The data sheets say use .1uf bypasses. The devices in question are a Wolfson W8740 dac and an AD797. I have a quantity of surplus .01 polystyrenes. The current bypasses are ceramics (and not npo/cog) Is there any downside to going to the .01 value and is there any upside to using the film caps? I would appreciate advice on both the digital and analogue devices. Thanks
I'd keep .1uf ceramics for bypass and use .01uf polystyrenes for signal, e.g. for RIAA correction, NFB networks in solid state amplifiers and such. It seems that ceramic caps in bypass networks has less influence on the sound - Cyril Bateman had a comprehensive review of that in, I believe, Electronics World.
I'm not an expert but here's what I think I know.
Typically it's better (i.e. safer) to use bypass caps that are slightly lossy, e.g. X7R ceramic instead of NPO or C0G, to damp any LC resonance that might be formed with the inductance of the supply rails. But you could instead, if necessary, add a small resistance in series with the supply rail, just before the bypass caps.
Think of the bypass caps as a small power supply that's there to meet a chip's sudden demands for current. The inductance of the supply-line conductors tends to slow down their (i.e. the supply lines') ability to "instantly" supply current steps or pulses that the chip needs. And that also causes voltage spikes on the rails (v = L x di/dt). But small capacitors very close to the chip's power pins can do it much more easily, and without disturbing the rail voltages nearly as much.
As a side effect, the bypass caps also act as low-pass filters (in combination with the supply line impedance), and tend to filter out externally-generated disturbances on the rails.
You can use a range of values in parallel, for bypass caps. But you have to be careful not to set up any LC resonances with the supply rail inductances. And adding more cap values raises that risk. For a chip that has very fast edge/rise times on the current steps or pulses that it needs, smaller capacitor values might be helpful, since they can discharge (supplying current) faster than larger caps. But having additional larger ones in parallel can't hurt, except for the potential resonance problem.
For "average" cases, or where no effort is put into researching chip specs or supply point impedance (and/or proper test instruments aren't available), it is customary to use a lower-quality 0.1 uF ceramic, often in parallel with a 10 uF electrolytic. Often, 0.01 uF is used instead of 0.1 uF, for digital IC bypassing.
The caps (especially the smaller one) should be connected as close as possible to the chip's power pin. But where the OTHER end of the cap is connected can be very important, too. You don't want the high frequency edges to have to go all the way to the power supply and back to get to the other end of the cap, for example. Often, the small cap should be placed directly across the pos and neg power pins, or the power and gnd pins.
For your current situation, you could always just try the polystyrenes, especially if you have a good oscilloscope and proper high-frequency probes with the proper accessories so you can check for ringing and can compare before/after waveforms, etc.
I would tend to want to use what the datasheet specifies. However, you could try adding the .01uF polystyrenes in parallel with the 0.1 uF, if you think there are high frequency requirements that warrant them.
The downside to going to only .01 might be that they can't supply sudden demands for current for a long-enough time, when it's needed, which might then result in spiky stuff on the supply rails, and/or errors in the ADC, or a slight degradation of mid or lower frequency response in the AD797, or something "unpredictable".
Keep in mind that this stuff is complicated and it's difficult to know which of several simultaneous "competing" phenomena might be important and which might be negligible, in a given circuit implementation, unless you have either a lot of experience or some good test equipment. i.e. "Your mileage may vary."
Cheers,
Tom
Typically it's better (i.e. safer) to use bypass caps that are slightly lossy, e.g. X7R ceramic instead of NPO or C0G, to damp any LC resonance that might be formed with the inductance of the supply rails. But you could instead, if necessary, add a small resistance in series with the supply rail, just before the bypass caps.
Think of the bypass caps as a small power supply that's there to meet a chip's sudden demands for current. The inductance of the supply-line conductors tends to slow down their (i.e. the supply lines') ability to "instantly" supply current steps or pulses that the chip needs. And that also causes voltage spikes on the rails (v = L x di/dt). But small capacitors very close to the chip's power pins can do it much more easily, and without disturbing the rail voltages nearly as much.
As a side effect, the bypass caps also act as low-pass filters (in combination with the supply line impedance), and tend to filter out externally-generated disturbances on the rails.
You can use a range of values in parallel, for bypass caps. But you have to be careful not to set up any LC resonances with the supply rail inductances. And adding more cap values raises that risk. For a chip that has very fast edge/rise times on the current steps or pulses that it needs, smaller capacitor values might be helpful, since they can discharge (supplying current) faster than larger caps. But having additional larger ones in parallel can't hurt, except for the potential resonance problem.
For "average" cases, or where no effort is put into researching chip specs or supply point impedance (and/or proper test instruments aren't available), it is customary to use a lower-quality 0.1 uF ceramic, often in parallel with a 10 uF electrolytic. Often, 0.01 uF is used instead of 0.1 uF, for digital IC bypassing.
The caps (especially the smaller one) should be connected as close as possible to the chip's power pin. But where the OTHER end of the cap is connected can be very important, too. You don't want the high frequency edges to have to go all the way to the power supply and back to get to the other end of the cap, for example. Often, the small cap should be placed directly across the pos and neg power pins, or the power and gnd pins.
For your current situation, you could always just try the polystyrenes, especially if you have a good oscilloscope and proper high-frequency probes with the proper accessories so you can check for ringing and can compare before/after waveforms, etc.
I would tend to want to use what the datasheet specifies. However, you could try adding the .01uF polystyrenes in parallel with the 0.1 uF, if you think there are high frequency requirements that warrant them.
The downside to going to only .01 might be that they can't supply sudden demands for current for a long-enough time, when it's needed, which might then result in spiky stuff on the supply rails, and/or errors in the ADC, or a slight degradation of mid or lower frequency response in the AD797, or something "unpredictable".
Keep in mind that this stuff is complicated and it's difficult to know which of several simultaneous "competing" phenomena might be important and which might be negligible, in a given circuit implementation, unless you have either a lot of experience or some good test equipment. i.e. "Your mileage may vary."
Cheers,
Tom
Last edited:
I probably should have added that the data sheets called for 10uf bypassed with .1uf ceramics. I am planning on increasing the 10 to a 47uf Oscon for the Wolfson chip and 47Uf Panasonic HFs for the AD797. Any comments or suggestions? Thanks for the help so far. I think I will leave the ceramics in place and use polys only in the signal path both digital and analogue.
The datasheet that says to use a 10uf and a 0.1uf is telling you true.
The big guy acts as the local power source for peak loads, while the 0.1 uf is the sheriff responsible for keeping transients and high frequency cr@p from sneaking into town.
I like polypros for the 10uf and, as GooTee points out, X7Rs for bypass.
10 uf is plenty for opamps, but for amps and buffers something bigger like your 47 uf are certainly useful. A friend uses motor start capacitors for amplifiers. Huge and ugly, but effective.
The big guy acts as the local power source for peak loads, while the 0.1 uf is the sheriff responsible for keeping transients and high frequency cr@p from sneaking into town.
I like polypros for the 10uf and, as GooTee points out, X7Rs for bypass.
10 uf is plenty for opamps, but for amps and buffers something bigger like your 47 uf are certainly useful. A friend uses motor start capacitors for amplifiers. Huge and ugly, but effective.
Hello,$0.02 worth.
Did you say what voltage PS you were bypassing?
I often ceramic or monolithic ceramics on PS to chips, right at the pins chips often have the bandwidth to oscillate into the RF.
I often also use ceramic caps to bypass the heater pins of tubes.
0.01or 0.1 uf whatever comes out of the box no preference.
Ceramic capacitors have better high frequency response to bypass RF then film. (not good for signal path application)
DT
All just for fun!
Hi,
I cannot recall ever reading a manufacturer's datasheet recommending polystyrene for decoupling duty.
Ceramic, Polycarbonate, Polyester, Tantalum Electrolytic and Aluminium Electrolytic all get recommended for IC decoupling.
Only use what each manufacturer recommends for their IC.
When you can prove that your replacement works as well or better, then consider doing something different. Until then do not change from recommendations.
I cannot recall ever reading a manufacturer's datasheet recommending polystyrene for decoupling duty.
Ceramic, Polycarbonate, Polyester, Tantalum Electrolytic and Aluminium Electrolytic all get recommended for IC decoupling.
Only use what each manufacturer recommends for their IC.
When you can prove that your replacement works as well or better, then consider doing something different. Until then do not change from recommendations.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.