Different capacitors and where to use them

[elaar]

Hi,

I'm building an amplifier out of one of Randy Sloane's books, i've successfully etched the pcbs and am now sourcing the components. I'm a little stuck on what type some of the capacitors are, and so was wondering what type of capacitors are best to use in what locations on the circuit.

Firstly there's the larger polarised caps (100uF->220uF), decoupling the rails to GND and similar in various locations, I assume these must all be electrolytic because of their size and the voltage level put through them.

Then there's 2 large tantalum looking capacitors (100nF), non polarised (so can't be tantalum) between the rail and GND right where the rail comes in, any idea what these might be?

Also there's 2 47uF capacitors in series, polarised and with their alternate polarities facing each other, which the low-level input goes into. I'm assuming these are tantalums, if so what voltage would they need to be for this low-level input? This amplifier has an 800mV input sensitivity and 10kohm input impedance.

Lastly, there's a large amount of ceramic looking (it's very hard to tell from the bad b&w photo), ranging from 68pF to 100nF, all non-polarised and used sometimes for decoupling and generally in the middle of the circuit.

Any ideas on what type these capacitors might be, or any advice on better alternatives than those?

Any help will be gratefully received

Thanks,
Andy

[myhrrhleine]

I'm building an amplifier out of one of Randy Sloane's books, i've successfully etched the pcbs and am now sourcing the components.

Congratulations.

I'm a little stuck on what type some of the capacitors are, and so was wondering what type of capacitors are best to use in what locations on the circuit.

Firstly there's the larger polarised caps (100uF->220uF), decoupling the rails to GND and similar in various locations, I assume these must all be electrolytic because of their size and the voltage level put through them.

I'd use electrolytics.

Then there's 2 large tantalum looking capacitors (100nF), non polarised (so can't be tantalum) between the rail and GND right where the rail comes in, any idea what these might be?

I'd use metalised Polypropylene or polystyrene.

Also there's 2 47uF capacitors in series, polarised and with their alternate polarities facing each other, which the low-level input goes into. I'm assuming these are tantalums, if so what voltage would they need to be for this low-level input? This amplifier has an 800mV input sensitivity and 10kohm input impedance.

I'd use electrolytics here.

Lastly, there's a large amount of ceramic looking (it's very hard to tell from the bad b&w photo), ranging from 68pF to 100nF, all non-polarised and used sometimes for decoupling and generally in the middle of the circuit.

I'd use metalised polystyrene or C0G ceramics.

Any ideas on what type these capacitors might be, or any advice on better alternatives than those?

Any help will be gratefully received

Thanks,
Andy

Expect as many answers as people here though.
Cyril Batemen did a lengthy article on capacitor distortion if you can find it.

[AndrewT]

Hi,
keeping to the same format

Firstly there's the larger polarised caps (100uF->220uF), decoupling the rails to GND and similar in various locations, I assume these must all be electrolytic because of their size and the voltage level put through them.

I'd use electrolytics. agreed

Then there's 2 large tantalum looking capacitors (100nF), non polarised (so can't be tantalum) between the rail and GND right where the rail comes in, any idea what these might be?

I'd use metalised Polypropylene or polystyrene.
I'd use pes or ceramic. PS are usually coiled and have raised inductance that affects VHF response.

Also there's 2 47uF capacitors in series, polarised and with their alternate polarities facing each other, which the low-level input goes into. I'm assuming these are tantalums, if so what voltage would they need to be for this low-level input? This amplifier has an 800mV input sensitivity and 10kohm input impedance.

I'd use electrolytics here.
I'd use polyproylene but reduce the value to 10uF instead of 22uF for a turn over frequency of 1.6Hz instead of 0.7Hz if no other DC blocking cap is on the source. If source has a lower value blocking cap it will determine the bass roll on frequency. Motor start PP are cheaper than almost all others at this value or recycle old PP from flourescent lamps.

Lastly, there's a large amount of ceramic looking (it's very hard to tell from the bad b&w photo), ranging from 68pF to 100nF, all non-polarised and used sometimes for decoupling and generally in the middle of the circuit.

I'd use metalised polystyrene or C0G ceramics.
I'd use polystyrene or polypropylene if audio coupling, or ceramic if bypassing to ground.

[traderbam]

Air core is the best. But practicalities of size and contruction make these rare except for some old-fashioned trimmers. There is a "roll your own" following to achieve the best audio caps too.

Polystyrene are the next best, aside from exotics like teflon and paper in oil. I've only used polystyrene. Probably too expensive for values larger than 1nF.

Most polypropylenes are next best. It depends on the model. Tend to be too large AND too expensive for values larger than 10uF.

Polyester are next. They are pretty good for values below 10uF. Above 1uF they get big and expensive, but not as much as polypropylene.

Polarized electrolytic caps are poor and should by avoided in signal paths. Also should be avoided where they aren't properly polarized. As input dc blocker and feeback dc blocker you are best using a non-polarized electrolytic or a tantalum or one of the film caps above.

Avoid ceramics and silvered mica in signal paths.

In general in a power amp the semiconductors make lots more distortion than the capacitors so you have to get the design really good before spending big money on caps is worthwhile...and even then there is a lot of debate.

I'm surprised if Slones's book doesn't recommend what caps to use for his designs.

[lndm]

Ceramic capacitors are usually very low in value and not typically used in the signal path. I believe their unsuitability for the signal path is fairly commonly held and so if one is specified, it might probably be due to the ceramic capacitors high frequency characteristics and should probably be left alone for the sake of stability, unless you know what you are doing.

Tantalums are a common choice for rail decoupling due to their low ESR, so any change might be a compromise.

Electrolytics are inexpensive and small in size for their value. If you can fit them, I feel there is no reason not to replace the electrolytics with something else. Probably starting with the input cap. If the question came down to the compromise of using 10uF instead of 24uF, I would choose the 10uF.

[jpnascim]

Hello Andy,

I've built one of the designs from Slone book (the one of figure 11.4) with success. The capacitors I've used were:

- The large polarized capacitors used in rail decoupling were good quality electrlytics. The smaller (100nF) ones used as the same function (rail decoupling) were metallized polyester.

- The signal input capacitors (appear connected as back-to-back in schematic) were supposed to be electros but I've substituted them with a good quality metallized polyester (that's the best I could find) with value 4.7 uF with no perceivable degradation in bass response. On the contrary, sound seemed cleaner (this is one of the good characteristics os Slone's amplifiers - clean sound).

- Small capacitors used in pole compensation (pF range): for this ones I've used polypropylene capacitors.

- Small capacitors used in feedback network (pF range): I've used a high quality ceramic type (I think it was a COG type).

- Large capacitor used in feedback network (uF range): I've used a Panasonic high quality electro but you could try to find higher quality materials if you can fit them on the board. Or you could put a smaller high quality capacitor in parallel to help compensate the distortions of the electros.

The results I've got were very good. I am happy with the amp and I've been running it almost everyday for more than a year. I hope I've helped. The other advices given here are excellent too (as there are more than one type of capacitor that fits for a particular application). If you wish, I can help you with my experience building a Slone amplifier just by following the book. Just e-mail me privately or simply ask here

Good luck !

João Pedro

[elaar]

Hi,

Once again thanks for the replies, I have included a picture I already made for another forum post to show the pcb pads (I've removed the main part of the circuit because I don't want to break any copyright laws or cause a reduction in the sales of his excellent books), as i'm not completely sure on the objective of some of the capacitors.

Traderbam: After reading more of the book and earlier designs I did find that sloane always uses tantalums for the input coupling capacitors, he also says because of problems with DC failure aluminized polypropylene with a value of approximately half the capacitance of the tantalums, unfortunately I don't understand the "turn over frequency" points you made AndrewT so thought it might be helpful if you could see the input section.Polypropylene are pretty expensive and very large, although saying that I found that tantalums 47uF 35v are also VERY expensive, are the plates made from unicorn horn?
Other than that, sloane does not mention what other capacitors to use.

Now, i'm not too sure what capacitors are in the signal path (apart from the input coupling caps), I assume the 2 pads to the left are the "dirty" GNDs, dealing with all of the rail decoupling, but the individual GND points coming fromt the transistors, are these in the signal path?

So, so far I have tantalums/polypropylene for the input coupling, electrolytics for parts, but to further distinguish what capacitors to use where I would like to know the difference between audio coupling and rail decoupling/bypassing, not so much the difference but how to tell what one is what.

Again, thanks to all and sorry i've probably forgot to mention something but when you waffle at this length as i tend to do there's always something you forget.

Cheers,
Andy

[AndrewT]

Hi,
C8 & C9 decoupling.
C19 & C20 decoupling or bypass depending on where they are located. Assume the return is dirty.
C21 Thiel (Zobel)
C1 & C2 input coupling (audio)
C3 radio frequency or VHF filter. I would use an NPO/COG ceramic here, but some others say you need a good audio cap (PP) to ensure an even response when voltages change. I think you need very low inductance to help the filter do it's job and NPO are very well behaved in this location.
C4 & C5, now the argument starts. I will state my stance. The combined cap & Zeners hold a fairly stable voltage that is modulated by the ClassA front end stage/s. The current transistions of the ClassA stages are all smooth and audio frequency related. There are NO ClassAB glitches here. My argument is that the returns are clean and additionally you want a clean voltage reference to feed the ClassA stage. the returns for C4 & C5 must be treated as if clean. The opposing argument is that R3 & R14 connect these caps to the modulated power rails and thus must have a dirty influence at their return points and should be connected to the power ground. If you were experimental you could build a PCB that allows short links to either ground (power or signal) and try to listen to/measure the difference.
The NFB caps are missing but must be considered clean and returned to signal ground. Some commentators suggest that the NFB caps and resistors should be of the highest quality of any on the PCB.

Finally, starting from the input. It is suggested that the best behaviour from the amp can be obtained if the bass and treble filters control the whole amplifer bandwidth You have R1/C3 controlling treble, RC~=3uS to 0.5uS. The bass filter, C1//C2&R2, should be about RC~= 30mS to 100mS. These time constant can be converted to frequency if it helps you understand the numbers. F=1/2Pi*R*C where C is in Farads. all are assumed to be single pole filters.
The next filter should then be NFB and it is suggested it should be at least half an octave lower than the input filter. Half an octave means that the RC timeconstant should be 1.4 times larger than the input filter. The last in the sequence is the power supply. It should have the lowest frequency to help ensure the bass stability is adequate and again half an octave below the NFB seems to ensure frequencies do not "motorboat" (low frequency oscillation). as an example if you choose 80mS for input this requires NFB=120mS and PSU>=160mS. Let's assume you have 8ohms speakers then C=160mS/8=20mF this is the capacitance required for the smoothing caps +-20mF on each channel. If 4ohm capabilty were required the caps need to be changed to +-40mF per channel (three parallel pairs of 15mF =+-45mF and have adequate ripple capacity for any 40hm amp).

I hope this helps. Signing off to go dancing.

[traderbam]

Andy,
It is hard to comment based on the partial diagram you posted, except that there are some potential problems even with the circuit you posted.

If you wish to email me the full diagram and pcb layout I'll give you some tips on parts and how to make the amp sound good.

[Curmudgeon]

Some random comments. I like to use HF electrolytics for supply decoupling; they're the ones with 100 kHz loss specifications.

Tantalums are relatively high distortion. Unfortunate, as they have good high frequency characteristics and indefinitely long service life.

Polypropylenes are definitely better than mylar/polyester. Metallized polypropylenes, as a generalization that has exceptions, are not as good as foil/polypropes.

Polystyrenes can get expensive in larger sizes, but are definitely better than polypropylenes.

You may see NPO ceramic capacitors; they are the same as C0G's and are fine. High capacitance ceramics, Y5, Z5's and such are to be avoided.

Higher voltage capacitors sound better than the lower voltage versions of the same type.

Finally, details like winding tension, end sealing, lead material and attachment method all affect the final sound, so evaluation has to be done on a model by model basis; therefore the material guidelines mentioned above are only rules of thumb. Price is an indicator that is not always reliable.

If you experiment, and compare cap types, break them in first. 40 hours is a good rule of thumb. Connecting a resistor across the output and breaking in silently will prevent your become used to a given cap's signature.

As well as the Bateman capacitor article mentioned above, Walt Jung's site has the full reprint of his seminal article with Richard Marsh on capacitor types and sound quality. It made general knowledge that which had been the proprietary trade secrets of a few high end manufacturers.