Joseph K said:
Hi Joseph,
Thanks for your encouragment - I went back and increased the ESL values as Thorston suggested and the plot still seems quite even.
To implement this, caps would have to be mounted on both sides of the board - it will be quite busy, but I think that it is definitely worth an experiment - if it sounds good I will think about a few different pcb layout elements that I could 'paste' into different applications.
So, might be complicatied first few times, but after that, quite simple.
My first application of this will be a CD player PSU's - digital & analogue.
I should start in a week or two
mike
Peranders, now I understood your question!
Yes, I forgot to show that. But I have shown the 1 ohm response, like a calibration / check. if you only plug in the test jig, it shows you a straight line at the top at 0 dB, which is the reference level.
Now, what is at the bottom, it is more difficoult to say. If you DON'T plug in anything into the ports, then port 2, the receiver in this case,
will show a noise spectrum from 5 to 200000 kHz. The receiver's noise floor is between -90 and -100 dB, that is, you see it at the bottom of the screen. And this is really the dynamics of this analyser, you cannot go below this. [My soundcard is a bit better in this respect..]
Now, to reach that level in bypassing...
Let's suppose -90 dB, that is equal to 0.8 mOhm, 0,0008 Ohm..
At this level I should NOT use lemo, but the original N-type connectors, which are another league..
So, in practice, I will dig it up for You, I have seen such a thing, once, a Sanyo Os-Con descending to about -80 dB. There it starts to show up, at the bottom of the curve, the anlyser noise. But I had to be sure about the connectors, their contact resistance is quite unstable at this levels.
If you put a short, then what you see is the inductance of it, and of the setup - and it's not zero at all!
I will make the photos.
Yes, I forgot to show that. But I have shown the 1 ohm response, like a calibration / check. if you only plug in the test jig, it shows you a straight line at the top at 0 dB, which is the reference level.
Now, what is at the bottom, it is more difficoult to say. If you DON'T plug in anything into the ports, then port 2, the receiver in this case,
will show a noise spectrum from 5 to 200000 kHz. The receiver's noise floor is between -90 and -100 dB, that is, you see it at the bottom of the screen. And this is really the dynamics of this analyser, you cannot go below this. [My soundcard is a bit better in this respect..]
Now, to reach that level in bypassing...
Let's suppose -90 dB, that is equal to 0.8 mOhm, 0,0008 Ohm..
At this level I should NOT use lemo, but the original N-type connectors, which are another league..
So, in practice, I will dig it up for You, I have seen such a thing, once, a Sanyo Os-Con descending to about -80 dB. There it starts to show up, at the bottom of the curve, the anlyser noise. But I had to be sure about the connectors, their contact resistance is quite unstable at this levels.
If you put a short, then what you see is the inductance of it, and of the setup - and it's not zero at all!
I will make the photos.
Thorsten, please, I did not want to sound orthodox, in fact, I WISHED to see something, a teeny little bit of L-cancellation, this is why I got carried away, [what is so well documented here..]
What I have seen, is that it is really hard to make this measurements, every move of yours just makes stray inductance sneeking in..
What I have seen, is that it is really hard to make this measurements, every move of yours just makes stray inductance sneeking in..
Konnichiwa,
Hey, I REALLY appreciate your work. It also illustrated why my indirect measurements often showed higher ESL that you find.
I was just curious if the L-Cancelling effect could exist or not. I guess we have illustrated it does not, at least not in ways we can conceive. Such a negative is a good thing. Now we know.
FWIW, it may still sound better or not, that is another story.
So again, many thanks for all these measurements.
Ciao T
Joseph K said:
Hey, I REALLY appreciate your work. It also illustrated why my indirect measurements often showed higher ESL that you find.
I was just curious if the L-Cancelling effect could exist or not. I guess we have illustrated it does not, at least not in ways we can conceive. Such a negative is a good thing. Now we know.
FWIW, it may still sound better or not, that is another story.
So again, many thanks for all these measurements.
Ciao T
May I take up some more bandwith and turn back to the question of Yoghurt?
You don't have to build two, with todays technology, you simply save the result then put on the snubber, save the result.
I did not do this, yet. But I feel I will not see anything. Not because there is no effect at all, but because it is more subtle that it would be possible to measure in a simple way. I have measured my GC, Peranders measured his GC, PMA measured Pavel's 4780 amp.
You can see the results, they are just bloody good!
In my measurements, I think I had seen a better distorsion spectrum than the baseline for the soundcard itself.. Then I was thinking about it, and concluded, that the GC presented a lighter load to the soundcard output IC, then it's [the soundcard's] own input!
To see some difference AT THE OUTPUT, you should think up sophisticated measurements, like the ones what Thorsten listed in some thread several years before. Like white noise comb filtered, and looking for the noise floor changing - but it is his territory.
In the meantime, we should thrust Carlos's ears.. and our own!
But, I do think that although the final, sum, effect might be complicated to measure, if you try to separate the different local mechanisms cooperating in the final result, and try to describe & understand these individually, then maybe get a picture.
Then, in these "simpler" sub-mechanisms, it can be even possible to measure & proove things.
Ciao!
You don't have to build two, with todays technology, you simply save the result then put on the snubber, save the result.
I did not do this, yet. But I feel I will not see anything. Not because there is no effect at all, but because it is more subtle that it would be possible to measure in a simple way. I have measured my GC, Peranders measured his GC, PMA measured Pavel's 4780 amp.
You can see the results, they are just bloody good!
In my measurements, I think I had seen a better distorsion spectrum than the baseline for the soundcard itself.. Then I was thinking about it, and concluded, that the GC presented a lighter load to the soundcard output IC, then it's [the soundcard's] own input!
To see some difference AT THE OUTPUT, you should think up sophisticated measurements, like the ones what Thorsten listed in some thread several years before. Like white noise comb filtered, and looking for the noise floor changing - but it is his territory.
In the meantime, we should thrust Carlos's ears.. and our own!
But, I do think that although the final, sum, effect might be complicated to measure, if you try to separate the different local mechanisms cooperating in the final result, and try to describe & understand these individually, then maybe get a picture.
Then, in these "simpler" sub-mechanisms, it can be even possible to measure & proove things.
Ciao!
I have a question... there are two versions I see of Carlos' snubber PSU. One has 3300 + 10000 uF while the other one just 10000uF..... I haven't found the original thread for the 10,000 uF one(and don't know if it's before or after the other one)... so can someone please tell me which one to use??? Thanks.. 😀
Hi Joseph K. Inspired by your work I hooked up some ASC paper/oil caps used in my amp to the Audio Precision Portable One at work (only good to 200 kHz) and ran some tests to determine their self-resonant frequency. The X386S 330 VAC 50 uF was in the range of 86 kHz and the the 370 VAC 60 uF around 73 kHz. I didn't pay enough attention to ESR to calculate a value, my purpose was to determine ESL for some RC shunts.
My question stems from how difficult I found it to devise test leads. The generator output is XLR male, the final arrangement was to solder <1" wires from the cap terminals to a bodiless XLR female. Leads any longer than that caused false low readings. How does the HP jig compensate for generator lead inductance?
My question stems from how difficult I found it to devise test leads. The generator output is XLR male, the final arrangement was to solder <1" wires from the cap terminals to a bodiless XLR female. Leads any longer than that caused false low readings. How does the HP jig compensate for generator lead inductance?
sardonx said:
That schematic was my inicial tests with a 2x3,300uf PSU, as I say on my first post of that thread.
You can find the final schematic on the last pages of my thread.
As I said, 10,000uf is a starting point, you can use a combination of equal value caps in parallel, like 2 or 3x4,700uf, 2x6,800uf, 2x10,000uf...
I couldn't fit all the possible combinations in one schematic, of course.
Also, I don't think you would gain anything in going to more than around 20,000uf per rail, in most implementations.
My suggestion is not comparison of two measures, but differential measure of near real situation. If differential measure shows nothing w.r.t. to noise floor shaping due to the chosen DUT input, then nothing is proven. But if something interesting raises, then bingo!
Rdf,
First of all, thanks for the link, it's good!
As regards of the measurement setup: the trick is in using a proper transmission line, terminated properly at both ends. This way you have a resistive measurement point.
For example, I have 50 ohm output - input ports on the analyser. If I connect these ports to my jig with 50 ohm cables, then at the lemo femail connectors "signal hot" on my test jig these appear like if them both were connected to the ground with a 50 ohm resistor. Even, a very good quality 50 ohm resistor, in Rf terms, because this is what effectively terminates the ports inside the instrument, and it's like if these resistors were " teletransported " to my lemo femails! Now, that wire with which I join the two connectors, is "breaking out" this transmission line, and effectively deteriorates it a bit, but not very much, as I see from the results. So, the net effect is that if you connect to that wire, it's really like if you were trying to short out the resulting 25 ohm pure resistive impedance there.
So, I think this is the main point. The analyser, in reality, is a beautiful instrument in itself, and has much more capabilities then are used here.
So, it's already some time that I was thinking, how would it be possible to repeat these results without the analyser. [and so eliminate my privilege in using it..] Now, I would say, it is not possible below a certain level - we are talking about RF measurements here. But, a good quality analog or digital scope and a good signal generator is available to a lot of people here.
So, lets say, we have a signal generator, not necessarily low distorsion, extending up to 40 - 50 MHz. It's output might already be terminated in 50 ohm, if not, then one should use a BNC T connector, a BNC 50 ohm good quality
termination, and put on the output.
At the other end, use the best scope available to You. It should have an input terminated in 50 ohm, it's usually not there, so use the same BNC T connector like at the generator.
Now you put together a small test jig like mine, the important thing is two connectors facing each other, not a long distance away, over a ground plane, that is, a raw PC board. You can use BNC-s, to be cost effective, or Lemo, to be more compact, or SMB, main point is the 50 ohm characteristic impedance.
I have just put it together here, at my job, and it works seamless!
Generator: Hp 8116A pulse/function generator 50 MHz
Scope: TEK 2465B 400 MHz
I will show the pic, they are nice 🙂
So as to be able to see some kind of a signal even at maximum attenuation, one should find a good value for test signal - a 6 Volts peak to peak symmetrical sinus, swept from 10 Hz to 50 MHz, with a Wima 1uF under test, gave me a still visible signal at 2 MHz, it's resonace point [this cap is really small, 63V].
An another one, 68 uF elco had shown a flat attenuation up to ~ 2MHz, then started to rise steeply.
You see very well the sinus getting to be distorted along the way, because of the phase shifts.
Well, it would be more difficoult to publish the results, but you get them, very precisely! And this all gave me an even better feeling about the parts behaviour, then the analyser! [It's linear, you really FEEL the meaning of 60 dB
attenuation.., then you see at once the signal in the time domain]
Then, if you have other cables / connectors, you could use other value of termination, 75 ohm, 93 ohm [connectors?] 110 ohm [good connectors?]
I wish you happy experimenting!
Ps.: Only not XLR, I'm afraid..
First of all, thanks for the link, it's good!
As regards of the measurement setup: the trick is in using a proper transmission line, terminated properly at both ends. This way you have a resistive measurement point.
For example, I have 50 ohm output - input ports on the analyser. If I connect these ports to my jig with 50 ohm cables, then at the lemo femail connectors "signal hot" on my test jig these appear like if them both were connected to the ground with a 50 ohm resistor. Even, a very good quality 50 ohm resistor, in Rf terms, because this is what effectively terminates the ports inside the instrument, and it's like if these resistors were " teletransported " to my lemo femails! Now, that wire with which I join the two connectors, is "breaking out" this transmission line, and effectively deteriorates it a bit, but not very much, as I see from the results. So, the net effect is that if you connect to that wire, it's really like if you were trying to short out the resulting 25 ohm pure resistive impedance there.
So, I think this is the main point. The analyser, in reality, is a beautiful instrument in itself, and has much more capabilities then are used here.
So, it's already some time that I was thinking, how would it be possible to repeat these results without the analyser. [and so eliminate my privilege in using it..] Now, I would say, it is not possible below a certain level - we are talking about RF measurements here. But, a good quality analog or digital scope and a good signal generator is available to a lot of people here.
So, lets say, we have a signal generator, not necessarily low distorsion, extending up to 40 - 50 MHz. It's output might already be terminated in 50 ohm, if not, then one should use a BNC T connector, a BNC 50 ohm good quality
termination, and put on the output.
At the other end, use the best scope available to You. It should have an input terminated in 50 ohm, it's usually not there, so use the same BNC T connector like at the generator.
Now you put together a small test jig like mine, the important thing is two connectors facing each other, not a long distance away, over a ground plane, that is, a raw PC board. You can use BNC-s, to be cost effective, or Lemo, to be more compact, or SMB, main point is the 50 ohm characteristic impedance.
I have just put it together here, at my job, and it works seamless!
Generator: Hp 8116A pulse/function generator 50 MHz
Scope: TEK 2465B 400 MHz
I will show the pic, they are nice 🙂
So as to be able to see some kind of a signal even at maximum attenuation, one should find a good value for test signal - a 6 Volts peak to peak symmetrical sinus, swept from 10 Hz to 50 MHz, with a Wima 1uF under test, gave me a still visible signal at 2 MHz, it's resonace point [this cap is really small, 63V].
An another one, 68 uF elco had shown a flat attenuation up to ~ 2MHz, then started to rise steeply.
You see very well the sinus getting to be distorted along the way, because of the phase shifts.
Well, it would be more difficoult to publish the results, but you get them, very precisely! And this all gave me an even better feeling about the parts behaviour, then the analyser! [It's linear, you really FEEL the meaning of 60 dB
attenuation.., then you see at once the signal in the time domain]
Then, if you have other cables / connectors, you could use other value of termination, 75 ohm, 93 ohm [connectors?] 110 ohm [good connectors?]
I wish you happy experimenting!
Ps.: Only not XLR, I'm afraid..
Low ESR powersupply caps
Second, I am puzzled by Jonathan Carr's experience with film caps vs. Nichicon Super Throughs as main filter caps. The film caps are sure low ESR but Jonathan did prefer the film caps. I have no idea about the Super Through's ESR but I believe it is not extremely low and not specified by Nichicon.
😕
First, Jocko I do believe you!Jocko Homo said:
Second, I am puzzled by Jonathan Carr's experience with film caps vs. Nichicon Super Throughs as main filter caps. The film caps are sure low ESR but Jonathan did prefer the film caps. I have no idea about the Super Through's ESR but I believe it is not extremely low and not specified by Nichicon.
😕
Note when Jocko says he has made his worst sounding amp ever, this means slightly worse than "Spitzenklasse" and a bit worse than top-of-the-line is extremely good.... yes?... so... the conclusion is:...?
ELso, did you refer to this?
On base of this, I would not be able to draw a clear conclusion. We would need more construction details, as regards of ESR. Maybe he used a configuration to mitigate the polyprops resonant behaviour? He was using them in a preamp!
There is much more to a capacitor's quality, then simply ESR / ESL. And a polyprop film/foil is a much better capacitor, then any electrolytic.
But still, I would insist, that in general, low ESR will cause stability problems, which should be considered, and dealt with.
And I think these are two different topics:
having better or worse sound because of the different capacitor distorsions;
or having an inconditionally stable / or not, amplifier, because of bypass structure phase shifts.
ciao, george
On base of this, I would not be able to draw a clear conclusion. We would need more construction details, as regards of ESR. Maybe he used a configuration to mitigate the polyprops resonant behaviour? He was using them in a preamp!
There is much more to a capacitor's quality, then simply ESR / ESL. And a polyprop film/foil is a much better capacitor, then any electrolytic.
But still, I would insist, that in general, low ESR will cause stability problems, which should be considered, and dealt with.
And I think these are two different topics:
having better or worse sound because of the different capacitor distorsions;
or having an inconditionally stable / or not, amplifier, because of bypass structure phase shifts.
ciao, george
Also, given the fact that he was using a pre, probably he did something, also, to limit the charge current peaks in the caps.
Low ESR caps in powersupply
Hello George,
That was exactly the text I was referring to.
Actually I feel a raw power supply for a preamp is not much different from that for a power amp.
At the moment I am deliberating whether I should use stinking normal electrolytics as I always did in the power supply or switch to Panasonic FC or Nichicon Muse KG Gold or perhaps Nichicon Muse FX in the preamp and Super Through in the poweramp......
😉
Hello George,
That was exactly the text I was referring to.
Actually I feel a raw power supply for a preamp is not much different from that for a power amp.
At the moment I am deliberating whether I should use stinking normal electrolytics as I always did in the power supply or switch to Panasonic FC or Nichicon Muse KG Gold or perhaps Nichicon Muse FX in the preamp and Super Through in the poweramp......
😉
Elso.
Although Nichicon doen't specify the ESR of their Muse caps, they do specify the max ripple current. Ohms law will allow you to back into an impedance calculation, ESI. Although this impedance also includes the inductive reactance in addition to the ESR, its contribution is quite small.....
Although Nichicon doen't specify the ESR of their Muse caps, they do specify the max ripple current. Ohms law will allow you to back into an impedance calculation, ESI. Although this impedance also includes the inductive reactance in addition to the ESR, its contribution is quite small.....
Elso Kwak said:
Of course it's not much different.
I was doing some comparisons in my ML380 preamp and the best caps in raw supply were 1000u BG N, greatly outperforming stinking normal caps and having edge over Nichicon FG, Panasonic for Audio and BG STD. I wouldn't even bother with FX, as FG was clearly superior.
Originally, there were 6800u Nippons in there.
- Status
- Not open for further replies.