Decoupling Caps Question

I like lots of 0.1 uf disc ceramics placed at 1-2 inch intervals along both axis. Cheap and effective even with transients from TTL logic gates.

The capacity is not so important. Specify by ESR value! For example use Tantal EPS from AVX - very low ESR.

regards - Henry

"linnix" schrieb im Newsbeitrag news: snipped-for-privacy@g14g2000cwa.googlegroups.com...

What, exactly are these capacitors bypassing or decoupling? Are these just across the DC bus feeding logic (paralleling some external supply) or are they part of an on board switching power supply? There is no way I can judge if the array of capacitors is at all appropriate for their intended use when I don't know what the use is.

Yes, they are for on-board switching regulators. There are also 0.01uF (x20) ,0.1uF (x60) and 1uF (x25). We don't need to change the rest, but only to reduce the expensiive SMDs ($2 to $3 for 100uF SMD vs. $0.25 for radials) . My question is whether I can replace the evenly placed 100uF with 470uF (closest to the regulators) and lowering the rest to 68uF and 4.7uF. Most of the caps are connected in parallel anyway. Thanks.

Tantakum bead caps used for power decoupling are sometimes known to explode / catch fire. I would discourage their use.

Graham

Sounds odd that there are no 0.1 uF in there. Large caps rarely make the best decouplers.

Hard to advise without knowing more ( computer board tells quite little ) and why perhaps the unusually large values you mention were originally specified. Does sound odd to me though.

Graham

The 100uF SMD devices may be something like the Panasonic ceramic ECJ4YB0J107M which has ESR in the order of

Let me rephase the problem.

There is a 100uF Cvr (Cap for voltage regulator output) and 100 uF Cuc (Cap for microcontroller). The power requirement is 1.5V @ 3A. We are proposing adding an Cext (Cap external) and running wires back to Cvr, or Cuc, or both (if it makes any difference). They are all 2" to 3" aparts.

The existing design use 100uF for all (1.5V, 2.5V and 3.3V) SMD caps. So, I don't think the designer spent too much time on performance and cost anyway. For next batch of PCB, we would change them to hole mounted. But until then, we need a creative solution.

We are trying to reduce Cvr and Cuc to 10uF, 6.8uF or 4.7uF and adding Cext with 220uF or 470uF to compensate. Hopfully, this would not change the low freq response too much. What do you think? Thanks.

For each of the paralleled sets, I would estimate the total RMS current rating of the existing combination, and also the effective parallel ESR and ESL. If I can come up with a cheaper combination that has at least as much total capacitance and RMS current rating (at the operating frequency), and no higher ESR and ESL, then there is little likelihood that any new problems will occur with that swap, providing I do at least as good a job with the layout as the original had.

If you want the very lowest cost, you would have to re-engineer the circuits to know if the total capacitance, RMS current rating, ESR or ESL are way better than they need to be, or if better layout or some other small design change could allow lower spec parts to do an adequate (this word refers to well informed engineering decisions, not guesses) job.

I have seen cases where a big batch of poorly suited capacitors could be replaced with a different set of choices that had lower total capacitance, cost and size but did a better job.

In some cases a better layout considerably reduced the need for large capacitor values.

The cap that is mounted on the board can be assumed to take all the high frequency components of the ripple current. Si if you are going to change a single, high value capacitor to a combination of a smaller one mounted where the original was, and a second, large, off board capacitor, you should assume that the smaller should have as low an ESR as the original. The larger, off board one will carry most of the fundamental frequency of the ripple, and will help a little with the harmonics.

The best advice I can give you, is to measure the ripple voltage of the original design, Then replace it a large, a good, low ESR ceramic multilayer capacitor (like a Panasonic X5R or X7R, with plenty of excess voltage rating, which lowers the ESR) and a good, low ESR electrolytic (like the Panasonic FM series) with a ripple current rating of a good fraction of the DC output current, to mount off board with the lead wires twisted. The input capacitors have a higher ripple current load than the output ones do, because of the filtering effect of the output inductors.

For example, the output cap for the 1.5 volt, 3 amp supply regulator might be a combination of a 10 uF, 10 volt, X5R 1210 size ceramic multilayer (ECJ-8YB1A106M) in parallel with a 560 uF, 6.3 volt EEU-FM0J561 (ripple current rating of .95A, ESR=.056 ohm).

Then measure the ripple voltages again and see if you have no more than there was with the original capacitor. I would want to see a significant improvement to allow for variations in other pairs and aging.

The correct place to attach the wires for the output filter caps is between the anode of the buck diode and the output end of the inductor.

The correct place to attach the wires for the regulator input cap is between the +5 volt connection on the switching mosfet, and the anode of the buck diode. Each switching regulator should have its own high frequency input capacitor, even though they may all be in parallel on the schematic. You may get away with a single large parallel capacitor, off board, if all the regulator mosfet inputs are close together, but remember that higher ripple current here, especially if the regulators can run synced (all drawing current at the same time, and then all off at the same time). There is little risk of having extra capacitance on the input side, since this does not get involved in the stability of the regulator.

Not if you load them correctly withinthe spec! There is a small possibility with tantalums to die very early within the burn-in.

Sorry, the correct name is TPS!

- Henry

"Pooh Bear" schrieb im Newsbeitrag news: snipped-for-privacy@hotmail.com...

explode /

Is the regulator a OTS module or not? If it is a module, then the datasheet will specify the capacitive loading necessary for it to remain stable. Check the datasheet, which should specify not only the capacitance, but also the ESR range required.

If it is a power supply designed onto the board, then the designer would have found it necessary to take loop compensation, input ripple and output ripple into account. The output capacitor is a fundamental part of the loop compensation, and if you wish to change it, you should make sure your replacement will maintain loop stability across the various loads, for which you will need at least a decent idea of the output load profile, amongst a number of other things. (Those things include mode [voltage or current], input voltage variation [for a voltage mode controller], output load variation [for a current mode controller], duty cycle, switching frequency and a thorough knowledge of the loop model of the controller).

The input capacitors are required to have certain ripple capacity, which requires a thorough knowledge of the operation of the supply (such things as duty cycle and load). Although they don't directly affect the loop, the wrong device can have spectacular failures (in particular, it's not generally a good idea to put tantalums on the input to a switcher).

What capacitors are used? There are numerous SMD solutions for 100uF, and they are all different.

With the Cuc cap only a couple of inches from the regulator, it affects it's loop compensation (yes, I'll harp on about this) as well as ripple performance. So the question remains - is it a drop in module or not? If not, you have more to consider than ripple performance. If it is, you should still abide by the module requirements (and you should have a datasheet showing the requirements).

Cheers

PeteS