Thorsten, I like your attitude. I know, but I won't tell. I'll suspect that there are no sensible answer.Kuei Yang Wang said:
1 cap//cap+res what is the purpose compared to only a cap? Can't see any.
4 Please do and show the sense in this arrangement... in a theoretical way that is.
5 I have no problems to derive it but what is the purpose. I can't see any and neither can anyone else except you and Carlos.
Note also that I have made _my_ implementation of a LM3886. Some likes it some don't. As you very well know there are dozens on designs.
About your very last remark, I say nowhere that I don't "like it". I just wondered if anyone had a good theoretical explanation of my very specific question. So far noone has.
Konnichiwa,
Actually, I have told you all anyone "skilled in the art" (as they so charmingly say in patent applications) needs to "get it".
If you fail to understand the concept of LC resonance tanks (and how to damp them), what business do you have designing electronic gear?
You suspect wrongly. People get tired of my regular comments to people to take EE101 rmidial in such situations, thus I shall abstain from recommending it here.
I repeat, analyse the circuit including full parasitic components assuming 1uH ESL for the large value capacitor.
You may find that the parasitic inductance forms a resosance circuit having very impedance at frequencies near the "marginal stability" regions of the Chip Amp.
So, bypassing using the values suggested in the Nat Semi Datasheet (0.1u) can create a supply that is high impedance exactly where it should be low impedance. This allows supply noise at these selected frequencies through but more importantly, it really louses up the stability margin of the Amp.
Adding a "snubber" damps that Tank circuit and significantly reduces the magnitude of that impedance peak, thus restoring most of the stability margin lost.
Clearly these very basic electrical facts escape you, I am seriosuly disappointed. Perhaps you should avoid designing electrical circuits in future, or at least untill you have caught up on such basic stuff....
Please analyse the circuit as directed above with & without RC. Consider this your homework if I need to tutor you.
If you see no purpose you do not know how to derive it, as you are obviusly ignorant of what electrical parameters are needed to achieve what it should achieve.
I suspect that Carlos FM's version is the result of simple empirical trial/error with parts that where at hand, rather than of an rigerous network analysis of the supply circuit, yet even so I suspect they minimise the problem sufficiently.
Sayonara
peranders said:
Actually, I have told you all anyone "skilled in the art" (as they so charmingly say in patent applications) needs to "get it".
If you fail to understand the concept of LC resonance tanks (and how to damp them), what business do you have designing electronic gear?
peranders said:
You suspect wrongly. People get tired of my regular comments to people to take EE101 rmidial in such situations, thus I shall abstain from recommending it here.
peranders said:
I repeat, analyse the circuit including full parasitic components assuming 1uH ESL for the large value capacitor.
You may find that the parasitic inductance forms a resosance circuit having very impedance at frequencies near the "marginal stability" regions of the Chip Amp.
So, bypassing using the values suggested in the Nat Semi Datasheet (0.1u) can create a supply that is high impedance exactly where it should be low impedance. This allows supply noise at these selected frequencies through but more importantly, it really louses up the stability margin of the Amp.
Adding a "snubber" damps that Tank circuit and significantly reduces the magnitude of that impedance peak, thus restoring most of the stability margin lost.
Clearly these very basic electrical facts escape you, I am seriosuly disappointed. Perhaps you should avoid designing electrical circuits in future, or at least untill you have caught up on such basic stuff....
peranders said:
Please analyse the circuit as directed above with & without RC. Consider this your homework if I need to tutor you.
peranders said:
If you see no purpose you do not know how to derive it, as you are obviusly ignorant of what electrical parameters are needed to achieve what it should achieve.
I suspect that Carlos FM's version is the result of simple empirical trial/error with parts that where at hand, rather than of an rigerous network analysis of the supply circuit, yet even so I suspect they minimise the problem sufficiently.
Sayonara
EE101 rmidial, which is?Kuei Yang Wang said:
This is not exactly the question which is answered here. I wondered about having a capacitor in parallel with an another capacitor which has a resistor in series.Kuei Yang Wang said:
Still the scenario you paint up here, is this a real problem? Do you really get signs of unstablility, resonances' etc? and with this snubber those signs are removed?
Konnichiwa,
EE101 Remidial - eg re-taking a basic electronics course.
Which isn't really the circuit that you discussed, is it now. The circuit you discussed had a large value electrolytic capacitor bypassed with a small value film capacitor and the whole combination snubbered (empirically done) which electrially results in a damping of the HF tank resonance circuit. All very basic, simple and verifiable electronics.
An amplifier with a marginal stability may not show any problems on the testbench, yet attach the right speaker load (or rather the wrong one) and feed it the right signal and you observe burst oscillation on peaks. Supply born noise can also cause audible problems at lower frequencies.
Therefore the addition of the snubber certainly has the POTENTIAL to cause audible differences.
It all goes back to square one, namely that a well designed powersupply is free from noise to a degree that allows the circuits PSRR to deal with any remainder and presents the circuit supplied with an essentially resistive supply impedance. Failure on either point tends to cause problems of some form or other.
Perversly, using a low impedance, low ESL 1,000uF Cap as in the gaincard and many gainclones approaches this much closer than a big 10,000uF PSU Capacitor with added bypassing in most cases.
Sayonara
peranders said:
EE101 Remidial - eg re-taking a basic electronics course.
peranders said:
Which isn't really the circuit that you discussed, is it now. The circuit you discussed had a large value electrolytic capacitor bypassed with a small value film capacitor and the whole combination snubbered (empirically done) which electrially results in a damping of the HF tank resonance circuit. All very basic, simple and verifiable electronics.
peranders said:
An amplifier with a marginal stability may not show any problems on the testbench, yet attach the right speaker load (or rather the wrong one) and feed it the right signal and you observe burst oscillation on peaks. Supply born noise can also cause audible problems at lower frequencies.
Therefore the addition of the snubber certainly has the POTENTIAL to cause audible differences.
It all goes back to square one, namely that a well designed powersupply is free from noise to a degree that allows the circuits PSRR to deal with any remainder and presents the circuit supplied with an essentially resistive supply impedance. Failure on either point tends to cause problems of some form or other.
Perversly, using a low impedance, low ESL 1,000uF Cap as in the gaincard and many gainclones approaches this much closer than a big 10,000uF PSU Capacitor with added bypassing in most cases.
Sayonara
Bypassing simulated....
Konnichiwa,
As we have complete blockheads around who could not think themselves out of a room that is open on three sides, here some proof. I have quickly installed P-Spice at work and run a basic impedance analysis of 4 possible networks:
1) Standard 10,000uF PSU Cap
2) Standard 10,000uF PSU Cap with 100nF bypass
3) Standard 10,000uF PSU Cap with 100nF bypass and 100n/1R snubber
4) Standard 10,000uF PSU Cap with optimised bypass and snubber
First here the various equivalent circuits employed in the simulation, these include a first order approximation of the parasitics in the capacitors, as John Watson (I think) showed in recent Wireless World issues this is a very inadequate solution but is sufficient to illustrate the problems, reality is a lot worse than my cute simulations BTW, not better....
Now, lets see what sort of impedance each combination offers, shall we?
The purple line is the plain, unbypassed capacitor, the impedance rises from 64mOhm at 10KHz to 6.3 Ohm at 1MHz.
The aquamarine (neon blue) line is the capacitor with a 100nF bypass capacitor, our impedance starts at 64mOhm and peaks >180 Ohm at 500KHz (hint, study the LM3875 Datasheet, check the pasemargin at 500KHz and also the PSRR....).
The green/yellowish line is the in the capacitor with bypass and snubber, our impedance starts again at 64mOhm at 10KHz but rises only to 19 Ohm at 360KHz, a nearly tenfold reduction of the peak impedance.
Now if we take my own ballpark formula of a 1:3 ratio and change the snubber capacitor to 330nF we drop the magnitude of the impedance peak to a little over 4 Ohm at 250Khz, so a hint for Carlos & co, change your 100nF Snubber capacitor to 330nF, this should be even better.
Finally, the bypassing/snubber I personally would suggest is shown in royal blue. We start again from 64mOhm at 10KHz, our impedance peaks with 1 Ohm at 140KHz and stays low to above and beyond 10MHz, assuming the 1uF bypass is placed close to the Amplifier Chip.
I think the above illustrates the issues and problems as well the usefullness and limits of the solution(s).
It is worthwhile to note also that many of the empirically arrived at bypass combinations (Stein Music, l'Audiophile) invariably use a fairly high ESR Film capacitor as the largest value bypass with the smaller additional bypasses with also progresiively lower ESR, resulting in a well behaved and low PSU line impedance.
Sayonara
Konnichiwa,
As we have complete blockheads around who could not think themselves out of a room that is open on three sides, here some proof. I have quickly installed P-Spice at work and run a basic impedance analysis of 4 possible networks:
1) Standard 10,000uF PSU Cap
2) Standard 10,000uF PSU Cap with 100nF bypass
3) Standard 10,000uF PSU Cap with 100nF bypass and 100n/1R snubber
4) Standard 10,000uF PSU Cap with optimised bypass and snubber
First here the various equivalent circuits employed in the simulation, these include a first order approximation of the parasitics in the capacitors, as John Watson (I think) showed in recent Wireless World issues this is a very inadequate solution but is sufficient to illustrate the problems, reality is a lot worse than my cute simulations BTW, not better....
An externally hosted image should be here but it was not working when we last tested it.
Now, lets see what sort of impedance each combination offers, shall we?
An externally hosted image should be here but it was not working when we last tested it.
The purple line is the plain, unbypassed capacitor, the impedance rises from 64mOhm at 10KHz to 6.3 Ohm at 1MHz.
The aquamarine (neon blue) line is the capacitor with a 100nF bypass capacitor, our impedance starts at 64mOhm and peaks >180 Ohm at 500KHz (hint, study the LM3875 Datasheet, check the pasemargin at 500KHz and also the PSRR....).
The green/yellowish line is the in the capacitor with bypass and snubber, our impedance starts again at 64mOhm at 10KHz but rises only to 19 Ohm at 360KHz, a nearly tenfold reduction of the peak impedance.
Now if we take my own ballpark formula of a 1:3 ratio and change the snubber capacitor to 330nF we drop the magnitude of the impedance peak to a little over 4 Ohm at 250Khz, so a hint for Carlos & co, change your 100nF Snubber capacitor to 330nF, this should be even better.
Finally, the bypassing/snubber I personally would suggest is shown in royal blue. We start again from 64mOhm at 10KHz, our impedance peaks with 1 Ohm at 140KHz and stays low to above and beyond 10MHz, assuming the 1uF bypass is placed close to the Amplifier Chip.
I think the above illustrates the issues and problems as well the usefullness and limits of the solution(s).
It is worthwhile to note also that many of the empirically arrived at bypass combinations (Stein Music, l'Audiophile) invariably use a fairly high ESR Film capacitor as the largest value bypass with the smaller additional bypasses with also progresiively lower ESR, resulting in a well behaved and low PSU line impedance.
Sayonara
Hi,
Where on earth do you get large cans with 1uH inductance? Recently I have measured and collected datasheets for 10000uF/450V caps for inverter design. Typical inductance od such cans (89mm diameter and 170mm length) is 20nH to 50nH, with some (Epcos) going even below 20nH).
Best regards,
Jaka Racman
Where on earth do you get large cans with 1uH inductance? Recently I have measured and collected datasheets for 10000uF/450V caps for inverter design. Typical inductance od such cans (89mm diameter and 170mm length) is 20nH to 50nH, with some (Epcos) going even below 20nH).
Best regards,
Jaka Racman
Re: Bypassing simulated....
Hi Kuei,
Thanks for your work.
It really beats me how some, instead of discussing the sex of the virgin, can't test this PSU, listen and measure it.
The fine-tuning was made with listening tests, and without changing the 1 ohm resistor, the amp started to sound better and better as I reduced the value of the snubber cap.
330nf was one of the values I tested.
100~120nf was about the ideal in my setup and it has been giving me consistent results with different trafos, rectifiers and even power op-amp chips.
I believe that this may/can be optimized, and you are all free to test.
Kuei Yang Wang said:
Hi Kuei,
Thanks for your work.
It really beats me how some, instead of discussing the sex of the virgin, can't test this PSU, listen and measure it.
The fine-tuning was made with listening tests, and without changing the 1 ohm resistor, the amp started to sound better and better as I reduced the value of the snubber cap.
330nf was one of the values I tested.
100~120nf was about the ideal in my setup and it has been giving me consistent results with different trafos, rectifiers and even power op-amp chips.
I believe that this may/can be optimized, and you are all free to test.
Konnichiwa,
I have come across similar inductances in generic capacitors of similar values to these ones used. I found the measured inductance was rather frequency dependent, with larger values observed as the frequency rises. Once you are past 100KHz the gloves are well off with many Elcap's....
Quite frankly, if the ESL was as low as you noted reliably, at high frequencies any bypassing would show virtually no effect (measured or otherwise). Could you suggest which specific capacitors offer as low an ESL and how you took the measurements?
Sayonara
Jaka Racman said:
I have come across similar inductances in generic capacitors of similar values to these ones used. I found the measured inductance was rather frequency dependent, with larger values observed as the frequency rises. Once you are past 100KHz the gloves are well off with many Elcap's....
Quite frankly, if the ESL was as low as you noted reliably, at high frequencies any bypassing would show virtually no effect (measured or otherwise). Could you suggest which specific capacitors offer as low an ESL and how you took the measurements?
Sayonara
Re: Re: Bypassing simulated....
Konnichiwa,
Point taken. I think the upshot is to realise that a powersupply is a collection of multiple resonat circuits and thus is BY DEFINITION a tuned circuit, which requires tuning either by measurements or by ear (if you are so inclinsed) for the actual components and circuit used, taking into account the actual circuit supplied and it's reaction to to powersupply impedance and noise effects.
Or as Kimura San did, just use a single Cap of good quality....
I think the above also illustrates the issues some may have when bypassing other Electrolytics (especially Black Gate).
Sayonara
Konnichiwa,
carlosfm said:
Point taken. I think the upshot is to realise that a powersupply is a collection of multiple resonat circuits and thus is BY DEFINITION a tuned circuit, which requires tuning either by measurements or by ear (if you are so inclinsed) for the actual components and circuit used, taking into account the actual circuit supplied and it's reaction to to powersupply impedance and noise effects.
Or as Kimura San did, just use a single Cap of good quality....
I think the above also illustrates the issues some may have when bypassing other Electrolytics (especially Black Gate).
Sayonara
*reality check ON/
- Audibility means that the movement of the speaker cone(s) changes so much in the middle of the complex movements resulting from the music reproduction that the change is audible;
- Thorsten's simulations show that the impedance of a power supply changes to a lower value upwards of 250kHz with those snubber things;
- Carlosfm maintains he can hear the differences introduced by those supply changes.
*reality check OFF/
Funny, all of a sudden I feel really at home with those that Don't Get it. Could it be that Those Who Think They Get It are also The Ones That Lost It
?
Jan Didden
- Audibility means that the movement of the speaker cone(s) changes so much in the middle of the complex movements resulting from the music reproduction that the change is audible;
- Thorsten's simulations show that the impedance of a power supply changes to a lower value upwards of 250kHz with those snubber things;
- Carlosfm maintains he can hear the differences introduced by those supply changes.
*reality check OFF/
Funny, all of a sudden I feel really at home with those that Don't Get it. Could it be that Those Who Think They Get It are also The Ones That Lost It
?Jan Didden
Just a warning when you're reading capacitor data sheets and doing simulations...
Only the capacitance lies within the "10%" written on the datasheet. Usually there's no tolerance given for ESL or ESR, just a minimum number - sometimes they can be up to an order of magnitude less than you expect them to be. If you're designing any circuit that's sensitive to the ESL or ESR in a capacitor, MEASURE IT YOURSELF!
I realized this quickly when I started doing switching power supply design. I just couldn't figure out why my perfectly-simulated, well-laid-out power supply wasn't stable - it turns out the output electrolytic ESR spec was far better than I thought it was
So the best thing you can do is measure your main capacitors, calculate a new snubber value and simulate. Hell, you can even measure your snubber while you're at it and calculate an even better one.
Or if you don't have access to equipment for measuring ESL/ESR, just take the carlosfm approach and start swapping things until it sounds good. I mean, that's the whole point, isn't it?
Only the capacitance lies within the "10%" written on the datasheet. Usually there's no tolerance given for ESL or ESR, just a minimum number - sometimes they can be up to an order of magnitude less than you expect them to be. If you're designing any circuit that's sensitive to the ESL or ESR in a capacitor, MEASURE IT YOURSELF!
I realized this quickly when I started doing switching power supply design. I just couldn't figure out why my perfectly-simulated, well-laid-out power supply wasn't stable - it turns out the output electrolytic ESR spec was far better than I thought it was
So the best thing you can do is measure your main capacitors, calculate a new snubber value and simulate. Hell, you can even measure your snubber while you're at it and calculate an even better one.
Or if you don't have access to equipment for measuring ESL/ESR, just take the carlosfm approach and start swapping things until it sounds good. I mean, that's the whole point, isn't it?
Konnichiwa,
*reality check ON/
LM3875 Datasheet - PSRR @ 250KHz = ~30db
LM3875 Datasheet - Phasemargin reduces above appx. 100KHz
LM3875 Datasheet - Distortion (and thus non-linearity) increases above 100Hz appx 12db/8ave, probably more above 20KHz (specification limit)
*reality check OFF/
If you feel that, you are probably right.
As with so many things, what is heard and what is measured are different thing. What is heard with all this monkeying around with the supply impedance outside the audio range has to do with noise levels and the (intermodulation product) folding back into the audio range plus issues around marginal stability. With amplifiers having limited power we also may more often than not hear clipping recovery artifacts when playing loud and NOT the actual clipping (usually to short to be audible).
If we remember the (practical) work Joe Rasmussen did on the stability issues with the LM3875 (he noted a tendency to instability in the 300KHz range IIRC) and we look at a high PSU impedance in the same range it should be easy to note that there is at least potenial for trouble and thus audible differences, quite drastically audible differences in fact.
Did you never have a regulator go oscillating on you and wondered why the sound had turned bad?
Sayonara
janneman said:
*reality check ON/
LM3875 Datasheet - PSRR @ 250KHz = ~30db
LM3875 Datasheet - Phasemargin reduces above appx. 100KHz
LM3875 Datasheet - Distortion (and thus non-linearity) increases above 100Hz appx 12db/8ave, probably more above 20KHz (specification limit)
*reality check OFF/
janneman said:
If you feel that, you are probably right.
As with so many things, what is heard and what is measured are different thing. What is heard with all this monkeying around with the supply impedance outside the audio range has to do with noise levels and the (intermodulation product) folding back into the audio range plus issues around marginal stability. With amplifiers having limited power we also may more often than not hear clipping recovery artifacts when playing loud and NOT the actual clipping (usually to short to be audible).
If we remember the (practical) work Joe Rasmussen did on the stability issues with the LM3875 (he noted a tendency to instability in the 300KHz range IIRC) and we look at a high PSU impedance in the same range it should be easy to note that there is at least potenial for trouble and thus audible differences, quite drastically audible differences in fact.
Did you never have a regulator go oscillating on you and wondered why the sound had turned bad?
Sayonara
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.