Hi, first time poster here. I'm trying to understand the role of non-polarized capacitors in the output network of a circuit. In Douglas Self's Electronics for Vinyl, chapter 14 on line outputs, I read that "output capacitors should be non-polarized types, as they may face external DC voltages of either polarity and should be rated at no less than 35V". The circuit below is a simple unbalanced line output from the book and presents such a capacitor.
Yet further down in chapter 17 on practical phono pre-amp designs, the first two circuits do not present such capacitors in the output network. They do present an output capacitor, but it is polarized and does not seem to require a high voltage rating:
So I guess my question is, can I use a polarized capacitor in the output network of my circuit? My circuit should send a line-level signal to the outside world. Also, does the voltage rating really need to be this high? I currently have some 100uF 25V polarized capacitors that I would like to use. 100uF 35V NP capacitors seem hard to find right now at my usual retailers and they are pricy. What's the risk of using the capacitors I have? A more elaborate explanation than the one cited above would be greatly appreciated!
Yet further down in chapter 17 on practical phono pre-amp designs, the first two circuits do not present such capacitors in the output network. They do present an output capacitor, but it is polarized and does not seem to require a high voltage rating:
So I guess my question is, can I use a polarized capacitor in the output network of my circuit? My circuit should send a line-level signal to the outside world. Also, does the voltage rating really need to be this high? I currently have some 100uF 25V polarized capacitors that I would like to use. 100uF 35V NP capacitors seem hard to find right now at my usual retailers and they are pricy. What's the risk of using the capacitors I have? A more elaborate explanation than the one cited above would be greatly appreciated!
Can you be sure that there will never be more than 1.5 V in reverse across the capacitor? If so, you can use an ordinary polarized aluminium electrolytic capacitor. If not, you can take two polarized electrolytic capacitors of twice the required value and connect them in anti-series.
In either case, the capacitor distortion will be quite small, but not as small as that of a non-polarized electrolytic capacitor.
I haven't got Douglas Self's book and the parts you quote don't make it clear where the potentially large voltages in his line interface come from, so I can't comment on that.
In either case, the capacitor distortion will be quite small, but not as small as that of a non-polarized electrolytic capacitor.
I haven't got Douglas Self's book and the parts you quote don't make it clear where the potentially large voltages in his line interface come from, so I can't comment on that.
When I still designed circuits for use in a local radio studio here, I often used non-polarized or anti-series connected electrolytic capacitors rated for 63 V. The idea behind that was to increase the probability that the circuit would survive if someone would accidentally connect a line output to a microphone input with 48 V (+/- 4 V) phantom supply. Doesn't match with Self's 35 V, though.
Not "reversed" .
In many cases, polarized capacitors can be used "improperly" simply because there is no significant DC present, at most only a few mV.
So in that schematic, right pointing or left pointing are both wrong, just they can get away with it.
By the way, both Cin and C0 are also wrong.
In many cases, polarized capacitors can be used "improperly" simply because there is no significant DC present, at most only a few mV.
So in that schematic, right pointing or left pointing are both wrong, just they can get away with it.
By the way, both Cin and C0 are also wrong.
The first circuit is also shown in Douglas Selfs "Small Signal Audio Design". The voltage rating is not commented in this case though, but the circuit is the same, described under "Line Outputs".
One possible explanation might be that, for a line output on an XLR connector, there is a chance of the output being subjected to a Phantom supply, so 48 V through 6k8. In this case there will be a voltage just over 35 V across the capacitor.
I don't know if this is the thought behind this, so we would need Douglas to explain the reason behind it.
One possible explanation might be that, for a line output on an XLR connector, there is a chance of the output being subjected to a Phantom supply, so 48 V through 6k8. In this case there will be a voltage just over 35 V across the capacitor.
I don't know if this is the thought behind this, so we would need Douglas to explain the reason behind it.
Okay so if I understand correctly one should in theory always use non-polarized capacitors in the signal path, but in practice voltages are so low that one can get away with using polarized ones. When used that way does the orientation of the capacitor matter? Both are wrong but is one way "less bad" than the other?
When your circuit has a significant amount of DC on the output (e.g. single supply circuits), a polarised capacitor should be better than a non-polar, and two back-to-back is even better. Remember to bias the capacitor correctly by using a pull-down as shown in the above schematics. When using back-to-back, another biasing resistor should be used at the midpoint between the two.
Usually for opamps with the de-facto differential input, the output will present a very small negative offset (few tens of uV). This is not really DC and there is no reason to use anything than a non-polar capacitor. If you absolutely must, it is better to use the negative terminal towards the opamp output, again as shown in the second schematic.
Usually for opamps with the de-facto differential input, the output will present a very small negative offset (few tens of uV). This is not really DC and there is no reason to use anything than a non-polar capacitor. If you absolutely must, it is better to use the negative terminal towards the opamp output, again as shown in the second schematic.
What I recall from Cyril Bateman's series about capacitor distortion is that non-polar electrolytic capacitors always outperform polarized ones.
Peter Walker (Quad) reputedly designed his circuits so that at turn on the electrolytic coupling capacitors were pulsed with a bias voltage to keep the dialectric layer in good condition, so even though in use there isn't significant DC across them the correct polarity was not arbitrary.
Bateman's measurements do support the use of bipolar capacitors for coupling on the basis of lower distortion and I occasionally use Nichicon Muse where a film capacitor is not practical. Film capacitors are not necessarily lower in distortion, but they are more stable over time and therefore to be preferred IMHO.
It seems to me that when audio circuitry is designed the designer rarely considers the coupling capacitors in preceding or post stages, which can include 1/2 dozen or more capacitors effectively in series in the audio path. For this reason I normally ensure that the fc of any coupling capacitor is ~1Hz or below. Even then there is still significant phase shift at 20Hz from each coupling capacitor.
Bateman's measurements do support the use of bipolar capacitors for coupling on the basis of lower distortion and I occasionally use Nichicon Muse where a film capacitor is not practical. Film capacitors are not necessarily lower in distortion, but they are more stable over time and therefore to be preferred IMHO.
It seems to me that when audio circuitry is designed the designer rarely considers the coupling capacitors in preceding or post stages, which can include 1/2 dozen or more capacitors effectively in series in the audio path. For this reason I normally ensure that the fc of any coupling capacitor is ~1Hz or below. Even then there is still significant phase shift at 20Hz from each coupling capacitor.
It doesn't really matter for such small voltages anyway, but...
The NE5534A's input bias currents always flow into the IC, because it has an NPN input stage without base current compensation. That means that in this circuit, the NE5534A's positive and negative inputs will always have a small negative DC bias voltage. I would therefore connect C0 the other way around. Cin is connected the best way around and it doesn't matter for Cout, because the sign of the output DC voltage is unpredictable.
1) Designers do not always take their own advice. Which may change decade to decade.
2) There "may" be a big difference between Pro Studio gear ( frequent patchings and high reliability needed), and a hi-fi hobbyist. If the radio station fails much revenue is lost. If the hi-fi dies, oh well, I wanted a new system anyway.
PRR, the key to understanding the difference is Mr. Douglas Self's word "external".
Post #1 Figure 1 is connected to the external world through a connector on the chassis rear panel.
Post #1 Figure 2 is connected to another circuit 2 centimeters away (the input selector), inside the same chassis, with no access to the external world.
Post #1 Figure 1 is connected to the external world through a connector on the chassis rear panel.
Post #1 Figure 2 is connected to another circuit 2 centimeters away (the input selector), inside the same chassis, with no access to the external world.
If the op amp output pin were stuck or shorted to +Vcc, that output coupling capacitor
would be in serious trouble. Just use the right part, a bipolar or nonpolar capacitor.
would be in serious trouble. Just use the right part, a bipolar or nonpolar capacitor.
Last edited:
I read it about the 44 on one of the Quad enthusiast sites, not sure which one. Nor do I know if it's correct, but it sounds feasible. I know on another enthusiast site there is a recommendation to 'correct' the polarity of the electrolytic capacitors (presumably based on measurement of one particular sample) and wondered if that might actually undo the design intent. A small reverse bias voltage of a few millivolts is tolerable for most electrolytic capacitors, but pulsing with rail level reverse polarity is not.
I don't see anything about the 44 that would lead to that conclusion. People have thought that the 34's asymmetric voltage rails are there to bias the coupling capacitor, but they would not have that effect. The real reason for that is to avoid TL071/2 latch-up on the negative rail.