That would be trafo_secondary -> rectifier -> resistor -> reservoir capacitor.
Another way to skin the cat is to move the resistor upstream one position: trafo_secondary -> resistor -> rectifier -> reservoir capacitor.
You could even move it further upstream, by installing the resistor in series with the transformer primary (don't forget to scale the impedance by the square of the turns ratio), and thereby get less HF content and inrush current limiting ("soft start") as well.
I think that the RC works best in this case when the R is right before the C, as the filtering action is after the diodes and any junk coming out of them. But, I have not tried the other R locations in a back to back test.
I do seem to remember finding out that any R or L in the primary affected the bass and dynamics, but that was some time ago. Some things that look good on paper, don't always sound good. Look at all the aftermarket power cords for sale, I still don't know for sure how they work to improve the sound, but some do. I wish they didn't.
I do seem to remember finding out that any R or L in the primary affected the bass and dynamics, but that was some time ago. Some things that look good on paper, don't always sound good. Look at all the aftermarket power cords for sale, I still don't know for sure how they work to improve the sound, but some do. I wish they didn't.
I don't have any of the nice testing equpiment you men in this thread have but my experience here in East London has consistently been in accordance with Rick; the lower the impedance before the transformer the better the sound from the loudspeakers; bass, dynamics, clarity, focus, imaging. And cleaning and minimising / deleting contacts helps a lot, especially with imaging and reducing harshness.
I've tried mains cable to the point that I now have a 50mm² pair of mains conductors going from the supply service head at the house entry point, to the transformers, or as near as I can get, and then use the thickest mains cable I can from the 50mm² terminus. I do it because for me with the stuff I'm using so far, it works, sounds better, I spent perhaps 15 testing and improving, not because I copied the idea from a forum, I started with big cables in the early 1990's after reading about it in about in a Hi-Fi magazine in 1984 where someone had the supply to their property replaced with larger gauge cable and had good results. A friend had it done in 1994 as part of a necessary modernisation, and we had impressive slam and weight there. His mains consumer terminal was the other side of the wall to the Hi-Fi plug sockets.
Last edited:
IanAS, I agree with everything you said. I think that the greater the current capability on the primary side is best. Ever put a scope across the secondary of a transformer with a cap input supply? After about 47uf of C for the main filter cap the waveform starts to flat top. All of our cap input supplies are clipping the power transformer, this can't be a good thing, lots of harmonics produced. Perhaps lower R and Z on the input lessens the flat topping, and gives the harmonics a better path back to the mains transformer outside. Just my idea.
But here we were talking about a regulated supply. That's a pretty good insulator between the load and anything the transformer does.
Jan
Noting that the super regulator has no film cap at its output (and Walt Jung wrote that he doesn't like the idea of such film caps), I would like to ask what is the purpose of such a cap for any regulator, in terms of:
(1) output impedance
(2) stability
Normally, I would expect that the film cap provides lower impedance for higher frequencies and demanding transients. Then, it should be advantageous if placed close to the load, and actually useless if not.
In terms of stability, I don't see how it could offer any benefits. But I also have limited experience. I am actually asking since Mike Sulzer noted that these caps are needed for stability under certain conditions in his design, but I am having a hard time understanding what he meant. Playing a bit with Spice simulations, I must admit that it usually worsens the phase margin (hoping that I do everything correct, of course).
So, seeing that the super regulator excels in output impedance, and reading that Walt Jung suggested that a film cap is more likely to bring stability problems, rather than cure them, I really don't understand the usefulness of installing such a cap at any regulator's output. Of course, I am talking about good regulator designs that should not depend on such a cap for low HF impedance - even Sulzer mentioned something about stability, nothing about improving impedance.
Opinions?
(1) output impedance
(2) stability
Normally, I would expect that the film cap provides lower impedance for higher frequencies and demanding transients. Then, it should be advantageous if placed close to the load, and actually useless if not.
In terms of stability, I don't see how it could offer any benefits. But I also have limited experience. I am actually asking since Mike Sulzer noted that these caps are needed for stability under certain conditions in his design, but I am having a hard time understanding what he meant. Playing a bit with Spice simulations, I must admit that it usually worsens the phase margin (hoping that I do everything correct, of course).
So, seeing that the super regulator excels in output impedance, and reading that Walt Jung suggested that a film cap is more likely to bring stability problems, rather than cure them, I really don't understand the usefulness of installing such a cap at any regulator's output. Of course, I am talking about good regulator designs that should not depend on such a cap for low HF impedance - even Sulzer mentioned something about stability, nothing about improving impedance.
Opinions?
@audiostrat: A very low ESR capacitor results in an output circuit with high-Q*. You get the desirable attenuation of high frequency signals, but there is a sharp decrease in the phase margin which will cause the regulator to oscillate.
A higher ESR capacitor on the output lowers the Q of the control loop. By examining the gain and phase of the output of the regulator you can optimize the value of capacitor so that the circuit is correctly damped.
Christophe Basso is far more expert than this layman:
http://cbasso.pagesperso-orange.fr/Downloads/Papers/Phase margin and quality factor.pdf
A higher ESR capacitor on the output lowers the Q of the control loop. By examining the gain and phase of the output of the regulator you can optimize the value of capacitor so that the circuit is correctly damped.
Christophe Basso is far more expert than this layman:
http://cbasso.pagesperso-orange.fr/Downloads/Papers/Phase margin and quality factor.pdf
Noting that the super regulator has no film cap at its output (and Walt Jung wrote that he doesn't like the idea of such film caps), I would like to ask what is the purpose of such a cap for any regulator, in terms of:
(1) output impedance
(2) stability
Normally, I would expect that the film cap provides lower impedance for higher frequencies and demanding transients. Then, it should be advantageous if placed close to the load, and actually useless if not.
In terms of stability, I don't see how it could offer any benefits. But I also have limited experience. I am actually asking since Mike Sulzer noted that these caps are needed for stability under certain conditions in his design, but I am having a hard time understanding what he meant. Playing a bit with Spice simulations, I must admit that it usually worsens the phase margin (hoping that I do everything correct, of course).
So, seeing that the super regulator excels in output impedance, and reading that Walt Jung suggested that a film cap is more likely to bring stability problems, rather than cure them, I really don't understand the usefulness of installing such a cap at any regulator's output. Of course, I am talking about good regulator designs that should not depend on such a cap for low HF impedance - even Sulzer mentioned something about stability, nothing about improving impedance.
Opinions?
Jack is absolutely right. But I'd like to note that you DO need *some* (lossy) capacitance to keep it stable. It is very hard to design a reg (which of course is a nfb loop from output to the pass device) without some output capacity for stability.
In fact, there's a lot of competition between integrated reg makers to get to the smallest required capacity because that saves cost & PCB real estate.
Some products out there get away with much less than 1uF at the output, but as a diy-er you don't have that constraint and a 10uF electrolytic is pretty small and cheap anyway. Electrolytics by nature have some ESR as well so they work perfect here.
But the output impedance of the reg is much less than the output cap so the bulk of the load current comes from the reg and not the output cap.
Jan
Jack, many thanks for your response and the link.
Jan, thank you. Of course and I will use an output cap (electrolytic) to provide a safety stability margin, and the original Sulzer circuit used a 47uF tantalum. I will use a common electrolytic type, maybe 68uF-220uF based on ESR.
In that manner, I understand that an additional film cap in parallel with the electrolytic capacitor will make things worse, in terms of stability. Given its low ESR, of course. So I conclude to the following, please correct me:
A film cap in parallel with the output cap will not aid in stabilising the circuit, it will actually endanger it, given the new phase margin. Following Jan's comment that the regulator will provide the current needed, I hardly see any reason why a film cap should be used across the output of the regulator. Maybe it could have some use directly at the load point, to improve transient response.
This means that there is hardly any reason a film cap (parallel to the main output cap) has any use at all across the output of a regulator. So I am willing to get rid of it in my Sulzer regulator pcb.
All this emerged from Mike Sulzer's obscure comment that such a film cap could aid in stability under certain conditions, which seems to be in contrast with usual conditions. Any idea why he could have said this?
Jan, thank you. Of course and I will use an output cap (electrolytic) to provide a safety stability margin, and the original Sulzer circuit used a 47uF tantalum. I will use a common electrolytic type, maybe 68uF-220uF based on ESR.
In that manner, I understand that an additional film cap in parallel with the electrolytic capacitor will make things worse, in terms of stability. Given its low ESR, of course. So I conclude to the following, please correct me:
A film cap in parallel with the output cap will not aid in stabilising the circuit, it will actually endanger it, given the new phase margin. Following Jan's comment that the regulator will provide the current needed, I hardly see any reason why a film cap should be used across the output of the regulator. Maybe it could have some use directly at the load point, to improve transient response.
This means that there is hardly any reason a film cap (parallel to the main output cap) has any use at all across the output of a regulator. So I am willing to get rid of it in my Sulzer regulator pcb.
All this emerged from Mike Sulzer's obscure comment that such a film cap could aid in stability under certain conditions, which seems to be in contrast with usual conditions. Any idea why he could have said this?
Jack, many thanks for your response and the link.
Jan, thank you. Of course and I will use an output cap (electrolytic) to provide a safety stability margin, and the original Sulzer circuit used a 47uF tantalum. I will use a common electrolytic type, maybe 68uF-220uF based on ESR.
In that manner, I understand that an additional film cap in parallel with the electrolytic capacitor will make things worse, in terms of stability. Given its low ESR, of course. So I conclude to the following, please correct me:
A film cap in parallel with the output cap will not aid in stabilising the circuit, it will actually endanger it, given the new phase margin. Following Jan's comment that the regulator will provide the current needed, I hardly see any reason why a film cap should be used across the output of the regulator. Maybe it could have some use directly at the load point, to improve transient response.
This means that there is hardly any reason a film cap (parallel to the main output cap) has any use at all across the output of a regulator. So I am willing to get rid of it in my Sulzer regulator pcb.
All this emerged from Mike Sulzer's obscure comment that such a film cap could aid in stability under certain conditions, which seems to be in contrast with usual conditions. Any idea why he could have said this?
Fully agree with you. You can use a film cap at the load because it is isolated from the reg itself by the wire inductance and resistance, and it might well have a positive effect.
Not sure where Mike's comment that you quoted last comes from.
Jan
Thanks for the response Jan! I will eventually go that route.
Yes, he mentioned it in the first of his two articles, not giving any further details. I assume such caps are used to keep HF garbage out where common electrolytics or tantalums fail to do so - and maybe HF garbage could cause oscillations for a high bandwith opamp, thus compromising the regulator's performance. I really don't know much on that.
Since it is proposed that the super regulator can be built with a NE5534, I would like to ask something about its bandwidth given the layout.
Since NE5534 comes uncompensated, the total stray capacitance from pin 5 to 8 will actually determine the frequency response of the loop. I played a bit with Spice for a Sulzer circuit that uses the NE5534, and when installed a 0.1pF capacitance as a compensation capacitor, frequency response went up to 100MHz accompanied by an outrageously dangerous phase margin - 10 degrees or less. When changed this value to 1pF, everything went normal: UGBW went a bit above 10MHz (which is what is specified for the NE5534), and phase margin was above 40 degrees for many output cap combinations, up to 2A DC output.
So, can one safely assume that stray capacitance between pins 5 and 8 will be at least 1pF, or something like that? Actually I don't know a practical way of calculating such strays, but simulation shows me that maybe this is an important factor when using externally compensated opamps.
Since NE5534 comes uncompensated, the total stray capacitance from pin 5 to 8 will actually determine the frequency response of the loop. I played a bit with Spice for a Sulzer circuit that uses the NE5534, and when installed a 0.1pF capacitance as a compensation capacitor, frequency response went up to 100MHz accompanied by an outrageously dangerous phase margin - 10 degrees or less. When changed this value to 1pF, everything went normal: UGBW went a bit above 10MHz (which is what is specified for the NE5534), and phase margin was above 40 degrees for many output cap combinations, up to 2A DC output.
So, can one safely assume that stray capacitance between pins 5 and 8 will be at least 1pF, or something like that? Actually I don't know a practical way of calculating such strays, but simulation shows me that maybe this is an important factor when using externally compensated opamps.
The NE5534 is stable without compensation for gains of 3 or more. In a typical superreg where the Vout is divided by two for the -input, that's basically a gain of two circuit.
I don't know whether this is influenced by the current source into the pass device, might be.
In your sim, can you make it gain of 3, by using say a 4V ref (battery) and a Vout of 12V?
How does it look then?
Do you have a cap across the top divider resistor - what happens when you remove the cap?
Anyway, a small 22pF cap on the underside of the PCB doesn't sound like an unsurmountable problem for a diy-er
Jan
I don't know whether this is influenced by the current source into the pass device, might be.
In your sim, can you make it gain of 3, by using say a 4V ref (battery) and a Vout of 12V?
How does it look then?
Do you have a cap across the top divider resistor - what happens when you remove the cap?
Anyway, a small 22pF cap on the underside of the PCB doesn't sound like an unsurmountable problem for a diy-er
Jan
Yes, I have the usual 4.7uF cap used in the Sulzer circuits. I use a 6.83V reference for a 9V output, so gain is not close to 3 at all.
In fact, the schematic I use in Pspice is the following:
I will see what I get for what you propose, when I have a bit more time.
I remember Mike Sulzer suggesting the use of an uncompensated opamp in order to obtain more feedback in terms of getting more bandwidth. Then, the output capacitor will aid in stabilizing the circuit. So I thought that it would be good if I wouldn't have to limit the opamp's BW by inserting a compensation cap. It turns out that things would be manageable if 1-2pF would exist as stray capacitance. In theory, at least.
As you say, I can fit a small cap (but would be obviously happy to avoid it: less parts + the above paragraph), but wouldn't this compromise the output impedance of the whole regulator, as Sulzer suggested?
In fact, the schematic I use in Pspice is the following:
I will see what I get for what you propose, when I have a bit more time.
I remember Mike Sulzer suggesting the use of an uncompensated opamp in order to obtain more feedback in terms of getting more bandwidth. Then, the output capacitor will aid in stabilizing the circuit. So I thought that it would be good if I wouldn't have to limit the opamp's BW by inserting a compensation cap. It turns out that things would be manageable if 1-2pF would exist as stray capacitance. In theory, at least.
As you say, I can fit a small cap (but would be obviously happy to avoid it: less parts + the above paragraph), but wouldn't this compromise the output impedance of the whole regulator, as Sulzer suggested?
Yes, it is low. Comes from a Panasonic FM datasheet, for a 220uF - 25V cap, plus some track resistance. Actually, this cap specifies for a 56mΩ impedance at 100kHz, which means that if it has approximately 20nH of inductance, then it has around 55mΩ at that frequency.
Anyway, as I said before, I tried to push the regulator to the limit. Maybe an FC cap with bigger ESR might be better, but at high frequencies I suppose it will too have quite a small ESR. I am experimenting with that value anyway, maybe I will test a 68uF for the output. Still with such an R3, phase margin did not fall below 40 degrees.
Glad to hear you had no issues with a NE5534 - it is certainly worth something. Maybe I go with no caps, and if I discover instability, add some pF's.
Is the 120uF value of the two caps used for feedback + output absolutely critical? Morgan Jones suggested that the former is made equal to the Thevenin resistance it sees.
I also remember Sulzer suggesting that for higher currents, the output capacitor should raise in value and fall in impedance. I also recall posts in this thread where we talked about a 2A output current, but no change in caps. Why would higher currents require the adjustment of the loop response? I assume that what changes with load current is the Vbe voltage of the pass transistor, and its beta. How would this affect the operation?
Anyway, as I said before, I tried to push the regulator to the limit. Maybe an FC cap with bigger ESR might be better, but at high frequencies I suppose it will too have quite a small ESR. I am experimenting with that value anyway, maybe I will test a 68uF for the output. Still with such an R3, phase margin did not fall below 40 degrees.
Glad to hear you had no issues with a NE5534 - it is certainly worth something.
Maybe I go with no caps, and if I discover instability, add some pF's.Is the 120uF value of the two caps used for feedback + output absolutely critical? Morgan Jones suggested that the former is made equal to the Thevenin resistance it sees.
I also remember Sulzer suggesting that for higher currents, the output capacitor should raise in value and fall in impedance. I also recall posts in this thread where we talked about a 2A output current, but no change in caps. Why would higher currents require the adjustment of the loop response? I assume that what changes with load current is the Vbe voltage of the pass transistor, and its beta. How would this affect the operation?
The cap across the feedback resistor could be smaller, then again using all the same value saves on parts orders with no adverse effect .
I can understand that loop gain and margins change with load current, but I'm not sure how that would work out on the output cap. Any reasonable output cap is more than required anyway so there is some leeway in that.
Jan
.I can understand that loop gain and margins change with load current, but I'm not sure how that would work out on the output cap. Any reasonable output cap is more than required anyway so there is some leeway in that.
Jan