I have subjectively compared different values / voltages / ESR / ESL of capacitors in my line level, correctly implemented (no resonance, etc) circuits in my very revealing system and found the following:
Panasonic FC:
They have clear and clean sound with a bit of "hands cupping the mouth" type of sound. The low frequency is awesome. The very high frequency is exaggerated a bit.
Rubycon ZL:
Bass is not as strong as the Panasonic but very clean. They are the clearest, cleanest, fastest, most musical caps by far within a large range of the audio bandwidth, however, with a terrible problem causing me to basically abandon the use of these caps - they have a very picky treble that appears to be caused by resonances that can not be completely corrected using filters. Rolling off the treble does not help.
Elna Silmic (and many other Elna audiophile caps):
They do not sound "bad" but with very strong colorations. They seem to add rich harmonics to the sound. Solid state sound was made "tube" like. It is not my cup of tea.
Xicon (General Purpose):
They are higher ESR / ESL type but with an overall tonal balance. However, they do not have the ultimate refinement in sound.
Now I am in despair! I am thinking about using Panasonic FC throughout then alter my passive XO to adjust!
Have you guys found any other caps (low ESR / ESL or not) that sound balanced without major colorations / distortions?
Regards,
Bill
Panasonic FC:
They have clear and clean sound with a bit of "hands cupping the mouth" type of sound. The low frequency is awesome. The very high frequency is exaggerated a bit.
Rubycon ZL:
Bass is not as strong as the Panasonic but very clean. They are the clearest, cleanest, fastest, most musical caps by far within a large range of the audio bandwidth, however, with a terrible problem causing me to basically abandon the use of these caps - they have a very picky treble that appears to be caused by resonances that can not be completely corrected using filters. Rolling off the treble does not help.
Elna Silmic (and many other Elna audiophile caps):
They do not sound "bad" but with very strong colorations. They seem to add rich harmonics to the sound. Solid state sound was made "tube" like. It is not my cup of tea.
Xicon (General Purpose):
They are higher ESR / ESL type but with an overall tonal balance. However, they do not have the ultimate refinement in sound.
Now I am in despair! I am thinking about using Panasonic FC throughout then alter my passive XO to adjust!
Have you guys found any other caps (low ESR / ESL or not) that sound balanced without major colorations / distortions?
Regards,
Bill
Signal path is one but I had a great surprise - the cap that bypasses the adjustment pin of the LM317/337, although what the regulator powers up are OPA627s that are claimed to have 120dB rejection. According to the datasheet, a 10uF will do but put on a 470uF and be ready for the dramatical improvement on sound quality.
I can't explain in theory but go and try it one will find the difference bigger than changing a power amp.
This is easily a good tip for the tweakers.
I guess it has got to do with DF, etc, but perhaps more importantly with the internal variations in the ESR at low frequencies and ESL at high frequencies, as well as internal resonances, etc.
I have tested many caps in the past 2 weeks and continuously for 3 days in the past long weekend.
I have not found the ideal cap.
I can't explain in theory but go and try it one will find the difference bigger than changing a power amp.
This is easily a good tip for the tweakers.
I guess it has got to do with DF, etc, but perhaps more importantly with the internal variations in the ESR at low frequencies and ESL at high frequencies, as well as internal resonances, etc.
I have tested many caps in the past 2 weeks and continuously for 3 days in the past long weekend.
I have not found the ideal cap.
I think you'll find the OPA627 doesn't have anywhere near that rejection at high frequencies, nor does any other opamp. If you put a large cap on the reg adjustment pin, you should use protection diodes per the app notes. Most people use about 240 ohms for the top resistor because that's what the app note shows, and because under many conditions it gives the reg enough idle current to function within nothing attached. There's no law that says you can't scale the resistors up a bit, making the adjustment bypass cap that much more effective. It's also desirable to use a large output cap with significant ESR, or even a series resistor, thus preventing a high frequency noise peak. A small output cap with low ESR, like a film, can result in a rather remarkable noise spectra from those regs, and the peak can be right in the upper audio range.
The Nichicon Muse electrolytics (4 lines I think?) have gotten good sonic reviews. Blackgate are also worth trying.
I trust you're breaking them in before trying to evaluate... I use 40 hours as a rule of thumb, but it's only that. More or less time may be needed... Were the Rubycons fully run in?
I trust you're breaking them in before trying to evaluate... I use 40 hours as a rule of thumb, but it's only that. More or less time may be needed... Were the Rubycons fully run in?
Hi,
Thank you very much for this information. Stuff like this is a LOT more useful, IMO, than charts and graphs from the mfr.
These are nasty in the treble because they are inaccurate.
I personally have found that bridging these with a 1uF or 2.2uF poly-anything (zillion dollar cap not needed here) cleans the treble up.
Keep up the awesome work!
Thank you very much for this information. Stuff like this is a LOT more useful, IMO, than charts and graphs from the mfr.
HiFiNutNut said:
These are nasty in the treble because they are inaccurate.
I personally have found that bridging these with a 1uF or 2.2uF poly-anything (zillion dollar cap not needed here) cleans the treble up.
Keep up the awesome work!
Here's a recent test made by dorkus, he tested the pana FC, FM, Silmic II, BG-N and Wima MKP10:
http://diyhifi.org/forums/viewtopic.php?f=13&t=985&start=62
Strange is his observation about the Silmic II, quite the opposite of yours.
Probably it's all about where the caps are you used (coupling-decoupling) and how the rest of your audio gear sounds.
http://diyhifi.org/forums/viewtopic.php?f=13&t=985&start=62
Strange is his observation about the Silmic II, quite the opposite of yours.
Probably it's all about where the caps are you used (coupling-decoupling) and how the rest of your audio gear sounds.
Hi,
nothing against subjective evaluation but may I kindly ask what is meant by
Without knowing where a C is located in a circuit, changing to a different cap the evaluation can only be subjective and hence general statements about the sonic character are rather meaningless. Slightest changes in the setup could lead to totally different impressions. In bad cases it leads even to a malfunction of the circuit by ringing, or becoming noisy. Especially ´fast´ OP-amps tend to be critical with the quality of the power suply. Its not for nothing that Op-amp manufacturers suggest that the PS-lines should be clean over the complete bandwidth of the OP. A point which You can´t fulfill with just a single blocking electrolytic somewhere.
Conrad already made some remarks about the whys and hows.
My experience with the ZLs are only very positive when implemented correctly and in the right position.
jauu
Calvin
nothing against subjective evaluation but may I kindly ask what is meant by
and how You prooved it? Were there any paralleled caps or just the electrolytics? How about the inductance of PS-traces, etc. etc.
Without knowing where a C is located in a circuit, changing to a different cap the evaluation can only be subjective and hence general statements about the sonic character are rather meaningless. Slightest changes in the setup could lead to totally different impressions. In bad cases it leads even to a malfunction of the circuit by ringing, or becoming noisy. Especially ´fast´ OP-amps tend to be critical with the quality of the power suply. Its not for nothing that Op-amp manufacturers suggest that the PS-lines should be clean over the complete bandwidth of the OP. A point which You can´t fulfill with just a single blocking electrolytic somewhere.
Conrad already made some remarks about the whys and hows.
My experience with the ZLs are only very positive when implemented correctly and in the right position.
jauu
Calvin
My +/-15V LM317/LM337 regulator is implemented in this way.
Full rectification schottkey diodes bypassed with 0.01uF polystyrene.
3300uF || 0.1uF disc ceramic - 8.2R - 3300uF || 0.12uF polypropylene before the regulator.
One diode is used across the input and output. One diode is used across the adj pin and output pin.
The upper resistor is chosen to be 150R. This is in consideration of providing a higher load to improve regulation.
1.62k is used to set the voltage to about +/-15V.
1.62k is bypassed by the TEST CAP (1). Currently it is the general purpose Xicon 100uF/100v.
On output, one bypass TEST CAP (2) is in series with a TEST RESISTOR (3). I tried various TEST CAPs ranging from 25uF to 2200uF with various impedance. The TEST RESISTOR ranged from 0R to 0.5R. The current (final?) setting is a general purpose Xicon 100uF/100V + 0.44R.
Because I use 120uF Rubycon ZL || 0.01uF MKP to bypass each rail immediately at the power supply pins on each opamp (6 dual opamps in total in my implementation of the NaO circuit), this would create a huge resonance on the power rails (it was very audible!).
So a TEST resistor (4) is connected in serious with the regulator to remove the resonance. The test value ranged from 0R to 1R. Currently at 0.94R.
Since I don't want 1R output impedance from the power supply, I added another TEST cap (5) (currently a Panasonic FC 4700uF/25v) to provide a low impedance (rated about 0.015R at 100kHz and should be low with ESR at lower frequencies due to the size of the cap) to the PSU.
The idea is that the regulator has a 0.5uH output inductance so it never wants to see capacitors without using some resistance (up to 1R value should be sufficient to dampen the resonances) so I created that environment for the regulator. I then use very big capacitors (I may increase it to 10,000uF) that will work better than batteries to supply a very low and flat impedance to the opamps for much improved transient response than without a large cap.
So far, this sounds the best. It blows another regulator I have (constant current source driving a shunt regulator) by a large margin.
The summary of my impressions in my original post came from my past 2 weeks of tests of using various caps and resistors in various places.
But I still have not found the right capacitors. These capacitors all have their obvious colourations to varous extents.
Full rectification schottkey diodes bypassed with 0.01uF polystyrene.
3300uF || 0.1uF disc ceramic - 8.2R - 3300uF || 0.12uF polypropylene before the regulator.
One diode is used across the input and output. One diode is used across the adj pin and output pin.
The upper resistor is chosen to be 150R. This is in consideration of providing a higher load to improve regulation.
1.62k is used to set the voltage to about +/-15V.
1.62k is bypassed by the TEST CAP (1). Currently it is the general purpose Xicon 100uF/100v.
On output, one bypass TEST CAP (2) is in series with a TEST RESISTOR (3). I tried various TEST CAPs ranging from 25uF to 2200uF with various impedance. The TEST RESISTOR ranged from 0R to 0.5R. The current (final?) setting is a general purpose Xicon 100uF/100V + 0.44R.
Because I use 120uF Rubycon ZL || 0.01uF MKP to bypass each rail immediately at the power supply pins on each opamp (6 dual opamps in total in my implementation of the NaO circuit), this would create a huge resonance on the power rails (it was very audible!).
So a TEST resistor (4) is connected in serious with the regulator to remove the resonance. The test value ranged from 0R to 1R. Currently at 0.94R.
Since I don't want 1R output impedance from the power supply, I added another TEST cap (5) (currently a Panasonic FC 4700uF/25v) to provide a low impedance (rated about 0.015R at 100kHz and should be low with ESR at lower frequencies due to the size of the cap) to the PSU.
The idea is that the regulator has a 0.5uH output inductance so it never wants to see capacitors without using some resistance (up to 1R value should be sufficient to dampen the resonances) so I created that environment for the regulator. I then use very big capacitors (I may increase it to 10,000uF) that will work better than batteries to supply a very low and flat impedance to the opamps for much improved transient response than without a large cap.
So far, this sounds the best. It blows another regulator I have (constant current source driving a shunt regulator) by a large margin.
The summary of my impressions in my original post came from my past 2 weeks of tests of using various caps and resistors in various places.
But I still have not found the right capacitors. These capacitors all have their obvious colourations to varous extents.
Comparing electrolytic capacitors
I perform a complete listening test about electrolytic capacitors : Rubycon ZL, Panasonic FC, Elna Silmic II, Black Gate, Nichicon UHD, Panasonic ECA...
I compare various parameters : softness, tone realism, definition, bass level, bass depth, space clarity...
I publish results on my web site : Comparing electrolytic capacitors
Eric
I perform a complete listening test about electrolytic capacitors : Rubycon ZL, Panasonic FC, Elna Silmic II, Black Gate, Nichicon UHD, Panasonic ECA...
I compare various parameters : softness, tone realism, definition, bass level, bass depth, space clarity...
I publish results on my web site : Comparing electrolytic capacitors
Eric
Hi HiFiNutNut,
LM317, understanding and listening
Each capacitors around LM317 greatly improve or degrade sound.
Eric
I wrote an another article about LM317 sounding.HiFiNutNut said:
LM317, understanding and listening
Each capacitors around LM317 greatly improve or degrade sound.
Eric
Eric and all, I'm completely skeptical of these listening results. Unless there is some RF problem because the PS diodes aren't bypassed right at the diode, or unless there is some ground return issue, i.e., layout, that's affected by different current waveforms resulting from the different esrs, I don't see how it's possible for caps that are bypassed with films, and live ahead of the regulator, to affect sound. If you put a scope on the regulator output, with a high gain preamp if necessary, what do you see when you play music? If there's no difference in the voltage being fed to the circuitry, can someone explain some mechanism by which the sound changes? OTOH, if there is some difference, it would seem wiser to investigate that, put some numbers on it, thereby having a more reliable and faster way to evaluate caps, or maybe design a better regulator circuit.
Please read the following articles.
There are many capacitor myths that deserve to be put to rest.
Cyril Bateman has gratiously made his capacitor measurement articles available for free on his own website..
http://uk.geocities.com/cyrilb2@btinternet.com/
(That is a limited bandwidth web host. If you cannot connect, wait an hour or so and try again)
ZAP
There are many capacitor myths that deserve to be put to rest.
Cyril Bateman has gratiously made his capacitor measurement articles available for free on his own website..
http://uk.geocities.com/cyrilb2@btinternet.com/
(That is a limited bandwidth web host. If you cannot connect, wait an hour or so and try again)
ZAP
Hi zapnspark,
Lest's have a summary:
FORR - Good ceramic capacitors Post #1725 - august 23, 2007
In his capacitor sound series, Cyril Bateman wrote : " ...COG : temperature coefficient : +/- 30 ppm [...]
COG ceramic provides the most stable capacitor value over long time periods and temperature excursions. It is frequently used as a capacitance transfer standard in calibration labatories [...]
For small, low distorsion up to 10 nF, my personal choice would be COG ceramic, also including discs up to N750, extended fois/PS and extended foil/PP, with the leadout wires soldered to the eletrodes..."
His measurements for a 10 nF class 1 COG ceramic at 100 or 1000 Hz, with a signal of 4 Vrms and 18 BV bias across it gave an total harmonic distorsion of 0.00009%. In the same conditions, a class 2 X7R gave 0.14495%.
REINHARD
How can he measure 0,00009% distortion?
STOOLPIGEON
The same way he measured 0.14495%
JOHN CURL
We discovered problems with distortion in ceramic caps about 34 years ago.
Tektronix accidently found that a measurement modification to the 577 curve tracer showed the ceramic problems clearly. They showed this to me in Feb. 1974. I told Bruel & Kjaer within the year, and a section was added to a technical paper showing the distortion. Later, in 1978, I presented an IEEE paper 'Omitted Factors in Audio Design' that showed distortion in both tantalum and ceramic caps. Walt Jung was at that IEEE conference and took note. It took many years for people to try, and even admit that there was a problem with typical ceramic caps. The better ceramic caps are EXPENSIVE, VERY LARGE, AND SMALL VALUED. Therefore, they are not often used. Their greatest asset is low tempco.
I have avoided using ceramic caps for the last 34 years, once that I realized that it was a source of 3 separate distortions: Normal IM or Harmonic, DA, and 'Non-return to zero' distortion. A person has to be almost reckless or non-caring to use these caps in critical audio pathways. The fix is to use a different type of cap with lower distortion.
The best caps are usually polystyrene and teflon, but they cost real money, and take space. Another alternative is to design the caps out of the series path and use servos to keep the output, DC stable.
Eric
There is funny discussion you can find in this forum about using ceramics capacitors in audio Post #1725 from Blowtorch preamplifierzapnspark said:
Lest's have a summary:
FORR - Good ceramic capacitors Post #1725 - august 23, 2007
In his capacitor sound series, Cyril Bateman wrote : " ...COG : temperature coefficient : +/- 30 ppm [...]
COG ceramic provides the most stable capacitor value over long time periods and temperature excursions. It is frequently used as a capacitance transfer standard in calibration labatories [...]
For small, low distorsion up to 10 nF, my personal choice would be COG ceramic, also including discs up to N750, extended fois/PS and extended foil/PP, with the leadout wires soldered to the eletrodes..."
His measurements for a 10 nF class 1 COG ceramic at 100 or 1000 Hz, with a signal of 4 Vrms and 18 BV bias across it gave an total harmonic distorsion of 0.00009%. In the same conditions, a class 2 X7R gave 0.14495%.
REINHARD
How can he measure 0,00009% distortion?
STOOLPIGEON
The same way he measured 0.14495%
JOHN CURL
We discovered problems with distortion in ceramic caps about 34 years ago.
Tektronix accidently found that a measurement modification to the 577 curve tracer showed the ceramic problems clearly. They showed this to me in Feb. 1974. I told Bruel & Kjaer within the year, and a section was added to a technical paper showing the distortion. Later, in 1978, I presented an IEEE paper 'Omitted Factors in Audio Design' that showed distortion in both tantalum and ceramic caps. Walt Jung was at that IEEE conference and took note. It took many years for people to try, and even admit that there was a problem with typical ceramic caps. The better ceramic caps are EXPENSIVE, VERY LARGE, AND SMALL VALUED. Therefore, they are not often used. Their greatest asset is low tempco.
I have avoided using ceramic caps for the last 34 years, once that I realized that it was a source of 3 separate distortions: Normal IM or Harmonic, DA, and 'Non-return to zero' distortion. A person has to be almost reckless or non-caring to use these caps in critical audio pathways. The fix is to use a different type of cap with lower distortion.
The best caps are usually polystyrene and teflon, but they cost real money, and take space. Another alternative is to design the caps out of the series path and use servos to keep the output, DC stable.
Eric
As John says, C0G ceramics can be big, expensive and only come in small values. They do, however, better their film counterparts on almost every parameter. IMO, the rest of the ceramic family is useless for anything other than very high frequency bypassing applications. Regardless, I try to keep caps out of the signal path, electrolytics in particular, but that's not what we're talking about here. Filter caps ahead of regulators are not coupling caps and are not in the signal path unless you use a very liberal definition of signal path. It's interesting that everyone is pretty much in agreement that simple distortion figures have near to nothing to do with sound quality, yet they pop right up in a discussion about the subtleties of different caps ;-) Bateman's studies were interesting, but I believe he wisely avoided broad statements about sound quality connected to the measurements. (I haven't re-read the articles since they were published) If there's an actual sound difference, there has to be a reason along one of the lines I mentioned earlier- RF problems, bad layout, or insufficient regulation. These are all things you can determine on the test bench, whereas assigning a sound quality description or rating to different brand names (where the design and manufacture could change at any time) is side stepping the real issues.
One thing Bateman didn't touch on was microphonics. Garden variety ceramics are terrible in this respect. Pretty much all caps have it to some degree. C0Gs are not nearly as bad as regular ceramics, because they don't use the same piezo ceramic as the garden variety. However much better than regular ceramics the C0Gs are in this respect, I have not seen any factual data or reliable sonic comparisons made, either against film caps or the other ceramics. I think the evidence from 34 years ago did not include C0Gs in the comparisons or other studies. If they did, it was not evident, so grouping them with regular ceramics would not seem to be fair.
Conrad,
I agree with your approach completely.
I used one 0.01uF polystyrene snubber bypassing each diode only because I did not have 0.01u ceramic at the time but had a bucket of polystyrene. The use of 0.1uF disc ceramic before the regulator was only to suppress RF if any, since RF would pass through the regulator unsuppressed. I did not expect these caps would effect within the audio bandwidth. They are not within the scope of my "cap test".
Stability is the number one issue here. Therefore, I ensured that the regulator output, which is inductive, would see some ESR/resistance to damp any resonances from the LCR circuit. 0.5ohm to 1ohm resistance should be sufficient to ensure stability.
I used 1ohm resistance at the output of the regulator to ensure stability. But since the line level XO circuit would like to see a low and flat impedance, after the resistor I had a 4700uF capacitor to serve that purpose. Or actually, the 1R resistor is added to ensure the proper use of a 4700uF cap. So the low impedance is not provided by the regulator but by the large capacitor. Good Transient response is provided by the large reservoir capacitor.
Any problems you can see that would invalidate the result of the cap tests? I am happy to listen to your opinion.
All,
Thank you for your valuable posts. They have been very helpful.
Regards,
Bill
I agree with your approach completely.
An externally hosted image should be here but it was not working when we last tested it.
I used one 0.01uF polystyrene snubber bypassing each diode only because I did not have 0.01u ceramic at the time but had a bucket of polystyrene. The use of 0.1uF disc ceramic before the regulator was only to suppress RF if any, since RF would pass through the regulator unsuppressed. I did not expect these caps would effect within the audio bandwidth. They are not within the scope of my "cap test".
Stability is the number one issue here. Therefore, I ensured that the regulator output, which is inductive, would see some ESR/resistance to damp any resonances from the LCR circuit. 0.5ohm to 1ohm resistance should be sufficient to ensure stability.
I used 1ohm resistance at the output of the regulator to ensure stability. But since the line level XO circuit would like to see a low and flat impedance, after the resistor I had a 4700uF capacitor to serve that purpose. Or actually, the 1R resistor is added to ensure the proper use of a 4700uF cap. So the low impedance is not provided by the regulator but by the large capacitor. Good Transient response is provided by the large reservoir capacitor.
Any problems you can see that would invalidate the result of the cap tests? I am happy to listen to your opinion.
All,
Thank you for your valuable posts. They have been very helpful.
Regards,
Bill
As you can see from the above diagram, I have 3 places for test caps.
Each brand/value of capacitor in each of the test place gives consistent pattern of their characteristic sound based on the brand. For example, the Elna Silmic has a smooth but "tube" like, strong colourations in each of the places. I can accept any challenge of any blind tests to identify the "colourful" Silmic. Other caps would have much less differences.
Each brand/value of capacitor in each of the test place gives consistent pattern of their characteristic sound based on the brand. For example, the Elna Silmic has a smooth but "tube" like, strong colourations in each of the places. I can accept any challenge of any blind tests to identify the "colourful" Silmic. Other caps would have much less differences.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.