Thanks Andy and Hoffmeyer, that is all very interesting indeed.
(Andy, I am also starting to read the thread in the other forum, but there are a lot of posts... )
But once more to the POOGE schematic:
Do you think it is possible to create this PSU to the same specs with "normal" components like a combination of BD135 / ZTX 653 / ZTX 753 / LM329 ? How critical is the value of C3 (820n) ? Is D1 a red LED ? All values for a 24V output ?
Are you going to publish your SMD PCB layout one day ?
Preamp rectification: Did you make noise measurements with different types of bridges, half bridges, fast diodes ? Is a big transformer
I am thrilled anyway. I think we (at least I ) never were so near to understand why different power supplies make so much different in sound with certain simple potentially good sounding circuits. And more and more I realize how important is the right combination in grounding, arrangement and power supply capabilities. Now slowly the people at N.A. will have something to think about, and they will have some new series of tests I suppose.
Serious work Andy !
Klaus
(Andy, I am also starting to read the thread in the other forum, but there are a lot of posts... )
But once more to the POOGE schematic:
Do you think it is possible to create this PSU to the same specs with "normal" components like a combination of BD135 / ZTX 653 / ZTX 753 / LM329 ? How critical is the value of C3 (820n) ? Is D1 a red LED ? All values for a 24V output ?
Are you going to publish your SMD PCB layout one day ?
Preamp rectification: Did you make noise measurements with different types of bridges, half bridges, fast diodes ? Is a big transformer
I am thrilled anyway. I think we (at least I ) never were so near to understand why different power supplies make so much different in sound with certain simple potentially good sounding circuits. And more and more I realize how important is the right combination in grounding, arrangement and power supply capabilities. Now slowly the people at N.A. will have something to think about, and they will have some new series of tests I suppose.
Serious work Andy !
Klaus
Lohk
Klaus,
Take a look at the Jung / Didden results though, they're awesome, but can be made even better, particularly in line rejection.
Pretty much, I'll post some suggestions once I've had chance to do a bit of modelling for you - the Sulzer reg is MUCH better though, and proabably easier to build and smaller, primarily because the capacitor values needed are smaller.
I actually think C3 could be eliminated if a good low-ESR cap was used (OS-Con, Rubycon ZA etc.).
The LED in mine was Yellow, because I had a load of them salvaged from old strip-downs, but RED or green is likely to be fine.
The values shown will give circa 23V.
Personally I'm beginning to form the opininion that the raw supply, subject to some basic requirements is much less critical with a really good regulator. With the performance constraints of standard regulators the raw supply being massively over-engineered brings benefits, but they are primarily to make the regulators life easier and enhance it's failing performance.
The raw supply needs to be able to supply the current requirements, with a sensible margin, I don't feel special rectifiers are required (their prime aim is to prevent the ringing present in the transformer when the diodes switch off sharply) when a simple, but properly designed snubbing network can do the job better and cheaper. Please note caps across the rectifiers are bad news and can make things worse - they just shift the resonant frequency (which can be calculated by measuring the transformer inductance and capacitance), making it lower and potentially more intrusive.
On a toroidal transformer I had it's self resonant frequency was a couple of Mhz, an RC snubber can damp this within a cycle or so. It's clearly visible on a 'scope if you look at the transformer secondary waveform on one trace, the raw DC on the other - as the diode goes out of conduction the transformer breaks into a self-resonant oscillation.
Here's a good article explaining here: -
http://131.109.59.51/technical.html.
I'd like to try some different transformers next - standard toroids are very wideband devices, offering little filtering to mains-borne noise. Electrostatic shields help, but not very effectively. Split windings, with no overlap (and therefore no inter-winding capacitance) helps a lot.
Things don't always sound better though, and I haven't tried ti yet.
Andy.
Klaus,
Take a look at the Jung / Didden results though, they're awesome, but can be made even better, particularly in line rejection.
Pretty much, I'll post some suggestions once I've had chance to do a bit of modelling for you - the Sulzer reg is MUCH better though, and proabably easier to build and smaller, primarily because the capacitor values needed are smaller.
I actually think C3 could be eliminated if a good low-ESR cap was used (OS-Con, Rubycon ZA etc.).
The LED in mine was Yellow, because I had a load of them salvaged from old strip-downs, but RED or green is likely to be fine.
The values shown will give circa 23V.
Personally I'm beginning to form the opininion that the raw supply, subject to some basic requirements is much less critical with a really good regulator. With the performance constraints of standard regulators the raw supply being massively over-engineered brings benefits, but they are primarily to make the regulators life easier and enhance it's failing performance.
The raw supply needs to be able to supply the current requirements, with a sensible margin, I don't feel special rectifiers are required (their prime aim is to prevent the ringing present in the transformer when the diodes switch off sharply) when a simple, but properly designed snubbing network can do the job better and cheaper. Please note caps across the rectifiers are bad news and can make things worse - they just shift the resonant frequency (which can be calculated by measuring the transformer inductance and capacitance), making it lower and potentially more intrusive.
On a toroidal transformer I had it's self resonant frequency was a couple of Mhz, an RC snubber can damp this within a cycle or so. It's clearly visible on a 'scope if you look at the transformer secondary waveform on one trace, the raw DC on the other - as the diode goes out of conduction the transformer breaks into a self-resonant oscillation.
Here's a good article explaining here: -
http://131.109.59.51/technical.html.
I'd like to try some different transformers next - standard toroids are very wideband devices, offering little filtering to mains-borne noise. Electrostatic shields help, but not very effectively. Split windings, with no overlap (and therefore no inter-winding capacitance) helps a lot.
Things don't always sound better though, and I haven't tried ti yet.
Andy.
Andy
Could you please check that the link in your last post is correct. I cannot get it to work (page not found error).
Thanks
Geoff
Could you please check that the link in your last post is correct. I cannot get it to work (page not found error).
Thanks
Geoff
Hello
I need to get good regulation for my Pass X series project's front-end. I am aiming at +/-47V, and just wonder which one of the circuits discussed previously in the thread would be best for my purpose. If any is.
Current needed is 400mA. The raw supply will be +/-64V with no load and +/-56V with 1A load (which I see is more than needed, so practically it will be at +/-60V or so).
Transformers will be two 63VA 20+20V in series, resulting in one 126VA 40+40V with 15% Uo, that is 46+46V with no load.
Or would I be just fine with a capacitance multiplier?
Also, I don't think I will hunt for any special parts, so ordinary polyester and electrolytic capacitors, 1% metal film resistors, ordinary diodes and zeners and commonly available transistors are in. Do you think I will miss much?
I am indeed a person who doesn't want to find out how to order electrolytic caps and special resistors of certain brandname, product line or model. Maybe it's this country, maybe it's money, maybe it's just that I'm too anxious to wait for such to arrive.
-Kimmo Sundqvist
I need to get good regulation for my Pass X series project's front-end. I am aiming at +/-47V, and just wonder which one of the circuits discussed previously in the thread would be best for my purpose. If any is.
Current needed is 400mA. The raw supply will be +/-64V with no load and +/-56V with 1A load (which I see is more than needed, so practically it will be at +/-60V or so).
Transformers will be two 63VA 20+20V in series, resulting in one 126VA 40+40V with 15% Uo, that is 46+46V with no load.
Or would I be just fine with a capacitance multiplier?
Also, I don't think I will hunt for any special parts, so ordinary polyester and electrolytic capacitors, 1% metal film resistors, ordinary diodes and zeners and commonly available transistors are in. Do you think I will miss much?
I am indeed a person who doesn't want to find out how to order electrolytic caps and special resistors of certain brandname, product line or model. Maybe it's this country, maybe it's money, maybe it's just that I'm too anxious to wait for such to arrive.
-Kimmo Sundqvist
Jung / Didden schematics + Sulzer oscillation?
Andy,
you keep referring to the Jung / Didden results. Do you have a link or a schematic? I had heard before about the Jung regulators but found no schematic. Have built and measured one?
Is there any literature on the Sulzer circuit? I am a little surprised that in won't oscillate. After all, the op amp is pretty wide-band compared to the time constant of any possible change in the output voltage. The non-inverting input is AC-grounded, but more so than the inverting input. Won't noise or transient cause wild oscillations?
Greetings,
Eric
Andy,
you keep referring to the Jung / Didden results. Do you have a link or a schematic? I had heard before about the Jung regulators but found no schematic. Have built and measured one?
Is there any literature on the Sulzer circuit? I am a little surprised that in won't oscillate. After all, the op amp is pretty wide-band compared to the time constant of any possible change in the output voltage. The non-inverting input is AC-grounded, but more so than the inverting input. Won't noise or transient cause wild oscillations?
Greetings,
Eric
The Op Amp is operating as a unity gain buffer at high frequencies, due to the AC feedback provided by C3. Loop gain is rolled off. These regulators have been around for eons; I used this same circuit in a musical instrument preamplifier power supply in 1977.
Best regards,
Jon
Best regards,
Jon
oscillations
Yeah, I remember seeing similar circuits in some app notes etc.
I realize those circuits work, but I'd like to develop an intuitive understanding why.
I can see that there is unity gain HF feedback to the inverting input. But what does the op amp do, once it detects an error signal? After all, there is a substantial capacitve load at its output...
Yeah, I remember seeing similar circuits in some app notes etc.
I realize those circuits work, but I'd like to develop an intuitive understanding why.
I can see that there is unity gain HF feedback to the inverting input. But what does the op amp do, once it detects an error signal? After all, there is a substantial capacitve load at its output...
Hello
Hello, this is a little off-topic but does not warrant new thread.
Would adding a PI filter to a power supply be good or bad for a low-noise circuit?
I've used them to great affect in high-power applications such as the SOZ but am a little worried about putting them in a small-signal application.
Thanks,
Aaron
Hello, this is a little off-topic but does not warrant new thread.
Would adding a PI filter to a power supply be good or bad for a low-noise circuit?
I've used them to great affect in high-power applications such as the SOZ but am a little worried about putting them in a small-signal application.
Thanks,
Aaron
LOW NOISE regulators
Hi All.,
If you want low noise don't use fixed or adjustable regulators like µA7815 or 7915 or LT1086 or LT1033. This is admitted by the manufacturers themselves.
If you use any of the above regulators it is difficult to get rid of the noise , once it has entered your system.
Start with a low noise LT1021 reference or AD 586. These are <B>BURRIED</B> zener references and have far lower noise than any bandgapreference or normal 'zener' diode or TL431. If you apply a low pass filter you can filter or attenuate the noise even more. If you use a Sulzer or Walt Jung like regulator after these a very low noise supply is obtained.
Only for low output moving coil cartridge amplifiers more ripple attenuation is needed. In this case I use a 22mH choke and a additional 4700 µF capacity after the main filter capacitors. More would do no harm, as is not the case if applied without the choke.
Hi All.,
If you want low noise don't use fixed or adjustable regulators like µA7815 or 7915 or LT1086 or LT1033. This is admitted by the manufacturers themselves.
If you use any of the above regulators it is difficult to get rid of the noise , once it has entered your system.
Start with a low noise LT1021 reference or AD 586. These are <B>BURRIED</B> zener references and have far lower noise than any bandgapreference or normal 'zener' diode or TL431. If you apply a low pass filter you can filter or attenuate the noise even more. If you use a Sulzer or Walt Jung like regulator after these a very low noise supply is obtained.
Only for low output moving coil cartridge amplifiers more ripple attenuation is needed. In this case I use a 22mH choke and a additional 4700 µF capacity after the main filter capacitors. More would do no harm, as is not the case if applied without the choke.
I have some questions about regulator noise from one of my current projects.
I'm using three AD8610 op amps (all used non-inverting, unity gain) in one section of a 4th order Linkwitz-Riley crossover. One op amp is the input buffer and the other two provide the high pass section. There are at least 2 resistors and 5 capacitors in this section besides the op amps, contributing noise. +/-12V regulation is provided by 7812 and 7912 regulators from a 20 VDC supply initially filtered with 440 microfarads; 200 microfarads filter the regulator outputs. Also, each op amp has 0.1 microfarad caps at the +/- voltage inputs.
The residual noise of this system is just over 100 microvolts (maybe 105). When I bypass the AC supply and connect two batteries suppling +/-12 VDC (actual voltage is about 13 VDC), the noise drops to 80 microvolts. Here I must note that the battery input is before, not after, the 12 V regulators. I did this in case the batteries were accidentally connected when the battery re-charger was being used as this can supply over 14 VDC and the AD8610 is only rated for 13 VDC; at $15 apiece, I don't want to fry any. This means that the regulators are still in the circuit contributing noise. (BTW, the battery->regulator gives +/- 11.2 VDC, whereas the AC supplied power is +/-12.2 VDC, so the supply is affected by the regulator drop-out, while not regulating anything.)
After reading the above comments in this thread, I'm wondering: (1) would the 80 microvolts of noise be lower with other types/designs of regulators; (2) would the noise be significantly lower if I put the battery power in _after_ the regulators; (3) is the 100 -> 80 microvolts of noise lower only because the of the lower power supply (11.2 vs 12.2 VDC); and finally, (4) is it even worth bothering with since 80 microvolts is very low (many commercial amps/preamps have easily 200 or more microvolts of noise)--and besides no audible noise added to my speakers when the circuit is used? Comments?
I'm using three AD8610 op amps (all used non-inverting, unity gain) in one section of a 4th order Linkwitz-Riley crossover. One op amp is the input buffer and the other two provide the high pass section. There are at least 2 resistors and 5 capacitors in this section besides the op amps, contributing noise. +/-12V regulation is provided by 7812 and 7912 regulators from a 20 VDC supply initially filtered with 440 microfarads; 200 microfarads filter the regulator outputs. Also, each op amp has 0.1 microfarad caps at the +/- voltage inputs.
The residual noise of this system is just over 100 microvolts (maybe 105). When I bypass the AC supply and connect two batteries suppling +/-12 VDC (actual voltage is about 13 VDC), the noise drops to 80 microvolts. Here I must note that the battery input is before, not after, the 12 V regulators. I did this in case the batteries were accidentally connected when the battery re-charger was being used as this can supply over 14 VDC and the AD8610 is only rated for 13 VDC; at $15 apiece, I don't want to fry any. This means that the regulators are still in the circuit contributing noise. (BTW, the battery->regulator gives +/- 11.2 VDC, whereas the AC supplied power is +/-12.2 VDC, so the supply is affected by the regulator drop-out, while not regulating anything.)
After reading the above comments in this thread, I'm wondering: (1) would the 80 microvolts of noise be lower with other types/designs of regulators; (2) would the noise be significantly lower if I put the battery power in _after_ the regulators; (3) is the 100 -> 80 microvolts of noise lower only because the of the lower power supply (11.2 vs 12.2 VDC); and finally, (4) is it even worth bothering with since 80 microvolts is very low (many commercial amps/preamps have easily 200 or more microvolts of noise)--and besides no audible noise added to my speakers when the circuit is used? Comments?
in most cases, why would a low noise supply be helpful?
I tend to agree with Nelson and I do use a lot of LC, RC or CLC filtering myself.
Apart from the fact that I still haven't grasped the reason why the Sulzer regulator does not oscillate, at least when faced with a transient and that I still am eager to see a schematic for the Jung / Didden amps, I wonder in what applications a low noise regulator is really helpful (maybe a somewhat heretical question in this forum...).
HF-noise can easily be filtered without introducing too much series resistance for damping and easier filtering. Any op-amp or similar circuit will easily suppress lower frequency supply noise.
There are some circuits where I can imagine a benefit, but I am not sure to what degree...
- a clock oscillator
- a single-ended, class A gain stage, i.e. a single common emitter transistor
- the voltage or current reference input of a DAC
- maybe even a DAC, depending on how the current sources are implemented
Any comments?
Eric
I tend to agree with Nelson and I do use a lot of LC, RC or CLC filtering myself.
Apart from the fact that I still haven't grasped the reason why the Sulzer regulator does not oscillate, at least when faced with a transient and that I still am eager to see a schematic for the Jung / Didden amps, I wonder in what applications a low noise regulator is really helpful (maybe a somewhat heretical question in this forum...).
HF-noise can easily be filtered without introducing too much series resistance for damping and easier filtering. Any op-amp or similar circuit will easily suppress lower frequency supply noise.
There are some circuits where I can imagine a benefit, but I am not sure to what degree...
- a clock oscillator
- a single-ended, class A gain stage, i.e. a single common emitter transistor
- the voltage or current reference input of a DAC
- maybe even a DAC, depending on how the current sources are implemented
Any comments?
Eric
capslock,
> HF-noise can easily be filtered ... op-amp or similar circuit will easily suppress lower frequency supply noise
'easily' is a realtive term, as is 'filtered' (by how many dBs?).
In clock oscillators, noise in the power rails ends up at the output as amplitude and phase noise, which is bad.
If you want a really low phase noise oscillator, you want to filter supply noise sufficiently that the oscillator phase noise performance is not influenced by the power supply noise.
20-60dB of filtering may be 'easy', but 120-160dB+ of filtering is not so easy.
Sud
> HF-noise can easily be filtered ... op-amp or similar circuit will easily suppress lower frequency supply noise
'easily' is a realtive term, as is 'filtered' (by how many dBs?).
In clock oscillators, noise in the power rails ends up at the output as amplitude and phase noise, which is bad.
If you want a really low phase noise oscillator, you want to filter supply noise sufficiently that the oscillator phase noise performance is not influenced by the power supply noise.
20-60dB of filtering may be 'easy', but 120-160dB+ of filtering is not so easy.
Sud
I've heard that the newer regulators, like the LT1086 or even the LM317 types, have less noise than the older 78xx/79xx. I decided to check these out.
Looking at spec sheets for the LT1086 and LM317, both give noise levels of 0.003% (or about 30 microvolts noise/V out. The spec sheets for the MC7800/7900 quote 10 microvolts/V out or about 0.001% (one sheet showed 40 microvolts at 15V out, or 0.0003%).
This level of 10 microvolts/V out would mean from my earlier post using 12V regulators that I should have around 120 microvolts of noise from the regulators alone. I measured around 100 for the entire circuit which has 3 op amps in the signal path. (I assume the op amps power supply rejection showhow reduced this?). This measurement appears consistent, or at least the same order of magnitude, with the spec sheets. In any event, the 7800/7900 don't seem that bad.
Looking at spec sheets for the LT1086 and LM317, both give noise levels of 0.003% (or about 30 microvolts noise/V out. The spec sheets for the MC7800/7900 quote 10 microvolts/V out or about 0.001% (one sheet showed 40 microvolts at 15V out, or 0.0003%).
This level of 10 microvolts/V out would mean from my earlier post using 12V regulators that I should have around 120 microvolts of noise from the regulators alone. I measured around 100 for the entire circuit which has 3 op amps in the signal path. (I assume the op amps power supply rejection showhow reduced this?). This measurement appears consistent, or at least the same order of magnitude, with the spec sheets. In any event, the 7800/7900 don't seem that bad.
Three terminal regulators
There is much more to regulators than a quick glance at noise specs. Transient response and rejection of input noise being two other equally important characteristic. The Linear Technolgy 1086 sounds as good as it gets for three terminal parts. I Like to use green leds for the voltage reference. Make sure you have good bypass caps on the input. I like an RC filter before any regulator also as the niose rejection is poor at high frequencies for alll three terminal regulators. Read the data sheet and ap notes very carefully to optimize your circuit. Many regulators like a load resistor on the output to pull a few tens of milliamps out them. Sort of like biasing for class A....
H.H.
http://www.linear-tech.com/prod/datasheet.html?datasheet=233
There is much more to regulators than a quick glance at noise specs. Transient response and rejection of input noise being two other equally important characteristic. The Linear Technolgy 1086 sounds as good as it gets for three terminal parts. I Like to use green leds for the voltage reference. Make sure you have good bypass caps on the input. I like an RC filter before any regulator also as the niose rejection is poor at high frequencies for alll three terminal regulators. Read the data sheet and ap notes very carefully to optimize your circuit. Many regulators like a load resistor on the output to pull a few tens of milliamps out them. Sort of like biasing for class A....
H.H.
http://www.linear-tech.com/prod/datasheet.html?datasheet=233
There was a comparison of the 78/7900 versus the 317/337 done a while ago in Audio Xpress (/ Audio Amateur / Audio Electronics). The big difference was that you could bypass the adjustment terminal on the 317/337s and this made a big difference in noise performance.
Not only do noise specs not tell the full story, they are typical or maximum values, and don't really tell what you will get in a real circuit. You have to plop the parts into a circuit to really know the difference.
A more recent article claimed that any sort of feedback in a voltage regulator is audibly bad. I believe their suggestion was to use a buffered reference instead.
Not only do noise specs not tell the full story, they are typical or maximum values, and don't really tell what you will get in a real circuit. You have to plop the parts into a circuit to really know the difference.
A more recent article claimed that any sort of feedback in a voltage regulator is audibly bad. I believe their suggestion was to use a buffered reference instead.
So many questions
Here's a few potted thoughts, evidence is littered around the forum in some of my other posts - you'll have to seek them out.
Adjustable 3 terminal reg's offer lower noise and much improved line rejection than fixed reg's, by a significant and audible margin. As stated above an LT1086 is as good as it gets in 3-terminal reg's.
The spec sheets don't show this, the difference though is measurably significant.
TL431's and bandgap references in general are noisy unless filtered, buried zeners (e.g. LM129 / 329) are much quieter.
All circuits have finite PSRR, and particularly with op-amps they deteriorate rapidly from the headline figures at LF. In most PSU's PSRR and PSU line rejection are deteriorating simultaneously, with detrimental sonic effect.
Filters (LC / RC etc) are effective at HF, but worsen regulation and supply impedance. They're useful for circuits with steady current drain, but a good low impedance supply will be better, particularly if these characteristics can be maintained over a wide bandwidth.
An LM317 for example has about 80dB of line rejection at LF, but only about 40dB or so at 100k. The Sulzer circuit can offer almost 120dB at LF, and 90dB @ 100k.
Most people won't notice the difference though, owing to non-ideal implementation (hence many of the comments above, I suspect).
Take a look at the 'Imperfect Kelvin Sensing' post on this thread : -
http://www.diyaudio.com/forums/showthread.php?threadid=985&perpage=15&pagenumber=3
rljones - I suspect that this is the primary reason why adding the battery in place of the mains supply reduces noise - I have posts elsewhere on the forum showing the Sulzer reg. to have lower noise than a battery, when current is being drawn .
They look awfully quiet until you ask them to work, wherupon they produce all sorts of non-corellated noise of significant levels.
Andy.
Here's a few potted thoughts, evidence is littered around the forum in some of my other posts - you'll have to seek them out.
Adjustable 3 terminal reg's offer lower noise and much improved line rejection than fixed reg's, by a significant and audible margin. As stated above an LT1086 is as good as it gets in 3-terminal reg's.
The spec sheets don't show this, the difference though is measurably significant.
TL431's and bandgap references in general are noisy unless filtered, buried zeners (e.g. LM129 / 329) are much quieter.
All circuits have finite PSRR, and particularly with op-amps they deteriorate rapidly from the headline figures at LF. In most PSU's PSRR and PSU line rejection are deteriorating simultaneously, with detrimental sonic effect.
Filters (LC / RC etc) are effective at HF, but worsen regulation and supply impedance. They're useful for circuits with steady current drain, but a good low impedance supply will be better, particularly if these characteristics can be maintained over a wide bandwidth.
An LM317 for example has about 80dB of line rejection at LF, but only about 40dB or so at 100k. The Sulzer circuit can offer almost 120dB at LF, and 90dB @ 100k.
Most people won't notice the difference though, owing to non-ideal implementation (hence many of the comments above, I suspect).
Take a look at the 'Imperfect Kelvin Sensing' post on this thread : -
http://www.diyaudio.com/forums/showthread.php?threadid=985&perpage=15&pagenumber=3
rljones - I suspect that this is the primary reason why adding the battery in place of the mains supply reduces noise - I have posts elsewhere on the forum showing the Sulzer reg. to have lower noise than a battery, when current is being drawn .
They look awfully quiet until you ask them to work, wherupon they produce all sorts of non-corellated noise of significant levels.
Andy.
- Status
- Not open for further replies.