Hi Johan,
In my project I decided to go for an A Class SE or PP amp,which will end up with continious current flow,so designing for a swinging choke will not be necessesarily.
Even if there'll be some peak fluctuations of the voltage/current, they will be averaged and equalized by the following electrolytic cap.
And having a low DCR PSU (or low impedance supply if you want) is a very important parametar and condition which must be fullfilled(especially for a high power amp),and will lead to a fast transient and motor-boating-free PSU.
Regards,
Yugovitz
Hi Yugo,
I saw some excellent design procedures in pages quoted some posts ago - was it from Morgan Jones? As said in the past, I did not so much use formulae; worked from graphs. Then came P.Spice and that was even easier. (I am lazy with formulae. Someone once said lazy people make good engineers. Then I was a very good engineer.)
I have tried measuring but it is a little difficult at >1KV and where capacitances can be an influence. I then simulated. I am unable to give a circuit here, but the description should suffice; perhaps of interest. (It is an L.C filter - followed by other stuff which is not of consequence here.)
My power supply voltage is 620 Vrms, transformer winding resistance=17 ohm, bridge rectification. The choke was 2,2H, resistance 21 ohm. I guesstimated an interwinding capacitance of 300pF. The load resistor (representing the amplifier) was 2.4K, presenting a load current of some 250 mA. (This is standing current, zero signal.)
The main C was 100uF, in series with a resistor of 0,1 ohm (internal resistance). The R.C over the choke was R1, C1. Then I also inserted another series R.C in parallel with the 100uF (+ internal resistance), called R2, C2.
1. Without R1, C1 and R2, C2, there was a spike after the 6th 50 Hz cycle after switch-on, of over 1 KV (as the main 100 uF charged up sufficiently to have the choke "turn-off" for the first time), and an oscillatory response at every subsequent choke "turn-off" of about 4 KHz.
2. Inserting R1=10K, C1=22nF, stopped the oscillations but did little to contain the first peak. (Voltage over C1 went to over 1 KV.) After several sets of values I found R1=39K, C1=220nF to suppress the peak to below the max. values of the half-wave voltage cycles after the rectifier. This also rendered a practical value of 2,5W for R1 and <400V for C1. These values are of course dependant on the choke parameters, but the above appeared good for a start.
3. The MJ(?) treatise indicated the use of R2, C2. (I must admit I have never seen that in practical circuits.) With R2=100 ohm and C2=100uF, and without R1, C1, the first "turn-off" peak was somewhat reduced, but choke turn-offs were still oscillatory. With again R1=10K and C1=22nF, the first peak was reduced and all oscillations stopped as before. Then with R1=39K and C1=220nF (now plus R2, C2), the first peak was just about invisible and the entire response very acceptable. Dissipation in R1 was now about 2W and V(C2) about 320V.
I though these analysis might give a practical insight into the tendency of things. R2, C2 has a good effect, but might be a somewhat expensive addition at over 500V - one more or less needs 2 capacitors in series there. (I believe it would be discussed in the quoted design procedure.)
Thus you can see that R1, C1 is essential to keep choke turn-off spikes under control, while R2, C2 contributes to better turn-on behaviour.
This was a rather qualitative discussion. I myself put such filters on Spice just to make sure. I don't however use L-input that much, but certainly recognises its attributes.
Regards.
I saw some excellent design procedures in pages quoted some posts ago - was it from Morgan Jones? As said in the past, I did not so much use formulae; worked from graphs. Then came P.Spice and that was even easier. (I am lazy with formulae. Someone once said lazy people make good engineers. Then I was a very good engineer.)
I have tried measuring but it is a little difficult at >1KV and where capacitances can be an influence. I then simulated. I am unable to give a circuit here, but the description should suffice; perhaps of interest. (It is an L.C filter - followed by other stuff which is not of consequence here.)
My power supply voltage is 620 Vrms, transformer winding resistance=17 ohm, bridge rectification. The choke was 2,2H, resistance 21 ohm. I guesstimated an interwinding capacitance of 300pF. The load resistor (representing the amplifier) was 2.4K, presenting a load current of some 250 mA. (This is standing current, zero signal.)
The main C was 100uF, in series with a resistor of 0,1 ohm (internal resistance). The R.C over the choke was R1, C1. Then I also inserted another series R.C in parallel with the 100uF (+ internal resistance), called R2, C2.
1. Without R1, C1 and R2, C2, there was a spike after the 6th 50 Hz cycle after switch-on, of over 1 KV (as the main 100 uF charged up sufficiently to have the choke "turn-off" for the first time), and an oscillatory response at every subsequent choke "turn-off" of about 4 KHz.
2. Inserting R1=10K, C1=22nF, stopped the oscillations but did little to contain the first peak. (Voltage over C1 went to over 1 KV.) After several sets of values I found R1=39K, C1=220nF to suppress the peak to below the max. values of the half-wave voltage cycles after the rectifier. This also rendered a practical value of 2,5W for R1 and <400V for C1. These values are of course dependant on the choke parameters, but the above appeared good for a start.
3. The MJ(?) treatise indicated the use of R2, C2. (I must admit I have never seen that in practical circuits.) With R2=100 ohm and C2=100uF, and without R1, C1, the first "turn-off" peak was somewhat reduced, but choke turn-offs were still oscillatory. With again R1=10K and C1=22nF, the first peak was reduced and all oscillations stopped as before. Then with R1=39K and C1=220nF (now plus R2, C2), the first peak was just about invisible and the entire response very acceptable. Dissipation in R1 was now about 2W and V(C2) about 320V.
I though these analysis might give a practical insight into the tendency of things. R2, C2 has a good effect, but might be a somewhat expensive addition at over 500V - one more or less needs 2 capacitors in series there. (I believe it would be discussed in the quoted design procedure.)
Thus you can see that R1, C1 is essential to keep choke turn-off spikes under control, while R2, C2 contributes to better turn-on behaviour.
This was a rather qualitative discussion. I myself put such filters on Spice just to make sure. I don't however use L-input that much, but certainly recognises its attributes.
Regards.
Hi Johan,
Thank you very,very much for your comprehensive work and testing.
I was always delighted with some members on this forum,including you of course,with your spirit,enthousiasm,never lasting energy and "childish" curiosity to help others,solve problems and to learn at the same time!
Johan,I guess that the first oscilation of about 4Khz w/o snubbers,was the intrinsic,self oscilation of the circuit with all parasitic series and parallel L's and C's.Adding R1C1 snubber did raise the impedance and suppress the first peak.Furthermore after changing the snubber with higher values increased Z of the L/CR element even more.In this case the major role plays R rather than C ,IMHO.
I only regret that you didn't check first snubber with 220 nF front/back of the L input instead of RC pair to examine if there is any evident difference.If you have still the same set up on your bench please Johan try again this testing.
Also I suggest you to try and see on the scope what will happen if you substitute C1 with 1.15 uF(maybe you have to trim this value).This is VERY important because you'll get this time a resonance of the LC parallel circuit and raise upmost the Z.
Then you can play again with different values of resistors in series with the C and check it out what will happen.With this values you'll get around 90-100K impedance of the LC circuit.
Regarding R2C2 pair I must admit that this was something new for me and never seen before intentionally made.But after a couple minuits I started to recall a discusion made 5-6 years ago about pro/cons of paralleling big elcos with small PP's in PSU.It seamed that all "clear highs" gained with this tweek were result of simple oscilation between components.So doing this way with big 100 ohm we simply raise ESR of the cap and also damp the peak.Probably good on the scope but with diminishing return and not very practical for me.
Anyway Johan thank you again for good company and of course constructive discussion!
Cheers,
Yugovitz
Thank you very,very much for your comprehensive work and testing.
I was always delighted with some members on this forum,including you of course,with your spirit,enthousiasm,never lasting energy and "childish" curiosity to help others,solve problems and to learn at the same time!
Johan,I guess that the first oscilation of about 4Khz w/o snubbers,was the intrinsic,self oscilation of the circuit with all parasitic series and parallel L's and C's.Adding R1C1 snubber did raise the impedance and suppress the first peak.Furthermore after changing the snubber with higher values increased Z of the L/CR element even more.In this case the major role plays R rather than C ,IMHO.
I only regret that you didn't check first snubber with 220 nF front/back of the L input instead of RC pair to examine if there is any evident difference.If you have still the same set up on your bench please Johan try again this testing.
Also I suggest you to try and see on the scope what will happen if you substitute C1 with 1.15 uF(maybe you have to trim this value).This is VERY important because you'll get this time a resonance of the LC parallel circuit and raise upmost the Z.
Then you can play again with different values of resistors in series with the C and check it out what will happen.With this values you'll get around 90-100K impedance of the LC circuit.
Regarding R2C2 pair I must admit that this was something new for me and never seen before intentionally made.But after a couple minuits I started to recall a discusion made 5-6 years ago about pro/cons of paralleling big elcos with small PP's in PSU.It seamed that all "clear highs" gained with this tweek were result of simple oscilation between components.So doing this way with big 100 ohm we simply raise ESR of the cap and also damp the peak.Probably good on the scope but with diminishing return and not very practical for me.
Anyway Johan thank you again for good company and of course constructive discussion!
Cheers,
Yugovitz
Hi Yogo,
I did that - but one must be careful of a misconception. The purpose of C1 is first and foremost to keep the power supply output impedance low. Any design with L-resonance after that comes in later. In that sense the value of 1,15uF is not at all good. Nevertheless I did that; the output remained an ac between 0 and about 800V - no real smoothing by C1 under my load conditions of 250mA. Also the initial peak where the "switching action" of L comes in (after the 5th cycle) was now in excess of 2 KV. (Resonance but in the wrong way.)
With a snubber of 220nF at the input side of L and C1 back to 100uF, the first peak rose to 1,5KV. I did not expect it do do much else than resonate there with L and make matters worse. The R.C over L is really only a damper for the first peak and to a lesser extent for subsequent ones under greatly varying loads. Remember that at that moment one has an open choke with a lot of current energy in it. It works rather like a motor car coil, where in conjunction with some C one gets 10s of KV for the spark. In this circuit the opposite is required. One also cannot only use the R of 39K. The series C is there to cancel some of the inductive current and assist in suppressing the peak.
Keep in mind that this peak is only there on turning the supply on. It really has no consequense on amplifier or power supply performance afterwards. The only problem might be that under certain circumstances this peak may be high enough to damage the choke ot the rectifier. But that is a real enough problem and must be addressed.
Also, the R2.C2 pair alters the initial charging current, flattening it out somewhat in value and phase. C2 charges in a delayed fashion to C1 until it is fully charged. As the power supply output impedance go below about 8 ohm after 100 Hz (the reactance of C1), C2 in series with 100 ohm no longer contributes a lot.
Yogo, I must presume that my values, especially the rather high h.t. of 600V that I need at a fair current, perhaps differ a lot from yours. All these arguments are very dependant on load current. I would suggest that you put a circuit diagram here of what you intend using, with component values and also annotated R1, R2....C1, C2... so that we can communicate better. If you do not have Spice I could analyse your case and try to optimise, and other members would be able to follow and comment better.
Regards.
I did that - but one must be careful of a misconception. The purpose of C1 is first and foremost to keep the power supply output impedance low. Any design with L-resonance after that comes in later. In that sense the value of 1,15uF is not at all good. Nevertheless I did that; the output remained an ac between 0 and about 800V - no real smoothing by C1 under my load conditions of 250mA. Also the initial peak where the "switching action" of L comes in (after the 5th cycle) was now in excess of 2 KV. (Resonance but in the wrong way.)
With a snubber of 220nF at the input side of L and C1 back to 100uF, the first peak rose to 1,5KV. I did not expect it do do much else than resonate there with L and make matters worse. The R.C over L is really only a damper for the first peak and to a lesser extent for subsequent ones under greatly varying loads. Remember that at that moment one has an open choke with a lot of current energy in it. It works rather like a motor car coil, where in conjunction with some C one gets 10s of KV for the spark. In this circuit the opposite is required. One also cannot only use the R of 39K. The series C is there to cancel some of the inductive current and assist in suppressing the peak.
Keep in mind that this peak is only there on turning the supply on. It really has no consequense on amplifier or power supply performance afterwards. The only problem might be that under certain circumstances this peak may be high enough to damage the choke ot the rectifier. But that is a real enough problem and must be addressed.
Also, the R2.C2 pair alters the initial charging current, flattening it out somewhat in value and phase. C2 charges in a delayed fashion to C1 until it is fully charged. As the power supply output impedance go below about 8 ohm after 100 Hz (the reactance of C1), C2 in series with 100 ohm no longer contributes a lot.
Yogo, I must presume that my values, especially the rather high h.t. of 600V that I need at a fair current, perhaps differ a lot from yours. All these arguments are very dependant on load current. I would suggest that you put a circuit diagram here of what you intend using, with component values and also annotated R1, R2....C1, C2... so that we can communicate better. If you do not have Spice I could analyse your case and try to optimise, and other members would be able to follow and comment better.
Regards.
Hi Johan,
Thank you again for your excellent and fast reply.You save me a lot of time and nerves!!!!
As far as C1(the snubber cap)is considered,I am little dissapointed with the results you got with your mesurements.I hope I've done right my windows calculator when I wrote 1,15 uF for the snubber cap,but it simply doesn't turn right in practice!!!
Magnetics for me were always nasty and worst thing I have to learn and very,very unpredictable!
I see that you made all possible combinations with snubbers and your results and findings don't even correlate with Tweekers' statement that (as per MJ),220nF pair in front/back of the choke showed better mesurement than the RC snubber in // with choke?!! That is really interesting again!
Simply Johan,as you write,your high voltage-high current based PSU is PROBABLY IMHO responsible for such different results than theoretical,someone will expect.
My idea was to learn and find the right and shortest way how to design resonant PSU and no one particular schematcs was on my mind,except that it will be Class A SE Amp or preamp with moderate currents and voltages.
Of course Johan I will ASAP post a certain schematics of an amp's PSU that is of my interest to have to discuss on smth. more precise.
Kindest regards,
Yugovitz
P.S. I don't see unfortunatelly more members interested on this topic willing to participate!!
Thank you again for your excellent and fast reply.You save me a lot of time and nerves!!!!
As far as C1(the snubber cap)is considered,I am little dissapointed with the results you got with your mesurements.I hope I've done right my windows calculator when I wrote 1,15 uF for the snubber cap,but it simply doesn't turn right in practice!!!
Magnetics for me were always nasty and worst thing I have to learn and very,very unpredictable!
I see that you made all possible combinations with snubbers and your results and findings don't even correlate with Tweekers' statement that (as per MJ),220nF pair in front/back of the choke showed better mesurement than the RC snubber in // with choke?!! That is really interesting again!
Simply Johan,as you write,your high voltage-high current based PSU is PROBABLY IMHO responsible for such different results than theoretical,someone will expect.
My idea was to learn and find the right and shortest way how to design resonant PSU and no one particular schematcs was on my mind,except that it will be Class A SE Amp or preamp with moderate currents and voltages.
Of course Johan I will ASAP post a certain schematics of an amp's PSU that is of my interest to have to discuss on smth. more precise.
Kindest regards,
Yugovitz
P.S. I don't see unfortunatelly more members interested on this topic willing to participate!!
Yogo,
My pleasure - others now long gone have helped me decades ago, and it is my duty to maintain that legacy. Also, my results are not physical measurements, they are from simulating. For measuring these effects one needs a fast memory (digital) scope with suitable probe, neither of which I have. I can only say that what I can see on an ordinary scope (e.g. the onset of the d.c. at switch-on) seems to correspond with the simulation, so I expect the rest not to be too far off. A lot will depend on the characteristics of the choke: inductance at a particular d.c., internal capacitance, .... those I could only guess, for my particular component.
I can perhaps add that I have experimented (in practice) with a low value cap at the input to the choke, to raise the d.c. slightly, but that does give peaks and poor regulation, and is not working for me. Then, a snubber after the choke: If I understand correctly, that will be in parallel with my C1 mentioned earlier? A 220nF in parallel with 100uF would not make the slightest difference; it would simply be 100.22uF. (By that I do not mean to exclude similar caps fitted inside an amplifier to keep power supply impedance low because electrolytics are not perfect - but that is not relevant in basic analysis.)
And if all this begins to sound rather perplexing, not to worry. It is perhaps against proper engineering practice, but one does not always design a power supply in greatest detail on paper. As said, some of the parameters are difficult to measure. Rather one designs as far as is practically useful, and judge from the practical result whether further development is necessary.
But feel welcome to submit a circuit and one can at least try to keep you out of serious trouble.
Regards.
My pleasure - others now long gone have helped me decades ago, and it is my duty to maintain that legacy. Also, my results are not physical measurements, they are from simulating. For measuring these effects one needs a fast memory (digital) scope with suitable probe, neither of which I have. I can only say that what I can see on an ordinary scope (e.g. the onset of the d.c. at switch-on) seems to correspond with the simulation, so I expect the rest not to be too far off. A lot will depend on the characteristics of the choke: inductance at a particular d.c., internal capacitance, .... those I could only guess, for my particular component.
I can perhaps add that I have experimented (in practice) with a low value cap at the input to the choke, to raise the d.c. slightly, but that does give peaks and poor regulation, and is not working for me. Then, a snubber after the choke: If I understand correctly, that will be in parallel with my C1 mentioned earlier? A 220nF in parallel with 100uF would not make the slightest difference; it would simply be 100.22uF. (By that I do not mean to exclude similar caps fitted inside an amplifier to keep power supply impedance low because electrolytics are not perfect - but that is not relevant in basic analysis.)
And if all this begins to sound rather perplexing, not to worry. It is perhaps against proper engineering practice, but one does not always design a power supply in greatest detail on paper. As said, some of the parameters are difficult to measure. Rather one designs as far as is practically useful, and judge from the practical result whether further development is necessary.
But feel welcome to submit a circuit and one can at least try to keep you out of serious trouble.
Regards.
Hi Johan,
Today while surfing on the net I found this application guide which I think might be usefull.
http://www.cde.com/catalogs/igbtAPPguide.pdf
Regards,
Yugovitz
Today while surfing on the net I found this application guide which I think might be usefull.
http://www.cde.com/catalogs/igbtAPPguide.pdf
Regards,
Yugovitz
....and regarding 220 nF combo caps,the position and their effect should be like this:
http://www.diyaudio.com/forums/showthread.php?postid=1178486#post1178486
...and like this:
http://www.diyaudio.com/forums/showthread.php?postid=1178327#post1178327
Regards,
Yugovitz
http://www.diyaudio.com/forums/showthread.php?postid=1178486#post1178486
...and like this:
http://www.diyaudio.com/forums/showthread.php?postid=1178327#post1178327
Regards,
Yugovitz
Yugo,
Thanks for the references.
We have a slightly different situation with power supplies. I have not studied Tweeker's earlier notes with care (time problems here now), but I do not seem to get the right results with the "twin 220nF" pair. As said before, I simply cannot see what the 220nF directly in parallel with 100uF will achieve, as it just adds up. (Also again, not with regard to other like bypasses with poly caps in the amplifier. Those are for h.f. bypassing; here we are talking of basic design.)
As to the 220 nF at rectifier-choke junction, that does cure spikes to a certain extent, but simply because the filter then begins to act like a C-L-C. Output voltage is increased and regulation is compromised. (One gets 490 V out compared to the design value of 450V with the L-C.)
I did a few further simulations, this time with increased choke internal resistance of 60 ohm (smaller choke) and at a current of 150mA at 450V. The series RC over the choke of 10K/10nF did not quite cure switching transients, 10K/100nF did. Changing the choke internal capacity did not have a lot of effect, but the internal resistance did.
I must reiterate that spikes at choke turn-off every half cycle is best cured by a snubber across the source of the spike viz. the choke itself, not involving the rest of the circuit.
In all this I did not really address your initial quiery regarding resonant design. I simply tried to point out the (non) effect of the twin 220nF as far as spikes are concerned, which is a real thing as voltages can rise to destructive values.
I must also bow out for a few days because of other pressing matters, but will be back.
Regards
Thanks for the references.
We have a slightly different situation with power supplies. I have not studied Tweeker's earlier notes with care (time problems here now), but I do not seem to get the right results with the "twin 220nF" pair. As said before, I simply cannot see what the 220nF directly in parallel with 100uF will achieve, as it just adds up. (Also again, not with regard to other like bypasses with poly caps in the amplifier. Those are for h.f. bypassing; here we are talking of basic design.)
As to the 220 nF at rectifier-choke junction, that does cure spikes to a certain extent, but simply because the filter then begins to act like a C-L-C. Output voltage is increased and regulation is compromised. (One gets 490 V out compared to the design value of 450V with the L-C.)
I did a few further simulations, this time with increased choke internal resistance of 60 ohm (smaller choke) and at a current of 150mA at 450V. The series RC over the choke of 10K/10nF did not quite cure switching transients, 10K/100nF did. Changing the choke internal capacity did not have a lot of effect, but the internal resistance did.
I must reiterate that spikes at choke turn-off every half cycle is best cured by a snubber across the source of the spike viz. the choke itself, not involving the rest of the circuit.
In all this I did not really address your initial quiery regarding resonant design. I simply tried to point out the (non) effect of the twin 220nF as far as spikes are concerned, which is a real thing as voltages can rise to destructive values.
I must also bow out for a few days because of other pressing matters, but will be back.
Regards
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.