Esaj's stupid electronics questions -thread

[esaj]

5 minutes ago, hobby16 said:

Never seen a stupid question from you. So far... lol.

Hmm, let's try: if I were to build a filter for a switching power supply for something that needs very steady voltage (ie. filter out ripple & swings), why it's usually said you should use multiple capacitors (preferably even different types?) instead of one big capacitor? The reactance should be something like

Xc = 1 / (2*pi*f*C)

Which would mean that the reactance drops with higher capacity and/or frequency (so higher capacity should actually have lower impedance)? Yet smaller capacitors pass high frequency "better"? Charge time? ESR? What the hell is going on here?   I better revisit some books...

EDIT: Oh, right on that topic, overkill much?

 

[hobby16]

4 minutes ago, esaj said:

Hmm, let's try: if I were to build a filter for a switching power supply for something that needs very steady voltage (ie. filter out ripple & swings), why it's usually said you should use multiple capacitors (preferably even different types?) instead of one big capacitor? The reactance should be something like

Xc = 1 / (2*pi*f*C)

Which would mean that the reactance drops with higher capacity and/or frequency (so higher capacity should actually have lower impedance)? Yet smaller capacitors pass high frequency "better"? Charge time? ESR? What the hell is going on here?   I better revisit some books...

EDIT: Oh, right on that topic, overkill much?

 

There is no merit whatsoever to use 2x100uF capacitors instead of a 200uF, it's a waste of space, materials and money. The guy who says otherwise is talking black magic. But there is a merit to parallel a 100uF with a 100nF because each capacitor is for filtering a different frequency domain. Capacitors are modelized as circuitry based on C, R **and** L, so a 100uF can do things a 100 nF can not do and vice-versa because of the big difference in L. Even capacitor technology entails differences in filtering (tantalum behaves differently than electrolytics or polyester, etc.).

For switching power filtering, where there is a lot of voltage ripple, what's important is not paralleling capacitors or not (never been a problem afaik) but the capacitor ESR, which is dependent on the nominal voltage (capacitor's higher nominal voltage => lower ESR => good). A high ESR capacitor is perfectly good for power filtering in an audio amplifier but put it in a switched power converted and it will explode !

P.S. The picture, lol !

 

 

[esaj]

9 minutes ago, hobby16 said:

There is no merit whatsoever to use 2x100uF capacitors instead of a 200uF, it's a waste of space, materials and money. The guy who says otherwise is talking black magic. But there is a merit to parallel a 100uF with a 100nF because each capacitor is for filtering a different frequency domain. Capacitors are modelized as circuitry based on C, R **and** L, so a 100uF can do things a 100 nF can not do and vice-versa because of the big difference in L. Even capacitor technology entails differences in filtering (tantalum behaves differently than electrolytics or polyester, etc.).

Oh right, that makes sense... I've preached here a couple of times on the differences between ideal and real-world components, and then trip over it myself   

 

Quote

For switching power filtering, where there is a lot of voltage ripple, what's important is not paralleling capacitors or not (never been a problem afaik) but the capacitor ESR, which is dependent on the nominal voltage (capacitor's higher nominal voltage => lower ESR => good). A high ESR capacitor is perfectly good for power filtering in an audio amplifier but put it in a switched power converted and it will explode !

I'm pretty sure my cheap Aliexpress-capacitors are anything BUT low ESR...

Quote

P.S. The picture, lol !

I was actually going to use that behind a computer ATX PSU (there are actually four rails there, +12V, +5V and two grounds running in the middle)... Haven't tried it yet though

If you sometime have extra time (which I'm pretty sure you don't ), you can check here for more laughs, it's mostly my ramblings on a zillion (unfinished) electronics projects and such: 

 

[hobby16]

2 minutes ago, esaj said:

I'm pretty sure my cheap Aliexpress-capacitors are anything BUT low ESR...

I was actually going to use that behind a computer ATX PSU (there are actually four rails there, +12V, +5V and two grounds running in the middle)... Haven't tried it yet though

If you sometime have extra time (which I'm pretty sure you don't ), you can check here for more laughs, it's mostly my ramblings on a zillion (unfinished) electronics projects and such:

Low ESR capacitor is becoming more common now (look at the price of a switched power supply module having at least 2 low ESR capacitor, it costs peanuts) so in fact, you can buy them cheap from Aliexpress. If you want zero risk and maximum reliability, buy a tantalum capacitor but it's more expensive.

Some years ago, there has been millions and millions of computer PSU affected by the capacitor ESR problem, which has caused an epic episode of computer hardware erratic failures (the is even a wiki article about it). They all originated from several capacitor manufacturers who all have sourced their parts from chemistry factory who had a faulty composition in some batch, resulting in the problem in all the supply chain.

I don't know the exact percentage but I would say a lot, a lot, a lot (did'nt I already said a lot) of electronic problems come from the capacitors, especially from the supposedly low ESR capacitor which ages badly and end up dying, with the consequences you can imagine on the rest of the device. The irony is that : power saving regulalations (Energy Star and other stuffs like that) have compelled switching power supply with always higher frequencies and (supposedly) higher efficiency. But it also means much more taxing demands on the low-ESR capacitor, the weak link in the chain. End result : less reliable devices requiring more changes thus more total... power consumption. Call it  the law of unintended consequences.

[esaj]

5 minutes ago, hobby16 said:

Low ESR capacitor is becoming more common now (look at the price of a switched power supply module having at least 2 low ESR capacitor, it costs peanuts) so in fact, you can buy them cheap from Aliexpress. If you want zero risk and maximum reliability, buy a tantalum capacitor but it's more expensive.

One part where I (likely) need to get rid of any extra noise in the voltage is a function generator (actually, two ), as I suspect that the ripple could affect the output. I've been eyeing some designs based on ICL8038 (a cheap VCO), and also bought a breakout-board with AD9851 (a 180MHz DDS/DAC synthesizer). The first won't go up to but about 400kHz, the latter can go up to around 70MHz, as the breakout board has a multistage-filter to remove noise (which alone could be enough, even if the supply wanders?). Anyway, as I planned on using a basic BJT-amplifier to boost the signal with AD9851, I'm pretty sure the ripple would show up there (yeah, I haven't actually tried it, but I guess )... So probably need some filtering for the PSU (or just run it from batteries)..? Same probably with the ICL8038-based version, as I planned on using an op-amp to boost the signal.

5 minutes ago, hobby16 said:

Some years ago, there has been millions and millions of computer PSU affected by the capacitor ESR problem, which has caused an epic episode of computer hardware erratic failures (the is even a wiki article about it). They all originated from several capacitor manufacturers who all have sourced their parts from chemistry factory who had a faulty composition in some batch, resulting in the problem in all the supply chain.

I don't know the exact percentage but I would say a lot, a lot, a lot (did'nt I already said a lot) of electronic problems come from the capacitors, especially from the supposedly low ESR capacitor which ages badly and end up dying, with the consequences you can imagine on the rest of the device. The irony is that : power saving regulalations (Energy Star and other stuffs like that) have compelled switching power supply with always higher frequencies and (supposedly) higher efficiency. But it also means much more taxing demands on the low-ESR capacitor, the weak link in the chain. End result : less reliable devices requiring more changes thus more total... power consumption. Call it  the law of unintended consequences.

I've had a couple of motherboards spew the guts out from some electrolytics (but they were old boards anyway, likely not related to the ESR-problem). Usually when the capacitors start to bulge, it's time start looking to upgrade

[hobby16]

6 hours ago, esaj said:

One part where I (likely) need to get rid of any extra noise in the voltage is a function generator (actually, two ), as I suspect that the ripple could affect the output. I've been eyeing some designs based on ICL8038 (a cheap VCO), and also bought a breakout-board with AD9851 (a 180MHz DDS/DAC synthesizer). The first won't go up to but about 400kHz, the latter can go up to around 70MHz, as the breakout board has a multistage-filter to remove noise (which alone could be enough, even if the supply wanders?). Anyway, as I planned on using a basic BJT-amplifier to boost the signal with AD9851, I'm pretty sure the ripple would show up there (yeah, I haven't actually tried it, but I guess )... So probably need some filtering for the PSU (or just run it from batteries)..? Same probably with the ICL8038-based version, as I planned on using an op-amp to boost the signal.

There are several additionnal ways apart from your multi-capacitor board to get the cleanest output signal, when possible : use an analog generator instead of a DDS, use a linear power supply instead of a switched power supply (you can also use a switched PS to get 12V followed by a linear PS to get a cleaner 5V), add an inductance filter.

When using an amplifier, the "power supply rejection" is a parameter to take into account (you can find it for any Operational Amplifier) if you don't want the ripples from the power supply to be injected into the signal. And finally, if you don't have appropriate instrument to measure the noise (eg a FFT oscilloscope), it would be hard to know if your signal is as clean as you want and if any improvement measure is efficient or not.

[esaj]

1 hour ago, hobby16 said:

There are several additionnal ways apart from your multi-capacitor board to get the cleanest output signal, when possible : use an analog generator instead of a DDS, use a linear power supply instead of a switched power supply (you can also use a switched PS to get 12V followed by a linear PS to get a cleaner 5V), add an inductance filter.

The VCO (ICL8038) is analog, I think, but needs a minimum of 10V, but probably using a low dropout linear regulator would work? Or pump up the voltage with a boost converter, then use regulator... I actually also did build a simple "rail-splitter" (don't know if that's the correct term) that regulates the "virtual ground" via a pair of adjustable regulators (LM337, LM317), so it can work with different voltages (ie. split the rail creating the virtual ground in the middle of positive input and real ground), in case I want to create "real" +- signal. Probably handy with circuits needing dual-supply op-amps too.

The DDS seems to produce a pretty clean signal up to around 40-50MHz, the breakout board has something like "7th-order LC-filter" and keeps the sine-wave peak-to-peak pretty much at 1V peak-to-peak. It's 32bit, so even though it's DAC-based, the sine-wave does look pretty clean on oscilloscope. The chip itself is 5V (absolute maximum 6V), so I probably can't boost the voltage going there, but for larger amplitude signal I was planning on using a BJT-amplifier, as the op-amps I have cannot keep up at high frequencies. That's probably the part where I need some filtering, but I admit that that filter-stage I posted the picture of is probably overkill...  I did post a question in the Headache-topic if using too many capacitors could actually be harmful to the circuitry (or maybe the PSU). Probably it at least pulls a large (short lived) current spike when the capacitors charge up...

 

1 hour ago, hobby16 said:

When using an amplifier, the "power supply rejection" is a parameter to take into account (you can find it for any Operational Amplifier) if you don't want the ripples from the power supply to be injected into the signal. And finally, if you don't have appropriate instrument to measure the noise (eg a FFT oscilloscope), it would be hard to know if your signal is as clean as you want and if any improvement measure is efficient or not.

Thanks for the tip, I really need to dig up some "cheat sheet" or similar which explains all the parameters given in the datasheets. I know the basics of the op-amps (common signal rejections, input offsets, swings, slew rates, GBW, differences between rail-to-rail and "normal" op-amps etc), but there are lots of parameters in datasheets (not just op-amps, but in general) which meaning I don't really know.

I did get a Rigol DS1054Z a couple of months back (a 4-channel digital scope), which actually has the same hardware as DS1074Z and DS1104Z. When bought, it has about 30+ hours of trial time, during which time it works like the DS1104Z (100MHz bandwidth, 24M memory depth, advanced triggers, RS232/SPI/I2C decoders and such) but ran out of my trial period, so it's now 50MHz/12M, no special triggers, no decoders etc. How stupid is that, the hardware is fully capable of 100MHZ etc, but they limit it in the firmware, and then you can buy the "options" separately. There are places ( http://www.gotroot.ca/rigol/riglol/  for example) that can produce the unlocking codes based on serial number, but haven't tried that yet, as I haven't had that much need for higher bandwidth and the extra memory, and likely will lose my 3-year warranty if I unlock it myself.

It does have FFT, but not a very good one. Someone said that tweaking the settings so that it always uses full memory depth instead of the "auto"-mode, where it picks memory depth based on timescale, it should produce better results with FFT, but probably not that good even then.

Now I'm taking this topic way off-topic (yeah, I'm getting good at that), so probably should move this conversation to off-topic or something at some point  

[Blechi]

As to why paralleling capacitors might be a good thing:

https://www.youtube.com/watch?v=BcJ6UdDx1vg

https://www.eevblog.com/2015/05/09/eevblog-742-why-electrolytic-capacitors-are-connected-in-parallel/