Using tantalium capacitors?

This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Tantalum capacitors

First, I don't think I know any bipolar tantalum capacitors. I don't think there are any, and believe me I have a made a thorough research on bipolar caps recently for a project I am working on.

Second, tantalum capacitors were reported as bad sounding on the capacitors article, by Jung and Marsh, published in Audio magazine years ago. Tantalums also have a problem that they have to be biased by some voltage applied directly to work properly. I have seen tantalum caps catching fire very easily, and they were just bypassing, not the main ones.

To solve chemical capacitors distortion problem, as that also happens with electrolytic types, the advise was to use two caps, back to back, with the joint point being polarized to V+ or V- (depending on that joint polarity) through a resistor. Arranged in such a way and bypassed by a film cap, these caps would they spec almost as a polyester cap.

Beg to differ

I've found Tantalums to offer some of the best sound for coupling in audio. They have measurably lower distortion than electrolytics, and to my ears sound better than anything I've tried, including OS-Con's etc. They're used by a number of high-end manufacturers in expensive amplifier designs.

As suggested above, many circuits suffer from incorrect biasing of capacitors, something that should be a done for all polarised capacitors, but is essential for tantalums, if only because they will fail catastrophically if this is not observed. The spec. for most tants is 0.2V reverse voltage MAX!. You need to ensure that when signals are passing through the cap this is not exceeded or even approached.

A pull down on one lead, and suitable DC biasing, preferebly via a potential divider will allow them to perform at their best.


P.S. I agree about ceramics they measure bad and are very microphonic.

[Edited by ALW on 08-04-2001 at 10:41 AM]
re: comments

first off all,
thanks for the various comments.

Let me make one thing clear: I've messed up major by naming tantalium capacitors 'bipolar'.

The Dutch short for a capacitor with an explicit + and - lead is ELCO (from ELektrolitische COndensator) but I do not know an appropriate English term.

I basically made a wrong translation.

Basic intention was to query about 'normal' tantalium capacitors with + and - leads.

Over the years I've blown up several: incorrect connection, slightly higher voltages etc, but their advantage in DC-regulating circuits still crept up.
So when finishing the JLH design (and power-supply), finding some capacitors needed for DC-coupling the use of tantaliums crossed my mind.

All in all, the response has been very welcome.
My JLH amp is now done without any tantaliums and it's nearly finished....

Using Tantalums


Here's an example circuit, consisting of a complimentary feedback pair with current source, running from a single 24V supply. Ignore the specific semiconductor types, as I had limited ones available in the SPICE software.

The input and output coupling capacitors are tantalums
(C8, C9). The input has a pull down of 100K (R20) to keep the input at 0V DC, the +ve terminal has a divider (R21,R22). The divider provides bias for the tantalum (12V), and the complimentary feedback pair.

At the output, the amplifer is nominally biased to sit at half rail (12V) for max swing, hence this provides bias for the tantalum. The -ve side of the tantalum has a pull-down (R16) to keep it at 0V D.C.

This circuit measures and sounds great, I hope it helps you understand the principle!


An externally hosted image should be here but it was not working when we last tested it.

Thanks very much for the schematic. Can the same principle be used when higher value electrolytic capacitors are used back-to-back in the gain/feedback arm of most conventional amplifiers. For example, a 470uF electrolytic and a film 470nF cap are used in Anthony Holton's N-channel amplifier. I used two 1000uF ELNA electros in series in place of the 470uF. Can I use the same principle to bias the junction of this series connected pair?

BTW, how do you upload schematics on to this site? I use Electronics Workbench V.5 for simulations. Would this be of any use in this regard?

Thanks in advance.
Hi Andy,

can you go on and post more NAIM circuits here.
(Your graphic is a part of most or all NAIM preamps since 20 years till now ...)
I think this circuits are NOT patented, and at least out for so many years now. We could eventually give them another name (another naim : "naim clone" could be a "claim" ).

Hi Jos,

I have absolutely positive experiences with correctly polarized biased tantalum caps in the audio path.

But at least there seem to be one drawback:
They seem to be not very reliable concerning constant quality over time. I have to change mine (in a naim-clone) every third year approx. They start to crackle with high volumes, maybe also leaking some DC making the volume and balance pots noisy.
One even exploded in a power supply !
(BTW, all tantalum caps had the right electrical tolerances).


ps: As I saw in a magazine recently, even NAIM changed the input caps in their NAP150 from tantalum pearls to another type of electrolytics.
Re: Tantalum capacitors

The company I work for spent a lot of time investigating capacitors for coupling low-level signals.

We found some interesting things I thought to share:

1. Electrolytics and, in general, most capacitors with "wet" dielectrics seem to have non-linear transfer functions with regard to signal amplitude linearity.
We noticed as you vary the charge applied across the capacitor, you get a different voltage/current transfer characteristics. Keep in mind this is low power, around 1mW or less experiments. So we decided to keep these types of capacitors for non-critical applications, that is out of the signal path.
With a DC bias the characteristics are better but still far from the ideal capacitor.
2. Then we looked at the Metallised capacitors MKT, MKP (Metallised Polyester and polypropylene etc). These are quite good compared to the ones above, in that their transfer functions are quite linear, but at small signals levels around the zero crossing of the audio signal these capacitors are highly non-linear. We first experience this problem when we discovered a non-linear operation of a VCO circuit's control loop response.
We contacted the manufacturer in Germany and we worked out the problem was to do with electron charge movement and the electrochemical potential of the compound needed to be reached in order to get the electrons to move back and fourth, this causing the non-linear response.

Putting a DC bias on the capacitors alleviates this problem.

But the noise of the Metallised component can be a big problem especially if you use it on the input of a high-gain amplifier.

Then we looked at Tantalum capacitors, because of their small size and high capacity; their impedance characteristics are quite good. Most Electrolytics, common garden types, present an inductive component around 1kHz which will usually dominate the impedance characteristics by 20Khz - making the capacitor quite useless. (That's why it's quite a good idea to shunt the electrolytic capacitors with a 0.1uF polypropylene, that is still predominantly capacitate at 20kHz and beyond, but eventually will be inductive, but outside the audio range. But you also have two noise sources, and more distortion - you don't get something for nothing I think.)

The Tantalums we bought from three manufacturers had characteristics better than all the Electrolytics but worse or average compared to the Metallised capacitors.
Also their power dissipation is poor since small surface area, this heats the capacitor, even just 0.5 degree C, will change the characteristics that we measured.
So this capacitor was OK, but expensive, and quite temperature sensitive, more so than you might think.
Solid tantalum dielectric is more stable than the cheaper wet electrolyte types.

Next we looked at the simpler dielectrics of polyester and polystyrene and polypropylene.

To make it short, the polyester is okay, but is non-linear a small amount but the nature of the dielectric means it is poor performing at 10Khz and above. Also it has high noise voltage that is temperature dependent. Similarly capacitor value stability is questionable.

The best by far, we think, is polypropylene. It has the best transfer function characteristics, quite low distortion, and is very temperature stable compared to others, and works well at High frequencies.
They are expensive, and large for their size.
You can get better performance by parallel use of these capacitors to achieve a certain value, rather than using one of those values, but be sure to apply a voltage across the "bank" so that the charge potential is equally placed across the capacitors. This is still good, even though the characteristic of the capacitor may vary. Also you get more noise with the bank, since the noise sources add in magnitude, sometimes it is better to use one cpacitor and put up with the performance if the noise is a problem.
Polystyrene is number two after polypropylene.

Even you can get this information if you read the databook from Wima, Rifa, or Beyschlag, but we wanted to be sure for ourselves.

We did not want our amplifiers changing their tone character with time, temperature, or voltage fluctuations from the mains power. We believe this is bad design and a bit lazy on the part of the designer, but if you use valves you can't help it because they begin to degrade the minute you turn on the heater. :(
Semiconductors do breakdown but mainly due to static electricity and or operating them outside of the SOAR characteristics, and dv/dt , di/dt.

The capacitor study was part of our mission to better understand the capacitor operation and get what we think is the best performance from our designs.

Hope this is OK.

Please excuse the English it's not my first language.
Andreas, thank you, it was very interesting to read your post. I have previously researched the issue of linearity but since the levels I was working with were not especially low I had not considered noise. Your comments about this factor were particularly interesting.

The agree with your ranking of capacitors, but unfortunately have only ever found polystyrene available for small values. What value of capacitors were you testing?


Pete Fleming
Hello Pete:

We were testing a broad range of capacitors from about 100pF to 10uF.

Polypropylene gets quite large at 2uF and beyond.
Also polystyrene is only really available at bellow 100pF normally, but our manufacturer made us some polypropylene capacitors for 22pF and 56pF that are almost the same performance for polystyrene - so maybe either or types is good depending on the value you want.

I forgot to mention, that heating these capcitors during testing, because of ESR and or exceeding voltage range can really give you some strange results, misleading if anything.

My colleague told me the other day he saw somewhere someone had a website where they were testing capcitors with a 70V RMS signal, eventhough the capcitor was DC rated 100VDC, so they were using a 198Vp-p signal, this isn't so good for the capcitor as your results are for extreme limits of operation, you won't get there in most audio work.

For our testing we used the maximum voltage of 10V RMS and change this depending on the capacitor we tested.

If you stress the capacitor too much with heat during tests, you can cause temporary plate separation at some point, kind of bubble, this has a big effect on the static charge dissipation over the plate surface and hence the power transfer characterisitc.

For the phono stage testing we used maximum voltage of 2V RMS at the input. We haven't seen a cartridge which has an output that would ever exceed this but you still had to have some way of handling the peaks, especially the ones made by dirt, scratch, or just handling the tone arm badly.
Also you have to have some kind of protection limit so as not to damage your tweeters, since the scratches have a lot of high frequency energy, 2kHz +.

Most of the capcitor testing was focussed on the low voltage signal 100uV and 2mV. The big concern was linearity, and noise performance.

We tried some hand made capacitors from the United States, which were very expensive, but we found that the quality varied alot, and we couldn't always get the same results from our tests, within a tolerance range. On one test we decided to unwind one capacitot because of the strange results we got from a test. We found that the foil had been stretched and had almost teared, also there were air bubbles trapped within several layers that were about 0.05mm air gap.
We thought this wasn't so good quality.

From a scientific perspective the capacitors wound by automated machine, in a controllesd atmosphere seem, and then sealed with resin, to give the best repeatability and also longevity performance.

We never considered paper oil, since the oil tends to concentrate in one area depending how you mount the capacitor, this has an effect on the performance and dielectric absorption, and dissipation factors, mesurements - sound changes to the user we think.
That's OK if you can live with it. Also heating of the capcitor, internal and external, tends to make the dielectric properties change alot.
This information is second hand from one our capacitors manufacturers who still makes paper oil capacitor for AC motor starting, since they can tolerate quite high ripple currents momentarily and are a quite a lot cheaper than plastic film since it's a legacy product for them.

Our feeling is that the amplifier should add as small as possible noise and no or very little non-linearity as is possible - practically of course.
There is already enough non-linearity from the cartridge and the surface noise from the record tracing. Tracing distortion is there.

Similarly, CD pressings change alot during manufacture stages and we found in one study that testing CD for errors, of the same batch but different times of production is bad and exists. So we found that one manufacturer made the best tracking system and oversamplig processor for dealing with bit/rate loss and therefore guessing what information is missing so that user doesn't think they missed something, no bad pops or clicks. If you have time I think Bob Stewart from Boothroyd Stewart wrote quite a good document from this, may be you can look at there web site.

So with all these things the company founder's philosphy is that our equipment should reproduce what was recorded by the artist/engineer as best as possible - it's not for us to decide how you should hear the music. Unless the CD has bad bad bit errors and we have to rely on the oversampling processor to guess what is missing.

Not everybody agrees with this so that's okay, and it's an ideal.

I hope this is helpful,

This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.