Not sure how physically big this is, but this should work well.

http://www.amazon.com/100MFD-100uF-Motor-Starting-Capacitor/dp/B0087ZB0LI

Need to check the caps on mine!

Even better - $3.19 with free shipping:

http://www.amazon.com/TOOGOO-100MFD-100uF-Starting-Capacitor/dp/B00YOKWYZQ/ref=pd_sim_sbs_328_14

OK, I have some really good news for us AFM owners: You can repair/rebuild your strobe board with normal, inexpensive electrolytic capacitors instead of impossible-to-find non-polarized caps. Each 100uF/100V NP capacitor can be replaced by a commodity 100uF/250V capacitor and a 1N4004 diode.

Here's the link to the how-to: https://pinside.com/pinball/forum/topic/replace-caps-on-afm-strobe-board#post-3184588

The rest of this post is background on this modification and proof-that-this-is-not-a-crazy-thing-to-do:

We know there was a big mistake in the original design: C1 is rated at 100V, but C1 sees about 145V in normal operation. But there was another "mistake" (really more a missed opportunity) in the design that will eliminate the need for these impossible-to-find non-polarized caps.

The problematic circuit is this:
Original Schematic

C1-C4 and D1-D4 form what is known as a "voltage doubler" (in this case a voltage quadrupler since they employ the technique twice in series). This circuit takes the 50Vac input and turns it into about 290Vdc. I'm not going to explain how voltage doublers work, but there's plenty of info here: http://www.allaboutcircuits.com/textbook/semiconductors/chpt-3/voltage-multipliers/ and on google.

Here's a simulation of that circuit (ignore diodes D5 and D6 for now):
Simulation of voltage quadrupler

I believe the designer of the original circuit used non-polarized capacitors because it seemed like the right thing to do in a doubler, particularly if you don't fully think-through the design. In the original design, C2 does need to be NP, because when the circuit is starting up (when power is first applied) the voltage across C2 (which is the voltage at point B minus the voltage at point A, or "V(b)-V(a)", the gray trace) does go to -30V before it goes positive and settles at about +70V:
Startup of original design

But note that except for the bottom of the first 60Hz (or 50Hz) cycle, the voltage across C2 is always positive. That means we could use a normal, polarized electrolytic capacitor for C2 if only we could get rid of that one-time -30V dip. Well it turns out that's not going to be hard to do. If we put a diode (D5) across C2 as shown, it will allow as much positive voltage as we need, but will clamp any negative voltage to about -0.7V. And -0.7V for 1/60th of a second every time you power on your AFM is not going to hurt a polarized cap (electrolytic caps are designed to withstand -1V to -1.5V specifically so they can be protected by a diode).

Here's the simulation with the diode:
Startup with Schottky diode clamp

I put a diode across C1 as well (D6), but it's really not necessary because the way the circuit works, C1 will always be positively charged.

When you're analyzing circuits with large capacitors, it's important to see what happens when you turn OFF the power as well, because unexpected things can happen. Fortunately in this circuit, everything just gradually discharges to 0V:
Power-On and Power-Off

So my recommendation is to replace C1 and C2 with 100uF / 250V polarized capacitors, add D5 (and D6 if you want), and enjoy a very robust Strobe Board.

Here's a good cap choice:
http://www.newark.com/illinois-capacitor/107tta250m/aluminum-electrolytic-capacitor/dp/28M9179

Use a standard 1N4004 diode for D5 and D6.

Hope this helps. If there are any questions let me know and I'll answer/update the post.

I couldn't have said it better myself!

I don't have time right now but will work on the TL;DR summary soon.

Yep - ltspice. http://www.linear.com/designtools/software/ Free and fairly easy to use. I believe it's really nice on Windows - the Mac version I used has a very painful GUI (you can memorize all the keyboard shortcuts to avoid).

Is pinwiki editable by anyone? I'll try to clean up the explanation here first...

Can anyone post or send me a highish-res photo of both the front and the back of the strobe board? That would speed up my how-to write-up (I won't have to pull mine from my machine).

No question, two caps back-to-back can work.

It has already been mentioned that, connected in series you HALVE the value of capacitance but ADD the voltage. In the case of the board we make below, we use 220mfd @ 250V so, with two in series, we end up with 110mfd @ 500V.

Agree on the halving of the capacitance, but not on the doubling of the voltage. Look at it this way: if you had two 250V caps back-to-back in series and you put 500V across the string, you could have 250V across each cap (in which case one cap would have +250V across its + to - terminals, but the other would have +250V across its *-* to *+* terminals, which is verbotten), or you could have +500V across one cap and 0V across the other (to prevent reverse biasing the cap), but then you'd obviously be exceeding the max 250V spec of the cap.

Adding voltages works when they are non-polarized caps (and can therefore split the voltage evenly), but not for polarized. At best you'll end up with a 110uF @ 250V cap.

What was I partially incorrect about? My only point was that a 250V electrolytic cap back-to-back with another 250V cap doesn't give you a 500V cap - it becomes a non-polarized 250V cap. I guess you could argue that it's 500Vp-p, but that's not how max capacitor voltages are spec'd.

In any case, the caps we use at 250V are each more than double the originals at 100V.

I agree. Though I'd rather use silicon diodes to control the current flow than rely on the characteristics of a reverse-biased electrolytic. If I *was* using the two caps back-to-back solution, I'd still put a 1N4004 across each cap to guarantee no reverse biasing and have a high-current path that doesn't go through a reverse-biased capacitor.

If you break it down, the back-to-back solution only ever (except for the bottom of the first 50/60Hz cycle on C2) uses 1 of the 2 caps used to form C1 and/or C2. The other capacitor in each pair is basically just acting as a diode, as well as cutting the total capacitance in half. The back-to-back solution apparently works, but it's certainly not as good from a cost, physical size, and design elegance perspective as this new modified design. I suspect the modified design (by not using electrolytic caps as diodes) is additionally also more reliable as the caps age over time, but I don't know enough about reverse-biased electrolytics to know for sure.

I'm not saying anyone should worry about their boards if they are currently working. But if you're repairing/restoring a board you might want to consider this modified design.

I just reread both notes for the fourth time and still can't find anything mentioning the total voltage being any higher than that of one cap. But I may have a blind spot - can you quote just the part where they say that? The closest I can find is:

Article #1: "When voltage is applied, the correct polarity capacitor gets the full voltage." That's exactly what I've been saying: with one polarity voltage, the cap in the "right" direction gets the full voltage (250V max in our example), then when the polarity is reversed, the voltage across that cap goes to zero (actually more like -1V) and the max of 250V is applied to the other (now the correctly-biased) capacitor. At no point do you ever get more than 251V across the two capacitors (the additional 1V comes from the reverse biased cap acting as a diode).

Article #2: the closest I can find to a relevant passage here is the exact same sentence that's in article #1: "When voltage is applied, the correct polarity capacitor gets the full voltage." (Almost makes me wonder if there's a little plagiarism going on.). Again, the "full voltage" is the max voltage of one capacitor, and it applies not only for the first half cycle but for any amount of time the voltage is that polarity.

BTW, what the last sentence of article #1 is saying is that the only reason the back-to-back capacitors don't overheat and explode in our AFM application is because they are seeing DC most of the time - one is always reverse-biased and acting like a diode, and the other just sees a large DC voltage. They only see any AC during startup, before they settle to their DC values, and after the strobe fires, where the voltage across C3 and C4 is depleted causing the C1/C2 voltage doubler to have to supply current again to charge them up. The modified circuit I'm proposing never subjects the capacitors to any of the excess heating and corresponding short life the article mentions, which is one of the reasons I consider it a better design.

We are nitpicking anyway as the application notes clearly suggest that it is a perfectly sound method of creating a non polarised part and that is the point I was making.

Sorry if I seem nitpicky - I just genuinely want to understand your point because if what you're saying is correct, it contradicts a lot of what I know about electronics/physics/mathematics and I'd like to correct my understanding if it's wrong.

Homepin: I just realized that the Homepin I was having the capacitor discussion with was *probably* the Homepin that made that beautiful replacement strobe board shown in https://pinside.com/pinball/forum/topic/replace-caps-on-afm-strobe-board#post-3178339 with the www.homepin.com URL printed on it. (I really need to start paying more attention to handles. )

I want to make sure everyone reading this thread knows that I agree that your board is a solid design and will work fine in their AFM. It's a much better board/better design than the original. It remains the best solution to the original problem: how to replace an unavailable NP capacitor with an even higher voltage (and even less available) NP capacitor.

When I analyzed the issue, I realized that the design didn't actually need NP capacitors, so I was able to suggest an alternative solution. I did not mean to suggest that your solution/board was inferior - we just had a little back-and-forth about how to make NP caps out of P ones. The design needs 160V caps, you believe you created 500V caps, I believe you created 250V caps, but whichever one of us is right you have at least 56% voltage margin (which is very safe) vs the -37% of the original design (which is really inexcusable).

TL;DR (which is what I should probably change my handle to): No one should have any qualms about purchasing the Homepin strobe board, and it would be what I recommend to people who don't do board work or just want to buy a well-engineered solution that works.

Great catch!!

Remember what I said in my PM about pinball electronics flagrantly ignoring the laws of physics? I just looked at the schematic, and if you'd asked me "What would happen if D7 was reversed?", I would have said "That's easy: *nothing* would happen. The board would simply never fire the strobe because a reversed D7 would just block the trigger signal for the opto isolator, so as far as the board could tell, players just never made it to Strobe Multiball. Very basic diode theory.

But you saw it working fine and shutting down the sound board.

My head just exploded. Again. From freakin pinball electronics.