Hi SY,
If one is shooting for very high performance in an amplifier using a DC servo, I think there are some caveats to arguing that the op amp is out of the way in the servo. It is ideally out of the way in that what little of the subsonic portion of the signal it passes will not have any first-order effect on the audio chain. However, if the op amp's output has noise or distortion (including crossover distortion), it can get injected into the signal path. I go through some of the numbers in my book, and it does not turn out to be as benign and insignificant as we'd like.
The other thing that can trip up some DC servo designers is servo clipping and headroom. The designer may with the best intentions use a very large resistor for injecting the servo correction signal back into the signal path at the input to get the servo out of the signal path as much as possible. However, depending on the worst-case numbers (and the lowest frequency at which the amplifier may be tested at full power), there are conditions under which the servo can run out of headroom and clip. This is a manageable problem if the designer is aware of it and is careful.
It is also tempting to put a passive LPF between the output of the servo and the injection point to further get rid of noise and distortion from the servo. This can be done, but one then has to be very careful of the stability of the second-order system thus formed. This arrangement actually gives the amplifier a 2nd order high-pass filter characteristic, whereas the simple servo gives the amplifier a first order high-pass.
Indeed, in a phono preamp, if one wishes, and uses very high quality components in the servo, one can use this to advantage to create a second order subsonic filter.
Cheers,
Bob
Actually, I think it's better to place the passive pole before the servo input. That way, the HF demands are low, and the pole can be placed safely a decade above the servo cutoff frequency. If the output is divided down (for example, in my preamp, it's divided down by a factor of >10,000 above 2 Hz or so), then any HF noise or distortion is knocked down by another 80dB or more.
Really, how many modern opamps have significant crossover distortion?
Really, how many modern opamps have significant crossover distortion?
Yes John, I would not try to convince hi end engineers and I am not going to go on a anti DC coupled vendetta here. I am simply pointing out reality.
Hi SY,
You're right - modern audio-grade op amps have little crossover distortion, especially when they are not driving a low load impedance. That's exactly why I recommend that designers of DC servos use audio-grade op amps. It is important that they not think that it is OK to use some 25-cent general-purpose JFET op amp like an old 356 or TL071-class device. That was my point. A modest audio-grade op amp like an OPA604, for example, is fine for a DC servo.
The use of the additional passive pole ahead of the integrator can alternatively be used, as you point out. Note that the placement of the pole after the servo integrator is equally easily placed well above the servo cutoff frequency.
Cheers,
Bob
Scott,
About 45 years ago my cousins used to play a version of Where's Waldo. Their father would bring home "photographic plates" (large films in reality) covered with small steaks of light. They (and sometimes I) would try to find a streak that was different from all the rest. That was state of the art nuclear physics!
It is not uncommon for high school physics classes to do demonstrations of the major experiments of the last century. What was once state of the art is now often available at any science store as a child's toy!
As to the difference in using capacitors as a series pass (called by some in the circuit path) or using them as shunts (and here we agree called by some incorrectly as not in the circuit path), there is a difference in what influences the capacitor.
As has been discussed some capacitors do "burn in." I think one of the major causes may be reduction in internal moisture due to some level of heating from the current flow.
In a properly sized series capacitor the voltage drop across it is quite low, so there is less energy to "burn in" the capacitor. In the parallel mode the capacitor sees higher voltage and most likely "burns in" faster.
These days capacitor distortion is, using contemporary analyzers, able to be measured. Burn in of capacitors has been observed and reported.
So it is not violating any reasonable engineering principles to support the idea that capacitors used in shunt mode contribute less distortion than when used in series mode to a well designed equalized preamp.
ES
Yes it is, my example was for when a capacitor has voltage across it such as in an equalization network. At the 6dB point the R and C have the same AC voltage and switching them from series to shunt makes no difference. The major point is simply that shunt capacitors ARE in the signal path as opposed to the add drivel that some folks publish. "Burn-in" of quality film caps "well known to everyone"? Please provide a link from one of the major manufaturers.
Guys, a servo amp needs to provide little or no current so the class A trick makes this point moot.
D.Self himself has published data about distortion changes in mylar caps in the Linear Audio Volume 1. After they had run in over night the distortion droped in half. Polypropylens did not show that effect. The distortion reduction was semi permanent. After some 90 days off duty the distortion came back somewhat but the cap needed less time the second round to drop again in distortion.
Scott,
We are in agreement that the nomenclature of calling a shunt capacitor not in circuit is wrong.
However as to the difference in use the energy comes from across the entire bandwidth not just at the crossover point, so a bit of math would be in order to compare circuit induced heating. I was considering capacitors used in a feedback loop in particular. It might be interesting to compare topologies used for equalization and the total energy level placed on the capacitors.
I assume you got your Linear Audio Vol. 1. It has a bit on capacitor burn it. I didn't know you considered that a volatile topic. (But I am not sure if D. Self is fat! sp!)
There are just some issues I though were usually recognized such as, tubes age, electrolytic capacitors dry out, wax paper capacitors are subject to humidity as are some plastic ones, iron is magnetic, electrets fade with time. etc.
ES
A capacitor only has heating due to series resistance losses. A coupling capacitor in a preamp probably is down to .000001 degree C at nominal levels. The stored energy has nothing to do with the heating.
Where did stored energy come from? But it will cause heating due to the leakage resistance (called G in classic lit.) as it discharges.
When you have moisture in the dielectric, it is slightly polar and is moved by the changing charge, doing work and generating excess loss and heat. Yes dissipation factor is lower in a capacitor that has picked up moisture.
Since we are talking about distortion at levels of 10 to the -5 the increase in heat is probably unmeasurable, but in THEORY that is the mechanism that reduces the moisture content.
In my "tweaky" test gear I have found leaving silica gel and silver anti-tarnish paper inside the case reduces turn on drift. Yes there is still some due to thermal drift, but I see less of the initial noisy stuff.
The new paradigm, "bate" and switch. Doug's numbers are at 9V AC across the capacitor. In an RIAA network you would rarely be over .1V AC and since the distortion goes as the square of the voltage we are back at -140dB numbers.
Who's theory? At best 10's of micro-degrees and considering the encapsulation, it would take hours or even days to equilibrate the humidity. The parallel resistance is listed at 30G Ohms or so so the ESR would be the major heat source. With RF caps resonated in 1M Watt transmitters heat might matter. The numbers are easily computed here and there is no way .00001 C temp rise drives out water.
Also it's not like Doug is doing an HTOL on these, they sit at normal room conditions and relax back a little. You guys could soak one under water and see what happens.
Unless you have eliminated surface contamination activated by moisture ingress as a cause for warm up problems there is no evidence for the claim that it's the capacitors. BTW we have made boards where this was capable of overcoming weak logic pullups and causing hard failures (flux residue in this case).
Please no bait and switch ever intended, unless real money is involved!
Yes we are at - 140db numbers maybe higher as an interesting issue is, does the % of distortion drop with increasing level? (I know you don't think so, but I'll try the experiment anyways.)
Equipment burn in is often at numbers as high as 200 hours. If I get the chance I will dunk a cap and see what happens. (Thanks for the really good simple test idea.)
Yes I have also seen dirty PC cards cause problems.
I though the issue was do capacitors perform differently between series and shunt mode? I know it is irritating that the reason for trying the change is that shunt mode is "out of the circuit." That of course is nonsense. The initial observation that the new topology sounded better is also better explained by perception issues rather than engineering ones. It is only after a number of people perceive it is a better design that it becomes worthwhile to see if there is actually a difference. It turns out in my OPINION there can be one due to actually measurable differences in the way the capacitors are used.
Last edited:
Fair enough.
Now this looks interesting...
Spurious signal generation in plastic film capacitors
Borough, J. W.; Burnham, J.; Simmons, W. J.; Webster, S. L.
(Electronic Industries Association and Institute of Electrical and Electronics Engineers, Electronic Components Conference, 26th, San Francisco, Calif., Apr. 26-28, 1976.) IEEE Transactions on Parts, Hybrids, and Packaging, vol. PHP-13, Dec. 1977, p. 402-406. Research sponsored by the International Telecommunications Satellite Organization.
Spurious signals generated in plastic film capacitors are discussed; the transients appear as small spikes in capacitor voltage in the 10 to 300 microvolt range which may repeat at intervals from fractions of a second to years. Dielectric absorption evidently provides the source of charge and the energy for the voltage transients. Thus materials with high dielectric absorption, such as Mylar, more frequently show this type of spurious signal generation than do nonpolar dielectrics such as styrene. Minimization of spurious signals through selection of materials, preconditioning or annealing and capacitor screening techniques is considered.
BTW my smoking gun on this is the high DA of Mylar. The voltage coefficient of capacitance could be getting changed by the AC voltage stress. Dupont has lot of info on Mylar on line.
EDIT - Anyone who reads that paper will never use Mylar capacitors again.
Now this looks interesting...
Spurious signal generation in plastic film capacitors
Borough, J. W.; Burnham, J.; Simmons, W. J.; Webster, S. L.
(Electronic Industries Association and Institute of Electrical and Electronics Engineers, Electronic Components Conference, 26th, San Francisco, Calif., Apr. 26-28, 1976.) IEEE Transactions on Parts, Hybrids, and Packaging, vol. PHP-13, Dec. 1977, p. 402-406. Research sponsored by the International Telecommunications Satellite Organization.
Spurious signals generated in plastic film capacitors are discussed; the transients appear as small spikes in capacitor voltage in the 10 to 300 microvolt range which may repeat at intervals from fractions of a second to years. Dielectric absorption evidently provides the source of charge and the energy for the voltage transients. Thus materials with high dielectric absorption, such as Mylar, more frequently show this type of spurious signal generation than do nonpolar dielectrics such as styrene. Minimization of spurious signals through selection of materials, preconditioning or annealing and capacitor screening techniques is considered.
BTW my smoking gun on this is the high DA of Mylar. The voltage coefficient of capacitance could be getting changed by the AC voltage stress. Dupont has lot of info on Mylar on line.
EDIT - Anyone who reads that paper will never use Mylar capacitors again.
Last edited:
Hi John,
I guess from the point of view of body pickup it is nice to be able to have one end of the capacitor at ground, but apart from that I see no reason for a preference for use of a capacitor in shunt rather than in series - I seems to me that the capacitor is equally in the signal path in either case.
Your philosophy of preferring to have the capacitor in an equalizer in shunt also suggests that fully passive RIAA equalization is preferable to use of feedback equalization. This seems inconsistent with your use of feedback equalization for the 3180/318 us portion of the RIAA curve in your preamp. What am I missing here?
Cheers,
Bob
Bob, you read too much into my input, and I will not parse about it. I almost ALWAYS use negative feedback. I have ALWAYS used negative feedback in the majority of my designs, both personal and commercial. ONLY when I can do about as good a job, WITHOUT negative feedback, such as the input gain stage of a MC cartridge preamp, or even a low gain line stage, do I NOT use negative feedback. I have constantly used negative feedback in my own personal designs for the last 45 years. I just don't PREFER negative feedback, if I can avoid it.
John,
I am not complaining about your use of negative feedback. That would certainly not be me.
All I'm saying is that when you use negative feedback to achieve the LF part of the RIAA curve, you have a capacitor in series with the signal instead of to ground. If you go with passive RIAA EQ, then you can have the capacitance in shunt, as you appear to prefer. I'm just noting that you are deviating from your philosophy that caps in series are not preferred (which I don't agree with anyway).
Cheers,
Bob
It is preferred to NOT put caps in series, even in a feedback loop, but sometimes it can't be helped. Then, using the very best caps can be even more useful in avoiding potential low level aberrations that might be generated by a cheaply made capacitor.
My main concern when it comes to general circuit design is to MINIMIZE the series caps, especially unnecessary coupling caps, that can be designed out by using servos, instead.
My main concern when it comes to general circuit design is to MINIMIZE the series caps, especially unnecessary coupling caps, that can be designed out by using servos, instead.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.