2
\$\begingroup\$

I had used some capacitors as decoupling capacitors for my USB powered device. Due to some chip issues, this resulted in using three 47 uF capacitors.

I used electrolytic capacitors. Three in parallel, between Vcc and Gnd. However, there were still problems.

This other group I am working with used ceramic capacitors. They gave me their modified part and everything worked.

I thought that the main difference between the different materials in capacitors was when they are used at high frequencies. Why is it that ceramic capacitors worked in this case and not the electrolytic type?

Some links I used as reference:

\$\endgroup\$
  • \$\begingroup\$ Something didn't work with electrolytics and it didn't work in some unknown way - consider the information you have in your question and ask your self if you believe anyone on stack exchange is a mind reader? \$\endgroup\$ – Andy aka Feb 2 '15 at 22:58
  • \$\begingroup\$ Sorry if I was not clear enough. I wasn't hoping for mind reading, and just wanted to see why electrolytic decoupling capacitors did not work as well as ceramic decoupling capacitors. \$\endgroup\$ – WayneDinh Feb 5 '15 at 18:56
  • \$\begingroup\$ If you're as poor as I was trying to learn the trade, it really helps to play with a decent simulator, several of which are free to use. \$\endgroup\$ – Sean Boddy Feb 8 '16 at 21:38
  • \$\begingroup\$ @SeanBoddy Most simulators use generic (ideal) models for capacitors. Yes, parasitics can be assigned in some, but out-of-the-box this probably will not help illustrate the technical differences between a ceramic and electrolytic cap. \$\endgroup\$ – rdtsc Sep 30 '18 at 17:21
  • \$\begingroup\$ @rdtsc, indeed - you know, I'm not sure the question looked like that two years ago. Regardless, simulation is literally the only way I could have afforded to get through my training. But yes, configuring your capacitors realistically can be tricky, and impossible if you do not first look up what it takes to make them real. \$\endgroup\$ – Sean Boddy Sep 30 '18 at 18:36
9
\$\begingroup\$

As you say, electrolytic and ceramic caps have different performance at high frequencies. Basically, electrolytic caps stop acting like caps at much lower frequencies than ceramics do.

Decoupling is a high frequency issue, so decoupling caps need to work at high frequencies. Electrolytics don't, but ceramics do.

\$\endgroup\$
7
\$\begingroup\$

OlinLathrop and SomeHardwareGuy's answers are both correct. And since you mentioned this was a USB Bus Powered Device, there's another gotcha. According to the Universal Serial Bus Specification 2.0 Chapter 7 Electrical, section 7.2.4.1 Inrush Current Limiting:

The maximum load (CRPB) that can be placed at the downstream end of a cable is 10 µF in parallel with 44 O. The 10 µF capacitance represents any bypass capacitor directly connected across the VBUS lines in the function plus any capacitive effects visible through the regulator in the device. The 44 O resistance represents one unit load of current drawn by the device during connect.

This is easily missed, since it's pretty deep in the rev 2.0 spec document -- many years ago, we accidentally violated this spec on some of our early USB powered boards by using a 10uF capacitor, and the result was that sometimes, some boards would have too much inrush current when plugged into USB port. Windows would report an error, and would power-off that USB port until the device was unpluggged.

I didn't see this section listed in the USB 3.1 specification, and I assume the ceramic capacitors you used were the same value as the electrolytics (3 x 47µF). USB 2.0 spec was finalized in April 2000, an era when ceramic capacitors generally weren't widely available in values higher than 1µF, so electrolytic capacitors would have been more commonly used. The physics hasn't changed, but the economics has -- here in the year 2015 it's possible to buy >100µF ceramic capacitors, and the characteristics of ceramic capacitors are generally closer to "ideal" in this kind of application.

If you must for some reason use electrolytic capacitors on a USB bus-powered device, the solution is to either keep the amount of capacitive load directly connected to VBUS less than 10 µF, or use an external power supply (i.e. USB Self-Powered Device instead of Bus Powered Device configuration.). The FTDIchip.com FT232 data sheet has examples of using a FET to isolate the USB VBUS supply from the rest of the circuit. When the device is plugged into USB, the FT232 first negotiates with the USB host, and only after the host gives permission, then FET switches on to power up your device.

Without this negotiation, the sudden inrush current of >10uF of fully discharged electrolytic capacitors would be indistinguishable from a short-circuit fault. The USB port gets powered down to protect the host computer. There will still be some inrush current when the host sends the command to enable the device, but by that time, the host has already negotiated with the peripheral and given it approval to switch on.

Compared to electrolytic capacitors, ceramic capacitors have lower inductance, lower effective series resistance, and higher self-resonant frequency. Generally more nearly ideal performance for local power supply bypassing (at least below microwave frequencies). Surface-mount packaging also has less inductance than through-hole packaging. You didn't mention specifically, but I assume the 47µF electrolytic capacitors were probably through-hole. Even surface-mountable electrolytic capacitors are essentially the same as radial-leaded parts, with modified leads and a plastic base.

Electrolytic capacitors are still useful for bulk power supply decoupling, typically where power enters a board assembly -- electrolytics generally give more capacitance per unit volume than ceramic, and since the power supply system leads already have some series inductance, the additional inductance of the electrolytics is tolerable. But for local bypass (close to each IC), ceramic capacitors are essential.

With electrolytic capacitors, the inrush current is noticeable and measurable, and usually listed on the component's data sheet. With ceramic capacitors -- especially surface-mount capacitors -- that inrush current event is much smaller due to the lower inductance and higher self-resonant frequency.

\$\endgroup\$
  • \$\begingroup\$ I'm confused. Surely the inrush current for a lower ESR cap is higher than that for a higher ESR? Lower ESR = less resistance = higher current ... Am I missing something? \$\endgroup\$ – brhans Feb 3 '15 at 13:44
  • \$\begingroup\$ ESR, inductance, and SRF are not independent. Electrolytic caps have higher inrush current mainly because of inductance, as explained in one of the other answers. I see what you mean about that last sentence though, will edit. \$\endgroup\$ – MarkU Feb 3 '15 at 19:08
  • \$\begingroup\$ Moved the comparison of ceramic vs electrolytic and inductance/ESR/SRF to a new paragraph; removed reference to ESR in the last sentence. \$\endgroup\$ – MarkU Feb 3 '15 at 19:23
  • \$\begingroup\$ Thank you for the help everyone. I have swapped from electrolytic to ceramic and I can see why ceramic is essential in this case. \$\endgroup\$ – WayneDinh Feb 5 '15 at 17:30
5
\$\begingroup\$

Inductance... Followed by impedance. Your caps are trying to provide a low impedance path for current to flow at your frequency of interest. One culprit will likely be how you connected your caps, and the second will be their impedance over frequency.

Look for this curve in the datasheets of both parts. enter image description here

You'll see how much lower the impedance of the other groups caps will be at higher frequency (where your device probably needs it).

Not to say electrolytics are not useful in a power distribution network design, they're just only really useful at low frequency. The added inductance of the package doesn't help.

\$\endgroup\$
1
\$\begingroup\$

Digital ICs do not draw current continuously. They draw current in spikes when they switch. The faster the IC the faster and more frequent these spikes will be.

Moving to the frequency domain these spikes become high frequency components in the current waveform. To prevent these high frequency components in the current waveform resulting in unacceptable deviation in the supply voltage the supply must have a low impedance even at high frequency (just how high depends on the chip).

Unfortunately your incoming power supply will not have a low impedance at high frequency because wires have inductance. So to lower the impedance we add bypass capacitors (also known as "decoupling capacitors" though I find that term can be somewhat confusing).

So to be effective your capacitors have to still work as effective capacitors at high frequencies. Elecrolytics don't.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.