it is hard to say how a parallel cap combo will behave in practice..... for example, the effect of lead length or pcb trace. This is not a long length -- as used inside speaker crossovers and power supply to circuit trace lengths. ETC.
View attachment Cap lead length affects.pdf
Thx-RNMarsh
View attachment Cap lead length affects.pdf
Thx-RNMarsh
Last edited:
And, another measured result of how the lead length dominated the circuit and made the frequency extension of low Z with a small parallel cap null:
View attachment Lead length affects-2.pdf
THx-RNMarsh
View attachment Lead length affects-2.pdf
THx-RNMarsh
Thinking deeper about paralleling caps, manufacturers of electrolytics are not so stupid and there is no reason we can achieve better results in 2D (pcb) than they can do in 3D (volume of an electrolytic) to minimize inductances.
Thank for this point of reflection, Richard.
Last edited:
Leader length primarily sets the inductance, that's why smaller components perform better in that respect. Worst is the BIG loosely wound teflon caps they even have bad mechanical resonances. But I don't agree to the fact that we can't better the ESR of big electrolytics, We can by adding a small good cap right next til the device, this also to my experience improve the OPS stability. Lateral mosfets really need a local cap and sufficient gatestopper not to oscillate.
If you compare a film RC with a lytic then the lytic must have the same ESR as in the RC. That information is missing from your chart, but it looks to me like the lytic just has large ESR.
That is why I specified the 47uF 100V Nichicon VR. It has .2-.3R ESR. If you can show this cap or one similar has less discharge ability than a film cap with a .2-.3R series resistor, then I would be interested.
Yes, lead inductance is most important above 500KHz or so, and the 1st 1uS where most of the discharge occurs tends to be dominated by ESL. Were the pin spacing of the lytic and film cap you tested equal?
I have made measurements. Not discharge measurements, but impedance and square wave tests. A small lytic has much less inductance and more capacitance than a discrete RC would have. When using a lytic like this to damp the parallel resonance of a film cap, the less inductance the better because after you damp the first resonance so much, the film cap begins to resonate with the loop inductance through the lytic. So you have to use a lytic with an ESR value that's not too high or too low. Using this approach you don't have to put any resistor in series with the film cap. Not only this, but you can have both capacitors with low pin spacing in parallel and this halves the total inductance.
That is why I specified the 47uF 100V Nichicon VR. It has .2-.3R ESR. If you can show this cap or one similar has less discharge ability than a film cap with a .2-.3R series resistor, then I would be interested.
Yes, lead inductance is most important above 500KHz or so, and the 1st 1uS where most of the discharge occurs tends to be dominated by ESL. Were the pin spacing of the lytic and film cap you tested equal?
I have made measurements. Not discharge measurements, but impedance and square wave tests. A small lytic has much less inductance and more capacitance than a discrete RC would have. When using a lytic like this to damp the parallel resonance of a film cap, the less inductance the better because after you damp the first resonance so much, the film cap begins to resonate with the loop inductance through the lytic. So you have to use a lytic with an ESR value that's not too high or too low. Using this approach you don't have to put any resistor in series with the film cap. Not only this, but you can have both capacitors with low pin spacing in parallel and this halves the total inductance.
lets not play his audiophool ad copy “energy release rate” game though – try showing a audio amp's actual current demand with audio signal, typical dynamic loudspeaker load – its not 5A/us
such glib puffery interferes with actually understanding the real issues in audio amp part selection, application
very garden variety bulk reservoir caps in linear audio power supplies don't have “energy release rate” limitations in use – because they aren't being shorted by the amp's output Q – they are delivering 20 Hz - 20 kHz audio signals
designing/controlling the power supply impedance at the output device pins for MHz local parasitic oscillation control is usually physically separated, the ps wire/trace routing wipes out any “paralleling a quality film cap with the bulk supply electro” typical audiophile tweak advice relevance to the actual issues
such glib puffery interferes with actually understanding the real issues in audio amp part selection, application
very garden variety bulk reservoir caps in linear audio power supplies don't have “energy release rate” limitations in use – because they aren't being shorted by the amp's output Q – they are delivering 20 Hz - 20 kHz audio signals
designing/controlling the power supply impedance at the output device pins for MHz local parasitic oscillation control is usually physically separated, the ps wire/trace routing wipes out any “paralleling a quality film cap with the bulk supply electro” typical audiophile tweak advice relevance to the actual issues
The Hypex class D amps use only bulk foil decoupling on their entire range of power amplifiers , no sign of any parallel plastic film caps anywhere and you would expect them to be used here more so than a linear amplifier .
Last edited:
More real data rather than conjecture. There is no better truth than a good measurement of what you have attempted to accomplish. You really dont know what you have done until you measure it;
You might note the comparison with film and bi-polar caps is being shown? This is for applications primarily related to speaker EQ/cross-overs. But, has generic info for viewing why stability is affected by caps of various values and types and also how bypassing doesn't always get you what you think you got (Z-wise). Cross-overs are notorious for needing some tweeking of values after the SIM has been done... This type of measured info indicates why that happens.
THx-RNMarsh
View attachment C phase 1.pdf
View attachment C phase 2.pdf
View attachment C phase 3.pdf
You might note the comparison with film and bi-polar caps is being shown? This is for applications primarily related to speaker EQ/cross-overs. But, has generic info for viewing why stability is affected by caps of various values and types and also how bypassing doesn't always get you what you think you got (Z-wise). Cross-overs are notorious for needing some tweeking of values after the SIM has been done... This type of measured info indicates why that happens.
THx-RNMarsh
View attachment C phase 1.pdf
View attachment C phase 2.pdf
View attachment C phase 3.pdf
Last edited:
Whether it is transient/impulse response, phase angle maintained at 90 degrees or low Z vs freq... all these performance issues are beyond the simple Xc/lZl used in SIM.
I hope it gives added perspective to the physical application and affects of small amounts of Rs (esr) and Ls.
THx-RNMarsh
I hope it gives added perspective to the physical application and affects of small amounts of Rs (esr) and Ls.
THx-RNMarsh
Last edited:
And, if one wishes to extrapolate this real world info... other than IC size circuits -- amp stability with small value C comp parts and almost any length of trace path will not get the SIM stability margins/results... thus, more tweeking and fudging values.
-RNM
-RNM
Last edited:
actually its a pain to use |Z| in sim - actual lumped components are much simpler
where continuous/distributed properties are important just a few sections of lumped element approximation to transmission line, skin, DA effects is usually quite a good fit to data at audio or even up to typical audio power amp loop gain intercept frequencies
if you look at real EE knowledge, IEEE papers, IEEE press books on the subjects instead of the strawman stupid audio engineer beloved of audio gurus as a foil
where continuous/distributed properties are important just a few sections of lumped element approximation to transmission line, skin, DA effects is usually quite a good fit to data at audio or even up to typical audio power amp loop gain intercept frequencies
if you look at real EE knowledge, IEEE papers, IEEE press books on the subjects instead of the strawman stupid audio engineer beloved of audio gurus as a foil
Last edited:
Can you explain why this information conflicts with anything I have said?
Yes, I have measured this stuff. I found it impossible to successfully damp a rail resonance without laboring through the signal generator testing to find which lytic from my junk box damped the resonance. I found that lytics can have very low ESL and that the ESR is effectively constant into the MHz region. Unless there are extreme nonlinear effects going on, then the discharge test will reflect the measured impedance. A 200mA 10nS risetime square wave should be enough to make any such effects visible, and that's more extreme than anything you would see through decouplers in an audio amp.
Furthermore, a discharge test operates into a low impedance. In real applications the output transistors are like leaky current sources. This is why capacitor SRF doesn't affect them, but parallel resonance peaks do - it is voltage rather than current that induces leakage across such a current source. For this reason I think a more applicable test is to inject a signal from the output of a 50R square wave generator into the rails and watch how they respond.
Yes, I have measured this stuff. I found it impossible to successfully damp a rail resonance without laboring through the signal generator testing to find which lytic from my junk box damped the resonance. I found that lytics can have very low ESL and that the ESR is effectively constant into the MHz region. Unless there are extreme nonlinear effects going on, then the discharge test will reflect the measured impedance. A 200mA 10nS risetime square wave should be enough to make any such effects visible, and that's more extreme than anything you would see through decouplers in an audio amp.
Furthermore, a discharge test operates into a low impedance. In real applications the output transistors are like leaky current sources. This is why capacitor SRF doesn't affect them, but parallel resonance peaks do - it is voltage rather than current that induces leakage across such a current source. For this reason I think a more applicable test is to inject a signal from the output of a 50R square wave generator into the rails and watch how they respond.
dont know that is does conflict. Rather, more explanation from other perspectives.
the cut and try method works very well, too. Your measurement based approach is a good one, once you have a physical unit to work on and can measure and then tweek on it to get the real world results you are looking for.
it's all good info.
THx-RNMarsh
Last edited:
Just somes are living in a pure and marvelous world of 'knowledge' (or what they believe to be), and others have experienced that the things are often not so simple than believed 🙂
Somes love to show-off their 'science' while others just use the simplest and fastest way to achieve the results they expect. Who deserve the ironic name of "Audio gurus' ?
Last edited: