Many years ago I bought a new Rotel RA820A for a second system, and it was a great little amp. Well it's been bagged up and unused for many years so thought it about time to see what if any mods could be done to this old classic. I couldn't even come up with the correct circuit but found an RA820B was near enough even if that did have multiple outputs and single opamps. It's a classic design, very simple, beautifully built, the PCB layout is a model of good design, and it's very easy to work on.
How good is this amp ?
Well my points of reference have moved on over the years and I have come to favour a very "distinct musical sound". Current reference for those that haven't seen it is still the affectionately named "Mooly Amp" which still sees off all comers from a subjective point of view.
http://www.diyaudio.com/forums/solid-state/119151-my-mosfet-amplifier-designed-music.html
So scoring the reference as 10 I would say realistically the RA820 is a 6.5 It's a bit like swapping your "reference" CD player for a £60 model. It's that subtle (not so subtle even) difference that means everything. It lacks the ability to portray width and depth to the sound stage and can "shout" and sound a little hard. I would call it very two dimensional, almost like a cardboard cut out. The speakers used were B&W 703's which are a pretty demanding load with a 3 ohm impedance minimum. But that has to be put in context... this is a 25 year old "budget" amp although at the time it was the amp of choice for many a system.
Music used was Nicola Benedetti's new CD "Fantasie" and the Ravel "Tzigane" track a trully ravishing piece that shouldn't sound at all harsh or have you heading to turn down the volume. The solo violin should really sing out... every nuance reproduced. If it sounds harsh or unmusical you just aren't hearing it properly.
Next up Andras Schiff and some Mozart sonatas recorded in the 1980's on the "London" label. A very "warm" sound.
And finally Alfred Brendel and Bernard Haitink with the Beethoven "Emperor" concerto. Brendels piano has a wonderful "pearly" tone that seems to really shimmer, and again there should be no harshness.
So on to the amp. It seemed an intriguing possibility to replace the output transistors with HEXFETS... and as I had some... so a pair of IRFP240/9240 were fitted to each channel. Direct replacement and no problem with pin outs.
The print that went to the base (now the gate) was cut and 47 ohm stopper resistors fitted.
The driver stage should have a bit less work to do now so increasing the value of R625/6/7 and 8 from 220 ohms to 680 ohm seemed reasonable.
These are the ones that are the emmiter resistors for the drivers.
The HEXFET's need more bias voltage than bjt's, around 4 volts plus per device so that meant altering the vbe multiplyer. R619/620 was reduced from 2k2 to 330 ohms. Bias was set at around 100ma this falling slightly after it had warmed, and was stable enough that I felt there was little point in
altering the vbe generator to more accurately match the HEXFETS. The important thing is that it overcorrects, and under stress testing with the heatsink very hot, reduces the bias to around half that value so thermal runaway is avoided. It's one area that could usefully be improved though.
There is no output inductor or zobel network on this amp. There is however a 0.22 ohms series resistor to the speaker outputs. Curiously the NFB take off point is at the speaker end ? ... well maybe not so strange, but using the self inductance of this resistor to isolate the HEXFETS from capacitive loading seemed a good idea and this was also something I did on the "Mooly" Mosfet amp. It was a technique that I picked up on that was used by the late John Linsley Hood in several of his designs too. Easily accomplished here, just a cut in the print and a very short link to the other end of the resistor.
The DC offset of this amp is appalling, around 80mv so while that was acceptable 25yrs ago it isn't now. The R820A has unbalanced resistor values at the differential amp input (Why ?) so a good fiddle would have been to tweak the input resistor to bring the offset down. I tried and it worked a treat but with one problem. The Rotel has a passive tone control and if this is in circuit then the response of that stage is critically dependant on the input impedance it works into. So that idea wouldn't do. I came up with adding a fixed 3meg resistor from the regulated (now 15 volt) supply to the "inverting" side of the long tailed pair... well actualy the top end of the cap in the feedback arm... same ish difference.
I also swapped the input transistors for some higher gain matched 2SA872's
That brought the offset down to around 4 mv worst case.
Oh yes the 15 volt rails. Well they were 22 (I think) but I swapped zeners D901/2 to 15 volt so I could use different opamps if I wished... 22volts is outside manys range.
The stock opamp/s are NE5532... arghhh so in went an OPA2604... just the one, the other is for the phono stage and I don't possess any vinyl. I also fited 6pf caps across the feedback resistors of the opamp as there was considerable overshoot on squarewave testing at low volume. I also fitted 43 ohm series resistors with the opamp outputs (easy as there are wire links that can be removed)
What else, oh yes, I swapped all the electroylitics for decent commercial grade parts (low esr types) and used quite a few Panasonic "ECA" series I think from memory.
SO WHATS IT LIKE...
Well certainly different to how it was before.
At low to medium volume it now has real ability, the sound stage does have a lot more depth, in particular the detail seemed better, however on turning the volume up (loud) the sound stage seemed to harden a little and close in. It also seemed to better capture that wonderful piano tone of Brendels, something which totally eluded it before.
If I had to put numbers in I would say it was nearer to an 8 or 8.5 now, a very significant improvement and you had to go to pretty high levels before the sound stage hardened up.
It was an interesting "exercise" non the less although ultimately the reference MOSFET amp has no worries, it's still my "ultimate" reference for musicality.
And more than that... I couldn't live now without at least a remote volume contol
Now there's an idea, I'm struggling trying to learn PIC programming with a PICKIT2 (how hard can it be to get an LED to light up) lol so the Rotel might be a useful experiment.
If you have made it to the end here "thanks for reading"
How good is this amp ?
Well my points of reference have moved on over the years and I have come to favour a very "distinct musical sound". Current reference for those that haven't seen it is still the affectionately named "Mooly Amp" which still sees off all comers from a subjective point of view.
http://www.diyaudio.com/forums/solid-state/119151-my-mosfet-amplifier-designed-music.html
So scoring the reference as 10 I would say realistically the RA820 is a 6.5 It's a bit like swapping your "reference" CD player for a £60 model. It's that subtle (not so subtle even) difference that means everything. It lacks the ability to portray width and depth to the sound stage and can "shout" and sound a little hard. I would call it very two dimensional, almost like a cardboard cut out. The speakers used were B&W 703's which are a pretty demanding load with a 3 ohm impedance minimum. But that has to be put in context... this is a 25 year old "budget" amp although at the time it was the amp of choice for many a system.
Music used was Nicola Benedetti's new CD "Fantasie" and the Ravel "Tzigane" track a trully ravishing piece that shouldn't sound at all harsh or have you heading to turn down the volume. The solo violin should really sing out... every nuance reproduced. If it sounds harsh or unmusical you just aren't hearing it properly.
Next up Andras Schiff and some Mozart sonatas recorded in the 1980's on the "London" label. A very "warm" sound.
And finally Alfred Brendel and Bernard Haitink with the Beethoven "Emperor" concerto. Brendels piano has a wonderful "pearly" tone that seems to really shimmer, and again there should be no harshness.
So on to the amp. It seemed an intriguing possibility to replace the output transistors with HEXFETS... and as I had some... so a pair of IRFP240/9240 were fitted to each channel. Direct replacement and no problem with pin outs.
The print that went to the base (now the gate) was cut and 47 ohm stopper resistors fitted.
The driver stage should have a bit less work to do now so increasing the value of R625/6/7 and 8 from 220 ohms to 680 ohm seemed reasonable.
These are the ones that are the emmiter resistors for the drivers.
The HEXFET's need more bias voltage than bjt's, around 4 volts plus per device so that meant altering the vbe multiplyer. R619/620 was reduced from 2k2 to 330 ohms. Bias was set at around 100ma this falling slightly after it had warmed, and was stable enough that I felt there was little point in
altering the vbe generator to more accurately match the HEXFETS. The important thing is that it overcorrects, and under stress testing with the heatsink very hot, reduces the bias to around half that value so thermal runaway is avoided. It's one area that could usefully be improved though.
There is no output inductor or zobel network on this amp. There is however a 0.22 ohms series resistor to the speaker outputs. Curiously the NFB take off point is at the speaker end ? ... well maybe not so strange, but using the self inductance of this resistor to isolate the HEXFETS from capacitive loading seemed a good idea and this was also something I did on the "Mooly" Mosfet amp. It was a technique that I picked up on that was used by the late John Linsley Hood in several of his designs too. Easily accomplished here, just a cut in the print and a very short link to the other end of the resistor.
The DC offset of this amp is appalling, around 80mv so while that was acceptable 25yrs ago it isn't now. The R820A has unbalanced resistor values at the differential amp input (Why ?) so a good fiddle would have been to tweak the input resistor to bring the offset down. I tried and it worked a treat but with one problem. The Rotel has a passive tone control and if this is in circuit then the response of that stage is critically dependant on the input impedance it works into. So that idea wouldn't do. I came up with adding a fixed 3meg resistor from the regulated (now 15 volt) supply to the "inverting" side of the long tailed pair... well actualy the top end of the cap in the feedback arm... same ish difference.
I also swapped the input transistors for some higher gain matched 2SA872's
That brought the offset down to around 4 mv worst case.
Oh yes the 15 volt rails. Well they were 22 (I think) but I swapped zeners D901/2 to 15 volt so I could use different opamps if I wished... 22volts is outside manys range.
The stock opamp/s are NE5532... arghhh so in went an OPA2604... just the one, the other is for the phono stage and I don't possess any vinyl. I also fited 6pf caps across the feedback resistors of the opamp as there was considerable overshoot on squarewave testing at low volume. I also fitted 43 ohm series resistors with the opamp outputs (easy as there are wire links that can be removed)
What else, oh yes, I swapped all the electroylitics for decent commercial grade parts (low esr types) and used quite a few Panasonic "ECA" series I think from memory.
SO WHATS IT LIKE...
Well certainly different to how it was before.
At low to medium volume it now has real ability, the sound stage does have a lot more depth, in particular the detail seemed better, however on turning the volume up (loud) the sound stage seemed to harden a little and close in. It also seemed to better capture that wonderful piano tone of Brendels, something which totally eluded it before.
If I had to put numbers in I would say it was nearer to an 8 or 8.5 now, a very significant improvement and you had to go to pretty high levels before the sound stage hardened up.
It was an interesting "exercise" non the less although ultimately the reference MOSFET amp has no worries, it's still my "ultimate" reference for musicality.
And more than that... I couldn't live now without at least a remote volume contol
Now there's an idea, I'm struggling trying to learn PIC programming with a PICKIT2 (how hard can it be to get an LED to light up) lol so the Rotel might be a useful experiment. If you have made it to the end here "thanks for reading"
Last edited:
Very interesting project for me.
Please excuse my following questions:
1) why a pair of hexfets and not a pair of better and more powerful bjt ? some 150W complementary pair from let's say toshiba (or even sanken)?
2) why not to try to feed the differential pair directly from the volume pot (with some values mods) instead of passing through a op-amp ?
3) have you thought about a power supply update ? bigger caps, more VA transformer ?
I am very interested in upgrade a little but sane integrated and to use it with a pair of high efficiency speakers I have in mind
I prefer the low power high current amp with high eff speakers approach.
Thanks a lot and kind regards,
gino
Please excuse my following questions:
1) why a pair of hexfets and not a pair of better and more powerful bjt ? some 150W complementary pair from let's say toshiba (or even sanken)?
2) why not to try to feed the differential pair directly from the volume pot (with some values mods) instead of passing through a op-amp ?
3) have you thought about a power supply update ? bigger caps, more VA transformer ?
I am very interested in upgrade a little but sane integrated and to use it with a pair of high efficiency speakers I have in mind
I prefer the low power high current amp with high eff speakers approach.
Thanks a lot and kind regards,
gino
Hi,
To answer your questions
Why HEXFETS... because I think the "ease of drive" with FET's is a big plus. It gives the driver a much easier time generally. To just replace the outputs with another bjt wouldn't have achieved as much... they still take a lot of driving. The FETS also don't suffer from gain droop (transconductance with FET's) as do bjt's. Now if I had a spare pair of laterals lying around...
Why not feed the volume pot directly into the amp ? Because I essentially wanted to keep the amp in some sort of original condition... and this has all been done as much as anything as a bit of fun... and it's certainly made a huge difference to the sonics.
The OPA2604 is an Opamp whose distortion characteristics are very desirable for audio... the even harmonics predominate... what I thought of trying, and it's easily implemented is changing the configuration to inverting. Although the absolute phase would be reversed it's only a a matter of swapping speaker leads around.
Reverting to the original NE5532 is interesting... it lacks "musicality". Statements like that may make some cringe but it's true
PSU. The original caps are 6800mfd and IMO up to the task for an amp of this output. Same goes for the transformer, it's a good quality part and is silent in operation.
Do you have an 820 ? Remember the circuit shown is for the 820B not the A
Well worth doing anyway... you might be surprised
To answer your questions
Why HEXFETS... because I think the "ease of drive" with FET's is a big plus. It gives the driver a much easier time generally. To just replace the outputs with another bjt wouldn't have achieved as much... they still take a lot of driving. The FETS also don't suffer from gain droop (transconductance with FET's) as do bjt's. Now if I had a spare pair of laterals lying around...
Why not feed the volume pot directly into the amp ? Because I essentially wanted to keep the amp in some sort of original condition... and this has all been done as much as anything as a bit of fun... and it's certainly made a huge difference to the sonics.
The OPA2604 is an Opamp whose distortion characteristics are very desirable for audio... the even harmonics predominate... what I thought of trying, and it's easily implemented is changing the configuration to inverting. Although the absolute phase would be reversed it's only a a matter of swapping speaker leads around.
Reverting to the original NE5532 is interesting... it lacks "musicality". Statements like that may make some cringe but it's true
PSU. The original caps are 6800mfd and IMO up to the task for an amp of this output. Same goes for the transformer, it's a good quality part and is silent in operation.
Do you have an 820 ? Remember the circuit shown is for the 820B not the A
Well worth doing anyway... you might be surprised
Thank you so much for your very kind and valuable reply !
To answer your question, no I do not have an 820
Nevertheless I have destroyed a RA-930AX (whose pieces are still on my table) in an attempt to lower the power stage gain.
My idea was to drive directly the power stage with a tube headphone amp that I have and whose sound I quite like, but the gain of the latter is about 10 times.
So my idea was to lower tje gain of the power stage to use it like a power buffer.
I replace the feedback resistors to get a gain of about 4, the bias changed abruptly and all went in smoke !
That stupid it's me ...
A very superficial approach I understand.
But all started from the idea to use a high gain headphone amp as a line stage
Is easy to find a good sounding HP amp ... even a tube based one.
I will follow your 3D with the greatest interest.
Thank you very much indeed.
Kind regards,
gino
To answer your question, no I do not have an 820
Nevertheless I have destroyed a RA-930AX (whose pieces are still on my table) in an attempt to lower the power stage gain.
My idea was to drive directly the power stage with a tube headphone amp that I have and whose sound I quite like, but the gain of the latter is about 10 times.
So my idea was to lower tje gain of the power stage to use it like a power buffer.
I replace the feedback resistors to get a gain of about 4, the bias changed abruptly and all went in smoke !
That stupid it's me ...
A very superficial approach I understand.
But all started from the idea to use a high gain headphone amp as a line stage
Is easy to find a good sounding HP amp ... even a tube based one.
I will follow your 3D with the greatest interest.
Thank you very much indeed.
Kind regards,
gino
Thank you very much for your valuable advice
It will not happen again
I hope to be able to restore the unit by replacing all the broken bjts in the near future
I could use it as a small power amp with another line stage I have
I think that Rotel is a brand that offers a remarkable quality/price ratio
Anyway I am interested in this RA820B version as well.
I think it has a great potential of good sound
Kind regards,
gino
Last edited:
Managed to come up with a circuit although this was for the RA810, but it seems essentially the same.
That passive preamp... the effect of the controls is minimal with the design as it stands. Even saw one on EBay that was described as perfect but for the tone controls "that don't seem to work" They do... they are just very subtle.
That passive preamp... the effect of the controls is minimal with the design as it stands. Even saw one on EBay that was described as perfect but for the tone controls "that don't seem to work"
They do... they are just very subtle.Attachments
Last edited:
Further update on this as I wanted to see how far it could be pushed sonically...
I tried (very easily done) adding a buffer transistor from the first stage to the VAS, however the results were dissapointing sonically, it just did not sound as good. Surprising that !
I also tested it with pure capacitive loads and found that there was a tendency to instability so have added output inductors now. It will now happily drive a 0.1mfd or a 2.2mfd cap up to the limits of the output stage on squarewave testing.
I also looked at the feedback, you'll notice the gain of this power amp is low (8.2k and 470 ohm) giving a voltage of around 18.5 I played around with this but it highlighted another issue, increase the gain and hum becomes intrusive, which to be honest was a little unexpected. Further investigation showed it to be coming from ripple on the negative rail and being injected into the voltage amplifier stages.
It would seem then that the amp as it stands/stood is a pretty perfect "quality match" and I can see why it is designed the way it is. It has been carefully designed as a "whole".
For interest and at the expense of a little loss of negative voltage excursion I have added a shunt resistor/zener (30 volts) supply to the driver and VAS/input stage with 2200mfd decoupling. This has dramatically reduced the level of injected hum
The HEXFETS's place little demand on the drivers so this shunt arrangement was an option. It still delivers around 25 watts rms plus into 8 ohms, and thanks to the FET ouputs now has no issues with low impedance loads.
That's probably as far as I will take this for now... it's way better than in standard form and to be honest a cracking little amp in all respects.
I tried (very easily done) adding a buffer transistor from the first stage to the VAS, however the results were dissapointing sonically, it just did not sound as good. Surprising that !
I also tested it with pure capacitive loads and found that there was a tendency to instability so have added output inductors now. It will now happily drive a 0.1mfd or a 2.2mfd cap up to the limits of the output stage on squarewave testing.
I also looked at the feedback, you'll notice the gain of this power amp is low (8.2k and 470 ohm) giving a voltage of around 18.5 I played around with this but it highlighted another issue, increase the gain and hum becomes intrusive, which to be honest was a little unexpected. Further investigation showed it to be coming from ripple on the negative rail and being injected into the voltage amplifier stages.
It would seem then that the amp as it stands/stood is a pretty perfect "quality match" and I can see why it is designed the way it is. It has been carefully designed as a "whole".
For interest and at the expense of a little loss of negative voltage excursion I have added a shunt resistor/zener (30 volts) supply to the driver and VAS/input stage with 2200mfd decoupling. This has dramatically reduced the level of injected hum
The HEXFETS's place little demand on the drivers so this shunt arrangement was an option. It still delivers around 25 watts rms plus into 8 ohms, and thanks to the FET ouputs now has no issues with low impedance loads.
That's probably as far as I will take this for now... it's way better than in standard form and to be honest a cracking little amp in all respects.
Last edited:
Some real world pictures of it all.
Scope shows 20 volts pk pk into 5 ohms, then into 2.2 mfd, and finally the assymetric clipping which as the amp is loaded equals out as the positive rail falls toward the 30 volt stabilised one.
Scope shows 20 volts pk pk into 5 ohms, then into 2.2 mfd, and finally the assymetric clipping which as the amp is loaded equals out as the positive rail falls toward the 30 volt stabilised one.
Attachments
Last edited:
Hi Mooly,
I liked the position of compensation capacitor there, which you could also try here, then omit C607/C609 (introducing noise). I would insert emitter resistors for Q601/Q603.
I`m always very interested in observations based on listening instead of reading the usual nonsense provided LTSpice.
It must be equally disappointing in your lateral amp as well. I suggested the folded cascode topology as an upgrade for that amp in the other thread.
I liked the position of compensation capacitor there, which you could also try here, then omit C607/C609 (introducing noise). I would insert emitter resistors for Q601/Q603.
You regrettably rejected the idea of separate power supply, allowing lower rail voltages and higher quiescent current in the output section.
I`m always very interested in observations based on listening instead of reading the usual nonsense provided LTSpice.
Hello Lumba, how are you ?
Yes, any amp (or any other equipment) stands or falls by it's sonics doesn't it
A little error has crept in to my description of the above scope traces... I wondered why I was drawing such an arc with capacitive testing it was a 4.7mfd metalised polyester cap, not 2.2 mfd I used... in error actually.
Also the 680pf cap C603 at the input is shown incorrectly in the schematic and is actually a 100pf (as fitted by Rotel).
Your ideas on C607/9 are interesting... it's very heavily "slugged" for want of a better word but performs surprisingly well all things considered.
Truth be told I never auditioned the lateral without a buffer on the VAS, you think I should ? ... to late now, but that amp as you know is a totally different topology, with it's single ended input stage etc. And it runs with much reduced NFB compared to this one.
Quite why it is as good as it is... and make no mistake it is... well it just all seemed to come together perfectly... and to be honest it performs so well sonically I seriously wonder where to go from there with any further design. We will see
I much prefer the hands on approach... and have never got to grips with simulation at a serious level.
Yes, any amp (or any other equipment) stands or falls by it's sonics doesn't it
A little error has crept in to my description of the above scope traces... I wondered why I was drawing such an arc with capacitive testing
it was a 4.7mfd metalised polyester cap, not 2.2 mfd I used... in error actually.Also the 680pf cap C603 at the input is shown incorrectly in the schematic and is actually a 100pf (as fitted by Rotel).
Your ideas on C607/9 are interesting... it's very heavily "slugged" for want of a better word but performs surprisingly well all things considered.
Truth be told I never auditioned the lateral without a buffer on the VAS, you think I should ? ... to late now, but that amp as you know is a totally different topology, with it's single ended input stage etc. And it runs with much reduced NFB compared to this one.
Quite why it is as good as it is... and make no mistake it is... well it just all seemed to come together perfectly... and to be honest it performs so well sonically I seriously wonder where to go from there with any further design. We will see
I much prefer the hands on approach... and have never got to grips with simulation at a serious level.
I know it's resurrecting an old thread, but a couple of suggestions based on some experience with similar mods...
1) Add zeners across G-S of the MOSFET (with series diodes to prevent it from clamping reverse voltages). Without the zeners a short on the output fries your MOSFETs by destroying the gate insulation when the drivers try to drive full rail voltage into the gate. Looking at the circuit one would expect this can't happen due to power supply droop, but it's enough that it lasts a very short time, such as needed to discharge the filter caps.
2) Try input pair degeneration. Also, it helps to replace the two 1N914 diodes in the input pair tail CCS with LEDs, while adjusting the current set resistor to get the right current. If you decide to increase the tail current, remember to adjust the collector resistor of the input pair for proper offset. A fellow DIY-er here has done several mods on Rotel amps along these basic guidelines, also following observed improvements with measurement. The degenerated versions offer one very important positive aspect to the distortion profile - the harmonic spread remains very linear with frequency and power, showing better open loop behavior. Under such circumstances the feedback factor can be lowered/optimized.
Lumba Ogir also suggested a separate power supply for the small signal stages of the power amp. This may not be easy to do but it is actually a plus, with regards to clipping. The extremely high peak to mean power ratio of musical signals means that almost any amplifier goes into clipping at normal listening levels at some point (unless you have truly efficient speakers and a truly enormous amp), so clipping behavior is one of the more important aspects of amp design, especially for lower powered amplifiers.
Although it is often customary to provide higher voltage small signal rails in order to get the last possible watt out of an amp, in this case I would recommend the opposite: somewhat lower rail voltage. In fact, it should be tailored so that the input stage clips before the output stage, thus offering clearly defined slipping levels and symmetric clipping. There is, however, a downside to this, especially with vertical MOSFETs - as the G-D voltage difference approaches some 10V and less, the effects of nonlinear capacitance become quite pronounced in VMOS. Setting a lower front end voltage actually makes for defined clipping at the expense of reduced headroom for nonlinear capacitance effects. The drivers do make this better, but the underlying problem still persists. You could use a IRFP9140 instead of 9240 to provide more symmetric input and reverse capacitances, or use something like Toshiba audio MOSFETs (2SJ200/201 and complementary). Laterals would offer somewhat better performance in this regard, but their low transconductance requires a higher G-S drive voltage so you end up with even less power and headroom.
Finally, compensation may well be different with MOSFETs on the output. The capacitor loading the VAS is quite large... but at least it's not Miller compensated!
1) Add zeners across G-S of the MOSFET (with series diodes to prevent it from clamping reverse voltages). Without the zeners a short on the output fries your MOSFETs by destroying the gate insulation when the drivers try to drive full rail voltage into the gate. Looking at the circuit one would expect this can't happen due to power supply droop, but it's enough that it lasts a very short time, such as needed to discharge the filter caps.
2) Try input pair degeneration. Also, it helps to replace the two 1N914 diodes in the input pair tail CCS with LEDs, while adjusting the current set resistor to get the right current. If you decide to increase the tail current, remember to adjust the collector resistor of the input pair for proper offset. A fellow DIY-er here has done several mods on Rotel amps along these basic guidelines, also following observed improvements with measurement. The degenerated versions offer one very important positive aspect to the distortion profile - the harmonic spread remains very linear with frequency and power, showing better open loop behavior. Under such circumstances the feedback factor can be lowered/optimized.
Lumba Ogir also suggested a separate power supply for the small signal stages of the power amp. This may not be easy to do but it is actually a plus, with regards to clipping. The extremely high peak to mean power ratio of musical signals means that almost any amplifier goes into clipping at normal listening levels at some point (unless you have truly efficient speakers and a truly enormous amp), so clipping behavior is one of the more important aspects of amp design, especially for lower powered amplifiers.
Although it is often customary to provide higher voltage small signal rails in order to get the last possible watt out of an amp, in this case I would recommend the opposite: somewhat lower rail voltage. In fact, it should be tailored so that the input stage clips before the output stage, thus offering clearly defined slipping levels and symmetric clipping. There is, however, a downside to this, especially with vertical MOSFETs - as the G-D voltage difference approaches some 10V and less, the effects of nonlinear capacitance become quite pronounced in VMOS. Setting a lower front end voltage actually makes for defined clipping at the expense of reduced headroom for nonlinear capacitance effects. The drivers do make this better, but the underlying problem still persists. You could use a IRFP9140 instead of 9240 to provide more symmetric input and reverse capacitances, or use something like Toshiba audio MOSFETs (2SJ200/201 and complementary). Laterals would offer somewhat better performance in this regard, but their low transconductance requires a higher G-S drive voltage so you end up with even less power and headroom.
Finally, compensation may well be different with MOSFETs on the output. The capacitor loading the VAS is quite large... but at least it's not Miller compensated!
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.