Hi
I know that electrolytic caps have about the worst characteristics for audio - especially in series with the signal - but...
I see highly respected manufacturers using what look like electros as coupling caps between gain blocks. Are these caps possibly the hybrid polymer types instead of straight electrolytics? I friend found that the leakage through the polymer caps was low enough he could use them in series with a feedback gain control since he did not have space for an equal value of plastic cap.
In any case, has anyone tested the hybrid polymers for audio?
I was puzzled when I did a circuit sim and the THD20 was lower than the THD1. Then I increased the value of the input coupling cap and it was much better. However, I had to make a fairly massive increase which makes using polypropylene (preferred) a lot more space consuming.
Thanks for any advice
I know that electrolytic caps have about the worst characteristics for audio - especially in series with the signal - but...
I see highly respected manufacturers using what look like electros as coupling caps between gain blocks. Are these caps possibly the hybrid polymer types instead of straight electrolytics? I friend found that the leakage through the polymer caps was low enough he could use them in series with a feedback gain control since he did not have space for an equal value of plastic cap.
In any case, has anyone tested the hybrid polymers for audio?
I was puzzled when I did a circuit sim and the THD20 was lower than the THD1. Then I increased the value of the input coupling cap and it was much better. However, I had to make a fairly massive increase which makes using polypropylene (preferred) a lot more space consuming.
Thanks for any advice
Type II ceramic caps are the worst for audio, far worse than electrolytics (other than for longevity).
Electrolytics used as coupling caps can perform extremely well, so long as they have adequate capacitance that the voltage drop across them is tiny across the audio band, i.e. that its roll-off frequency is well below 20Hz (maybe 0.1 to 0.5Hz). With only a tiny voltage across the cap its non-linearity is greatly diluted. Bipolar types typically are the best for linearity when there isn't any DC bias across the cap.
Why would hybrid electrolytics not be suitable for audio - they have good specifications...
Electrolytics used as coupling caps can perform extremely well, so long as they have adequate capacitance that the voltage drop across them is tiny across the audio band, i.e. that its roll-off frequency is well below 20Hz (maybe 0.1 to 0.5Hz). With only a tiny voltage across the cap its non-linearity is greatly diluted. Bipolar types typically are the best for linearity when there isn't any DC bias across the cap.
Why would hybrid electrolytics not be suitable for audio - they have good specifications...
Yes. For VLN work, use the highest voltage cheapo electrolytics you can fit in the space.
DON'T USE TANTS. OR POLYMERS or anything fancy.
Check out this huge thread for evidence
richard-lees-ultra-low-noise-mc-head-amp
Hybrid polymers and similar capacitors have higher leakage currents. This may or may not be important in your circuit. I would stick with electrolytic capacitors, good manufacturers from authorised distribution networks. They fake everything there days!
I have tried them in some audio signal locations and noticed no difference in distortion. In power supply bypass applications they shine. Give a lot of voltage headroom. I don't trust them until they have been in service for a while. Memories of solid Tantalum capacitors, wet slug tantalum caps are excellent.
Solid tantalum capacitors are terrible for audio, right down there with high K ceramics. However, NP0 / C0G ceramic capacitors are excellent.
I have tried them in some audio signal locations and noticed no difference in distortion. In power supply bypass applications they shine. Give a lot of voltage headroom. I don't trust them until they have been in service for a while. Memories of solid Tantalum capacitors, wet slug tantalum caps are excellent.
Solid tantalum capacitors are terrible for audio, right down there with high K ceramics. However, NP0 / C0G ceramic capacitors are excellent.
I mean, the ear decides - diy audio. All that remains is to buy and listen.
At the moment I have two three electrolytic capacitors, which I prefer to any audiophile capacitors. The advantage lies, for example, in the smaller size - since current is not taken into account when developing capacitors, the current deviations are smaller, which I think is audible;-)
At the moment I have two three electrolytic capacitors, which I prefer to any audiophile capacitors. The advantage lies, for example, in the smaller size - since current is not taken into account when developing capacitors, the current deviations are smaller, which I think is audible;-)
cumbb, this may explain your confusion.
Sure the ear can decide - on a day to day basis I guess. I could go on about why some folks prefer this and that. We can argue about it. Obviously if something does sound good, you don't listen to it. But things that sound good measure well also.
Your ears and measured response all make up information. To ignore one is foolish. They keep each other on track. Not unless you don't like what one is telling you I guess, and it can go either way.
Sure the ear can decide - on a day to day basis I guess. I could go on about why some folks prefer this and that. We can argue about it. Obviously if something does sound good, you don't listen to it. But things that sound good measure well also.
Your ears and measured response all make up information. To ignore one is foolish. They keep each other on track. Not unless you don't like what one is telling you I guess, and it can go either way.
Only motherboards seem to use them in switching applications, very high frequency compared to audio.
So finding them in small quantities, and how the result is a matter that can be experimented with.
For a small quantity..well it is your money, and your time, go right ahead and tell us your results, with details of the test equipment used.
I have seen them only in small values and small voltages, typically 5V, 6.3V 16V or 25V, good for pre amp.
Costs are not known.
So finding them in small quantities, and how the result is a matter that can be experimented with.
For a small quantity..well it is your money, and your time, go right ahead and tell us your results, with details of the test equipment used.
I have seen them only in small values and small voltages, typically 5V, 6.3V 16V or 25V, good for pre amp.
Costs are not known.
Use them where you need low ESR or very good high frequency ripple current capability. I use them extensively in my power supply designs, but I have never used them as coupling caps in audio circuits. (I am still debating the issue and leakage current may be a concern,)
I generally use Panasonic, Nichicon, Kemet or Wurth polymer or hybrid/polymer up to about 100uF. I like them for power rail decoupling in my discrete op-amps and power amplifier designs. No complaints.
Kemet and Wurth are very cost effective (not much more than standard electrolytics from U.S. resellers) and offer good performance, where I need the highest current ratings and lowest ESR I tend to use Panasonic or Nichicons which are a bit more expensive. Some through hole cap families are being discontinued, but I am happy to use SMD types as well when that fits my design objectives.
I generally use Panasonic, Nichicon, Kemet or Wurth polymer or hybrid/polymer up to about 100uF. I like them for power rail decoupling in my discrete op-amps and power amplifier designs. No complaints.
Kemet and Wurth are very cost effective (not much more than standard electrolytics from U.S. resellers) and offer good performance, where I need the highest current ratings and lowest ESR I tend to use Panasonic or Nichicons which are a bit more expensive. Some through hole cap families are being discontinued, but I am happy to use SMD types as well when that fits my design objectives.
Hi Guys
Thank you very much for all the replies 😀
I found some info from Doug Self where he made a low-pass filter using an eletrolytic in series with the signal and a 1k to ground following it. Using a low value of 47uF THD was pretty terrible, although he was using a 10Vrms test signal. Higher amplitudes caused quick jumps in THD.
The important factor is as mentioned above in post-2, keeping the voltage across the cap as low as possible. D.S. does not say this, but I think the lost signal voltage over the cap could be thought of as an "error voltage", determined primarily by the ESR of the cap.
The easiest way to reduce the signal voltage across the cap is to increase the value, just as my sim showed. This pushes the low-frequency roll-off being created by the cap and load to way below the audio limit. It is not good enough to set the -3dB point at 20Hz, as the circuit is not flat until 200Hz. So, you could set it to 2Hz and it will be flat at 20Hz, but even this is not good enough to reduce the THD from the cap. I began with a low value of 10uF (to be able to use polypropylene) and ended up with 1mF. When Self went to 1mF the cap voltage was 80mVrms and THD was at the limit of the AP test system! That's pretty cool and explains why some of the big-name manufacturers I respect are using electros in the signal path.
I did search on Mouser and Digikey and found 400uF and higher value PP caps intended for DC-link duty. These are $150+ each. To get the THD1 down to the THD20 level in my sim, I needed 1mF - 800uF was close, but not quite there. So, now I am saved by what seems to be a nice cheap alternative. Or am I?
The service life of electrolytic caps is a maximum of 14 years if you want the best performance, regardless of what the life cycle calculation for your app suggests. The cap can function for decades if used regularly, but its DA and DF will go crazy. Not a problem in a guitar amp maybe, but could be for hifi.
My understanding is that electrolytics need regular application of voltage across them to maintain their dielectric. Is 80mV enough? I know the cap will reform over time to whatever voltage it is actually being subjected to. If I use an even higher value cap and possibly find one with super-low ESR, will there be enough mV to maintain the cap's health?
Thanks
Thank you very much for all the replies 😀
I found some info from Doug Self where he made a low-pass filter using an eletrolytic in series with the signal and a 1k to ground following it. Using a low value of 47uF THD was pretty terrible, although he was using a 10Vrms test signal. Higher amplitudes caused quick jumps in THD.
The important factor is as mentioned above in post-2, keeping the voltage across the cap as low as possible. D.S. does not say this, but I think the lost signal voltage over the cap could be thought of as an "error voltage", determined primarily by the ESR of the cap.
The easiest way to reduce the signal voltage across the cap is to increase the value, just as my sim showed. This pushes the low-frequency roll-off being created by the cap and load to way below the audio limit. It is not good enough to set the -3dB point at 20Hz, as the circuit is not flat until 200Hz. So, you could set it to 2Hz and it will be flat at 20Hz, but even this is not good enough to reduce the THD from the cap. I began with a low value of 10uF (to be able to use polypropylene) and ended up with 1mF. When Self went to 1mF the cap voltage was 80mVrms and THD was at the limit of the AP test system! That's pretty cool and explains why some of the big-name manufacturers I respect are using electros in the signal path.
I did search on Mouser and Digikey and found 400uF and higher value PP caps intended for DC-link duty. These are $150+ each. To get the THD1 down to the THD20 level in my sim, I needed 1mF - 800uF was close, but not quite there. So, now I am saved by what seems to be a nice cheap alternative. Or am I?
The service life of electrolytic caps is a maximum of 14 years if you want the best performance, regardless of what the life cycle calculation for your app suggests. The cap can function for decades if used regularly, but its DA and DF will go crazy. Not a problem in a guitar amp maybe, but could be for hifi.
My understanding is that electrolytics need regular application of voltage across them to maintain their dielectric. Is 80mV enough? I know the cap will reform over time to whatever voltage it is actually being subjected to. If I use an even higher value cap and possibly find one with super-low ESR, will there be enough mV to maintain the cap's health?
Thanks
I've just quickly swapped my input electrolytic capacitors for these:
https://de.rs-online.com/web/p/poly...HNVns2Uo1oJRdDWWoD2a3x0WsK0GcHnWTNziTkIXXtuLX
They sound a bit doughy, artificial, covered, somewhat lifeless and boomy - similar to the ones I use. They're not the worst I've heard, but I wouldn't use them for audio purposes. That's not to say that there aren't polymers suitable for audio. As always, just buy a handful and listen: diy audio;-)
https://de.rs-online.com/web/p/poly...HNVns2Uo1oJRdDWWoD2a3x0WsK0GcHnWTNziTkIXXtuLX
They sound a bit doughy, artificial, covered, somewhat lifeless and boomy - similar to the ones I use. They're not the worst I've heard, but I wouldn't use them for audio purposes. That's not to say that there aren't polymers suitable for audio. As always, just buy a handful and listen: diy audio;-)
Lore is that polymer capacitors have a very high leakage current.
Funny part is that modern hybrid polymer capacitors have several times lower leakage than high quality standard electrolytic capacitors. Though, manufacturers still specify outrageously high worst case leakage currents in their data sheets.
https://www.diyaudio.com/community/threads/super-regulator.247281/post-7299600
For rails decoupling or voltage reference filtering, I use only hybrid polymer capacitors, as they provide best performance.
As already said, I would not use them for signal coupling.
Funny part is that modern hybrid polymer capacitors have several times lower leakage than high quality standard electrolytic capacitors. Though, manufacturers still specify outrageously high worst case leakage currents in their data sheets.
https://www.diyaudio.com/community/threads/super-regulator.247281/post-7299600
For rails decoupling or voltage reference filtering, I use only hybrid polymer capacitors, as they provide best performance.
As already said, I would not use them for signal coupling.
Yes, their leakage is lower than we were lead to believe, but I'm not sure over time and temperature. I'm not as concerned but still keep it in mind.
Hi nauta,
Signal levels at that point in the circuit at in the mV level and impedances are normally 1K or higher. As for other capacitor types, the very large physical sizes bring other issues. Finally, at 20 Hz, the distortion for the speaker is pretty high and most roll off at well over 30 Hz. So worrying about THD at 20Hz isn't a practical concern within reason. Then we have very little musical content down there anyway. I wonder what the distortion for the mic is?
Hi nauta,
Signal levels at that point in the circuit at in the mV level and impedances are normally 1K or higher. As for other capacitor types, the very large physical sizes bring other issues. Finally, at 20 Hz, the distortion for the speaker is pretty high and most roll off at well over 30 Hz. So worrying about THD at 20Hz isn't a practical concern within reason. Then we have very little musical content down there anyway. I wonder what the distortion for the mic is?
I use the wurth hybrid polymer a lot, particularly when building up Paradises for people. They are great on the shunt, small size, high temp, long life, leakage has never been an issue. I prefer FC or FM for the rail decoupling on the current mirrors though. I can't pin down why, electrically at least, they just seem to sound better built this way.
Hi
THD1 is THD at 1kHz
THD20 is THD at 20kHz
These seem to be standard references used by Cordell and Self.
In my sim the THD at 20kHz was much lower than at 1kHz until I made the input cap value 1mF (1,000uF). Input leak resistor is 1k to ground.
Yes, leakage currents were concerning me last night looking at specs for 10mF and 33mF** caps. I wondered if using anti-parallel pairs might improve things as it does in DC mains blockers? or just the use of parallel caps instead of one big one? Would the leakage currents add together?
I assume leakage current is not fixed, that maybe it depends on the voltage across the cap? and temperature? If there is a 1Vrms signal feeding the cap, then the signal current would be 1mArms, and I think this would be equivalent to the ripple current in a filtering application, but at multiple frequencies and varying amplitude. One-milliamp won't raise the internal temp of the cap noticeably if at all, so the temperature part of leakage current would be insignificant hopefully?
As I said in my first post, it was leakage current that caused an issue in a circuit with a gain pot, where the pot was "scratchy" with the electro but quiet with a plastic cap and also with a polymer cap. I believe Self has demonstrated that electros can be used in a way that does not effect THD, but he never mentioned anything about leakage current.
**Since a power amp is pretty much like a high-power opamp, I figure there should be enough DC gain to keep output offset voltage in the uV range, so tend to roll off the amp gain using a cap in the shunt part of the feedback network. However, I do not use it to set the low-frequency audio limit, and use 33mF more or less as my standard. I used to use 10mF but then began reducing the resistor values in the feedback network and 1kHz THD would rise without the shunt cap being increased to compensate.
Self decries the use of electros as audio roll-offs, yet he did that in all of his designs up until the late 1990s.
THD1 is THD at 1kHz
THD20 is THD at 20kHz
These seem to be standard references used by Cordell and Self.
In my sim the THD at 20kHz was much lower than at 1kHz until I made the input cap value 1mF (1,000uF). Input leak resistor is 1k to ground.
Yes, leakage currents were concerning me last night looking at specs for 10mF and 33mF** caps. I wondered if using anti-parallel pairs might improve things as it does in DC mains blockers? or just the use of parallel caps instead of one big one? Would the leakage currents add together?
I assume leakage current is not fixed, that maybe it depends on the voltage across the cap? and temperature? If there is a 1Vrms signal feeding the cap, then the signal current would be 1mArms, and I think this would be equivalent to the ripple current in a filtering application, but at multiple frequencies and varying amplitude. One-milliamp won't raise the internal temp of the cap noticeably if at all, so the temperature part of leakage current would be insignificant hopefully?
As I said in my first post, it was leakage current that caused an issue in a circuit with a gain pot, where the pot was "scratchy" with the electro but quiet with a plastic cap and also with a polymer cap. I believe Self has demonstrated that electros can be used in a way that does not effect THD, but he never mentioned anything about leakage current.
**Since a power amp is pretty much like a high-power opamp, I figure there should be enough DC gain to keep output offset voltage in the uV range, so tend to roll off the amp gain using a cap in the shunt part of the feedback network. However, I do not use it to set the low-frequency audio limit, and use 33mF more or less as my standard. I used to use 10mF but then began reducing the resistor values in the feedback network and 1kHz THD would rise without the shunt cap being increased to compensate.
Self decries the use of electros as audio roll-offs, yet he did that in all of his designs up until the late 1990s.
When you are talking about such high values of capacitance and such low corner frequencies why not DC couple and use a DC servo if output offset is a concern? (This has generally been my approach, you can use DC offset detection if you are concerned about fry speaker voice coils.)
Hi
The low roll-off frequencies have never presented any issues, at least so far. I am not trying to send subaudio signals to the speakers; rather, moving the rolloffs out of the way to where the caps do not add distortion.
And I am concerned about DC killing speakers, so have DC detect relays set for way below the usual "failure-mode " thresholds. (Most DC protection circuits are set to trip at multiple diode drops, assuming there is a full-rail swing. I want to protect against too much drift, so less than a few hundred millivolts. I use two comparators as a window detector) The first solid-state amp I built killed a brand new speaker. I had no test equipment and basically knew squat about what I was doing. Now I know squat plus an infinitesimal bit.
My concern with the input cap is that I want to block any DC from preceding stages and not have the amp amplifying DC.
The low roll-off frequencies have never presented any issues, at least so far. I am not trying to send subaudio signals to the speakers; rather, moving the rolloffs out of the way to where the caps do not add distortion.
And I am concerned about DC killing speakers, so have DC detect relays set for way below the usual "failure-mode " thresholds. (Most DC protection circuits are set to trip at multiple diode drops, assuming there is a full-rail swing. I want to protect against too much drift, so less than a few hundred millivolts. I use two comparators as a window detector) The first solid-state amp I built killed a brand new speaker. I had no test equipment and basically knew squat about what I was doing. Now I know squat plus an infinitesimal bit.
My concern with the input cap is that I want to block any DC from preceding stages and not have the amp amplifying DC.
Ordinary electrolytics have worked for decades in that application.
Replace on a working amp, hear it, then decide if it does make a difference...technically, same uF and volts, so not much change will happen.
Only factor is how will a high frequency (400 kHz) capacitor work in the 4 kHz area, where most music is played (see your spectrum analyser), and most people are deaf above 14 kHz or so by age 55.
Then decide, I feel it is more a perception thing, like the brand is more important than the content...like B & O used Samhwa capacitors, and the public thought the sound was great...
No ties, just an example.
Replace on a working amp, hear it, then decide if it does make a difference...technically, same uF and volts, so not much change will happen.
Only factor is how will a high frequency (400 kHz) capacitor work in the 4 kHz area, where most music is played (see your spectrum analyser), and most people are deaf above 14 kHz or so by age 55.
Then decide, I feel it is more a perception thing, like the brand is more important than the content...like B & O used Samhwa capacitors, and the public thought the sound was great...
No ties, just an example.
Electrolytic caps as coupling capacitors are fine. Brand (unless defective or junk) means nothing in tht application.
When the capacitor is used to set frequency rolloff, you have, by definition a signal voltage across the capacitor. Everyone agrees this is where capacitor characteristics are far more important. I agree and have measured differences between capacitors in the roll-off area. So no surprises.
I prefer capacitors in the low arm of the feedback network over DC servos any day. One, your roll-off is generally below the program content. Any increased distortion you may imagine or measure is lower than the distortion generated by loudspeakers. Chances of exciting those frequencies are low, and if you do you very probably will not hear distortion. So lot's of worry about a thing that isn't a practical concern. Maybe on paper it is doom and gloom, but not in your listening room.
Servos have a few catches, and this is why I don't like them. If your input stage is well balanced, your DC offset will be less than 50 mV (much less). No need for a servo. If your input stage is well balanced with a capacitor in the feedback to common, DC offsets should be lower than 10 mV without going crazy, lower with good matches. Servos have a time constant to correct offset depending on your low frequency set point. So without a relay you will get a thump on turn-on. If you use a relay and the time constant is longer, you may get a large bang sound. Try putting a relay on most Adcom amps and you'll see what I mean. A DC servo works by injecting a current to unbalance the diff pair. Probably upsetting distortion performance (I haven't checked). That and without a filter on the output of your DC servo, you are getting any noise from that op amp circuit injected in just like a signal. How bad depends on the op amps used.
DC servos were introduced to save manufacturing costs. Firstly they allowed for unmatched transistors in diff pairs (compromising distortion performance), and eliminated the DC offset adjustment step on the manufacturing line. It also eliminated warranty technicians adjustments. Older designs using DC offset trim perform very well and some times better than designs using DC servos.
I can get less than 5 mV DC offset in a Hafler DH200 (no DC offset trim), and this is a complimentary diff pair design. This is just by matching the differential pairs - and you need to do that to get the best performance anyway. What does this tell you about DC servos? They don't do any practical good in a balanced circuit, the can inject noise or may not behave the way you intended if they aren't designed correctly.
When the capacitor is used to set frequency rolloff, you have, by definition a signal voltage across the capacitor. Everyone agrees this is where capacitor characteristics are far more important. I agree and have measured differences between capacitors in the roll-off area. So no surprises.
I prefer capacitors in the low arm of the feedback network over DC servos any day. One, your roll-off is generally below the program content. Any increased distortion you may imagine or measure is lower than the distortion generated by loudspeakers. Chances of exciting those frequencies are low, and if you do you very probably will not hear distortion. So lot's of worry about a thing that isn't a practical concern. Maybe on paper it is doom and gloom, but not in your listening room.
Servos have a few catches, and this is why I don't like them. If your input stage is well balanced, your DC offset will be less than 50 mV (much less). No need for a servo. If your input stage is well balanced with a capacitor in the feedback to common, DC offsets should be lower than 10 mV without going crazy, lower with good matches. Servos have a time constant to correct offset depending on your low frequency set point. So without a relay you will get a thump on turn-on. If you use a relay and the time constant is longer, you may get a large bang sound. Try putting a relay on most Adcom amps and you'll see what I mean. A DC servo works by injecting a current to unbalance the diff pair. Probably upsetting distortion performance (I haven't checked). That and without a filter on the output of your DC servo, you are getting any noise from that op amp circuit injected in just like a signal. How bad depends on the op amps used.
DC servos were introduced to save manufacturing costs. Firstly they allowed for unmatched transistors in diff pairs (compromising distortion performance), and eliminated the DC offset adjustment step on the manufacturing line. It also eliminated warranty technicians adjustments. Older designs using DC offset trim perform very well and some times better than designs using DC servos.
I can get less than 5 mV DC offset in a Hafler DH200 (no DC offset trim), and this is a complimentary diff pair design. This is just by matching the differential pairs - and you need to do that to get the best performance anyway. What does this tell you about DC servos? They don't do any practical good in a balanced circuit, the can inject noise or may not behave the way you intended if they aren't designed correctly.
I only use the solid polymer capacitors in digital applications, and am cautious to avoid interactions with other ceramic parts or inductance on the same rail. Their low esr can really be a deal breaker sometimes, but when they work, they work well.