Hi,
many of the cap manufacturers publish impedance Vs frequency plots for their electrolytics.
I have seen many/all that show a flattening out of the impedance as frequency rises and then rising again further up the frequency range.
Some have a pronounced flattening over a wide range of frequency. I guess that is telling me they have become resistive in character.
I also note that these minimum impedances are generally only a decade or so above audio bandwidth, i.e. ~<=200kHz.
I cannot recall a bigger (non speciallised) cap, >=2200uF, that had the minimum into a few Mhz.
What happens with advertised as "low ESR" or "High frequency" types at these higher frequencies?
many of the cap manufacturers publish impedance Vs frequency plots for their electrolytics.
I have seen many/all that show a flattening out of the impedance as frequency rises and then rising again further up the frequency range.
Some have a pronounced flattening over a wide range of frequency. I guess that is telling me they have become resistive in character.
I also note that these minimum impedances are generally only a decade or so above audio bandwidth, i.e. ~<=200kHz.
I cannot recall a bigger (non speciallised) cap, >=2200uF, that had the minimum into a few Mhz.
What happens with advertised as "low ESR" or "High frequency" types at these higher frequencies?
As far as I can tell, low ESR caps intended for switchers have, in fact, lower ESR, and thus, higher ripple current ratings, but they do not have substantially better high frequency performance so far as remaining a capacitor. They still become resistive at sub Mhz frequencies. Some of my favorite caps are the Illinois KXM series, and the OS-CONs, but though they have high ripple ratings, they follow the same pattern as the higher ESR caps. As a side note, I haven't seen any really low ESR high voltage electrolytic caps, making high frequency switchers for tube amps more difficult than one might first think. Illinois does make some really fantastic high voltage polypropylenes, up to several kV, but they're not cheap or easy to get. I've yet to find my holy grail of a 100uF cap that shows better than -80 degrees at 100khz.
edit/addition- The impedances are so low that what I seek may be impossible to realize or impossible to measure even if it did exist. Or maybe they all do it, but I can't prove it- gotta think about this some more.
edit/addition- The impedances are so low that what I seek may be impossible to realize or impossible to measure even if it did exist. Or maybe they all do it, but I can't prove it- gotta think about this some more.
Best capacitors I've ever seen for high voltage switchers are the power rings, made by Sprague-Barre: http://www.sbelectronics.com/powerring/index.htm
I used a pair of 500 uF versions this past year in a compact 20 kW boost converter for a hybrid racecar...certainly overkill for any (most?) audio application, but they performed exceptionally well at high frequencies. I don't remember how they measured, but if I can dig up a datasheet later on I'll give some specs.
The downside? They start at around $400 a pop, and they're not exactly small for an amp.
I used a pair of 500 uF versions this past year in a compact 20 kW boost converter for a hybrid racecar...certainly overkill for any (most?) audio application, but they performed exceptionally well at high frequencies. I don't remember how they measured, but if I can dig up a datasheet later on I'll give some specs.
The downside? They start at around $400 a pop, and they're not exactly small for an amp.
Great idea! I actually have one and will test it, though it may be a few days before I can get to it
FWIW, my reality check is to parallel a bunch of film caps together, just to confirm that whatever bridge, vector analyzer, or LCR meter I'm using, is telling me the truth. Invariably, the parallel bunch gives the expected low DF, so the problem lies with the electrolytics, not the measurements, at least not in the 100uF regime. Above that, it probably gets more difficult.
FWIW, my reality check is to parallel a bunch of film caps together, just to confirm that whatever bridge, vector analyzer, or LCR meter I'm using, is telling me the truth. Invariably, the parallel bunch gives the expected low DF, so the problem lies with the electrolytics, not the measurements, at least not in the 100uF regime. Above that, it probably gets more difficult.
Conrad Hoffman said:
High capacitance, low ESL, reasonable cost: pick two. You pretty well have to bypass with some serious film caps to null out the ESL in a low-esr capacitor. Look at a typical OS-CON 10 uF 25 V capacitor in:
http://us.sanyo.com/industrial/electronic_components/capacitors/os_con/downloads/oscon_chars.pdf
The diagram in section 2 indicates it has about 30 milliohms ESR and 1.8 nHenries ESL, so by the formula in a previous post you want to bypass it with 1.8 nh / (.03 ^2), or about 2 uF of higher frequency capable capacitor. You can reduce that to 1 uF without too much effect on the impedance curve, but lower than that will show parallel resonance.
In any event, power supply impedances in tube amps are way higher than for sand-state, so higher ESRs at higher voltage ratings shouldn't be an enormous problem. Smaller capacitors tend to show lower ESLs as well (it's not a linear relationship, unfortunately, more like a square root), so failing all else you could parallel a bunch of them together. It's an accepted technique in swithing power supply design.
And here's an inspiration for you low-ESL fanatics out there, straight from Siberia:
http://www.sandia.gov/news/resources/releases/2007/rapid-fire-pulse.html
http://www.sandia.gov/news/resources/releases/2007/rapid-fire-pulse.html
Hi,
Do a simple little test, connect a square wave generator with lets say 50Ohm output impedance to the end of the uncut leads on a new radial low esr electrolytic cap. Set for suitable frequency and plenty of juice.
Now connect a scope near the generator connections and watch the result, switch glitches from the square ramps should bee seen. Move your scope connections (ground and probe) closer to cap, a decrease in glitches can now easily be seen as we move closer to cap.
Note this is just a few centimetres of the capacitors own leads!
Measuring over high current pcb traces on an amp delivering high frequency is quite revealing.
IMO, Eva has right about the great importance of good pcb layout and the performance of today’s low esr electrolytic caps.
Do a simple little test, connect a square wave generator with lets say 50Ohm output impedance to the end of the uncut leads on a new radial low esr electrolytic cap. Set for suitable frequency and plenty of juice.
Now connect a scope near the generator connections and watch the result, switch glitches from the square ramps should bee seen. Move your scope connections (ground and probe) closer to cap, a decrease in glitches can now easily be seen as we move closer to cap.
Note this is just a few centimetres of the capacitors own leads!
Measuring over high current pcb traces on an amp delivering high frequency is quite revealing.
IMO, Eva has right about the great importance of good pcb layout and the performance of today’s low esr electrolytic caps.
Hi,
Allow me to point to another situation where paralleling film and electrolytic caps could be worthy, to my inexpert eyes: SMPS.
Particularly, I have modded a cheap multiformat player, with a 0u1 polypropilene cap across the leads of the "big" (around 150uF) main SMPS electrolytic cap, with good effect (but I also swapped the rectifier diodes to low recovery types).
Any lights about this will be welcome
Thanks,
M
Allow me to point to another situation where paralleling film and electrolytic caps could be worthy, to my inexpert eyes: SMPS.
Particularly, I have modded a cheap multiformat player, with a 0u1 polypropilene cap across the leads of the "big" (around 150uF) main SMPS electrolytic cap, with good effect (but I also swapped the rectifier diodes to low recovery types).
Any lights about this will be welcome
Thanks,
M
Conrad Hoffman said:
H Conrad. Take this one for a spin. Consider a cap with, let's say, two bypasses. A general rule of thumb I work with is any bypass much smaller than a 1/10th the main cap risks resonating against the main's parasitic. Now preceed each of the three caps with an inductor sized on the same order as the following cap's inductance. Take a look at the total network's effect on PS filtering and the effective source impedance looking back from the load.
As an example, a 100uF with 30nH primary inductance bypassed with 10uf/20nH and 1uF/15nH. Adjust nH to taste to account for traces, etc.. So the power supply filtering (after the rectification cap) becomes:
30nH > 100uF > 20nH > 10uF > 15nH > 1uF > load
That looks dangerously like a circuit dreamed up by a tube guy! The Ls solves some problems I was having with simple bypassing, if the DCR is kept very low. Shouldn't be any problem at these values. I have everything handy to actually measure this, and will try it in the next few days. Thanks!
maxlorenz said:
Particularly, I have modded a cheap multiformat player, with a 0u1 polypropilene cap across the leads of the "big" (around 150uF) main SMPS electrolytic cap, with good effect (but I also swapped the rectifier diodes to low recovery types).
Any lights about this will be welcome
I'd recommend a larger poly cap, something like 0.47 uF to 1.0 uF. Bypass ratios of 1500:1 make me nervous about parallel resonances.
Conrad Hoffman said:
Guilty as charged. I'll take that as a complement.
Naturally as such I didn't think about large currents but it should work well for preamps, low power amps, etc. If the target is lowering HF PS impedance for reasons of stability no reason a small 'RF' bypass won't work with a large main cap jumps out. I wouldn't build a glass circuit without using this technique, to my ....perception it always brings clear benefits.these two latter inductances look quite small.rdf said:
Can the inductances be the wire interconnecting the main electros to the main circuit board? 1uf at the circuit and 10uF half way along the PSU to circuit board cable pair.
The snubbered Carlos PSU may mimic this topology where he has shown an RC at the output of the PSU and a further C at the circuit board.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.