47nF capacitor doubles THD when put across a 4ohm load? The capacitive reactance is 3.4kohm.
Confirmed, the problem was really in the 1st soundcard used. It had similar H2 as the amplifier under test and depending on phase of the distortion component there was a partial or complete H2 cancellation that grossly affected THD vs. amplitude plots. I have re-measured and re-posted most of them with a soundcard that has 5-10x lower distortion than the amp, but has no balanced input, so I had to use my analogue MSYS instrument that has balanced inputs and is intended as a front end for measurements with SE input soundcards. I should have done it in the beginning, but I was too lazy .....
The balanced input is absolutely necessary to prevent measuring groundloop, similarly as is done in AP and QA401.
Last edited:
Not if you can float everything on batteries, but life it much easier with balanced inputs. Also an isolated floating source helps a lot.The balanced input is absolutely necessary to prevent measuring groundloop, similarly as is done in AP and QA401.
Even if everything was powered from batteries, there would be a groundloop if there the amp under test has SE input and speaker output with internal common signal ground. Measuring cables would bring additional loop if the measuring system has common ground for input and output. Yes the floating output of the measuring system prevents this.
Last edited:
Keantoken, it seems to be obvious that the abrupt changes in EPDR impedance plot from REW are exactly at the points where the phase response crosses zero and changes sign. Don’t you think it is absolutely logical?
I must admit to having a bit of difficulty equating this SOA-related EPDR to actual speaker impedance. So if this EPDR calculated value is actually peak instantaneous VA in the PA, can we say the EPDR value is the instantaneous value of the speaker impedance with that particular signal at that instant??
Granted we do not have the formula used to calculate EPDR in REW,maybe I could ask John Mulcahy to chime in here...
Howie
I must admit to having a bit of difficulty equating this SOA-related EPDR to actual speaker impedance. So if this EPDR calculated value is actually peak instantaneous VA in the PA, can we say the EPDR value is the instantaneous value of the speaker impedance with that particular signal at that instant??
I think these EPDR data look reliable to me. If I compare the results with my dummy load
http://pmacura.cz/dummy_load_EPDR2.gif
obtained from full SOA simulation in Microcap and replace the dummy load with the EPDR resistance read from the EPDR plot at the appropriate frequency, I get almost same results. So quite easily, a resistance value at the given frequency and make Ic vs. Vce plot, of course in instantaneous I and V values, no averages, no rms.
Could somebody please explain what EPDR really brings.
I see a nice curve, but what doest it tell and how should it be interpreted.
Looking at a Soa curve it’s easy to see that at a certain frequency phase shift can easily push the amp outside the Soa, with much less as the maximum permitted power.
I still fail to see this with an EPDR.
Hans
I see a nice curve, but what doest it tell and how should it be interpreted.
Looking at a Soa curve it’s easy to see that at a certain frequency phase shift can easily push the amp outside the Soa, with much less as the maximum permitted power.
I still fail to see this with an EPDR.
Hans
I'm not sure what I can add beyond what has already been covered in this thread or in the Audio Science Review thread linked earlier, which has a formula to calculate it. The EPDR value is the resistance that would result in the same device peak power dissipation as the speaker load at a given frequency, under the assumption of class B amplification. Keith Howard's original intent was to come up with a measure that better reflected how a given speaker might tax an amplifier than the raw plot of its impedance.Granted we do not have the formula used to calculate EPDR in REW, maybe I could ask John Mulcahy to chime in here...
This EPDR thing is giving me deja vu. Teh Definitive EPDR Thread is at this link, and what I was reading here was, I thought, a revitalization of that thread:
Is speaker impedance ever less than DC resistance, even under transient conditions?
Is speaker impedance ever less than DC resistance, even under transient conditions?
I'm not sure what I can add beyond what has already been covered in this thread or in the Audio Science Review thread linked earlier, which has a formula to calculate it. The EPDR value is the resistance that would result in the same device peak power dissipation as the speaker load at a given frequency, under the assumption of class B amplification. Keith Howard's original intent was to come up with a measure that better reflected how a given speaker might tax an amplifier than the raw plot of its impedance.
I understand wthe articles describing EPDR, but have yet to see a test showing the speaker itself as the sink for the extra power EPDR predicts, as opposed to extra dissipation purely in the output devices. Also I'm not sure REW uses the same algorithm used in the ASR article.
Cheers!
Howie
Howie
This EPDR thing is giving me deja vu. Teh Definitive EPDR Thread is at this link, and what I was reading here was, I thought, a revitalization of that thread:
Is speaker impedance ever less than DC resistance, even under transient conditions?
Thanks for that link, benb, I missed that thread earlier.
Howie
REW calculates EPDR in the same way. The dissipation is in the output devices, the EPDR figure is the purely resistive load that would give the same output device peak power dissipation at a particular frequency.I understand wthe articles describing EPDR, but have yet to see a test showing the speaker itself as the sink for the extra power EPDR predicts, as opposed to extra dissipation purely in the output devices. Also I'm not sure REW uses the same algorithm used in the ASR article.
@Howie: I do note that he has his own 'technical ground'
Would be funny if that was the culprit...
I'm quite confident it is the problem.
And worse yet, a lightning strike nearby will absolutely toast his equipment. And, it has the potential to generate a flashover in his house as well.
All single bushing transformer distribution systems ground the neutral roughly every other pole. A strike at a pole will be very dangerous in his house because of that rod.
John
John, that’s the easy part of EPDR, at all places where the phase deviates from zero, a drop in impedance is suggested.
My point is, how do you know whether the amp you have in mind to drive this can cope with this load ?
That’s my point.
Hans
My point is, how do you know whether the amp you have in mind to drive this can cope with this load ?
That’s my point.
Hans
Last edited:
Draw Ic(Vce) trajectory for the whole audio band, potentially add square input. Plot the SOA from datasheet into the simulated set. That’s the only way.
Well here is the result of my foray into EPDR:
Calculating EADR (similar to EPDR) | Audio Science Review (ASR) Forum
It's a fun story I guess. I thought about posting it here too.
Calculating EADR (similar to EPDR) | Audio Science Review (ASR) Forum
It's a fun story I guess. I thought about posting it here too.
Turns out that Nelson Pass and Matthew Polk published this idea in Audio magazine in 1980. Nelson archived a scan of the article HERE
Matthew Polk even applied for, and was granted, a US Patent on the idea. .pdf document attached below. Cowabunga!
_
The article was my solo effort, but Matthew's cables created the motivation.
😉
Mark, I don't imagine that I would ever fault you.
And to be fair, the application on the patent was earlier than my article.
🙂
And to be fair, the application on the patent was earlier than my article.
🙂
The Polk Cobra cable was notorious for frying amps. I measured it to have approx 8 Ohms impedance. I also heard stories of the insulation breaking down and shorting amps.
The Monster M2 series speaker cable had 100 Ohm 25W Caddock resistors terminating at the speaker end. It seemed to make a difference pretty consistently.
The Monster M2 series speaker cable had 100 Ohm 25W Caddock resistors terminating at the speaker end. It seemed to make a difference pretty consistently.
Polk Cobra is a flat, but braided cable. These and similar construction have high capacitance and very low inductance.
================================
The main effect of low inductance is to minimise environmental RFI pickup, which itself is an issue if it gets into the feedback loop of a power amp.
The associated high capacitance can cause problems with amps that do not contain a series inductor and Zobel network. Some designers historically (old NAIM gear) and modern (NVA) and possibly others, omit these essential parts. The problem is that at high frequency, the output impedance of any semiconductor power amp becomes inductive. The idea of the Zobel is to ensure that the amp is loaded with a pure 8 or 10 ohm resistor at high frequency, and the series inductor isolates the loudspeaker cable and loudspeaker.
The net effect is that amps that omit these critical parts (and usually demand use of their own cable) can indeed oscillate and self destruct when the inductive output impedance resonates with the cable capacitance. But that is down to inadequate amp design, not the cable.
FWIW I use woven cable (Kimber) with my multichannel amps for my active crossover speakers (LS521) without any problem whatever. But they have Zobel and inductor fitted. Kimber 8TC has 346pF/m capacitive and 0.08uH/m inductive. So 3 metre cables have about 1nF capacitance. The old "unconditionally stable" criterion was to feed the amp into a simulated electrostatic loudspeaker load; I cannot recall precisely what that was, but it was of order 1uF in parallel with 8 ohms without it breaking loose.
================================
The main effect of low inductance is to minimise environmental RFI pickup, which itself is an issue if it gets into the feedback loop of a power amp.
The associated high capacitance can cause problems with amps that do not contain a series inductor and Zobel network. Some designers historically (old NAIM gear) and modern (NVA) and possibly others, omit these essential parts. The problem is that at high frequency, the output impedance of any semiconductor power amp becomes inductive. The idea of the Zobel is to ensure that the amp is loaded with a pure 8 or 10 ohm resistor at high frequency, and the series inductor isolates the loudspeaker cable and loudspeaker.
The net effect is that amps that omit these critical parts (and usually demand use of their own cable) can indeed oscillate and self destruct when the inductive output impedance resonates with the cable capacitance. But that is down to inadequate amp design, not the cable.
FWIW I use woven cable (Kimber) with my multichannel amps for my active crossover speakers (LS521) without any problem whatever. But they have Zobel and inductor fitted. Kimber 8TC has 346pF/m capacitive and 0.08uH/m inductive. So 3 metre cables have about 1nF capacitance. The old "unconditionally stable" criterion was to feed the amp into a simulated electrostatic loudspeaker load; I cannot recall precisely what that was, but it was of order 1uF in parallel with 8 ohms without it breaking loose.
Last edited:
- Home
- Member Areas
- The Lounge
- The Black Hole......