Correct orientation of feedback cap

[Greg Erskine]

Hi everyone,

I have been looking at a lot of schematics over the last few months and have noticed there is an inconsistency in the orientation of the polarised feedback capacitors.

I have always thought that the +ve of the cap should go to earth but I have seen a number of circuits with this cap reversed. I also came across an article saying the schematic is correct (i.e. +ve to earth) but the pcb layout is incorrect (-ve to earth) but the amp would function OK with cap reversed.

Can anyone enlighten me. I know this cap can be a relative low voltage cap and caps can take a small reversed voltage (without damage?) If the orientation isn't important then why can't we use a non-polar cap in this position.

Thanks

[Jan Dupont]

If you use polarised cap, the orientation doesn't matter. Either way you place it, it will have to handle the same amount of reverse voltage/current.

I think that the lack of use of non-polar caps are due to two facts:
- non-polar caps are rearly audio grade
- the can be a bit difficult to get in high uF values
However you can use many small non-polar film caps to make one large non-polar cap

[MikeB]

Hi Greg !

In most cases the orientation of this cap does matter ! (if polarized)
Depending on the topology, if the asymetric input-LTP is using pnp-devices like an AKSA,
the -pin of this cap needs to be connected to gnd, if using npn-devices like my symasym,
the +pin needs to go to gnd. This is caused by the inputoffsets generated by the basecurrents
of the LTP-devices. A diffamp with lowgain npn-devices like the 2n5551 can create an inputoffset
of ~-200mv, this is the voltage this cap is charged with. With highgain devices like mpsa18,
this offset is only ~27mv.

If you have a symetric input-stage or if you use jfets, this inputoffset is ~0 and no polarity
is defined. In this case you need nonpolarized/double caps. (Like in the Leach)

Mike

EDIT: The fact that this cap is charged with the inputoffset makes the cap rarely see a reversevoltage, this can only happen with very low frequencies at high swings.

[sss]

if u really want to make it right all u gatto do is to measure the output offset voltage of your amp , if its positive then + to gnd and vise versa.....
the purpose of that cap is to act as act as a short (or almost short) circuit when signal is applied .it will do the job no matter how u r gonna place it

[MikeB]

Hi sss, this is only valid for topologies not having this inputoffset, typically the output dc-offset is much smaller than the inputoffset. (Or should be)
The purpose of this cap is to reduce amp's gain to 1 for DC, and balancing the inputstage by creating the same inputoffset on both sides of the input-LTP.

Mike

[Greg Erskine]

Thanks all,

I just hadn't noticed the correlation between the orientation of the feedback cap and the use of pnp or npn TLP input stage plus the implication of positive or negative offset.

It all makes sense. Another small piece of the puzzle falls into place.

(It looks like one of the schematics I was studying was wrong and when I copied another schematic into Eagle I reversed the orientation of the feedback cap.)

Regards

[Christer]

Or perhaps it actually doesn't matter?

As has already been said, electrolytics can take some reverse voltage. As far as I understand from Cornell Dubiliers App note http://www.cornell-dubilier.com/tech/appguide.pdf an electrolytic capacitor works exactly like a non-polarised capacitor as long as the reverse voltage does not exceed some 1.5 V. We should not get even close to this value during normal operation. On the other hand, during fault conditions we could get a very high DC on the output, so this cap actually gets more than 1.5 V reverse voltage (depending on rail voltages and closed loop gain), which is probably why some designs have a protection diode in parallel. However, such fault condition do hopefully not persist very long, and may thus perhaps not need to be considered (electrolytics can take reverse voltage for short periods of time).

The obvious follow-up question then is whether the cap will behave exactly the same under a small reverse voltage bias as under a positive one, or none at all. According to the model in the app note it should, but models are always approximations, so maybe the might be differences, like higher distorsion? One might also ask if the threshold voltage is 1.5 V for all electrolytics, independent of type and brand.

So, it probably doesn't matter, but why not connect it in the righ direction anyway? Some pedantry hardly hurts. The more important question is perhaps whether to add that protection diode or not?

[MikeB]

I'd say, even if it does not matter at these low voltages, as long as the polarity of the voltage to block is known, use the correct orientation !
I did not experience blown feedbackcaps yet (did anyone else ?), but the diode can't hurt...

Mike

[Christer]

Probably a blown feedback cap won't appear as an isolated fault, but as a consequence of some more serious fault condition. My guess at why some add the protection diode is that a damaged feedback cap may go unnoticed when an amp is repaired. If an output transistor is blown so the output sticks to one of the rails, for instance, the cap may experience a too high reverse voltage. Probably it won't really blow up from it, but it may be damaged, perhaps getting a decreased capacitance or increased ESR. This may not be detected and thus only the blown transistor is replaced and then the amp seems to work again, while it actually may not perform as well as it used to.

[wintermute]

I might have faulty memory, but doesn't doug self say on his web site that you should use a non-polar cap with the highest voltage that will physically fit on the circuit board for the nfb cap??

The reason for this I beleive is due to bipolars (and higher voltages) having better dialectric absorbtion characteristics (which reduces distortion).... I've considered trying them but always wondered about the quality of the generic non-polars available to me.....

Tony.