Low voltage high capacitance polyprop caps- do they exist?

[Deke609]

I'm so pleased with the results of replacing the electrolytic cathode resistor bypass cap with a polypropylene film cap that I'd like to do something similar in the BeePre. Problem #1: the BP cap is 10,000uF 16V. So I'm thinking about adding a film cap in parallel @1%, or 100uF. Problem #2: I can't find any 100uF polypro caps under 100V. WIMA make some at 100V, but no one seems to stock them.

Any leads or ideas?

cheers and many thanks, Derek

[Paul Birkeland]

The 100uF cap in parallel with the 10,000uF cap is going to struggle to get any share of any kind of signal or noise current at audio frequencies. You could look for one of those vintage mid century console stereos, gut it, then build the BeePre in where the turntable was from the factory.  The rest of the console could be used to hold 100uF film caps.  You can do it!  (We want to see pictures)

[Jamier]

You could probably get 100, 100uf film caps into a good sized mini fridge which would reduce the footprint of the whole thing. You could just put the Beepre on top.

Jamie

[Deke609]

You could look for one of those vintage mid century console stereos, gut it, then build the BeePre in where the turntable was from the factory.  The rest of the console could be used to hold 100uF film caps.  You can do it!  (We want to see pictures)

You could probably get 100, 100uf film caps into a good sized mini fridge which would reduce the footprint of the whole thing. You could just put the Beepre on top.

  I never lost faith that you guys would come around to my way of thinking!

But seriously, i think there's something worth exploring here. I have 2 thoughts:

(1) Can I just replace the 10,000uF cap with, say, a 50uF super low ESR cap? The absurdly large capacitance of the cathode bypass cap in the Beepre has been a mystery to me. But just a few moments ago I did some ESR calculations to see if a 100uF film cap might actually make an effective "bypass" of the bypass electrolytic (more about this below at (2)) - and while crunching the numbers it occurred to me: maybe the only reason for the huge capacitance is to get super low ESR. Since ESR = Dissipation Factor (DF) * Capacitive Reactance (Xc) = DF/6.28*f*C. So the larger the C, the smaller the ESR.

Is this the case - is the only reason for 10,000uF to get ESR super low? In which case I can use a much smaller value film cap with even lower ESR?

(2) Assuming, for sake of this thought experiment, that there is some other reason besides low ESR for the cap  and that we really need 10,000uF - wouldn't it still be the case that a lower ESR film cap in parallel would do the lion's share of the work of clearing ac signal from the cathode?  By my (quite possibly wrong) calculations - a Nichicon 10,000uF 16V audio grade cap has 4 X the ESR of a Solen 100uF 250Vdc film cap (the latter was chosen arbitrarily just to see).

Here are my calc's, based on an arbitrarily chosen 1000Hx frequency:

Nicihcon Electrolytic 10,000uF

I don't remember what brand/series the 10,000uF cap is, but I figure the Nichicon UKA series is a good proxy.  The datasheet for the UKA series states that the 16V caps have a tangent of loss (DF) of 0.22 + an additional 0.02 for every additional 1000uF above 1000uF - so that makes 0.22 + 9*0.02 = 0.4

ESR = DF*Xc = 0.4*1/(6.28*1000Hz*0.01F) = 0.4/62.8 = 0.0064

Solen Film Cap 100uF

DF = 0.001 (from datasheet)
Xc = 1/6.28*1000*0.0001 = 1/0.628
ESR = 0.001/0.628 = 0.0016

So Nichicon electrolytic has 4X the ESR. If we think of the caps as just resistors from cathode to ground, then the 10,000uF electrolyic is 4 times more resistive than the 100uF film cap cap - shouldn't the film cap be the path of least resistance for ac signal?

Very curious/interested in this - please let me know if any of this make sense, and if not, where I've gone wrong.

cheers and thanks,

Derek

[Paul Birkeland]

(1) Can I just replace the 10,000uF cap with, say, a 50uF super low ESR cap?

You can, but the preamp will be significantly noisier.  ESR and reactive impedance are separate quantities. 

I don't remember what brand/series the 10,000uF cap is, but I figure the Nichicon UKA series is a good proxy.  The datasheet for the UKA series states that the 16V caps have a tangent of loss (DF) of 0.22 + an additional 0.02 for every additional 1000uF above 1000uF - so that makes 0.22 + 9*0.02 = 0.4

ESR = DF*Xc = 0.4*1/(6.28*1000Hz*0.01F) = 0.4/62.8 = 0.0064

Solen Film Cap 100uF

DF = 0.001 (from datasheet)
Xc = 1/6.28*1000*0.0001 = 1/0.628
ESR = 0.001/0.628 = 0.0016

So Nichicon electrolytic has 4X the ESR. If we think of the caps as just resistors from cathode to ground, then the 10,000uF electrolyic is 4 times more resistive than the 100uF film cap cap - shouldn't the film cap be the path of least resistance for ac signal?

10,000uF cap reactance at 120Hz: 0.13 ohms
100uF cap reactance at 120Hz: 13 ohms

Adjusted values: 10,000uF cap is 0.1364 ohms
                         100uF cap is 13.0016 ohms

[Deke609]

I got the ESR = DF*Xc equation from the attached "white paper" by Illinois Capacitor.

But now that I look at it, I see that DF is spec'ed at a certai frequency and I failed to factor that in.

But, more importantly, is disipation factor irrelevant? Is the equation wrong?

cheers and thanks, Derek

[Paul Birkeland]

The graph on the second page tells the story reasonably well.  The capacitive reactance line is way, way above the ESR line until the self resonance of the cap.  While there is no scale on that graph, this is a safe assumption for our 50/60/100/120Hz power supply noise issues.

The equation you want to use is the one for Z, which is impedance.  Xl and ESR are more or less zero at 60Hz, 120Hz, etc., so if you substitute zero in for Xl and ESR, you're left with Xc.  If you start working with switch mode power supplies that operate at high frequencies, all of this goes out the window.  That's why you'll see those power supplies have something like a 10,000uF cap and a 0.1uF ceramic cap in parallel with it.  The impedance of the big cap can get screwy at the switching frequency of the power supply, but a ceramic cap will be rock steady up there.

At the low frequencies we are working with, capacitive reactance is the dominant term in the impedance calculation.

[Deke609]

Many thanks for this PB. This is very helpful.

So the 10,000uF needs to stay for lower frequencies (120Hz hum). But then there's no harm in adding a 100uF film cap in parallel for any higher frequencies that might be there (assuming no oscillation problems between the two caps), right? Or would adding the higher impedance at at low freq. cap (100uF film) in parallel  produce more noise at low freq?

That's it for tonight - I promise!  Derek

[Paul Birkeland]

I don't think you're going to make any issues putting it in parallel, but I also don't believe it will provide much in the way of assistance.

[Deke609]

Ah! I see now. My (biggest) mistake was thinking that ESR took into account Xc, when in fact ESR is a separate factor as you stated. So ESR is not really relevant except when frequency is super low (e..g, 1-10Hz) or close to the self-resonant frequency. And in those cases, a low dissipation factor could make a difference, and would in fact make a difference were you comparing an electrolytic cap and a film cap with the same capacitance - e.g., 330uF (which is the largest film cap value I could find).  But not 10,000uF lytic versus 100uF film.

So I see your point about Xc dominating: capacitance is the name of the game for the Beepre cathode bypass cap. That being the case, more capacitance might make a difference.  So I think I'll try a 22,000uF Nichicon audio grade cap rated @ 80V (only b/c according to the datasheet the dissipation factor goes down a bit as the voltage rating goes up - "just in case" it might matter at super low or self-resonant frequencies).  This should reduce the already tiny Xc at 120Hz by a factor of 2.2.

To me, anything that can be done to reduce potential noise in the BeePre is worth the effort b/c that's where the signal is weakest. As it stands, the detail/clarity and dynamics from the combo of Beepre and dialed-in Kaiju is just astounding. I'm curious to know whether even more can be had.  At the same time, I've already reached the point where the detail and dynamics can be a bit fatiguing after 2-3 hours of listening. So more may not be better. But I'll cross that bridge if and when i get to it.

cheers and thanks again, Derek

[Paul Birkeland]

You can use the trim pots on the BeePre to trim the input down 3-6dB, which will allow you to turn the BeePre up 3-6dB more than normal, and that will also increase SNR.

[Deke609]

Many thanks! I wouldn't have thought of that. I only thought of using the pots for loudness imbalance between channels.  I'll try it.

[Deke609]

That actually makes a positive difference. I adjusted the Kaiju pec pots so that I went from -18dB on the Beepre coarse pot to -9dB. the difference is subtle but noticeable on cello (higher notes) and violin. There's a reduction in "glare" (what I thought was just recorded resonance from the recording space/concert hall) and some added focus.  Nice! it's small, but every bit helps.

many thanks for the suggestion.  cheers, Derek

[Deke609]

I installed the 22,000uF 80V Nichicon cathode resistor bypass caps in the Beepre. I decided to leave the stock 10,000uF caps in place. I also added a 0.1uF film cap - just in case ESR might make a difference at some frequencies. So in total I have 32,000 uF bypassing each 300B cathode resistor.

Sound impressions: I believe it makes a noticeable improvement: punchier and clearer. It's much easier to pick out individual instruments and voices. I wasn't totally convinced that it made a difference until I listened to "Obvious Child" by Paul Simon. That song always sounded congested to me, with Simon's vocals, the drumming and what sounds like a 12 string guitar smearing into/across each other. It's much less congested now.

In retrospect, going with the 80V rating was pointless and a complete waste of money. My only reason for doing so  was the slightly lower ESR of the 80V compared to lower voltage-rated equivalents. But b/c of the size of the caps, the only spot where they'd fit was down the middle where the shielded signal wiring runs - i.e., about 4 inches from the terminals to which they need to connect. The 8 inches of 24 ga wiring to each cap (4 in each for +ve and -ve) will negate much if not all of the lower ESR - hence the addition of the 0.1uF film caps. [Edit 2: clarification - as I *understand* things,  the additional wiring will contribute to both ESR and, potentially more importantly, parasitic inductance. The 0.01uF cap, which has very short lead lengths and negligible ESR/self-inductance was added to address that - particularly b/c I wasn't sure whether I'd need to remove the 10mF cap in the event that it didn't get along with the 22mF cap)

Were I to do it again, I'd try 16V or 25V rated caps - they're small and a fraction of the price of the 80V rated caps.

[edit 1: corrected my arithmetic: 10K + 22K = 32K, not 33K]

cheers, Derek