We normally see .1uF MLCC NP0 used as power bypass. Using film tends to be controversial, but is there a sound benefit, for example using MKS or MKP or would a tight tolerance MLCC NP0 still be best for this application?
I'm interested in this feedback too. I've experimented and have heard a difference. In my test case, I went from MLCC to mica. Instantly I heard a greater level of clarity - opamp bypass, but on extended listening I thought it was leaning towards harsh. I let it go 24/7 for a couple of weeks, and it normalized keeping an elevated level of clarity but without edge. I've seen MKS used in various pieces of source equipment, and wondering consensus around using more classically MLCC in contrast to MKS or other types per above - PIO, MKP, etc.
The proportions of individual cap values has to be distributed in a specific ratio for the most linear behavior throughout the frequency spectrum to be covered by the filter.
Just adding a tiny bypass cap isn't going to do much good, considering the small cap wil throw linearity of the relatively "loose" electrolytic filter portion out of whack with a sharp peak in the HF. Its more effective to use a combination of choke and cap (LC) for smoother, less peaking attenuation of all unwanted frequencies past 100+ hz. This also avoids ringing from the diode cutoff adding higher current noise.
The use of a snubber is advised either way. Higher quality smaller PP film caps are more than sufficient for this purpose.
Just adding a tiny bypass cap isn't going to do much good, considering the small cap wil throw linearity of the relatively "loose" electrolytic filter portion out of whack with a sharp peak in the HF. Its more effective to use a combination of choke and cap (LC) for smoother, less peaking attenuation of all unwanted frequencies past 100+ hz. This also avoids ringing from the diode cutoff adding higher current noise.
The use of a snubber is advised either way. Higher quality smaller PP film caps are more than sufficient for this purpose.
Well that's exactly my thinking Tom.. Looking at these designs, you have to question it. I mentioned in other threads, I've been on an anything-but-WIMA kick, and it really hasn't let me down. Like, they aren't great sounding caps, and maybe also not so great for this bypass purpose.
An easy way to design in the snubber is to use an electrolytic cap where the power enters the circuit board. 10-100 uF will usually work well. It'll have some ESR, typically 0.5-1 Ω which is sufficient to dampen out ringing. Then 100 nF at each supply pin of analog ICs. Maybe add 0.1-1 nF in parallel if you're dealing with a >100 MHz part.
If you want to get fancier than that, I suggest basing your decisions on measurements of the supply ringing and/or EM emissions with the circuit in use ... unless you have access to an EM field solver and can run a whole-board simulation with parasitics included.
I'll definitely recommend that those who geek out over decoupling to run a simulation of the implemented network with some reasonable estimates for the parasitic inductance included. I think you'll find that adding a whole school of capacitors in parallel might actually be worse than having a few capacitors in parallel.
Tom
If you want to get fancier than that, I suggest basing your decisions on measurements of the supply ringing and/or EM emissions with the circuit in use ... unless you have access to an EM field solver and can run a whole-board simulation with parasitics included.
I'll definitely recommend that those who geek out over decoupling to run a simulation of the implemented network with some reasonable estimates for the parasitic inductance included. I think you'll find that adding a whole school of capacitors in parallel might actually be worse than having a few capacitors in parallel.
Tom
The WIMA MKP and MKF capacitors are great film capacitors. I'd recommend them if you need a polypropylene film capacitor. But for decoupling I'd go with ceramic caps. NP0/C0G preferred, but X7R works just as well as long as you take the voltage coefficient into account. Even with C0G/NP0 it's worthwhile to go with 100 V types rather than the commonly available 25 or 50 V ones.
Many forget that the temperature coefficient of the various film types is actually quite large (100s of ppm/ºC). Versus zero for C0G/NP0.
The voltage coefficient of larger (> 0603), 100 V, C0G/NP0 is basically zero as well, whereas for some film caps (PIO I'm looking at you!!) the voltage coefficient can be rather horrid. Horowitz & Hill speculate that the voltage coefficient of film caps comes from the 'squishyness' of the dielectric. Paper is pretty squishy. As is some types of plastic. That makes the dielectric change thickness due to the electrostatic force between the two capacitor plates -> higher capacitance with higher applied voltage. Ceramic is not very squishy... The voltage coefficient of the high-k dielectrics (X7R, Y5V, etc.) comes from the piezoelectric properties of the material. C0G/NP0 is not piezoelectric...
Tom
Maybe these articles will keep you up at night.
Of course, without a well designed pcb with a ground plane, things will get even worse.
https://www.edn.com/bypass-capacitor-resonances/
https://fscdn.rohm.com/en/products/..._capacitor_impedance_characteristics_an-e.pdf
https://www.kyocera-avx.com/docs/techinfo/DecouplingLowInductance/mlcbypas.pdf
https://www.ti.com/lit/an/sloa069/sloa069.pdf?ts=1722349530395
https://www.analog.com/media/en/training-seminars/tutorials/MT-101.pdf
https://www.compuphase.com/electronics/ParallelCapacitors_Antiresonance.pdf
https://www.signalintegrityjournal.com/articles/1589-the-myth-of-three-capacitor-values
Of course, without a well designed pcb with a ground plane, things will get even worse.
https://www.edn.com/bypass-capacitor-resonances/
https://fscdn.rohm.com/en/products/..._capacitor_impedance_characteristics_an-e.pdf
https://www.kyocera-avx.com/docs/techinfo/DecouplingLowInductance/mlcbypas.pdf
https://www.ti.com/lit/an/sloa069/sloa069.pdf?ts=1722349530395
https://www.analog.com/media/en/training-seminars/tutorials/MT-101.pdf
https://www.compuphase.com/electronics/ParallelCapacitors_Antiresonance.pdf
https://www.signalintegrityjournal.com/articles/1589-the-myth-of-three-capacitor-values
And don't forget the X2Y capacitors. It's two capacitors in a single SMD package. Brilliant as it drastically reduces ESL. The decoupling example shows just a single rail. The real advantage (IMO) is for decoupling spit/bipolar supplies.
Tom
Tom
Me and many others did between PIO and some others way back when, but they were large values for big horn/woofer speakers where motor run caps were/are the norm at least for some of us, though smaller ones for multiway, small FR depending a lot on tube Vs SS and later found that making small/cheap cap grids were the 'hot ticket' when PIO got so expensive.