Hi all,
I am new to this site and will first of all apologise if my shearching has not been as good as it could be in regard to previous threads,however, I cant' find a great deal of info on really good regulated supplies for power amps. What I have in mind is something of similar performance to the Jung/Diddens, Sulzer designs but this does not seem to exist !,most designs I have seen are either far to slow or have far to high output impedance.
The obvious solution would be a Sulzer type with level translator which offers maybe x3 gain after the op-amp followed by the pass transistor/darlington/mosfet but this would almost certainly oscillate unless it was compensated to the point where it was no faster than an LM317 with series pass device...
Maybe such luminaries on this site as ALW or Janneman have some pearls of wisdom on the subject.... maybe I'll have to design a high speed,high voltage discrete op-amp unless someone knows of a suitable design elsewhere
Jez.
p.s I know many feel that simple capacitor supplies are superior but I have a suspicion that a suitably quick,low impedence,stable regulator would surprise a lot of people !.
I am new to this site and will first of all apologise if my shearching has not been as good as it could be in regard to previous threads,however, I cant' find a great deal of info on really good regulated supplies for power amps. What I have in mind is something of similar performance to the Jung/Diddens, Sulzer designs but this does not seem to exist !,most designs I have seen are either far to slow or have far to high output impedance.
The obvious solution would be a Sulzer type with level translator which offers maybe x3 gain after the op-amp followed by the pass transistor/darlington/mosfet but this would almost certainly oscillate unless it was compensated to the point where it was no faster than an LM317 with series pass device...
Maybe such luminaries on this site as ALW or Janneman have some pearls of wisdom on the subject.... maybe I'll have to design a high speed,high voltage discrete op-amp unless someone knows of a suitable design elsewhere
Jez.
p.s I know many feel that simple capacitor supplies are superior but I have a suspicion that a suitably quick,low impedence,stable regulator would surprise a lot of people !.
Here's a 90 Volt version of the "super reg" topology. Watch the dissipation in some of the resistors! Also the dissipation in the shunt reg for the op-amp is too high to use a TO-92 for the TL431. You'll need a DIP or surface mount package here.
I haven't built this yet, but stability and recovery from transient load currents have been simulated to death. A key issue in the stability is that it depends strongly on the non-ideal characteristics of the 100 uF load capacitor. Both the series resistance of around 0.25 Ohms and the series inductance of ~30 nH come strongly into play here. The load capacitor's series resonance is at about 93kHz which forms a low-Q notch filter which helps stabilize the circuit. Don't put a high Q 0.1 uF capacitor for a load! It will oscillate.
There's been lots of discussion of this and related topics in this thread: http://www.diyaudio.com/forums/showthread.php?s=&threadid=20937&highlight=
It's not capable of powering the output stage of a power amp, but it's suitable for MOSFET power amp designs in which the output stage supply is separate from that of the rest of the circuitry.
I haven't built this yet, but stability and recovery from transient load currents have been simulated to death. A key issue in the stability is that it depends strongly on the non-ideal characteristics of the 100 uF load capacitor. Both the series resistance of around 0.25 Ohms and the series inductance of ~30 nH come strongly into play here. The load capacitor's series resonance is at about 93kHz which forms a low-Q notch filter which helps stabilize the circuit. Don't put a high Q 0.1 uF capacitor for a load! It will oscillate.
There's been lots of discussion of this and related topics in this thread: http://www.diyaudio.com/forums/showthread.php?s=&threadid=20937&highlight=
It's not capable of powering the output stage of a power amp, but it's suitable for MOSFET power amp designs in which the output stage supply is separate from that of the rest of the circuitry.
Attachments
Hi,
An interesting option would be to use a couple of gainclones for each channel, DC coupled, putting a DC on the input that gives the required DC (+ or - supply) at the output as the supply. That will take care of stability problems which, as you say, will be probable and expensive. The gainclone should be stable in this application, using the customary zobel at the output.
I am pretty sure that you can put say 10.000uF at the gainclone's output to supply transient load currents. The inherent current limit of the gainclone would protect it at startup.
Since you can set the raw supply for the gainclone just a bit (say 5 volts or so) above the supply required, SOA wouldn't be an issue.
For a stereo power amp you would take 4 gainclones to have a separate +/- supply for each amp. You need only two raw supplies for the gainclones. Suppose your power amp need +/- 30VDC. The gainclone supplies would then be say +35VDC and gnd for the pos supply, and -35VDC and gnd for the neg one.
(I also like the irony of using a gainclone just to power a power amp, but that shouldn't bother you)
What do the other experts think about this idea? Am I off my rocker, or what?
Jan Didden
An interesting option would be to use a couple of gainclones for each channel, DC coupled, putting a DC on the input that gives the required DC (+ or - supply) at the output as the supply. That will take care of stability problems which, as you say, will be probable and expensive. The gainclone should be stable in this application, using the customary zobel at the output.
I am pretty sure that you can put say 10.000uF at the gainclone's output to supply transient load currents. The inherent current limit of the gainclone would protect it at startup.
Since you can set the raw supply for the gainclone just a bit (say 5 volts or so) above the supply required, SOA wouldn't be an issue.
For a stereo power amp you would take 4 gainclones to have a separate +/- supply for each amp. You need only two raw supplies for the gainclones. Suppose your power amp need +/- 30VDC. The gainclone supplies would then be say +35VDC and gnd for the pos supply, and -35VDC and gnd for the neg one.
(I also like the irony of using a gainclone just to power a power amp, but that shouldn't bother you
) What do the other experts think about this idea? Am I off my rocker, or what?
Jan Didden
Andy C,
Your mods to the super reg look most interesting !, I had'nt thought of losing the voltage across those zeners whilst keeping a low impedance drive for the next stage with minimal phase shift as I would think this would do.It could be interesting to replace Q4 with a push-pull follower with a touch of forward bias and then change Q5 for a low-on-resistance P-MOSFET of suitable current rating(the op-amp inverting and non inverting inputs would have to be reversed to maintain the correct feedback polarity). If this proved to be stable, I suspect this could be as close as I am likely to get to the ideal power amp supply, especially if a much faster op-amp remained stable and in conjunction with remote sensing.
I also found the output capacitor to be critical in high performance regulators, simulation shows oscillation with ideal output cap which can be tamed with a small series resistance, about 0.007 R was about optimum in my simulations of other , lower voltage regs (tend to look inductive at output).
Jan, I dont' know what a gainclone is yet !, there seems to be many threads on the subject but non of the ones I have seen so-far shows a diagram or explains exactly what peole mean by "gainclone", I'm sure further searching will produce enlightenment on the subject. I am,however,trying to avoid solutions which rely on a huge output cap to supply load transients and to give low ac output impedance.... or am I totally misunderstanding some subtlety of your approach here?.
I will check out the above link to an earlier thread.
Thanks for the ideas and "watch this space" as they say !
Jez.
Your mods to the super reg look most interesting !, I had'nt thought of losing the voltage across those zeners whilst keeping a low impedance drive for the next stage with minimal phase shift as I would think this would do.It could be interesting to replace Q4 with a push-pull follower with a touch of forward bias and then change Q5 for a low-on-resistance P-MOSFET of suitable current rating(the op-amp inverting and non inverting inputs would have to be reversed to maintain the correct feedback polarity). If this proved to be stable, I suspect this could be as close as I am likely to get to the ideal power amp supply, especially if a much faster op-amp remained stable and in conjunction with remote sensing.
I also found the output capacitor to be critical in high performance regulators, simulation shows oscillation with ideal output cap which can be tamed with a small series resistance, about 0.007 R was about optimum in my simulations of other , lower voltage regs (tend to look inductive at output).
Jan, I dont' know what a gainclone is yet !, there seems to be many threads on the subject but non of the ones I have seen so-far shows a diagram or explains exactly what peole mean by "gainclone", I'm sure further searching will produce enlightenment on the subject. I am,however,trying to avoid solutions which rely on a huge output cap to supply load transients and to give low ac output impedance.... or am I totally misunderstanding some subtlety of your approach here?.
I will check out the above link to an earlier thread.
Thanks for the ideas and "watch this space" as they say !
Jez.
janneman said:I also like the irony of using a gainclone just to power a power amp, but that shouldn't bother you
What do the other experts think about this idea? Am I off my rocker, or what?
Jan Didden
no expert but I think the gainclone idea should work.
However, would the reverse be true? Has anyone tried to use a regulator as amp? either a linear regulator or smps regulator?
jez said:
gainclones are IC-based amps. gainclone ICs are usually LM chips (LM3875 and 3886) or OPA (541/549 etc.). some also use TDAs as well.
Keep in mind that some writers on the subject such as D. Self contend that since a regulated power supply neccessarily involves a negative feedback network, using one with a power amplifier which also has a negative feedback network has the may cause serious problems. Potentially the two NFB cicuits could cause oscillations as they react to each other.
I can neither affirm or contradict the above assertion from my own experiences. I just think anyone contemplating the the use of a regulated PS with a power amp should be aware that the concern exists. There may well be circumstance specific to individual situations the either increase or decrease the likelyhood of a problem ocurring.
I can neither affirm or contradict the above assertion from my own experiences. I just think anyone contemplating the the use of a regulated PS with a power amp should be aware that the concern exists. There may well be circumstance specific to individual situations the either increase or decrease the likelyhood of a problem ocurring.
Yes, I had realised that a gainclone is a chip based power amp,it was as to how it differs from a standard application note design in such a way as to be called a gainclone that I am unsure of, I should have explained my self better...
I would have thought that a "power op-amp" solution, which is what is being suggested here, would not have the slew-rate,gain or GBP to rival the super regulator.
As the type,quantity and make of smoothing caps seems to have such a profound effect on the sound of an amp, and as they are most definately in the signal path of the output stage, it seems to me that if we could replace them with something better that would be A GOOD THING. The reg would need to have very low output impedance and be faster than the amp it was powering, I reckon that using slow error amps and following them with huge capacitors is defeating the point ! (easy for stability though)
Jez.
I would have thought that a "power op-amp" solution, which is what is being suggested here, would not have the slew-rate,gain or GBP to rival the super regulator.
As the type,quantity and make of smoothing caps seems to have such a profound effect on the sound of an amp, and as they are most definately in the signal path of the output stage, it seems to me that if we could replace them with something better that would be A GOOD THING. The reg would need to have very low output impedance and be faster than the amp it was powering, I reckon that using slow error amps and following them with huge capacitors is defeating the point ! (easy for stability though)
Jez.
millwood said:
millwood said:
Since you can power the gainclones as supplies just a few volts above the required output voltage, in practise the only spec of concern is the output current. Using a cap at the output would increase the transient performance (6-8 amps may be not enough in the peaks).
Jan Didden
[[snip]
Well, if you build that high power supply showed in the beginning of the thread, I am not convinced that it would be faster than a chip amp supply. It is VERY DIFFICULT to build a very fast supply that is stable. I have designed several, and found it to be even more difficult to design than a good, fast power amp.
And if you need high power pass elements, your problems multiply. It can be done (anything can, given enough time and $), but I think the chip amp supply would come very close if not better in specs than a discrete one.
Jan Didden
jez said:
Well, if you build that high power supply showed in the beginning of the thread, I am not convinced that it would be faster than a chip amp supply. It is VERY DIFFICULT to build a very fast supply that is stable. I have designed several, and found it to be even more difficult to design than a good, fast power amp.
And if you need high power pass elements, your problems multiply. It can be done (anything can, given enough time and $), but I think the chip amp supply would come very close if not better in specs than a discrete one.
Jan Didden
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.