this feedback network is different from the ones mosten often to be seen. It adresses the noise problem a high Z version of the "standard" network has, and the electrolytic "problems" that come with the usual approach.
Maybe the component values can be adjusted to reliably use this also with a BJT input.
Maybe the component values can be adjusted to reliably use this also with a BJT input.
Ultimative knowledge about caps - Mr. Jens Both: 25 Years Experience
have a look to this website (unfortunately only in German)
Elko-Consulting:
Elko-Consulting
good description about a wide range of problems of caps and the reasons;
also many pics arround exploded caps
have a look to this website (unfortunately only in German)
Elko-Consulting:
Elko-Consulting
good description about a wide range of problems of caps and the reasons;
also many pics arround exploded caps
Statement from Douglas Self about Capacitors - online for reading
Very interesting facts arround capacitors in audio applications -
go to page 51 from the read sample about
Small Signal Audio Design - Google Bücher
(new Book: "small signal audio design")
Very interesting facts arround capacitors in audio applications -
go to page 51 from the read sample about
Small Signal Audio Design - Google Bücher
(new Book: "small signal audio design")
May i resume some of what we know for sure, and good practices, about electrolytic caps?
Good ones can live for decades. But, when it comes to service a defect equipment, you will find that, most of the time, the faulty piece is an electrolytic capacitor and lateral consequences of its failure (dead transistor or so).
they can die suddenly in short circuit (rare if no over-voltage ) or slowly goes to 0 µf if they dry out (you can be sure if you see traces). They can explode if high tension, reverse voltage is applied to them: Care their sens as you do for diodes.
Her values are not precise (lets say -50+100% is tolerance of cheap ones) and can decrease with time (depending of their quality, you cannot be sure about). So, for sure, you cannot use them in filters.
Don't spend time in calculations; X2 their needed values.
Their capacitance varies with the frequency. (decrease with high frequencies above 1000hz), can have some selfic characteristics.
They need to be polarized, and never see an inverse tension to their wires.
But, they have very good performance, regarding their size and price in front of their capacity.
Knowing all that can help-you to properly use them, and design your circuits in consequence.
There is no other choice when it comes to design power supply's, filtering and DC power reserve. Never forget to add some polypropylene cap in parallel for keeping fast current call to low impedance. A good practice is to use 1/100 values.
Let's take,as an example, a 10 000 µf value required. I suggest you use a 100µf, electrolytic in parallel. Then, a 100nf polypro. (Even if you use 100nf for decoupling power rails near each OP amp + or - pins.).
You can use a combination of 10X1000µf if it is less expensive or better for volume.
It will be slightly better because it parallels serial resistances inside caps, and random reliability and values of each. Money will be your major requirement.
It is a good practice for several reason to limit high current shocks during power on when using over-sized transformer and huge µf values. Insert a well calculated (both in value and power) power resistance in serial between diode bridge and caps, and use a low resistive relay to bypass this resistance after a well calculated delay (2s to reach 80% of the tension value ?).
Forget to use electrolytic caps in audio filters, baxendals, passive Hp filters, whatever the price.
When it comes to use an electrolytic cap in the signal path (it is ridiculous to buy some huge 40$ paper cap just to stop DC in your output), make sure polarization plus or minus peak signal never apply a inverse polarity on the cap. If you cannot, use non polarized electrolytic caps.
Again, use 1/10values in parallel, finishing with a 100nf polypro.
Never use low values, when you can use affordable polypro instead.
Never use them in very low signal levels. (input stage of mike preamps). Change your design if needed.
Don't spend your money with very expensive esoteric kind of black stuff, Don't spend your money with expensive aluminum caps rated at high temperature if you equipment will not reach those temps. Save your money for beers.
Never believe in magic.
Take something like a 30% margin in service's voltage.
Change your electrolytic caps after 10 years.
Listen to good music on your well designed, high end, low distortion, high fidelity but low price equipment with electrolytic caps inside. ;-)
Good ones can live for decades. But, when it comes to service a defect equipment, you will find that, most of the time, the faulty piece is an electrolytic capacitor and lateral consequences of its failure (dead transistor or so).
they can die suddenly in short circuit (rare if no over-voltage ) or slowly goes to 0 µf if they dry out (you can be sure if you see traces). They can explode if high tension, reverse voltage is applied to them: Care their sens as you do for diodes.
Her values are not precise (lets say -50+100% is tolerance of cheap ones) and can decrease with time (depending of their quality, you cannot be sure about). So, for sure, you cannot use them in filters.
Don't spend time in calculations; X2 their needed values.
Their capacitance varies with the frequency. (decrease with high frequencies above 1000hz), can have some selfic characteristics.
They need to be polarized, and never see an inverse tension to their wires.
But, they have very good performance, regarding their size and price in front of their capacity.
Knowing all that can help-you to properly use them, and design your circuits in consequence.
There is no other choice when it comes to design power supply's, filtering and DC power reserve. Never forget to add some polypropylene cap in parallel for keeping fast current call to low impedance. A good practice is to use 1/100 values.
Let's take,as an example, a 10 000 µf value required. I suggest you use a 100µf, electrolytic in parallel. Then, a 100nf polypro. (Even if you use 100nf for decoupling power rails near each OP amp + or - pins.).
You can use a combination of 10X1000µf if it is less expensive or better for volume.
It will be slightly better because it parallels serial resistances inside caps, and random reliability and values of each. Money will be your major requirement.
It is a good practice for several reason to limit high current shocks during power on when using over-sized transformer and huge µf values. Insert a well calculated (both in value and power) power resistance in serial between diode bridge and caps, and use a low resistive relay to bypass this resistance after a well calculated delay (2s to reach 80% of the tension value ?).
Forget to use electrolytic caps in audio filters, baxendals, passive Hp filters, whatever the price.
When it comes to use an electrolytic cap in the signal path (it is ridiculous to buy some huge 40$ paper cap just to stop DC in your output), make sure polarization plus or minus peak signal never apply a inverse polarity on the cap. If you cannot, use non polarized electrolytic caps.
Again, use 1/10values in parallel, finishing with a 100nf polypro.
Never use low values, when you can use affordable polypro instead.
Never use them in very low signal levels. (input stage of mike preamps). Change your design if needed.
Don't spend your money with very expensive esoteric kind of black stuff, Don't spend your money with expensive aluminum caps rated at high temperature if you equipment will not reach those temps. Save your money for beers.
Never believe in magic.
Take something like a 30% margin in service's voltage.
Change your electrolytic caps after 10 years.
Listen to good music on your well designed, high end, low distortion, high fidelity but low price equipment with electrolytic caps inside. ;-)
Last edited:
That's a bad practice based on urban legends, as has been demonstrated again and again.
Adding small value ceramic or film caps in parallel with larger ones brings nothing else than resonances.
Here are some examples:
http://www.diyaudio.com/forums/powe...lm-caps-electrolytic-caps-17.html#post2257381
Other can be found on the forum.
May i resume some of what we know for sure, and good practices
Having tested a quite great number of electrolytics, I have to disagree with some of what Esperado exposed.
It is quite curious that the fact of resonance remarked above by Elvee to be still unknown by audiophiles.
For performances with time, read this Safco document (from my experience, the products of this manufacturer are first class)
sic safco catalogue 2009 pdf free ebook download from www.alliedinter.com
In a general manner, I'never found the capacitance value of electrolytics to decrease with time, what is affected (not that much with good caps) is the tan-delta which I measure with Cyril Bateman's instrument. Here, some of my findings (in french) :
Des condensateurs chimiques
More recently, I tested passive components for loudspeakers.
As stated by Bateman, bipolar caps are remarkable components. I noticed that many of caps used in passive crossover, be they bipolar or non-polarised, may have for greater than expected (in french too) :
Des composants passifs pour filtres | A comme Audio
Having tested a quite great number of electrolytics, I have to disagree with some of what Esperado exposed.
It is quite curious that the fact of resonance remarked above by Elvee to be still unknown by audiophiles.
For performances with time, read this Safco document (from my experience, the products of this manufacturer are first class)
sic safco catalogue 2009 pdf free ebook download from www.alliedinter.com
In a general manner, I'never found the capacitance value of electrolytics to decrease with time, what is affected (not that much with good caps) is the tan-delta which I measure with Cyril Bateman's instrument. Here, some of my findings (in french) :
Des condensateurs chimiques
More recently, I tested passive components for loudspeakers.
As stated by Bateman, bipolar caps are remarkable components. I noticed that many of caps used in passive crossover, be they bipolar or non-polarised, may have for greater than expected (in french too) :
Des composants passifs pour filtres | A comme Audio
You are perfectly true. You will have a resonance peak. Reason why i suggest to use a minimum of 1/100 of the values oh the highest capacitance, and better: 1/10. This will bring several little resonance peaks of little level, and a global lower impedance at high frequencies. And i agree, sometimes the cure can be worse than the disease.
In fact all depends of the charge. Do the measurements on the caps themself, at heavy and normal loads, and real conditions, and try to minimize the noise peaks.Damping resistances in serial with the little capacitors can help.
Do the measurement yourself, it is so easy...
If you are in concern with high slew rates amps, or fast switching at high frequency, like class D amps, you canl see, and listen, improvement with adequate values... or not....
it is a very controversial subject, as it depends a lot of the real environment. Charges are very complex equivalent circuits. Even a litle printed circuit line is selfic.
I was sure it will be remarks at this matter. A way to bring precisions about that ?
Last edited:
About the way electrolityc caps varies with time, it varies a lot with the way they are build and even each sample. My remark is build on the time i was working in the research and development office of a audio equipment manufacturer. Based on the Sic Safco and Sef caps we used, measurements made on the machines returned for service.
Sef ones where worse than Sic ones.
May-be i cannot generalize. Anyway, it is a safe practice to double the required value, don't you agree ?
As a matter of fact, good practices will not be the same if, as a manufacturer, you have to deal with statistics, or if you are working on a single prototype.
Sef ones where worse than Sic ones.
May-be i cannot generalize. Anyway, it is a safe practice to double the required value, don't you agree ?
As a matter of fact, good practices will not be the same if, as a manufacturer, you have to deal with statistics, or if you are working on a single prototype.
In the same spirit, if you are building for yoursself your poor man loudspeaker passive system, you can get good result using non polarized electrolytic caps, as long you make sure its value is as expected (tuning with parallel values or choosing between samples). And verify, if you have the feeling that something have changed in your listening experience after some times , that your filter still cut the same way it was supposed.
Based on my experience, the only reason why i chose expensive non electrolytic caps for passive filtering is low tolerance, and stability with the (very) long time. Not real listening differences.
I wonder why electrolitics caps have such a bad reputation in the audiophile market.
Based on my experience, the only reason why i chose expensive non electrolytic caps for passive filtering is low tolerance, and stability with the (very) long time. Not real listening differences.
I wonder why electrolitics caps have such a bad reputation in the audiophile market.
For lowest distortion, Self advocates using a value 2 to 4X larger than the "calculated" value
If a component has a non-linear beheviour, the less AC voltage across it, the less the distorsion it introduces. So a low impedance is always preferable. This is not possible in analog filters which are the audio circuits where the highest AC voltages are developped across caps, they require to be of high quality.
I wonder why electrolitics caps have such a bad reputation in the audiophile market.
Just as me. Not expensive enough ?
If a component has a non-linear beheviour, the less AC voltage across it, the less the distorsion it introduces. So a low impedance is always preferable. This is not possible in analog filters which are the audio circuits where the highest AC voltages are developped across caps, they require to be of high quality.
I wonder why electrolitics caps have such a bad reputation in the audiophile market.
Just as me. Not expensive enough ?
I do not understand this sentence. If it means that, for selfs, using magnetic cores, or air coils it is better to X2 or X4 the max power you will apply to them than the real one, i agree.
More you stays far away from the core saturation, better it is, and, low serial resistances (big wires) matters considering Q factor of your filter and damping factor of your speakers.
In the same spirit, oversizing resistances will decrease their temperatures then reduce their thermic noise. Can change things in low level stages. Poor carbon resistances will change their value if heated too.
Because Q, in some filters, the level of the signal can, as you says, be larger than the applied signal, it is better, using non polarized electrolitycs caps, to oversize their tension, indeed.
better to stay away from electrolytic caps in filters, as i said, for all those reasons,
As a matter of facts, in audio, it is better to oversize the power and tension margin of each component as well as the cost you imagine for your prototypes ;-).
[edit]I had made a typo several times, the correct name of the caps manufacturer i was talking of was CEF, not SEF, i spend too much time developing web applications ;-)
Last edited:
The Self results refer to the measurements Doug Self did on the distortion added by using a DC blocking cap in the lower leg of the NFB loop of his Blameless Amplifier. These are in his book and on the web.
Basically if the DC blocking capacitor acts as a LF high pass filter then that capacitor must have an AC voltage across it. Measure the distortion to prove that this AC voltage creates an extra distortion.
Increase the size of the DC blocking capacitor such that the Input DC blocking capacitor now acts as the LF bandwidth limiting filter and the NFB DC blocker has only a tiny AC voltage across it. The bigger the cap the less the AC voltage. Measure the distortion to show that the added distortion is less when the cap is an octave or two lower than necessary for the passband of the amplifier.
Self's test and measurement has been in the public domain for nearly two decades and is referred to in this Forum regularly.
I suspect a number of Designers before then also were aware of what the AC voltage across the DC blocking electrolytic did to the amplifier performance. Few if any would make their knowledge public. They have expertise and product to sell to the highest bidder.
Folk like Self & Pass & JLH & Cordell and many others give away their knowledge for next to nothing.
We should learn to read and benefit from that freely available information.
Basically, if the DC blocker has to be an electrolytic, then the AC voltage across it must be reduced to as low a value as is practicable for least added capacitor distortion.
Basically if the DC blocking capacitor acts as a LF high pass filter then that capacitor must have an AC voltage across it. Measure the distortion to prove that this AC voltage creates an extra distortion.
Increase the size of the DC blocking capacitor such that the Input DC blocking capacitor now acts as the LF bandwidth limiting filter and the NFB DC blocker has only a tiny AC voltage across it. The bigger the cap the less the AC voltage. Measure the distortion to show that the added distortion is less when the cap is an octave or two lower than necessary for the passband of the amplifier.
Self's test and measurement has been in the public domain for nearly two decades and is referred to in this Forum regularly.
I suspect a number of Designers before then also were aware of what the AC voltage across the DC blocking electrolytic did to the amplifier performance. Few if any would make their knowledge public. They have expertise and product to sell to the highest bidder.
Folk like Self & Pass & JLH & Cordell and many others give away their knowledge for next to nothing.
We should learn to read and benefit from that freely available information.
Basically, if the DC blocker has to be an electrolytic, then the AC voltage across it must be reduced to as low a value as is practicable for least added capacitor distortion.
Last edited:
My concern with an electrolytic in the feedback leg is that if the amp is presented an asymmetric signal, a small DC bias will develop across this cap. As the signal asymmetry goes from one polarity to another, this DC balance will change polarity as well. Could this possibly be a source of IMD?
Esperado,
For lowest distortion, Self advocates using a value 2 to 4X larger than the "calculated" value
This was a quote, hence the italics, I should have mentionned the author, Stellavox. Andrew gives a good explanation what it was meant, I think.
I suggest you read Cyril Bateman works, he has done a lot of investigations on all kinds of caps, making many myths fall.
Another interesting text comes from Mike Renardson :
Capacitor Distortion
A fact I've never seen mentionned elsewhere is found by Mike :
The 2.2uF input coupling capacitor is more of a problem, polypropylene at this value are both large and expensive. My tests revealed that a large physical size for this component can lead to a large increase in interference pickup, particularly if the signal source impedance is high.
For lowest distortion, Self advocates using a value 2 to 4X larger than the "calculated" value
This was a quote, hence the italics, I should have mentionned the author, Stellavox. Andrew gives a good explanation what it was meant, I think.
I suggest you read Cyril Bateman works, he has done a lot of investigations on all kinds of caps, making many myths fall.
Another interesting text comes from Mike Renardson :
Capacitor Distortion
A fact I've never seen mentionned elsewhere is found by Mike :
The 2.2uF input coupling capacitor is more of a problem, polypropylene at this value are both large and expensive. My tests revealed that a large physical size for this component can lead to a large increase in interference pickup, particularly if the signal source impedance is high.
That makes sense to me, making the C larger will lower its AC impedance, and thus the distortion. The thought that I now have is this; given that the C (let’s say 100u, scaled up to 220u) has to be an electrolytic one, should we now design amplifiers to have a defined offset (let’s say positive) to make sure it operates at the correct polarity?
Frans de Wit.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.