Hi all,
I am working on a kind of universal channel amplifier. One of the issues is local power regulation for each active element. I thought about using an opamp with a reference voltage at its input and the output serving as the local supply voltage. Kind of like a super regulator without the output transistor. Performance would be somewhere between a super regulator and a standard 3-terminal regulator, for a low price, especially using dual opamps.
My unknown is the capacitive load on the opamp. Normally, this is a no-no. But using the opamp output to supply power to an active element that includes supply decoupling will lead to cap load. Using a small resistor in series with the opamp output can help, but will increase output impedance of the "regulator".
I think that "overloading" the opamp output with large (> 100nF) cap would probably give a low enough pole to restore stable operation, but I don't know for sure.
Anybody has an opinion, experience, or comment on this idea?
Jan Didden
I am working on a kind of universal channel amplifier. One of the issues is local power regulation for each active element. I thought about using an opamp with a reference voltage at its input and the output serving as the local supply voltage. Kind of like a super regulator without the output transistor. Performance would be somewhere between a super regulator and a standard 3-terminal regulator, for a low price, especially using dual opamps.
My unknown is the capacitive load on the opamp. Normally, this is a no-no. But using the opamp output to supply power to an active element that includes supply decoupling will lead to cap load. Using a small resistor in series with the opamp output can help, but will increase output impedance of the "regulator".
I think that "overloading" the opamp output with large (> 100nF) cap would probably give a low enough pole to restore stable operation, but I don't know for sure.
Anybody has an opinion, experience, or comment on this idea?
Jan Didden
Koinichiwa,
Well, some notes first. The "idea" of using the same Op-Amp (power or not) as regulator came to me from Dennis Morecroft (aka DNM), who, for his most prestigeous Amp (if can call these cute little plastic boxes in any way "prestigeous") used tow more of the same Amp's as external regulated supply.
On experiemntation I found this to sound very good and to be stable as long as the the circuit was suitably optimised. I used it only with CFB Op-Amp's (aka AD811/LM6181/LM6182) and took care to set the bandwidth of the supplied Op-Amp to more than two octaves less than the the regulator, in effect making the regulator faster than the main Amplifier (15 to 20MHz bandwidth is still plenty fast for Audio anyway).
I figured, with tight low capacitance and low inductance layout (Amplifier on top of PCB, Regs below so that the reg out & and chip +/-V pins almost touched) the shooting match was stable and so it was.
Now to capacitive loading. Please remember that ANY integrated regulator as well as such circuits as Jung/Sulzer (and Didden
) regulators all contain an Op-Amp for the AC loop.
The insertion of an emitter follower does not do all that much to the phase margin for stability (except making it worse). So as long as your Op-Amp will supply sufficient current and the capacitive load is either below or above the critical range the circuit will be stable.
My main argument for a "capacitor less" solution is philosphical. The Capacitor over certain ranges will dominate the PSU's behaviour, while across others theOp-Amp dominates. The cap often dominates in the Audio range, if we do something like that, what is the point of investing in a "good regulator".
If we start playing with capacitors we may as well do our homework and use correct capacitive decoupling with a choke to decouple the whole schebang from the "bad" regulators at higher frequencies (I like using that approach in existing gear being pooge'd), preventing the reg from doing anything "bad". Just watch the Q of the LC filter.
Sayonara
PS, attached a rendering of an experimental linestage from a few years ago, drawn by Carlos M. from a description by me, no guarantee for accuracy, I never drew a schema of this, all this is from memory and my memory is notoriously bad.
I think I had a 1k resistor (maybe 10k?) between reference Voltage and Op-Amp input for the reg.
It used LM317/337 Pre-regulators with very low resistance voltage set resistors and around 20V output, 10R - 1W & 150R or 180R - 5W. Sounded jolly decent, but I drifted over to Valves about that time, better sound.
janneman said:
Well, some notes first. The "idea" of using the same Op-Amp (power or not) as regulator came to me from Dennis Morecroft (aka DNM), who, for his most prestigeous Amp (if can call these cute little plastic boxes in any way "prestigeous") used tow more of the same Amp's as external regulated supply.
On experiemntation I found this to sound very good and to be stable as long as the the circuit was suitably optimised. I used it only with CFB Op-Amp's (aka AD811/LM6181/LM6182) and took care to set the bandwidth of the supplied Op-Amp to more than two octaves less than the the regulator, in effect making the regulator faster than the main Amplifier (15 to 20MHz bandwidth is still plenty fast for Audio anyway).
I figured, with tight low capacitance and low inductance layout (Amplifier on top of PCB, Regs below so that the reg out & and chip +/-V pins almost touched) the shooting match was stable and so it was.
Now to capacitive loading. Please remember that ANY integrated regulator as well as such circuits as Jung/Sulzer (and Didden
) regulators all contain an Op-Amp for the AC loop. The insertion of an emitter follower does not do all that much to the phase margin for stability (except making it worse). So as long as your Op-Amp will supply sufficient current and the capacitive load is either below or above the critical range the circuit will be stable.
My main argument for a "capacitor less" solution is philosphical. The Capacitor over certain ranges will dominate the PSU's behaviour, while across others theOp-Amp dominates. The cap often dominates in the Audio range, if we do something like that, what is the point of investing in a "good regulator".
If we start playing with capacitors we may as well do our homework and use correct capacitive decoupling with a choke to decouple the whole schebang from the "bad" regulators at higher frequencies (I like using that approach in existing gear being pooge'd), preventing the reg from doing anything "bad". Just watch the Q of the LC filter.
Sayonara
PS, attached a rendering of an experimental linestage from a few years ago, drawn by Carlos M. from a description by me, no guarantee for accuracy, I never drew a schema of this, all this is from memory and my memory is notoriously bad.
I think I had a 1k resistor (maybe 10k?) between reference Voltage and Op-Amp input for the reg.
It used LM317/337 Pre-regulators with very low resistance voltage set resistors and around 20V output, 10R - 1W & 150R or 180R - 5W. Sounded jolly decent, but I drifted over to Valves about that time, better sound.
Attachments
local regs
Sayonara,
Thanks for the info. I was looking at the AD8019 as the reg, which has very good PSRR and a low Zout (10mOhm I think) up to 100kHz or so. It can also source/sink a couple of 100 mAs (it's a DSL line driver really). And you are right of course on the emitter follower as worsening the stability margins. But you need it to get enough current. Since this opamp has enough current capacity, I'll like to try it without the follower.
So I should be able to run that without any additional local bypass cap on the load opamp. I can lower the bandwidth of the load opamp if necessary.
I'm not sure I agree on the cap issue. A good cap makes life easier for the reg, especially at the higher freqs. The LC filter you discuss would make it easier for the reg, but worse for the load.
Maybe the best way is to use a good quality cap with a value as low as possible for stability, say 10 or 100nF film. What is your experience with that?
The idea of low-value voltage set resistors is a good one to bias the prereg, I will use that also. It may also be a good idea for the opamp regs, to bias them in class A, although the static load (about 50mA) might already be enough.
Jan Didden
Sayonara,
Thanks for the info. I was looking at the AD8019 as the reg, which has very good PSRR and a low Zout (10mOhm I think) up to 100kHz or so. It can also source/sink a couple of 100 mAs (it's a DSL line driver really). And you are right of course on the emitter follower as worsening the stability margins. But you need it to get enough current. Since this opamp has enough current capacity, I'll like to try it without the follower.
So I should be able to run that without any additional local bypass cap on the load opamp. I can lower the bandwidth of the load opamp if necessary.
I'm not sure I agree on the cap issue. A good cap makes life easier for the reg, especially at the higher freqs. The LC filter you discuss would make it easier for the reg, but worse for the load.
Maybe the best way is to use a good quality cap with a value as low as possible for stability, say 10 or 100nF film. What is your experience with that?
The idea of low-value voltage set resistors is a good one to bias the prereg, I will use that also. It may also be a good idea for the opamp regs, to bias them in class A, although the static load (about 50mA) might already be enough.
Jan Didden
You are about to build an overkill I tink, but if you care about part count an expanse you may check LM723 or equivalent you will find a voltage reference, an OP amp an a power transistor in a chip.
About the capacitor at the output of regulator I think it is the way to go and the most important about this one is the layout around the OP amp for groung loops and current loop. Analog Device had on is site an interresting text on the subject but they moved with TEXAS and I dont know if it is still on the web
About the capacitor at the output of regulator I think it is the way to go and the most important about this one is the layout around the OP amp for groung loops and current loop. Analog Device had on is site an interresting text on the subject but they moved with TEXAS and I dont know if it is still on the web
local power
Yes, thanks, I know about the 723 (in fact used it 20 years or more ago), but for me it is too slow to give a good wideband low noise and low dynamic output impedance.
It is overkill, in the sense that the improvement is small compared to the extra cost. But it will be fun, so I can absorb the cost (book it under "entertainment", mama).
Jan Didden
Yes, thanks, I know about the 723 (in fact used it 20 years or more ago), but for me it is too slow to give a good wideband low noise and low dynamic output impedance.
It is overkill, in the sense that the improvement is small compared to the extra cost. But it will be fun, so I can absorb the cost (book it under "entertainment", mama).
Jan Didden
I find that puzzling, looking at any frequency over the audio band the regulator (assuming it is 'good) will dominate impedance.
Xc for a 100u cap at 1k = 1.6R, at 10k = 0.16R
Zout for an decent reg with reasonable open loop gain will be microOhms at 1k, maybe several hundred uOhms at 10k.
The cap brings benefits at HF as open-loop response falls, where it then dominates output Z.
Other parameters vary along similar lines.
Or are you talking subjectively dominant - there's no doubt the caps around the reg affect the sound, not sure I'd consider it dominant though, although it may be with less 'good' regulators?
Andy.
If you are havig fun and you know what you are about to do it allrigt with me I will not blame you, amateur audio is for entertainement listening to it or building it!
Puzzling Evidence*
'I find that puzzling, looking at any frequency over the audio band the regulator (assuming it is 'good) will dominate impedance."
If you look at the scenario with a few 1000 uF on the output of three terminal regulator and the capacitor will be the lower impedance down into the audio band. Three terminal regulators and followers with large output caps are very common approaches and the capacitor will dominate. Good is a little bit vague a term for the regulator, "very low impedance" might be more descriptive. There are cases where it is desired for the cap to be dominate over the regulator in the audio band. When using good caps they the may sound better than the regulator at certain frequencies. The reason for choosing a certain value cap for the regulator becomes clearer in light the relative capacitor and regulator impedance at a given frequency range.
You do point out a very good reason why some caps can sound very different in a regulator application than as a coupling cap. Their low frequency characteristic are not being swamped by lower impedance in parallel when used as a coupling cap.
*I hope you get ev'rything you need
'I find that puzzling, looking at any frequency over the audio band the regulator (assuming it is 'good) will dominate impedance."
If you look at the scenario with a few 1000 uF on the output of three terminal regulator and the capacitor will be the lower impedance down into the audio band. Three terminal regulators and followers with large output caps are very common approaches and the capacitor will dominate. Good is a little bit vague a term for the regulator, "very low impedance" might be more descriptive. There are cases where it is desired for the cap to be dominate over the regulator in the audio band. When using good caps they the may sound better than the regulator at certain frequencies. The reason for choosing a certain value cap for the regulator becomes clearer in light the relative capacitor and regulator impedance at a given frequency range.
You do point out a very good reason why some caps can sound very different in a regulator application than as a coupling cap. Their low frequency characteristic are not being swamped by lower impedance in parallel when used as a coupling cap.
*I hope you get ev'rything you need
Attachments
local regs
Fred,
The large electrolytics you quote will have so much ESR (and ESL) that they don't really lower the output Z compared to a 3-pin regulator. Most 3-pin regulators rise in output Z above a few kHz, and the way to battle that is with a good film cap of several uF (10-20 preferably), not electrolytics! You may need the electrolytic for stability, but not for performance as such.
If you have a regulator with very low Z like ALW's, I don't know of any physical cap that can better that this side of 1MHz.
Jan Didden
Fred,
The large electrolytics you quote will have so much ESR (and ESL) that they don't really lower the output Z compared to a 3-pin regulator. Most 3-pin regulators rise in output Z above a few kHz, and the way to battle that is with a good film cap of several uF (10-20 preferably), not electrolytics! You may need the electrolytic for stability, but not for performance as such.
If you have a regulator with very low Z like ALW's, I don't know of any physical cap that can better that this side of 1MHz.
Jan Didden
I have a feeling
I must admit to not considering 3 terminal reg's when I made my post. They are frankly anathema to good quality analogue stages, IME. Not that they cannot sound good (I've plenty of things that sound great using them) but simply that they are far, far away from being optimal IME. I use them simply because the circuits haven't been updated with newer developments I've made since.
Digital stuff may be different though.
There was a good reason for putting 'good' in quotes and hence being a little vague, since if I define what I know to be 'good' I'd give away more than I'm prepared to at present.
You as always though make good points, and as with everything in life, a system is usually as good as it's worst bit, not it's best. A balanced outlook is always sensible.
That's my...
anyway.
Andy.
I must admit to not considering 3 terminal reg's when I made my post. They are frankly anathema to good quality analogue stages, IME. Not that they cannot sound good (I've plenty of things that sound great using them) but simply that they are far, far away from being optimal IME. I use them simply because the circuits haven't been updated with newer developments I've made since.
Digital stuff may be different though.
There was a good reason for putting 'good' in quotes and hence being a little vague, since if I define what I know to be 'good' I'd give away more than I'm prepared to at present.
You as always though make good points, and as with everything in life, a system is usually as good as it's worst bit, not it's best. A balanced outlook is always sensible.
That's my...
An externally hosted image should be here but it was not working when we last tested it.
anyway.
Andy.
About large caps...
I once tried to upgrade my MicroMega CD...and replaced caps after regulators...no go!
I tried several types / brands in the 1000-3300uF range, no go.
I then put in a 100uF os-con, and all was well.
Did not have time to check what really happened, but it just would not read discs with large caps...
Arne K
I once tried to upgrade my MicroMega CD...and replaced caps after regulators...no go!
I tried several types / brands in the 1000-3300uF range, no go.
I then put in a 100uF os-con, and all was well.
Did not have time to check what really happened, but it just would not read discs with large caps...
Arne K
A Panasonic 3300uF 25 V FC type electrolytic has an impedance of 18 milliohms at 100KHz. At 10 KHz the capacitive reactance (the impedance for an ideal 3300uF cap) is about 5 milliohms and at 100 KHz is 0.5 milliohms. It is reasonable to estimate the ESR as about 18 milliohms.
From the graph of the output impedance of a LM317 at 10 KHz with bypassed adj. terminal is about 30 milliohms and with an unbypassed adj. terminal about 300 milliohms. At 10Khz the capacitor impedance is 2/3s of the regulator's impedance in the first case and is about 16 times lower in the case of the unbypassed regulator's impedance. This impedance graph is for a 500 mA load current. The regulator will most likely have a higher output impedance at a few 10s of mA of load current.
Three 1000 uF FCs in parallel with get you 10 milliohms and the caps will start to compete with the regulator impedance above a few KHz. Most 7800 regulators will have higher output impedance than LM317 with a bypassed adj. terminal. As I stated earlier the caps will have an influence in the audio frequency range. I think some three terminal regulators may not be very stable with only a 20 uF film cap on the output.
Sorry for the math, data and graphs but I don't like to shoot from the hip when someone makes an assertion that I question and that can be easily researched. Also it gives a little better feel for the actual impedances involved. I guess I still stand by my original statement.
http://www.schuro.de/Daten/Panasonic/FC.pdf
http://www.microsemi.com/datasheets/SG117A.pdf
From the graph of the output impedance of a LM317 at 10 KHz with bypassed adj. terminal is about 30 milliohms and with an unbypassed adj. terminal about 300 milliohms. At 10Khz the capacitor impedance is 2/3s of the regulator's impedance in the first case and is about 16 times lower in the case of the unbypassed regulator's impedance. This impedance graph is for a 500 mA load current. The regulator will most likely have a higher output impedance at a few 10s of mA of load current.
Three 1000 uF FCs in parallel with get you 10 milliohms and the caps will start to compete with the regulator impedance above a few KHz. Most 7800 regulators will have higher output impedance than LM317 with a bypassed adj. terminal. As I stated earlier the caps will have an influence in the audio frequency range. I think some three terminal regulators may not be very stable with only a 20 uF film cap on the output.
Sorry for the math, data and graphs but I don't like to shoot from the hip when someone makes an assertion that I question and that can be easily researched. Also it gives a little better feel for the actual impedances involved. I guess I still stand by my original statement.
http://www.schuro.de/Daten/Panasonic/FC.pdf
http://www.microsemi.com/datasheets/SG117A.pdf
Attachments
local power stuff
Fred,
Thanks for that, I wasn't aware that these caps are so good at higher frequencies.
In case of the example you mentioned (3 x 1000 uF FC caps) that would give the possibility to (physically) istribute the caps as close as possible to the loading stages.
Coming back to ALW's (and others) super regulators, since that Zout is an order of magnitude lower, would you agree that here the advantage of the electrolytics is questionable?
Jan Didden
Fred,
Thanks for that, I wasn't aware that these caps are so good at higher frequencies.
In case of the example you mentioned (3 x 1000 uF FC caps) that would give the possibility to (physically) istribute the caps as close as possible to the loading stages.
Coming back to ALW's (and others) super regulators, since that Zout is an order of magnitude lower, would you agree that here the advantage of the electrolytics is questionable?
Jan Didden
everybody is being much too agreeable
I would if Andy and others hadn't stated how different small value electrolytics sound different. There is a lot the the high frequency transient response has to do with how regulators sound. The regulator is dealing with a lot of RF noise on it's input. I have found that the input cap on the regulator can have a great influence on the sonics, as well as how close to the regulator the caps are. I but a Black Gate about an inch and a half away from the output of a LT1085 once and wasn't at all pleased with the sound. I moved it as close to the regulator as possible and it sounded great. I often put a electroliytic by the circuit and another one VERY close to the regulator.
I would if Andy and others hadn't stated how different small value electrolytics sound different. There is a lot the the high frequency transient response has to do with how regulators sound. The regulator is dealing with a lot of RF noise on it's input. I have found that the input cap on the regulator can have a great influence on the sonics, as well as how close to the regulator the caps are. I but a Black Gate about an inch and a half away from the output of a LT1085 once and wasn't at all pleased with the sound. I moved it as close to the regulator as possible and it sounded great. I often put a electroliytic by the circuit and another one VERY close to the regulator.
Koinichiwa,
For a REALLY GOOD regulator - yes. Like the Jung/Sulzer Circuit, which achieves a very low output impedance. However, this must be ofset against the fairly marginally stability (even without the AD797 that thing likes to oscillate).
But even with the Jung Reg, once a few mm PCB trace exist it's exceptionally low output impedance is swamped out by the PCB traces.
Subjectively.
Let's take an example I suggested in the 3-Pin regulator thread.
We take a generic 7805 which supplies our digital section in a CD-Player (or DAC or such). The 7805 has a Z-out of 0.008 Ohm up to 1KHz, from where the output impedance rises to 0.025Ohm @ 10KHz (and about 0.2Ohm @ 100KHz). We now place a Panasonic FC series 10,000uF/10V capacitor after the Regulator. This has got an ESR of < 0.015Ohm and an Impedance of around 0.03 Ohm @ 10KHz and 0.015Ohm @ 100KHz.
Instead of a PSU driving the various (usually long) traces to the supplied nodes from a fairly inductive supply, we now have a supply that remains < 0.015 Ohm for the whole audio range and out to 100KHz.
Should any PSU line modulation make it back through the PSU line from our supplied pin this modulation will be "snubbed out" by in effect an 0.01...0.02Ohm impedance to ground.
IF we have paid attention to the PCB layout and we "fan" all supply traces in star wring form out from the 10,000uF terminal (sometimes when modding CD-Players it may worthwhile simply running wires and cut away certain PCB traces) any supply line noise making it through the circuit will be minimised and hopefully sunk at the PSU.
If we then take care to minimise the PSU Impedance at the apropriate supplied pin by suitable capacitive bypassing (which can remain effectiive into the 10s of MHz region) we have a most excellently quiet supply.
I would argue ANY DAY, that in same situation just replacing the 7805 with a "super reg" of any particular description (including the use of 3-Pin PCB that can drop in as previously discussed) and not attending to the various local bypass issues will bring much less of an improvement.
And yes, if you combine a 7805 or 7815/7915 with such very large value bypass capacitors (and suitable local decoupling) I found that these capacitors dominate the percieved sound quite strongly.
Sayonara
ALW said:
For a REALLY GOOD regulator - yes. Like the Jung/Sulzer Circuit, which achieves a very low output impedance. However, this must be ofset against the fairly marginally stability (even without the AD797 that thing likes to oscillate).
But even with the Jung Reg, once a few mm PCB trace exist it's exceptionally low output impedance is swamped out by the PCB traces.
ALW said:
Subjectively.
Let's take an example I suggested in the 3-Pin regulator thread.
We take a generic 7805 which supplies our digital section in a CD-Player (or DAC or such). The 7805 has a Z-out of 0.008 Ohm up to 1KHz, from where the output impedance rises to 0.025Ohm @ 10KHz (and about 0.2Ohm @ 100KHz). We now place a Panasonic FC series 10,000uF/10V capacitor after the Regulator. This has got an ESR of < 0.015Ohm and an Impedance of around 0.03 Ohm @ 10KHz and 0.015Ohm @ 100KHz.
Instead of a PSU driving the various (usually long) traces to the supplied nodes from a fairly inductive supply, we now have a supply that remains < 0.015 Ohm for the whole audio range and out to 100KHz.
Should any PSU line modulation make it back through the PSU line from our supplied pin this modulation will be "snubbed out" by in effect an 0.01...0.02Ohm impedance to ground.
IF we have paid attention to the PCB layout and we "fan" all supply traces in star wring form out from the 10,000uF terminal (sometimes when modding CD-Players it may worthwhile simply running wires and cut away certain PCB traces) any supply line noise making it through the circuit will be minimised and hopefully sunk at the PSU.
If we then take care to minimise the PSU Impedance at the apropriate supplied pin by suitable capacitive bypassing (which can remain effectiive into the 10s of MHz region) we have a most excellently quiet supply.
I would argue ANY DAY, that in same situation just replacing the 7805 with a "super reg" of any particular description (including the use of 3-Pin PCB that can drop in as previously discussed) and not attending to the various local bypass issues will bring much less of an improvement.
And yes, if you combine a 7805 or 7815/7915 with such very large value bypass capacitors (and suitable local decoupling) I found that these capacitors dominate the percieved sound quite strongly.
Sayonara
Re: local regs
Koinichiwa,
Yes, this looks like a good choice. If you arrange the PCB right this could be easily attached to a suitable heatsink. TI also has some similar ADSL Drivers whith cases that I think are better thermally, if you draw more than a few mA from the reg.
IF YOU KEEP THE LAYOUT TIGHT. Without load capacitor the layout becomes extremely critical. The reg must be very close to the supplied pin, or circuit stability will suffer.
True, but with a circuit similar to what we are talking here you don't as such need a capacitor, you need an RC circuit to introduce a suitable zero in the phase-response. I think a Film Cap may have too low an ESR to work well.
I tried a small film cap (10nF I think) across +V to -V, the regs started oscillating.
Sayonara
Koinichiwa,
janneman said:
Yes, this looks like a good choice. If you arrange the PCB right this could be easily attached to a suitable heatsink. TI also has some similar ADSL Drivers whith cases that I think are better thermally, if you draw more than a few mA from the reg.
janneman said:
IF YOU KEEP THE LAYOUT TIGHT. Without load capacitor the layout becomes extremely critical. The reg must be very close to the supplied pin, or circuit stability will suffer.
janneman said:
True, but with a circuit similar to what we are talking here you don't as such need a capacitor, you need an RC circuit to introduce a suitable zero in the phase-response. I think a Film Cap may have too low an ESR to work well.
janneman said:
I tried a small film cap (10nF I think) across +V to -V, the regs started oscillating.
Sayonara
Again I have to take issue with this, using an AD825 as the error amplifier, there's no marginal stability involved with my implementations of the circuit. Yes it can be made to oscillate, but the conditions for oscillation are easily definable and easy to deal with.
This is true of many circuits, the key with everything is to control the system as a whole, not just analyse one single element.
They are not though, I readily admit, a universal drop-in in every circumstance.
I don't wish to get into arguments any more, but have to say that in analogue supplies, having built some 10-15 super-reg's for my own use, and sold in excess of 90 for others, the sonic benefits have been massive in everyone's case, my own inlcuded. That's with no change in local decoupling arrangements.
I have a mail box full of ecstatic comments from everyone who has tried them, not a single disappointed user. From this I tend to draw the conclusion that others who've had different experiences either: -
a. listen to different things
b. have imperfect implementations
c. have vastly different topologies in their circuits that make these things a non-issue
Your view would seem to come from the view that the conventional measurements by which PSU's are assessed (impedance / noise / line rejection) are the only ones that are relevant.
As I've stated before, I cannot at present say much more at present, but really do encourage others to think a bit outside the box with regard to PSU's, I will say though that the AC performance is critical.
Can I say that again - AC performance is critical.
You have though presented an easy-to-try alternative (the large-cap, star-wired) and I will endeavour, with an open mind, to try this soon for a comparison with an existing set-up.
Andy.
local regs
Sayonara,
I agree with most of what you stated above, BUT (there always is a but, isn't there) the point with the super regulators is the remote sensing.
The curve that ALW showed earlier, measured by me, shows the Zout at the end of a 6 inch lead to the load. I measured both my original Jung/Didden prototype as well as ALW's board, with his board doing just a bit better above a few kHz.
The Zout is about 60 MICRO Ohms at low freq, rising to 1 milli Ohms above audio. So, this is available at the load pin!
As i said, I don't know of any physical capacitor that can match this.
Jan Didden
Kuei Yang Wang said:
Sayonara,
I agree with most of what you stated above, BUT (there always is a but, isn't there) the point with the super regulators is the remote sensing.
The curve that ALW showed earlier, measured by me, shows the Zout at the end of a 6 inch lead to the load. I measured both my original Jung/Didden prototype as well as ALW's board, with his board doing just a bit better above a few kHz.
The Zout is about 60 MICRO Ohms at low freq, rising to 1 milli Ohms above audio. So, this is available at the load pin!
As i said, I don't know of any physical capacitor that can match this.
Jan Didden
Koinichiwa,
I completely agree. And I happen to "trust" the AC behaviour of a Capacitor a little more than that of active, high feedback circuit.
I will readily admit however that there are situations where a given circuit supplied has a very poor PSRR and thusly the effect can be drastic.
I notice again that your "Super Regs" seem to be at home in a Naim envoironment. Naim Gear has many circuits with really **** poor PSRR, so improving the Reg above and beyond LM317 levels makes a lot of sense.
But I'd personally probably would use local Shunt Regs (TL431 with added shunt transistor) with a CCS as feed rather than series regs, except I prefer to have nothing much to do with Naim gear.
You should try shunt regs or at least bi-directional (push-pull) regs, the results may surprise you.
Sayonara
ALW said:
I completely agree. And I happen to "trust" the AC behaviour of a Capacitor a little more than that of active, high feedback circuit.
I will readily admit however that there are situations where a given circuit supplied has a very poor PSRR and thusly the effect can be drastic.
I notice again that your "Super Regs" seem to be at home in a Naim envoironment. Naim Gear has many circuits with really **** poor PSRR, so improving the Reg above and beyond LM317 levels makes a lot of sense.
But I'd personally probably would use local Shunt Regs (TL431 with added shunt transistor) with a CCS as feed rather than series regs, except I prefer to have nothing much to do with Naim gear.
You should try shunt regs or at least bi-directional (push-pull) regs, the results may surprise you.
Sayonara
maybe not
The Jung regulator has remote sense lines available to take the feedback for the regulator at the load. The trace impedance is part of the open loop output impedance and is reduced by the feedback to about the same levels. The open loop output impedance of the regulator is a few ohms and you are adding a few 10s of milliohms to that before the loop feedback.
I have tried about a dozen of some of the more popular AD and Burr Brown Op Amps in a very similar circuit who no stability problems. Even the AD 797 is rock stable in mine.
I will not argue that capacitors have big influence on the sound of regulators but aren't the whole story. The Jung op amp based regulator has orders of magnitude better PSSR, noise, and output impedance than a three terminal at audio frequencies where the regulator is doing the most work. Large caps will not fix all these issues. There are also other approaches like discreet transistor circuits where one can trade off lowest output impedance for a flat output impedance across the audio band. For really good performance a three terminal regulator has too many compromises in performance for many people. you can get away with there use in digital circuits but the type and brand is audible there as well.
AWL has done a very nice implementation of this regulator and a SNAPS regulator upgrade with the best written manuals
I have seen in a long time. The manuals by themselves are a works of art that fill me with jealousy. Very nice job Andy!
http://www.alw.audio.dsl.pipex.com/products.htm
Kuei Yang Wang said:
The Jung regulator has remote sense lines available to take the feedback for the regulator at the load. The trace impedance is part of the open loop output impedance and is reduced by the feedback to about the same levels. The open loop output impedance of the regulator is a few ohms and you are adding a few 10s of milliohms to that before the loop feedback.
I have tried about a dozen of some of the more popular AD and Burr Brown Op Amps in a very similar circuit who no stability problems. Even the AD 797 is rock stable in mine.
I will not argue that capacitors have big influence on the sound of regulators but aren't the whole story. The Jung op amp based regulator has orders of magnitude better PSSR, noise, and output impedance than a three terminal at audio frequencies where the regulator is doing the most work. Large caps will not fix all these issues. There are also other approaches like discreet transistor circuits where one can trade off lowest output impedance for a flat output impedance across the audio band. For really good performance a three terminal regulator has too many compromises in performance for many people. you can get away with there use in digital circuits but the type and brand is audible there as well.
AWL has done a very nice implementation of this regulator and a SNAPS regulator upgrade with the best written manuals
I have seen in a long time. The manuals by themselves are a works of art that fill me with jealousy. Very nice job Andy!
http://www.alw.audio.dsl.pipex.com/products.htm
Attachments
- Status
- Not open for further replies.