# Back to back aluminum electrolytics

Discussion in 'Electronic Design' started by John Smith, Feb 15, 2005.

1. ### John SmithGuest

The thread on back to back tantalums aroused my curiosity. I had no
tantalums, so I used a couple of aluminum electrolytics. They are 22
microfarads at 16V each, Mouser PN 555-16V22 (Elna RVS16V220M-R).

o-----------> -C+ -- +C--|
|
HP 204D Osc 100 Ohms
|
o------------------------|

Using an HP 331A Distortion Analyzer:

Set 3V at 20 +- 0.5 Hz on 204D output.
Read .395 V across the resistor.
Read .193% distortion out of 204D.
Read .365% distortion across the resistor.

So, what can I learn from this? Is there a way to verify that the increase
in distortion is due to the caps rather than my naivety with these instruments?

If these measurements are accurate, is the distortion introduced by the caps
equal to the difference in measurements (.365 - .193)?

And here's a fly in the ointment-- the oscillator with no load measures
..295% distortion.

I appreciate the education you guys are giving me. Thanks.

John

2. ### Roger LascellesGuest

Unfortunately, combined percent distortions do not add arithmetically. For
example, the capacitor might have a nonlinearity which is equal and opposite
to the generator nonlinearity - giving a low distortion reading, and leading
you to believe that the capacitor has *negative* distortion! Since the
distortion readings you quote are all in the same ballpark, I would'nt draw
any conclusions about the capacitor distortion.

Also, the capacitor is dropping most of the voltage, so it it is phase
shifting the oscillator distortion (harmonics) and that can change the way
the oscillator distortion reads on the analyser.

Also, the capacitor - resistor acts as a high pass filter which accentuates
the oscillator harmonics and thus increases the distortion appearing across
the resistor.

My guess is that your capacitor distortion is small compared to the

I don't know what the lower limit of the 331A analyser is, but it is
probably 0.1% or better. With a better oscillator, you can get your test
setup residual down.

I would pick my resistor to get the same voltage across the resistor as
across the capacitor - at this point, the capacitor is dropping plenty of
voltage, but there is still plenty of output across the resistor. This
seems to be a standard setup used for capacitors - that way you can compare
results with others.

Roger

3. ### John SmithGuest

Thanks, Roger. That's the kind of information I was looking for. I have an
HP 651 Test Oscillator which, if I recall correctly, has lower distortion.
It also has a 600 Ohm output and a 50 Ohm output.

The 331A's lowest switch setting is .1%, so the meter will read .1% at full
scale on that range.

I will repeat the test with the better oscillator and an integrating RC with
a resistor of about 36 Ohms. At 402 Hz, the capacitor should then look like
36 Ohms and the combination would then have an impedance magnitude of 50 ohms.

Does this sound like a better try at it?

Thanks immensely.

John

4. ### Robert BaerGuest

Might be better to measure magnitudes of frequency components
(fundamental, 2nd, 3rd, etc harmonics) at the output of the oscillator
and at the resistor.
If the RC is far away from the fundamental, one could get: a) low
frequency would give apparent high "distortion" at the resistor due to
RC rolloff; b) high frequency would give same "distortion" due to low
drop across the caps.
Hence the suggestion. One should be able to *calculate* the level of
each frequency component at the resistor, from the RC value and the
oscillator level of each frequency component.

5. ### Guy MaconGuest

Could you try it at 1V and 0.1V and tell us what results you get?

6. ### John SmithGuest

I do not have and instrument capable of these measurements.
The RC will be such that Xc = R = 36 Ohms at the test frequency of 402 Hz.
Is this what a frequency-selective voltmeter is for (or, better, a spectrum
analyzer)?

Thanks, Robert.

John

7. ### John SmithGuest

With this setup, the 331A does not have enough sensitivity at these lower
voltages to calibrate the reference level for a reading. That's the main
reason I used 3V. However, I also tried to use enough voltage to exceed a
possible 1V threshold reverse voltage for the caps (if that's important).

But, yes, I can repeat the measurements using some suggestions from Roger
and Robert and try to design the setup to for all the requirements as I now
see them. If I can't get to it tomorrow, it will have to wait until I get
back into town (the 28th).

John

8. ### Robert BaerGuest

Yes, one use.

9. ### Roger LascellesGuest

I get the 36R and 402Hz numbers too, so looks good.

Maxim have an interesting app note :

http://www.maxim-ic.com/appnotes.cfm/appnote_number/3171

In this app note tantalums, aluminium electros and polyester caps are
measured. In the app note, the graphs show distortion rising as you go
lower in frequ, which increases the signal voltage across the cap. There is
plenty of distortion at the frequency where the R and C have equal voltages
(99.5 Hz on figure 8), so this should be a good measurement point.

According to fig 8, below the -3db frequency, the distortion rises, but the
output signal decreases, so the audible effect won't be worse at lower
frequency.

Roger

10. ### John SmithGuest

Okay, here's what I did...

o-------C+--+C----|
|
402 Hz 50 Ohms
|
o-----------------|

At 3V and at 1V input, .32% distortion at input and .38% across the resistor.

Swapped the C and R positions (made it an integrator), changed the R to
33.33 Ohms, and changed the frequency to 440 Hz:

o-------33.33 Ohms----|
|
C
440 Hz +
+
C
|
o---------------------|

3V applied, .31% input distortion, 2.01 V @ .182% across C
1V applied, .31% input distortion, .67 V @ .176% across C
..422 V applied, .32% input distortion, .3V @ .185% across C

The last measurement was the lowest input the 331A could handle and I'm
thinking there may be more error in it than in the first two.

I don't know where I got the idea that the 651A had less distortion than the
204D. A bit disappointing.

I'll go read that Maxim page to which you referred and try to learn
something. Thanks.

John

11. ### Phil HobbsGuest

One interesting thing to try would be to apply a variable dc bias to the
point between the two capacitors, varying it from a moderate positive
voltage to a very small negative one, and see what that does to your
distortion measurements. A second interesting thing would be to make
one capacitor twice as large in value as the other (or use a different
technology, e.g. ceramic or tantalum). Your circuit will tend to make
odd-order harmonics cancel, while even-order ones will add. Making the
capacitors asymmetrical intentionally will eliminate this cancellation.

Cheers,

Phil

12. ### Guy MaconGuest

Excellent idea! If the distortion changes, it pretty much has to
be distortion introduced by the capacitors.

13. ### John SmithGuest

Yes, I agree. I just haven't had time. I've got to go pack now. I'll try to
get back on this when I get back in town about March 1.

John

14. ### John PopelishGuest

I think another interesting experiment would be to make a bridge with
a resistor and back to back electrolytics in one side, and a (possibly
higher value) resistor and film capacitor in the other side, and
amplify the difference voltage between the two sides, after adjusting
the resistors for best phase balance. This should emphasize the
distortion difference between the two kinds of caps and low pass
filter and cancel most of the distortion from the source. You would
need an instrumentation amplifier.

15. ### ddGuest

There is a series of articles in UK Electronics World written by a
capacitor manufacturing specialist on distortion due to capacitors and a
dedicated measurement system / circuit (pcb layout included) articles
published 1/2 years ago most easily available by buying CDs of the last
few years of Eletronics World

16. ### Robert BaerGuest

"Distortion" is *not* the term to use!
"Filtering" is the accurate and correct term.