Capacitor tolerance at +20% (Recapping a PC Engine Duo)

[jsteel]

I'm in the process of recapping a PC Engine Duo. I bought a bunch of Panasonic caps, thinking I didn't want to just buy the cheapest, these were a brand I've heard of and were easily available (in the UK).

While the tolerance on them (and the originals) states 20%, this seems like it could be pretty high, eg 470uF could read 564uF and still be in spec? Well, testing them, all of them read very close the advertised values apart from my 470uF, they read about 520 to 530uF. I wondered in practice what would happen if all the caps for example were 20% higher than the advertised value; nothing? interference? fire? I'm curious.

Cheers

[NewSchoolBoxer]

I like Panasonic electrolytics too. I buy the ones rated for at least 5000 hours. Right, after the X hours of use they're rated for, they're guaranteed to be within 20% of the advertised (nominal) value. The 20%, rated for X hours or the tolerance on any passive component, really, is the worst possible case.

If you got 560uF instead and they were labeled as 470uF, I'd assume them to be defective and wouldn't use. The thing is, electrolytics really aren't placed in circuits where the exact value is important. As a bypass/filtering capacitor to reduce ripple voltage on a DC supply, 470uF is just about the same thing as 560uF. They filter the same band of low frequencies. You may want a 10uF to 47uF capacitor in parallel to filter higher frequencies. Common to use factors of 10.

For the main capacitor on a DC power supply hooked to the AC power outlet, the ripple voltage reduces by approximately (2 x 50 or 60 Hz x C), so more capacitance is basically always better and the limitation is physical size and cost. I like seeing 4700uF but 3300uF is okay. The 2200uF on NES is too low imo, meaning you should power it with DC instead.

Where 20% components shouldn't be used is passive and active filters needing a specific corner frequency. Like a 8.5 MHz low pass filter for Standard Definition video. I'd say 2% is the practical tolerance limit here. Too much capacitance lowers the cutoff and softens the video while too little increases the cutoff and reduces the amount of noise that is filtered. The basic filter strategy is you pick a high resistor value to get a low capacitor value that comes in 1% ceramic or film form. You can typically get the same corner frequency by multiplying all resistors and dividing all capacitors by the same number.

I know you didn't ask all that but I wanted to give practical examples. If you want to delve more into this, I suggest learning about capacitor derating due being run close to their maximum voltage and/or temperature.

[jsteel]

Thanks for the detailed reply! On the Duo I'm not sure what caps these are exactly, my guess is the power supply. From your reply it sounds like this is a non-issue then, but possibly the caps should be much closer to the advertised value from the start. I'll contact the seller and see about swapping these ones out. Cheers

[jsteel]

I just double checked my other caps and I think I didn't notice the discrepancy as much with the lower values, as they were only say 13uF out, but on a 100uF that's 13%. Looking at percentages, my 470uF are about 12% higher than this, my 100uF are 13% higher and 47uF 11% higher. The other values are all about 5% higher. It's quite possible my multimeter is not particularly accurate, it was only a cheap one but I bought it specifically for its ability to measure a capacitor (out of circuit). Is there another way to measure a capacitor to verify my multimeter is correct? I'm not sure whether to be fussy and chase caps that give me a more accurate reading, or just not fuss and go ahead with the recap with these. I'd appreciate to know what someone more experienced would do here.

Thanks

[VEGETA]

NewSchoolBoxer did a nice reply. Basically if it is power related stuff then don't worry about it since more is better. Brand new caps should read very accurately, like 470u should be 470u with some minor reading margins.

What I would like to see in such applications, is the use of solid polymer capacitors instead of electrolytic ones, and I mean solid polymer not hybrid polymer. these are a lot better and do not need to be replaced as they don't have electrolyte inside of them. they also has orders of magnitude better ESR which is a huge factor in pretty much all applications especially ripple rejection and filtering. swapping your elec caps with solid polymer of same size and value in power supply filtering circuit is enough to give nice performance boost.

such retro systems are already costly, and getting nice elec. caps is not pennies... thus I suggest moving to solid polymer ones for a bit more extra.

as for brands, I suggest Lelon brand for both elec. caps and solid polymer ones. it is based in Taiwan\China but of high quality and being used on professional and expensive equipment, so it is not like other Chinese brands. you could go for Panasonic and Rubicon if you want to and accept the extra cost.


Examples:

Kemet elec. cap - 0.25$ for 1.
Kemet solid polymer cap - 0.45$ for 1.

Lelon elec. cap - 0.49$ for 10.
Lelon solid polymer cap - 1.2$ for 10.

as you can see for Kemet polymer cap it has mere 18 mOhms of ESR compared to high ESR of elec. cap which can go beyond 0.5-1 Ohms. While that Lelon polymer cap has mere 7 mOhms of ESR! Other brands like this Nichicon solid polymer cap is 5 mOhms ESR but at 1$ price for 1 unit, not worthy compared to Lelon one.

[NewSchoolBoxer]

Yeah thanks you two! Brands, I like Panasonic capacitors because they're pretty and I assume Panasonic uses their own. Unreal Engine stays better than Unity imo because Valve uses it for their own Fortnite. Sometimes I make fun of the retro gaming Nichicon cult. They're fine but several other top tier brands exist. I appreciate the Lelon tipoff!

I've never seen solid polymer caps discussed in retro gaming but they should be. Replacing leaky high ESR liquid electrolytic with the same thing is following the sold for a loss / just survive to the next console generation logic. I bought pairs of Kemet 10-15-22-33-47 uF solid tantalum because LDO regulator datasheets kept recommending them. Some ESR is actually necessary for stability so can't use MLCC ceramic without a low ohm series resistor and that takes up more space.

Now that I think about it, I'd like to see solid electrolytics in RGB cables and the 220uF is really just the minimum acceptable value and any lower from age and wear and tear on is going to darken the screen from left to right and top to bottom.

So I did a science experiment with measuring two sets of 10x Panasonic 220uF ±8% I bought last year. It's a nice hobbyist meter for the low price point. Capacitor accuracy is ±8%. While the 16V readings settled in about 5 seconds, the much larger 63V capacitors took about 5x as long to settle from ticking upward.

Purchased from official distributor DigiKey as can be seen. The multimeter IC hooks, I should have bought sooner:

Spoiler

There is the concept of precision. If you buy a bulk rate of capacitors/resistors/whatever, they're probably from the same manufacturing lot. Components cluster around the same value from the same lot, whether it's a little high or low. That the 16V were a little high and the 63V a little low, just how it was that day. Factor in 8% worst case measurement error and I got what I paid for.

There is also the concept of counterfeit components. I was surprised to learn people sell counterfeit capacitors and logic gates when the real things costs 20 cents, but they do. If you bought from an official distributor and the values are within the stated tolerance stacked with multimeter tolerance, you should assume they're fine unless actual use proves otherwise.

Is there another way to measure a capacitor to verify my multimeter is correct?

Yes, the voltage tolerance on the multimeter is probably much better than the capacitance. You need an AC supply. If you don't have a function generator, the NES AC supply is simply a transformer with turns ratio to output 9-10V AC at 50/60 Hz. A 555 timer could work, oscillator, whatever. You want a input voltage maybe 10-35% of what the capacitor is rated for and low power (from low current) to avoid derating or burning the resistor.

Edit: Wait sorry you're using polarized capacitors, -300mV is about the max safe limit for negative voltage. Can try voltage that low but could instead add a DC offset so the voltage doesn't swing below 0.

Now make an RC high pass or low pass filter. Put both in series and measure the voltage drop across the component you place second and tie to ground. Calculation a little easier to place the capacitor first. DC offset gets removed from equations and a simpler quadratic. Measure Vout as voltage across the resistor. Vout = Vin * R / (sqrt(Xc^2 + R^2)), where you measured Vin, Vout and R. Can autosolve with Wolfram Alpha. Xc = 2 * pi * f * C, so divide Xc you calculated by (2pi * 50 or 60 Hz) for C.

Multimeter will measure rms values. If the generator gives you the Vpeak value, divide it by sqrt(2).

[NewSchoolBoxer]

Well, I see on my meter that the AC voltage tolerance is 1.5% and AC current tolerance is 1.8%, so the current strat is possible. Just feed AC sine wave to a capacitor with a high enough DC offset! That's the whole circuit. If you haven't measured current before, you do so in series by passing it through the multimeter. A 9V AC supply is 9Vrms and the meter would measure rms or correct itself to. Irms = Vrms * 2 * pi * f * C. If the generator tells the you peak voltage then divide that by sqrt(2) to get Vrms.

So a measured 9.8Vrms AC input to a capacitor for a measured current of 450mArms yields: 0.45 = 9.8 * 2 * pi * (60 Hz) * C -> C = 121.8 uF. Capacitor should be rated considerably higher than 9V and 450mA ripple current at low frequency. If you need to add a resistor to attenuate the AC supply then you're better off with the voltage strat.

[VEGETA]

I like Panasonic capacitors because they're pretty

yes they look nice with that purple color.

Sometimes I make fun of the retro gaming Nichicon cult. They're fine but several other top tier brands exist. I appreciate the Lelon tipoff!

you got Nichicon, Nippon chemicon, Rubicon, Kemet, Panasonic, and others. Lelon are nice as I said before and you've got also Kyocera AVX, SamXon, Epcos...etc. Just don't get a no-name brand from China.

I've never seen solid polymer caps discussed in retro gaming but they should be. Replacing leaky high ESR liquid electrolytic with the same thing is following the sold for a loss / just survive to the next console generation logic.

yes, I saw people using Tantalum caps but not Aluminum solid polymer like I recommended.

Some ESR is actually necessary for stability so can't use MLCC ceramic without a low ohm series resistor and that takes up more space.

yes this is in power supply circuits, especially switching power supply since ceramics rings a lot if kept alone. traditional wisdom says you use all ceramics in all your stages + put one elec. cap since it has high ESR which dampens everything.

if you replace elec. caps with polymer you will still have some esr but significantly lower... i don't think anything bad will occur in terms of stability. actually, most retro power supply circuits either use linear regulation or very low frequency switching regulation which both won't suffer from low esr stability problem. if you are extra paranoid you can add a small elec cap as an extra to ensure stability.

Now that I think about it, I'd like to see solid electrolytics in RGB cables and the 220uF is really just the minimum acceptable value and any lower from age and wear and tear on is going to darken the screen from left to right and top to bottom.

this should be a no brainer. using solid polymer caps here is way much better than elec caps due to esr and the fact that their life span is way longer. solid polymer = solid electrolyte while elec. caps use liquid.

____

In my upcoming universal dreamcast power supply (100-230v AC input) I used all sorts of caps. ceramics everywhere + 1 elec. cap each stage for dampening and stability + one big polymer at the rail output for fast energy dispatch and lower the ripple. I will make a sample soon.

[NewSchoolBoxer]

I like your brand overview. I never thought of a brand as legit if DigiKey, Mouser or Jameco didn't stock it but you're showing me I was mistaken. LukeEvansSimon who probably has more CRT knowledge than anyone likes Cornell Dubilier based in the US. They supplied capacitors to the US military during WWII and later NASA. I don't suppose everyone likes a company that sells directly to the military but good enough for me.

I compared tantalum prices to solid electrolytics and tantalum is at least 50% more experience. The solids have much lower ESR and we don't need tantalum reliability in 1Vpeak video signals. I want solid polymer video cables. Wouldn't be hard to swap in a set and measure the difference. Even if minimal, real gain is outlasting the console.

In my upcoming universal dreamcast power supply (100-230v AC input) I used all sorts of caps. ceramics everywhere + 1 elec. cap each stage for dampening and stability + one big polymer at the rail output for fast energy dispatch and lower the ripple. I will make a sample soon.

You know what you're talking about! I'm slow rolling a linear regulated supply for 9V/1.5A to cover 16-bit consoles. See if passive power factor correction reduces current distortion without the inductor making things worse. If I can't compete with medical grade supplies then at least I learned something.

Measuring capacitance from voltage or current

Here is my deriving capacitance from voltage and current experiment: https://imgur.com/a/y7DiuJU
Idea is drop the voltage enough with R to not blow up the electrolytic. I knew -300mVpeak at positive terminal was fine long-term from video sync voltages and amped would be double that. Why the 9VAC supply reads 12V under load I think is due to the 2mA current being too low for normal operation on a 1.33A supply.

Using a 330uF solid polymer that I ordered after reading VEGETA's reply. I was saying calc was easier to put the capacitor first and solve for resistor but can auto-solve regardless. With C and R in series, they can go in either order. C doesn't blow up going first.

Say we doubt the 8% tolerance multimeter reading of the capacitor. Let's solve for it with the voltage strat:

Spoiler

AC voltage divider from R and C being in series: Vc = Vs * Zc / (Zc + R) where R = 471.7 ohms
0.207 = 12.04 * Xc / (Xc + 471.7)
Xc = 8.25166 = 1/(2*pi*f*C) where f = 60 Hz
C = 0.00032146 = 321 uF, compares nicely to 338.6 uF from multimeter.

Alternatively solve with current strat with slightly less accurate AC current measurement:

Spoiler

V/Z = I => 12.04/Z = 0.00014 => Z = 86000 ohms
Z = sqrt(R^2 + Xc^2) where R = 471.7 ohms
Xc = 86000 = 1/(2*pi*f*C) where f = 60 Hz
C = 0.0003084 = 308 uF, compares less nicely to 338.6 uF from multimeter. 15mA versus measured 14mA would have yielded 330uF.

We can presume the multimeter reading is a little high. Factoring in source resistance and ESR would reduce calculated C.
The 1/4 watt resistor was extremely hot. I used 1 kohm after that out of precaution.