distortion is less relevant
.Now if I want to increase class A portion, I can change 60°C PTC resistors to lets say 80°C . More heat and shorter life of components. Or to increase cooling, that costs in size and mechanical complications...
As above, other options are to increase NFB, just then this is not low feedback amp any more... compromises, compromises...
Might be the best compromise would be to change T1 (both +and-) IRF620's to something more linear in hard job that they have?
do some TLC, leave it as is - that amp deserved in all this years to stay as it is
with accumulated knowledge and mileage, make new one .... whatever choice is
I mean whatever you want to change, except changing FE mosfets to some scarce Toshibas, would involve mechanical changes - bigger heatsinks etc. just leading to too much work - meaning - it's easier to make new house from scratch than to rework old one to new
two ways of improving performance of input LTP:
with accumulated knowledge and mileage, make new one .... whatever choice is
I mean whatever you want to change, except changing FE mosfets to some scarce Toshibas, would involve mechanical changes - bigger heatsinks etc. just leading to too much work - meaning - it's easier to make new house from scratch than to rework old one to new
two ways of improving performance of input LTP:
- introduce JFet buffers in front of mosfets, or
- change to Toshiba mosfets, or
- introduce JFet buffers in front of mosfets and change to Toshiba mosfets
a 1KHz squarewave into 8ohm ( a very easy test ) should be clean without ringing