replacing ecap w/film cap on speaker crossover
- Thread starter oldman55
- Start date
Is this the article:http://www.mit.edu/~6.331/an47fa.pdf
I read AN47-12 on parallel bypassing caps, I don't see anything about zero ESL causing trouble. He was more talking about poor bypass caps causing problem.
I do parallel bypass. When I designed microwave circuits, I have 10uF in the general location, then 0.1uF 0603 close to the power pin of IC or transistor. Then 1000pF 0402 very close to the pin. The final step is an island of plane right under the component right on top( or bottom) of the ground plane to form an ideal parallel plate cap of say 100pF or so as the last line. I use a VIA right next to the power pin of the IC to get in very close. This parallel plate cap should be as close to zero ESL and ESR as it can.
I did not work on anything over 5GHz, but I have absolute luck with this. Waveforms were picture perfect.
One thing, I don't think you can use scope probe after 100MHz or so, Probe is not very flat at high frequency. I did all coax and SMA connection, not even BNC. My experience is in RF, if you don't see the waveform you want, 50% is how you measure it. It's not like these kind of audio amps, you literally can measure waveform with probes without the ground lead.
I read AN47-12 on parallel bypassing caps, I don't see anything about zero ESL causing trouble. He was more talking about poor bypass caps causing problem.
I do parallel bypass. When I designed microwave circuits, I have 10uF in the general location, then 0.1uF 0603 close to the power pin of IC or transistor. Then 1000pF 0402 very close to the pin. The final step is an island of plane right under the component right on top( or bottom) of the ground plane to form an ideal parallel plate cap of say 100pF or so as the last line. I use a VIA right next to the power pin of the IC to get in very close. This parallel plate cap should be as close to zero ESL and ESR as it can.
I did not work on anything over 5GHz, but I have absolute luck with this. Waveforms were picture perfect.
One thing, I don't think you can use scope probe after 100MHz or so, Probe is not very flat at high frequency. I did all coax and SMA connection, not even BNC. My experience is in RF, if you don't see the waveform you want, 50% is how you measure it. It's not like these kind of audio amps, you literally can measure waveform with probes without the ground lead.
Last edited:
Hipocrates
Super Member
I am interested what make them different. So we know you can get cheap film caps that have ESR and ESL down to <20mohm and <20nH. So what gives? You can only get down to ESL=ESR=0 to get a perfect cap, you cannot do better than a perfect cap.
Could it be they pray on psychosomatic nature of people. Like the $10K speaker cable!!!
Could it be they pray on psychosomatic nature of people. Like the $10K speaker cable!!!
Could it be? Naw, we're not susceptible to that stuff. 
Why would zero esr be bad? Because the cap would form a high Q tank circuit with any inductance, and there's always inductance. A small esr keeps Q low and insures that most accidental resonances are negligible. A big reason tantalum capacitors were popular for filtering voltage regulators wasn't some magic property of tantalum capacitors as filters, but the fact that they have a moderate and consistant esr that's about right to keep a regulator stable without forming a noise peak.
I wish people would stop using esr, but now they've got me doing it. Esr is an absolute number in ohms, which, though useful under some circumstances, is also misleading. One is much better off using dissipation factor (like engineers have used for decades) because it's the ratio of esr to reactance. If you know the dissipation factor at any given frequency is low, you know it's a small part of the reactance and you don't have to worry about it.
Why would zero esr be bad? Because the cap would form a high Q tank circuit with any inductance, and there's always inductance. A small esr keeps Q low and insures that most accidental resonances are negligible. A big reason tantalum capacitors were popular for filtering voltage regulators wasn't some magic property of tantalum capacitors as filters, but the fact that they have a moderate and consistant esr that's about right to keep a regulator stable without forming a noise peak.
I wish people would stop using esr, but now they've got me doing it. Esr is an absolute number in ohms, which, though useful under some circumstances, is also misleading. One is much better off using dissipation factor (like engineers have used for decades) because it's the ratio of esr to reactance. If you know the dissipation factor at any given frequency is low, you know it's a small part of the reactance and you don't have to worry about it.
True, BUT.............Zero ESR and ESL only happen in theoretical world, there is always parasitic resistance and inductance. Just how small you can get it. Also it is layout dependent also. It is a complete system. I always use ground and power plane......yes, even in this amplifier circuit to keep all the parasitic low. I have 0.1uF bypass cap within 0.1" from the collector of each emitter follower transistor on the board from ground to power plane. Then I have 10uF electrolytic close by on every transistor also. I have been doing this even on transistor circuits with frequency response to 5GHz for many years.
A good design has to be looked at in a system point of view, ESR and ESL of capacitors is only one part of the equation. Point is you want to keep all the parasitic low as much as possible so you don't have to deal with them. That the circuit becomes simpler and more predictable.
In microwave, we talk about wavelength. Every small distance becomes LRC element as in the Smith Chart. It's just how small you can get so it is out of the picture of the circuit. We use the guide line of if the distance is less than 1/20 of wavelength, we ignore it. Even SMD, chip components has parasitic lead length and capacitance, never zero. That's a given already, we just want to keep it as low as possible.
Another way to look at it is you make the parasitic low enough that resonance frequency is way beyond the circuit can respond so it becomes a non issue. So you design the circuit and choose components accordingly. I use leaded 0.1uF bypass cap that I know would work to over 100MHz, so for audio power amp, it's an ideal cap.
From experience, leaded components cannot work reliably beyond like 300MHz. Yes, you can monkey around and make it work, but it's how you move the parasitic to just make it work. You change the layout, the same circuit might burst into oscillation.
A good design has to be looked at in a system point of view, ESR and ESL of capacitors is only one part of the equation. Point is you want to keep all the parasitic low as much as possible so you don't have to deal with them. That the circuit becomes simpler and more predictable.
In microwave, we talk about wavelength. Every small distance becomes LRC element as in the Smith Chart. It's just how small you can get so it is out of the picture of the circuit. We use the guide line of if the distance is less than 1/20 of wavelength, we ignore it. Even SMD, chip components has parasitic lead length and capacitance, never zero. That's a given already, we just want to keep it as low as possible.
Another way to look at it is you make the parasitic low enough that resonance frequency is way beyond the circuit can respond so it becomes a non issue. So you design the circuit and choose components accordingly. I use leaded 0.1uF bypass cap that I know would work to over 100MHz, so for audio power amp, it's an ideal cap.
From experience, leaded components cannot work reliably beyond like 300MHz. Yes, you can monkey around and make it work, but it's how you move the parasitic to just make it work. You change the layout, the same circuit might burst into oscillation.
Last edited:
Any good designer has to think in terms of system and circuit location. As a cap example, take a 0.01 uF Teflon cap (PP is very similar) with D=.000023 at 1 kHz. At 1 kHz the total impedance is 15915 ohms. The esr is 0.366 ohms. Q is 43478. That cap would be a horrible choice for the output of a voltage regulator, creating a bad HF noise peak when combined with the pseudo-inductance of the regulator, but a great choice for an RIAA network if it were in series with some other impedance.
I was really talking about bypass cap now. Signal cap is a different animal. I agree for signal cap, as long as the impedance( ESR or ESL) is not insignificant in the whole signal chain, it's not that important. But I have to caution for crossover where impedance is down to 2ohm or below, so even if the parasitic is 0.1%, it matters.
I have to say, the $300 cap for audio frequencies is horse manure ( $hit).
I have to say, the $300 cap for audio frequencies is horse manure ( $hit).
