Coupling Caps - what sound the best?

[GrayDigger]

Some time ago, I considered a small test fixture that would allow me to quickly and easily compare various coupling capacitors in “in vivo”. After re-reading Martin’s “passive component poll” – and the spirited discussion it spawned – from last year, I finally got around to building something.

The fixture consists of two independent banks, each containing the same 12 capacitors of equal value. It’s made of simple bits and (other than some eye-wateringly expensive boutique caps) cost about $20.

The connecting leads use RG-174 wire, shield is “grounded” to the box. Shielding clips to chassis ground at a single point (per bank) on the DUT, preventing the box and leads from forming a ground loop or picking up stray noise. Super quiet.

Rotary switches make cap selection a breeze. It’s also (relatively) easy to swap caps on the boards.

All I have to do is lift one end of a coupling cap, then clip one bank to sub for that cap. I can test two sections of a pre-amp, or simultaneously test both outputs of a phase inverter and “mix and match”.

A really illuminating experiment so far, allowing my ears to quell a number of doubts that no longer need to swirl about. I can see this saving me a great deal of development time as well.

What fixtures do y’all use to evaluate caps for your designs? What are your thoughts on the tonal qualities of different coupling caps in different positions?

[nuke]

I admire the effort!

I just usually read the data sheet if there's anything important the circuit is sensitive to.

In general, I will use polypropylene coupling caps between tube stages where the value/voltage needed will fit the size. Mylar caps if I need something cheaper and smaller. Mica, ceramic or polystyrene for the pf range.

Most of all, I'll use the ones I have on hand first and foremost if they are suitable material and proper value/voltage.

[xtian]

Pretty!

Make sure you get some accurate test values on those caps so you know whether you're just hearing capacitance differences or materials.

[pdf64]

Is the switching single pole or DP?
If single, is the outer or inner foil connected to the switch common (as far as can be ascertained)?
It might be desirable to use DP switching, so as to isolate as far as feasible to cap under test from the others.
That should help to minimise field effects affecting the test.

[GrayDigger]

xtian -

Make sure you get some accurate test values on those caps so you know whether you're just hearing capacitance differences or materials.

Absolutely! (These values are from the first batch installed, measured with DE-5000). So far, the caps were within claimed tolerances for the brands evaluated… For grins and giggles I have since evaluated various other capacitors, including some vintage PIO Solars and others from the 1940’s and 50’s measuring within a few % of nominal.

IMG_4233.jpeg

Nuke, thanks for the reply.

In general, I will use polypropylene coupling caps between tube stages where the value/voltage needed will fit the size. Mylar caps if I need something cheaper and smaller. Mica, ceramic or polystyrene for the pf range

My practical SOP as well, but I always wondered if I could hear the difference, and whether it would be something to consider…

Pdf64, thanks for the suggestions!

It might be desirable to use DP switching, so as to isolate as far as feasible to cap under test from the others.

Yes…I mulled about isolating the cap under test a fair bit. I ended up using SP12T MBB’s. For better or worse, I wired the foil ends (when present) to a common rail on the bank, the switch common, then via a shielded lead to the low impedance side of whatever I’m looking at in the DUT (Hi Z and Low Z leads I labeled box so I could tell them apart). The “non-foil” leads are all in close proximity, and leads to switch running roughly parallel, but I couldn’t imagine stray signals within the shielded box from adjacent open-ended (non test) caps being much? Would they be? I figured the shielding scheme should well isolate the test caps from the DUT; in use at least, there is no discernible additional noise added.

Its been fun to compile some experimental (if only observational) data-driven observations, and where to rank order findings in the overall consideration of tonal impact.

[martin manning]

Can you share what you've observed?

[cdemike]

If you're up to do some blind recordings, I'd be curious about a lower-rated ceramic capacitor (e.g., Z5U as opposed to C0G) getting thrown in as a kind of well-controlled independent variable to assess for validity. The effective capacitance of those is well-documented (including in manufacturer datasheets) as decreasing with a considerable DC bias across them. As a result, they'll function as meaningfully smaller capacitors than their nominal rating if, for example, used as a coupling capacitor with the significant DC potential difference across its legs. As far as I'm aware, that's the only uncontroversial, well-documented instance of a capacitor performing differently within the audio frequency range than those made with other materials.

If you recorded completely identical signals isolating the capacitor as the independent variable (e.g., using a looper pedal), you could record the results. The sample with the ceramic capacitor could then get mixed in with clips of the other capacitors, blinding the listener to which sample they're hearing. If listeners can't hear the difference between the film caps and the ceramic, that seems like it'd be a meaningful finding.

[nuke]

Absolutely! (These values are from the first batch installed, measured with DE-5000). So far, the caps were within claimed tolerances for the brands evaluated… For grins and giggles I have since evaluated various other capacitors, including some vintage PIO Solars and others from the 1940’s and 50’s measuring within a few % of nominal.

Since you have a DE5000 in hand, you might want to measure all the things you can measure with it!

Here's a link an English manual, and on page 3, you'll find a nice discussion about what it does and how it does it.

https://ietlabs.com/pdf/Manuals/DE_5000_im.pdf

Coupling capacitors *CAN* sound different, but those effects *ARE* measurable. You have a powerful tool, which can provide a lot of measurements, not just capacitance. You will find the DE5000 can measure differences between mylar and polypropylene dielectric capacitors and you can tell them apart with the meter.

ESR (equivalent series resistance) is another easily measured value, which may affect how a capacitor might behave in a particular circuit.

That is how dissipation factor is calculated, ratio of ESR/Xc and also shown as tan δ on datasheets. Probably more applicable to filter caps, but it is one of the things that get considered in circuit design.

Some harder things to measure are properties like piezo-electric effects, aka, microphonics. Also, sometimes capacitors make audible sounds too, because of that. You might see that in some ceramics, and especially with MLCC surface mount caps.

Some dielectrics also exhibit value changes with applied voltage, you may find this on datasheets as voltage coefficient of capacitance.

All kinds of properties affect how well a capacitor might work in some proposed circuit or system. Knowing the properties and the pro's and con's of different dielectrics and construction methods can be important.

[Roe]

In plexis, the mustards sound too dark for the normal channel, so there I prefer something else (e.g., jensen PIO)

[GrayDigger]

Roe, There are any number of other tweaks you could use to brighten your Plexi’s normal channel… The fundamental design of the tone stack and coupling capacitor values have always been a bone of contention for these amps (why else would one automatically dime the treble and zero the bass before playing your first note through a Plexi?). I’m sure the Marshall forum has a plethora of suggestions. I would agree that the type of coupling caps used can make a difference, but I would suggest that fundamental changes to cap values will make a much bigger difference.

Nuke, You are an inveterate OCD enabler! Thank you posting the English version of the DE-5000 manual. Since I got the gadget, it’s been an intermittently used – and not to its potential – primarily because I couldn’t read the manual it came with (which was in Japanese). Now, thanks to you - I am testing everything in my stash. (My dog has even started to look at me very warily when I have test leads in hand.)

Martin, not ghosting you. (I’ve experienced repeated issues with “too many requests” trying to post a response, maybe this one will get through) I’ve gone through a number of iterations in how I try to categorize the differences I hear between different coupling caps into a numerical schema I can reproduce. I’ve also been side-tracked trying to recreate some of the findings in the Jung and Marsh 1980 Audio articles. I promise I’ll post something more comprehensive when I amalgamate my scat.

Cdemike - thank you. I’ve decided to keep the evaluations analog (listening to a dry “standard” recorded through a looper with one DUT and same speaker/cab), and avoid introducing additional variables inherent in recording.

The audible differences have been intriguing, not only between (as expected) different dielectric types (PIO, Mylar, PP, PS, etc) but (unexpectedly perhaps) betwixt different makes using the same dielectric type. (So far, I’m very much bemoaning my not- insubstantial investment in my stash of “boutique” Mylar and PIO caps).

[nuke]

Nuke, You are an inveterate OCD enabler! Thank you posting the English version of the DE-5000 manual. Since I got the gadget, it’s been an intermittently used – and not to its potential – primarily because I couldn’t read the manual it came with (which was in Japanese). Now, thanks to you - I am testing everything in my stash. (My dog has even started to look at me very warily when I have test leads in hand.)

You're welcome to the next level of nerdiness.

However, I'll take no responsibility for any lack of female companionship you might experience as a result.

[Roe]

Roe, There are any number of other tweaks you could use to brighten your Plexi’s normal channel… The fundamental design of the tone stack and coupling capacitor values have always been a bone of contention for these amps (why else would one automatically dime the treble and zero the bass before playing your first note through a Plexi?). I’m sure the Marshall forum has a plethora of suggestions. I would agree that the type of coupling caps used can make a difference, but I would suggest that fundamental changes to cap values will make a much bigger difference.

This thread is about the type of coupling caps, not all those other issues. changing the tonestack will affect both channels, not just one, in any case.

[martin manning]

Martin, not ghosting you. (I’ve experienced repeated issues with “too many requests” trying to post a response, maybe this one will get through) I’ve gone through a number of iterations in how I try to categorize the differences I hear between different coupling caps into a numerical schema I can reproduce. I’ve also been side-tracked trying to recreate some of the findings in the Jung and Marsh 1980 Audio articles. I promise I’ll post something more comprehensive when I amalgamate my scat.
...
The audible differences have been intriguing, not only between (as expected) different dielectric types (PIO, Mylar, PP, PS, etc) but (unexpectedly perhaps) betwixt different makes using the same dielectric type. (So far, I’m very much bemoaning my not- insubstantial investment in my stash of “boutique” Mylar and PIO caps).

I think spectral analysis is the only objective way to quantify the differences. Dielectric behavior is very complex, and I would not be surprised if the formulations have changed over time, and the properties also change with time. Mylar is not Mylar, and old Mylar is different from new Mylar.

[GrayDigger]

~~~~~~~~~~~~~~~~~~

Martin, in the end, I found nothing earth shattering. The effects of different types of coupling caps on sound is just as the hi-fi guys characterized for PP, PE, PIO, electrolytic, tantalum, and silver mica.

Essentially, “Polystyrene rules, Polyester drools”. After all my goofing around, I ended up replacing every signal path (PP and PE) and tone (PE and Silver Mica) cap in my DUT with polystyrene, and have been practicing and gigging with such since. The difference in the amp has been… Astounding. Not subtle. Not even close.

Jung and Marsh claimed: “With clean capacitors, you begin to perceive the music in a new light, one that closely resembles the sound of the original source. In fact, you’ll be able to discern subtle nuances that you never even noticed before.” Yup. 100%. No doubt.

https://milbert.com/Files/articles/Pick ... tors_2.pdf p62

(When I first got corrective lenses as a kid, I remember the difference was like a layer of gauze being lifted from my eyes. The flat, pleasantly muted world I was used to became three dimensional, vivid with so much visual detail. Colors popped, edges were so sharp it seemed I could see around them. There were layers and layers of clouds, the stars had colors (!), and everything was just … alive. Recapping with PS did the same for my amp.)

I previously thought the hi-fi guys’ world view – being all about eliminating distortion – would not apply to guitar amps, where I figured it was all about manipulating distortion. Honestly – until now – I thought those audiophiles were just chasing esoterica, not relevant to the dirty world of guitar amps. Now, I don’t think there is a disconnect at all – both worldviews are consistent and unified. Our distortion should come from our tubes (and of course transformers and speakers). But the signal path between tubes should be without distortion – “hi-fi” if you will – and certainly not clouded by Mylar, unless absolutely needed for specific tone shaping.

Jung and Marsh also noted, “… one single polyester (cap) … can be heard.” I totally agree. However, since the effect of a single “color” cap (i.e. PIO, PE, etc) is readily audible, this does provide us an opportunity to go back and judiciously tone tweak an all-PS amp in one or two places if needed. In the end, this experiment just convinced me that basically everything J&M assert applies to a guitar amp, and I will not doubt their gospel in the future.

https://milbert.com/Files/articles/Pick ... tors_1.pdf p60


~~~~~

My little caper ended up unfolding in several phases.

First, I was simply curious to see if I could hear a difference between the types of coupling caps. Once I realized the differences between types of caps was clearly audible, I wanted to know where in the signal chain the differences were most noticeable, and if the effect seemed simply additive or geometric. My fixture allowed me to test in two sections of the signal chain or both outputs of a PI at the same time.

I then needed to come up with some kind of rating scheme. Since I was testing by myself over many days, often separated by some time, I needed the scheme to be somewhat resistant to both Type I error (e.g. null: expensive/fancy/pretty caps sound better than cheap ones) or Type 2 error (e.g. null: cheap caps sound every bit as good, or better, than expensive pretty ones). The scheme also needed to be more or less repeatable.

So I recorded a few clips of a couple of guitars on a looper and fed this (dry) signal into one of my own amp designs. Initially, it was challenging to focus on specific tonal qualities and consistently rate them across a range of clean, saturated, distorted phrases, and focus in on bass, middle, and treble sections. I started with a long list of tonal qualities but ultimately narrowed it down to a few arguably independent attributes that I believed I could reliably discern. For positives, I rated “vibe, clarity, and overtones,” and for negatives, “harshness, compression, and tubby/flabby/woofy.” Each was rated on a 5-point scale, and a normalized composite weighted rating, emphasizing the positives over the negatives, resulted in a single numerical score. Although my final rating schema was pretty arbitrary and simplistic, it proved surprisingly repeatable over time, and I repeated listening across all the samples on separate days. Once I realized you could hear the effect of a single cap in the chain regardless of where it was, the testing got a bit more refined, in that I used PS on one position as a control.

All the caps in both banks of my test fixture were 10nF rated for > 400V.

I found Polystyrene caps clearly provided the best vibe, clarity and overtones with no added harshness, compression or woofiness, regardless of position in the signal chain.

Next to PS, but slightly harsher, grittier, or granular was Polypropylene. However, not all PP caps sounded equal. For example, the 715P and 716P consistently sounded better than a generic radial PP, which sounded pretty sterile and devoid of any vibe. The Solen Fast PP 1000V was pretty transparent, but added some unpleasant grittiness. (I measured the CDE/IC MPW, which is much more compact than the 715/716P and Solen Fast PP, on my LCR, but I didn’t test it in my fixture.)

Polyester caps – regardless of the manufacturer or whether they were foil, metallized film, or oil filled, were disappointing. Especially considering just about every amp I have has a bunch of them. Polyester essentially wrapped my amp’s sound in a layer of haziness, resulting in a loss of vibe, clarity, and overtones, and also added some unpleasant harshness. After listening and measuring with the DE-5000, I wouldn’t be a bit surprised if many of the fancy and expensive ones I have were simply OEM’d and labeled with pretty wrappers.

Among the “least bad” PE caps I tried, I guess I’d have to say the Vishay/Roederstein 1813s sounded the least disappointing, followed by the $0.35 Vishay/Philips MKT368 and Red Astron repros by Jupiter. Without naming names, all the rest of the PE caps (which represented a majority of all caps I listened to) were all equally disappointing.

The oil-filled PE caps were consistently (but marginally) more vibey, at the cost of more woofiness. I suppose if I had an amp where I needed to “add some gauze” or muddiness, I might throw an 1813 or an oil-filled PE in the chain and see what resulted.

PIOs for me were all quite different, which I guess shouldn’t be surprising given the dielectric and the variability on LCR measurements. For example, I didn’t find either vintage or reproduction Bumblebees or 1940’s Solars to my liking, but the vintage Black Beauties I tested had lots of vibe, at the expense of clarity and woofiness. I haven’t tested some of the Russian K40Y/K42Ys I have, but mean to get to them at some point.

For fun, I also tested a ceramic and an electrolytic, and - yuk. No need to go there.

The effect of two of the same type of capacitor in different places along the signal chain seemed additive rather than geometric. I could not convince myself that the magnitude of effect was significantly greater (or lesser) regardless of placement in the signal chain.

Thanks to nuke for showing me what my DE-5000 can really do, I got quite a bit sidetracked measuring a bunch of different capacitors across all the frequencies the gadget was capable of. I compiled the data from a bunch of the caps I tested into this spreadsheet.

Capacitor Measurements DE-5000 copy.xlsx

I also encountered a bit of a dilemma. Having designed and voiced my amp to the capacitors I used (716Ps and Vishay Roederstein 1813s), I wondered what would be an appropriate baseline for comparison. In other words, to eliminate false positives and false negatives, I needed a gold standard for both positive and negative attributes, but the very nature of the test didn’t start with zero. Ultimately, it kinda resolved itself in a Bayesian-ish way – The final phase involved implementing my findings into my DUT and recapping the whole signal (including tone stack) with polystyrenes.

~~~

The vast majority of hallowed designs and vintage amps seem to have various colored (yellow, mustard, blue, orange, etc.) PE caps “baked into” their tone lore. This has very deeply influenced my worldview, at least until now. If “PE is an inherent part of vintage sound”, I’m now thinking, why the heck should it be?

(I always spend a great deal of time simulating and carefully crafting each stage of tube distortion for each of my designs and mods as well. But I just use the standard CAP or POLCAP models. It seems LTSpice incorporates a number of factors into its standard cap model, but can it or the nonlinear model be used to reflect frequency dependent characteristics for PE caps reflecting measured ESR, DA or θ? I studied the math in a couple Marsh articles, but I’m no way smart enough to turn them into models reflecting what I measured for PIO, PE, and other “unclean” caps.

https://milbert.com/Files/articles/Pick ... tors_1.pdf
https://www.engr.colostate.edu/ECE562/l ... 8_supp.pdf. )

The last week or so, I’ve been A/B’ing two of my amps (identical builds) through an amp switcher. One had my usual complement of PP/PE caps, and the other recapped completely in PS. You can guess the “yowza what a difference” winner.

In my new (heretical? messianic?) view after this experiment: I’m in whole hog for PS on everything. Only after listening might I consider sprinkling in a “flavor cap” (maybe a PIO) somewhere, but only if the final voicing needs tweaking to chase a certain vintage thing. Sadly for me, my (not insubstantial) collection of vintage, repro, and otherwise cool PE caps will be looking for a new home.

All this being said, I don’t think PS caps will make a fundamentally poor design sound good, but could help a really good design sound better or even great.

[GrayDigger]

Roe,

In plexis, the mustards sound too dark for the normal channel, so there I prefer something else (e.g., jensen PIO)

I get what you’re saying, sorry for mis-interpreting. Based on my testing, to see what effect those mustards are having might require replacing all the caps in that channel’s signal chain with PS (or PP for the larger values) to get a “gold standard” baseline, as even one PE cap along the signal path will be audible. The PIOs I tested in my amp were quite vibey, but at the cost of being woofy and loosing some clarity as well. Unfortunately, each of the PIOs I tried were also different, and each tested quite differently across different frequencies on my LCR.