Polycarbonate Film Caps

Help Support GroupDIY:

abbey road d enfer

Well-known member
Moderator
Joined
Jan 22, 2008
Messages
13,030
Location
Marcelland
Matt Nolan said:
Parasitic inductance of the interconnect / wires / leads at high frequencies.
Actually, for some applications (RF decoupling), multiple caps are preferred to a single (larger) cap, for this very reason. It must be evaluated case by case.
 

abbey road d enfer

Well-known member
Moderator
Joined
Jan 22, 2008
Messages
13,030
Location
Marcelland
There is one element missing in this discussion: the application.
If I understand correctly, the OP wants ocupling caps for tube amps, a position where good practice imposes very little AC voltage across the caps (as opposed to filters where some caps are submitted to voltages almost equal to the incoming signal).
The traditional approach has considered only the LF response, resulting in the 10x rule of thumb (make every capacitor 10 times biggere than the -3dB calculation result).
This is an ancient rule that did not take distortion into account.
Let's take the case of a 10k load. The cap value for -3dB at 20Hz is 0.8uF.
The rule of 10x says 8uF.
The voltage across the capacaitor at 20Hz is 1/10 of the incoming voltage. This is enough to generate significant distortion, indeed with variable results depending on the technology.
This is a case where I would use a 100uF electrolytic capacitor. Measurements have shown that the capacitor's contribution to distortion is hardly measurable.
Now in the particular case of the coupling capacitors between the PI and the output tubes, the typical value for the grid leak resistors is 220k.  [email protected] results in 36nF. Applying the 10x RoT gives 0.36uF.
Applying another 10x coefficient would result in 3.6uF, which is likely to be a huge size and create other problems (parasitic coupling).
Apparently, Cyril Bateman, in his seminal paper, didn't publish any measurements for PC capacitors, but he clearly wrote that PPS is capable of much lower distortion.
Apparently, the highest value in 400V is 1.8uF, so putting two in parallels would just clinch the deal.
Now there may be some consequences regarding the stability of the amp; ground circulation would need critical attention.

 

rackmonkey

Well-known member
Joined
Sep 20, 2016
Messages
919
Location
Texas
I got it off track by mentioning that I like PC in filters, I think. It just happens to be where the differences can be heard most clearly. But the intent was to inform of an opinion that the OP could take or leave where coupling caps are concerned.

Very nice summary above, Abby. The dumbed-down line on coupling caps has long been "there's no voltage across the cap, so its affect on audio will be negligible." But the part usually left off that line is, "...provided the cap is sufficiently large." As anyone who's ever replaced 100nF or 2.2uF coupling caps with a same-value cap of different dielectric material in, say, a tube preamp knows, the cap construction has an effect on audio. Those old metallized paper in oil caps have a sound of their own.

To John's response above, yes, DA, DF and tempco in polycarbonate caps is significantly worse than PP or PS. But it's considerably better than polyester. If lowest distortion is the goal, then obviously use the parts with better numbers there - no argument. But I think the OPs question is more along the lines of subjective qualities, which to my ears PC is pleasing. People love the 1073's EQ, but Neve often used Mullard mustard caps in those, which are just polyester film/foil. A no-no on paper (polyester caps are not recommended for filters by any manufacturers guidance that I've seen), but subjectively something attractive to a lot of people over the years.

So you can use traditionally specced smaller values for coupling caps and use the objectively best measuring caps for lowest distortion in that situation; you can use caps for their flaws and the audio effects they impart; or you can just use sufficiently large electrolytics for coupling and not worry about it. It's all about goals.
 

PermO

Well-known member
Joined
Mar 16, 2011
Messages
359
Location
Hilversum, Netherlands
8)

Ok, some more info on the application, here's the schematic for the compressor I'm planning to build;
salxIcm.jpg


All caps 400V rated

C1.. 1uF
C2, C3, C5..0,1uF
C6.. 0.01uF

I usually just grab what I have, but I wonder, if you did not have any stock and you would need to source these caps, what would you use in this schematic ?

I do have some MKC for all these positions, but now I'm just curious and I think it's an interesting subject.
 

abbey road d enfer

Well-known member
Moderator
Joined
Jan 22, 2008
Messages
13,030
Location
Marcelland
PermO said:
8)

Ok, some more info on the application, here's the schematic for the compressor I'm planning to build;
salxIcm.jpg


All caps 400V rated

C1.. 1uF
C2, C3, C5..0,1uF
C6.. 0.01uF

I usually just grab what I have, but I wonder, if you did not have any stock and you would need to source these caps, what would you use in this schematic ?

I do have some MKC for all these positions, but now I'm just curious and I think it's an interesting subject.
C1, C2 & C3 are those that matter the most. C5 is not critical, C6 even less.
 

PermO

Well-known member
Joined
Mar 16, 2011
Messages
359
Location
Hilversum, Netherlands
Yes, from what I understand C2 and C3 pass audio and C1 is in the control voltage line and has an influence on the releasetime, from what I understand, if stereo linked the releasetime doubles... so maybe go for 0.47 instead ?

Any specific types that would work best in these positions ?
 

scott2000

Well-known member
Joined
Feb 21, 2015
Messages
2,562
Location
Sunny...Sometimes Florida- USA
Kingston has mentioned liking the Russian PIO..Not specifically for the Fed but still...... Been wanting to try them.

Kingston said:
I have a rather large amount of soviet union PIO caps. These are roughly from the seventies. Their capacitance is also spot on and do not change after years of use in tube gear with significant heat involved. These are all hermetically sealed and as far as I know designed for even fighter jet radio and radar use.

Type examples that you can still probably find on ebay if the modern Russians have not completely decimated the old-world soviet warehouses located in Ukraine/Donbass region yet:
K40Y-9
K42Y-2
K73N-2
K75-10
MBG4-1
MBGCH

But I'd probably use whatever polys I had until I got it up an running and familiar with it... People still using the 1uf?? Thought there were some mods for the release time...



 

dom

Member
Joined
Jul 13, 2021
Messages
22
Location
Austin, TX
Parasitic inductance of the interconnect / wires / leads at high frequencies.
Inductance would decrease, however, with addition of more parallel paths, not increase. Less lead inductance is generally a good thing for a cap.

One disadvantage off the top of my head with paralleling signal path caps would be an increase in layout area. More of a concern for high-impedance nets. But, generally, paralleling caps is common practice and there isn't any inherent disadvantage. Always depends on the particular application though.
 

abbey road d enfer

Well-known member
Moderator
Joined
Jan 22, 2008
Messages
13,030
Location
Marcelland
Inductance would decrease, however, with addition of more parallel paths, not increase. Less lead inductance is generally a good thing for a cap.

One disadvantage off the top of my head with paralleling signal path caps would be an increase in layout area.
And that's precisely for this reason that the interconnecting traces length increases, so here comes the law of diminishing returns.
But, generally, paralleling caps is common practice and there isn't any inherent disadvantage.
Paralleling caps is almost unavoidable when it comes to decouple (by-pass) rails for a large capacitance is needed at LF and very low ESR at HF. Mixing technologies (electrolytic and ceramic) is necessary.
 

Matt Nolan

Well-known member
Joined
Apr 29, 2015
Messages
399
Location
Bath, England, UK
Inductance would decrease, however, with addition of more parallel paths, not increase. Less lead inductance is generally a good thing for a cap.

One disadvantage off the top of my head with paralleling signal path caps would be an increase in layout area. More of a concern for high-impedance nets. But, generally, paralleling caps is common practice and there isn't any inherent disadvantage. Always depends on the particular application though.
At audio frequencies, yes, you can look at things in bulk terms. At much higher frequencies, everything is a transmission line. There is a point at which adding more parallel capacitors becomes futile.
 

JohnRoberts

Well-known member
Staff member
Moderator
Joined
Nov 30, 2006
Messages
21,012
Location
Hickory, MS
At audio frequencies, yes, you can look at things in bulk terms. At much higher frequencies, everything is a transmission line. There is a point at which adding more parallel capacitors becomes futile.
This has been well discussed and indeed for audio frequency only applications like simple (in band) audio filters we should be able to source completely adequate capacitors for use alone.

Power supply decoupling is a classic example where we commonly mix dielectrics in parallel. Aluminum electrolytic to supply in band current, and ceramic discs to provide low impedance at higher than audio frequencies. That HF impedance is important because modern active devices routinely have gain bandwidth well above the audio band so stable PS rails help improve passband linearity.

While opinions vary, a popular practice that I don't embrace is paralleling audio caps in the direct audio path with smaller high quality caps. I invested some bench work back in the 70s to quantify a difficult circuit node where a single aluminum electrolytic capacitor was showing evidence of non ideal behavior. This capacitor was in series with a resistor to ground establish the closed loop gain of a phono preamp. (IIRC it was a 22uF in series with 360 ohms).

The deviation from ideal I noticed was extra phase shift at 20kHz. Apparently the ESL (equivalent series inductance) interacted with the 360 ohm resistance causing the transfer function to deviate from ideal for a perfect capacitor. I found that a 22uF tantalum capacitors delivered much less phase shift than same value aluminum. To get a parallel film capacitor to dominate the transfer function required that the film cap be 10% or more than the electrolytic. A 2uF film cap was (is) impractical. I have read too many anecdotal reports making remarkable claims for parallel film caps too small to be significant. Most likely the audiophool version of placebo effect (expectation bias).

My response was to avoid using that topology for my several phono preamps designs after that. Caveat Lector, the aluminum electrolytic capacitors I had on my bench back in the 1970s have been much improved since then. Note: another similar circuit node susceptible to that issues is the large electrolytic cap in series with tiny gain resistors setting maximum gain for typical mic preamp designs. Over the years I designed many mic preamps using large caps in series with the gain pot. Those value customers would pretty much never hear the theoretical degradation, while they would all hear the scratchy gain pot and reject the SKU for "poor" sound quality. :unsure:

JR
 

Winston OBoogie

Well-known member
Joined
Jun 3, 2004
Messages
2,662
Location
UK.
I usually just grab what I have, but I wonder, if you did not have any stock and you would need to source these caps, what would you use in this schematic ?

As Scott said, I'd initially go for the indicated values in PP that aren't too hi-fi tweaky and go from there. Wima MKP would be fine with me for initial tests. There are better, but that sort of stuff would happen later.
I'd also start out with the stock value of 1uF for C1. For now.

Not really on topic but somewhat related: these types of compressors often have a delicate balance between the time constant of the control voltage cap and the value of the inter-stage coupling caps.
Blocking and its recovery time due to C2, C3 would be a greater audible issue than any distortion due to dielectric type, so I'd do all my playing around with values for release timing and inter-stage coupling until I was happy with things.

Actually, much like I'd probably now do with an RS124 type compressor with similar basic topology - with large voltage swings coming off the GR tube's plates - I'd shove in a couple of followers between stages and put the cap on the input side of the followers. Using a bigger R bias for the input, hence, a lower value cap. Bottom end of this R is tied to a negative rail.
Follower outputs (also biased from the same negative rail) direct coupled to the 6SN7 grids. Cap blocking now gone.

I do think it's well worth trying better caps than the Wima MKP's once you get it all happening.

Other than that, for small pF value caps for stability etc in high voltage tube stuff, I use mica.
 
Last edited:

Winston OBoogie

Well-known member
Joined
Jun 3, 2004
Messages
2,662
Location
UK.
One thing I meant to say: I've zero problems with using electrolytics for coupling in any device, and power supply de-coupling/smoothing etc. too.
Except when they're inside a device and positioned close to something that relies on reaching high temperatures to work as advertised.


I've often gone for big Solen PP for local H.T. de-coupling near a particularly hot tube. I also have zero problems using solid state to help things along and realise a bigger apparent value of cap if needed. Failure rate due to MOSFETs has so far, touch wood, been nil. Failure rate, or declining performance, due to an aged electrolytic after 7 -8 years daily use in a studio: 4 or 5 times that I can recall.
 

dom

Member
Joined
Jul 13, 2021
Messages
22
Location
Austin, TX
Proper selection of manufacturer, temperature rating, ripple current rating, voltage rating, and placement relative to power dissipation sources are key to electrolytic reliability.

Since electrolytics have an inherently high equivalent series resistance (Often 10's or 100's of mΩ), you don't get much benefit from placing them directly adjacent to the load you're decoupling. They can live further away to reduce their local ambient temperature, and if you need good high frequency AC decoupling, it's better to supplement with a smaller value, low ESR cap near the load, like a ceramic or film type.

Electrolytics are unfortunately avoidable in a lot of applications because nothing matches their energy density. There have been some good advancements in electrolytic technology over the past decade or so. So called hybrid electrolytic and polymer electrolytic types are gaining popularity and have significant advantages in lifetime and lower ESR compared to traditional electrolytics. These sort of caps are widely available now (They have their own categories on digikey for example), however prices are still at somewhat of a premium compared to old school 'lytics. The cost is often worth it though, especially in high-temperature, high-ripple current applications.
 

scott2000

Well-known member
Joined
Feb 21, 2015
Messages
2,562
Location
Sunny...Sometimes Florida- USA
Actually, much like I'd probably now do with an RS124 type compressor with similar basic topology - with large voltage swings coming off the GR tube's plates - I'd shove in a couple of followers between stages and put the cap on the input side of the followers. Using a bigger R bias for the input, hence, a lower value cap. Bottom end of this R is tied to a negative rail.
Follower outputs (also biased from the same negative rail) direct coupled to the 6SN7 grids. Cap blocking now gone.
Dang...way less than 1000 words so no need for a picture... :)
 

Winston OBoogie

Well-known member
Joined
Jun 3, 2004
Messages
2,662
Location
UK.
Proper selection of manufacturer, temperature rating, ripple current rating, voltage rating, and placement relative to power dissipation sources are key to electrolytic reliability.

Since electrolytics have an inherently high equivalent series resistance (Often 10's or 100's of mΩ), you don't get much benefit from placing them directly adjacent to the load you're decoupling.

All good advice Dom.
Against this stuff I've generally tried to weigh up the impact of increased distance of the decoupling cap against the desire to keep the current loop of a particuar stage to as small an area as possible. Somewhat topology dependent but, always an important goal for me.

In a box that's fairly price conscious and might possibly sell in the 1000's, an extra couple of dollars spent on one part really add up, maybe it even amounts to a fair chunk of the salary you're paid for the job so...

In a DIY or a boutique small production run gizmo, I'd not really sweat it too much if I spent the 2 dollars extra.
 

PermO

Well-known member
Joined
Mar 16, 2011
Messages
359
Location
Hilversum, Netherlands
@Winston OBoogie, thanks for your reply...

I just finnished the mechanical side of this thing... I'll take a little break from it now, do some more study, and wire it up...
I'll stick with CJ's 864 schematic for now, as that's all I have and sort of understand...

I went for a choke filtered PSU with less capacity.

Rackmonkey send me some nice caps with the transformers and I have some MKC caps @400V so all my parts are in.

DSCF3255.JPG

DSCF3202.JPG
 
Last edited:

Winston OBoogie

Well-known member
Joined
Jun 3, 2004
Messages
2,662
Location
UK.
@Winston OBoogie, thanks for your reply...

I just finnished the mechanical side of this thing... I'll take a little break from it now, do some more study, and wire it up...
I'll stick with CJ's 864 schematic for now, as that's all I have and sort of understand...

Cool :) I recognise those transformers as the ones that took a vacation on their way from Rackmonkey to you.

I like your test points too.

I spout crap sometimes anyway, the stock circuit is perfectly fine 👍
If you were looking to make it to do stuff much out of its comfort zone then that's when you'd look at overcoming pitfalls.
 
Top