NC252MP (class D) vs. A250W4R (classAB) burst measurements into 4ohm//2.2uF load

[Deleted member 48726]

I think the Stereophile simulated load is here.

I assume you would need a circuit simulator to work out the Cap vs frequency.

I'm most certain you can calculate it. The data is in the measurements from speaker review.

 

[amirm]

This is the real effect of the SINAD cult that is supported here. Completely inaudible phenomena are debated and main failures are overlooked, due to lack of knowledge and also lack of explaining of less popular issues. It is understandable, it is a popular forum and is closer to social network discussions like facebook then to engineering and scientific view.

You are arguing against yourself. So what we are doing wrong is bad and we need to jump in and use a load that is out of spec for amps and you can only point to one speaker that comes close to it? And even then, you are not remotely simulating actual response of that speaker.

So I go and do your test and say the Purifi amp is terrible. People listen to it and say it sounds great. How do I/you explain and justify this test?

You are also ignoring that a big component of SINAD at higher end of amplification is noise. Your test does nothing to characterize noise which is far more audible than distortion.

 

[Ruffy]

Has anyone ever tested speakers at the individual driver level? Eg for a two way taking the jumper lead out and testing the tweeter and woofer impedances separately. I wonder if some of these gnarly loads are in part because of the tweeter/woofer interaction, and these speakers present easier loads when separate.

I know when stereo power amps offer a bridged mono setting, it actually makes the amp worse at handling low impedance loads. I wonder if I may be better off biamping in these instances. This is an aside from headroom, I don’t want to get into that argument here.

Cheers,
Andrew

 

[LTig]

Yes, they are a single driver both. One in a cabinet and one in free air.

Then please explain where the impedance / phase trace from speaker reviews is manifested. This is clearly what the amplifier sees, no? And if it is, you can calculate the "seemingly" inductive and capacitive components at different frequencies.

I already did in my 2nd paragraph: crossover feeding multiple drivers.

 

[antcollinet]

Real life matter.

Today I took my LCR meter and measured my Polk R150 at the banana plug level (amp disconnected).
The R 150 is a simple 2 way speaker with a 1st order crossover, a 0.75" tweeter and a 5.25" midrange/woofer.
With the LCR meter I can select some measurement frequencies.
Here are the results regarding the speaker capacitance seen by the amplifier:
145.5 uF at 100 Hz.
25.0 uF at 120 Hz.
9.6 uF at 1 Khz.
1.7 uF at 10 Khz.

Well, the capacitance seen by the amplifier is significant, thus the usual measurements using a purely resistive load are just standardized procedure, nothing close to reality.
The capacitance should be higher with large tower speakers.

I remember in the old time of Hifi, some smart laboratory designed a complex load dedicated to simulate the speaker characteristics.
This complex load was used to benchmark the amplifiers.
Amir, do you think that is possible to measure the amplifiers using a complex load?

It would be a great improvement for ASR benchmarks.

Doesn't the fact that the capacitance value you measure is changing significantly (By 85x) with frequency tell you that your LCR meter is not an appropriate tool for finding the capacitance of a speaker?

(Think about it - if you were measuring a simple LCR circuit - would you expect the measured values of fixed value components to be frequency dependant? If I have a 1uF capacitor - it is 1uF regardless of applied frequency)

 

[IAtaman]

Doesn't the fact that the capacitance value you measure is changing significantly (By 85x) with frequency tell you that your LCR meter is not an appropriate tool for finding the capacitance of a speaker?

(Think about it - if you were measuring a simple LCR circuit - would you expect the measured values of fixed value components to be frequency dependant? If I have a 1uF capacitor - it is 1uF regardless of applied frequency)

If you have a mix of active and reactive components in the form of C and L, would you not expect the measured impedance to change as frequency changes as Xc is inversely rateled to frequency and Xl is directly related?

 

[jimk1963]

Doesn't the fact that the capacitance value you measure is changing significantly (By 85x) with frequency tell you that your LCR meter is not an appropriate tool for finding the capacitance of a speaker?

(Think about it - if you were measuring a simple LCR circuit - would you expect the measured values of fixed value components to be frequency dependant? If I have a 1uF capacitor - it is 1uF regardless of applied frequency)

What is needed is an equivalent electrical circuit, which mimics the impedance vs. frequency of the speaker load. Someone posted a Stereophile equivalent circuit of a specific speaker - that approach is exactly what’s needed to properly model a speaker’s load impedance. The capacitance, inductance and resistance values are not frequency dependent - rather, the reactances of the L’s and C’s are, i.e., jwL and -j/wC. The reason an LCR meter shows different “capacitance” values at different frequencies is because it’s just measuring the reactance of the combined circuit (a single value) and calculating the “equivalent capacitance” (or inductance) at that specific frequency. Without diving into filter theory, if we have an impedance plot (resistance and reactance or angle), we can generate an equivalent circuit using L’s, R’s and C’s. This of course assumes the speaker components all behave linearly, meaning its equivalent circuit doesn’t change with applied power or frequency. I imagine that assumption breaks down eventually, at high enough drive levels.

 

[antcollinet]

If you have a mix of active and reactive components in the form of C and L, would you not expect the measured impedance to change as frequency changes as Xc is inversely rateled to frequency and Xl is directly related?

Sure - impdeance changes. Capacitance and inductance do not.

In PMA's load he has a 2.2uF capacitor. That does not change reagardless of the frequency.

PS - capacitors, inductors and resistors are all passive. Active components are typically semiconductors.

 

[IAtaman]

What is needed is an equivalent electrical circuit, which mimics the impedance vs. frequency of the speaker load. Someone posted a Stereophile equivalent circuit of a specific speaker - that approach is exactly what’s needed to properly model a speaker’s load impedance. The capacitance, inductance and resistance values are not frequency dependent - rather, the reactances of the L’s and C’s are, i.e., jwL and -j/wC. The reason an LCR meter shows different “capacitance” values at different frequencies is because it’s just measuring the reactance of the combined circuit (a single value) and calculating the “equivalent capacitance” (or inductance) at that specific frequency. Without diving into filter theory, if we have an impedance plot (resistance and reactance or angle), we can generate an equivalent circuit using L’s, R’s and C’s. This of course assumes the speaker components all behave linearly, meaning its equivalent circuit doesn’t change with applied power or frequency. I imagine that assumption breaks down eventually, at high enough drive levels.

I have been reading about it recently, modelling of speakers with LCR components that is, and it gets even more complicated when the mechanical effects starts to have an impact on the electrical characteristics. For electrostats, which is the edge case being discussed here, there is even an article that says for power and distortion measurements, electrostats can not be accurately modeled with linear components. I have shared the article in the other thread, but guess the discussion has moved on the social sciences side of things with cults and such, rather than the engineering one.

 

[IAtaman]

Sure - impdeance changes. Capacitance and inductance do not.

In PMA's load he has a 2.2uF capacitor. That does not change reagardless of the frequency.

PS - capacitors, inductors and resistors are all passive. Active components are typically semiconductors.

Yes agree, I believe that is the point. I have not seen any electrical model of a speaker that does not have any inductance in it which changes the impedance differently than a single cap would, so is 4ohm parallel with 2.2uF a realistic load to test an amp with? There are quite a lot of smart people here, I am sure they can evaluate an amp design and come up with a load that might make the amp behave out of spec. But would that matter if that load is not something that an amp is never expected to work with?

 

[antcollinet]

What is needed is an equivalent electrical circuit, which mimics the impedance vs. frequency of the speaker load. Someone posted a Stereophile equivalent circuit of a specific speaker - that approach is exactly what’s needed to properly model a speaker’s load impedance. The capacitance, inductance and resistance values are not frequency dependent - rather, the reactances of the L’s and C’s are, i.e., jwL and -j/wC. The reason an LCR meter shows different “capacitance” values at different frequencies is because it’s just measuring the reactance of the combined circuit (a single value) and calculating the “equivalent capacitance” (or inductance) at that specific frequency. Without diving into filter theory, if we have an impedance plot (resistance and reactance or angle), we can generate an equivalent circuit using L’s, R’s and C’s. This of course assumes the speaker components all behave linearly, meaning its equivalent circuit doesn’t change with applied power or frequency. I imagine that assumption breaks down eventually, at high enough drive levels.

Isn't this also all messed up because the speaker mechanical movement (Causing back emf and hence disturbing the measurement) is getting into the mix.

 

[antcollinet]

Yes agree, I believe that is the point. I have not seen any electrical model of a speaker that does not have any inductance in it which changes the impedance differently than a single cap would, so is 4ohm parallel with 2.2uF a realistic load to test an amp with?

All convntional drivers have a voice coil and hence inductors. I know little about electorstatic speakers, but they are primiarily capacitive rather than inductive.

 

[IAtaman]

All convntional drivers have a voice coil and hence inductors. I know little about electorstatic speakers, but they are primiarily capacitive rather than inductive.

Yes indeed. Despite having a large transformer in the middle, I understand they are generally modeled as parallel RC circuits. Here is article that claims this model is not accurate however for power and distortion measurements.

 

[PeterOo]

Yes indeed. Despite having a large transformer in the middle, I understand they are generally modeled as parallel RC circuits. Here is article that claims this model is not accurate however for power and distortion measurements.

And from the linked paper I understand they only model the electrostatic speaker itself. In that case the model does not include the step-up transformer that the amplifier is connected to, or additional Rs and Cs that the manufacturer may have added. Or did I miss that?

It is still not clear to me that the 2.2uF is representative for electrostats even.

 

[Deleted member 48726]

Isn't this also all messed up because the speaker mechanical movement (Causing back emf and hence disturbing the measurement) is getting into the mix.

It is not messed up. What matters is what the amplifier sees. Which is the impedance / phase trace. This contains both the electrical and mechanical side of things. So at any given frequency the equivalent components (electrical + mechanical behaviour) can be broken down.
That's why it seems off. But look at the links reg. equivalent circuits. Look at the quite large values the components have to simulate the mechanical driver data.

 

[pma]

What matters is what the amplifier sees. Which is the impedance / phase trace. This contains both the electrical and mechanical side of things.

I use this circuit as an amplifier dummy load to simulate "usual" 2-way passive speaker with dynamic drivers

Impedance magnitude and phase in the image below, but please do not forget the EPDR - most important as a measure of amplifier load

REW is now clever enough to tell you R+C, R//C or R+L, R//L at any point of the measured impedance. See the data in oval bottom left, valid for 80.1Hz. Please take into account that these values are valid only at the selected frequency point.

 

[jimk1963]

It is not messed up. What matters is what the amplifier sees. Which is the impedance / phase trace. This contains both the electrical and mechanical side of things. So at any given frequency the equivalent components (electrical + mechanical behaviour) can be broken down.
That's why it seems off. But look at the links reg. equivalent circuits. Look at the quite large values the components have to simulate the mechanical driver data.

Right, should have clarified this. The equivalent circuit of course accounts for “everything” including mechanical elements because they are attached to the crossover. Again however, equivalent passive circuits assume an inherently linear system, because the circuit is made up of ideal passive components. Is this valid? Yes, for most speakers, because when a speaker’s audio quality is checked at levels below its rating, the distortion is usually well below -10 dB, which means nonlinearities contribute very little to the behavior and can be ignored for purposes of modeling. if one wants to model the effects of distortion at high power handling, then the model has to include elements which distort, I.e., nonlinear elements. As an example, a variable capacitor (varactor, varicap, etc) whose capacitance varies with applied voltage is a nonlinear element. These are useful for modeling power-based nonlinearities in an otherwise linear system. Same idea for inductors - once their cores saturate, they can be modeled using similar nonlinear techniques.

 

[Deleted member 48726]

I use this circuit as an amplifier dummy load to simulate "usual" 2-way passive speaker with dynamic drivers

View attachment 275127

Impedance magnitude and phase in the image below, but please do not forget the EPDR - most important as a measure of amplifier load
View attachment 275131

REW is now clever enough to tell you R+C, R//C or R+L, R//L at any point of the measured impedance. See the data in oval bottom left, valid for 80.1Hz. Please take into account that these values are valid only at the selected frequency point.

We seem to agree then. -That we can in fact do some vector break down from a measured speakers' impedance / phase traces to assess it's capacitance and inductance loads at single frequency points. Also this apparently looks valid from the live measurements done by @valerianf here.

BUT; how can you then claim to simulate capacitive load with a single capacitor? You know, to claim the operation limits of your tested class D amplifiers? It seems to me, that it goes against what we seem to have established in terms of what is real load / simulated real load and what is not.

 

[fpitas]

Has anyone ever tested speakers at the individual driver level? Eg for a two way taking the jumper lead out and testing the tweeter and woofer impedances separately. I wonder if some of these gnarly loads are in part because of the tweeter/woofer interaction, and these speakers present easier loads when separate.

I know when stereo power amps offer a bridged mono setting, it actually makes the amp worse at handling low impedance loads. I wonder if I may be better off biamping in these instances. This is an aside from headroom, I don’t want to get into that argument here.

Cheers,
Andrew

For a passive speaker, the crossover has such a drastic effect the individual driver response alone is meaningless. Now for an active speaker where the amp directly drives each driver, it's important.

 

[Deleted member 48726]

Right, should have clarified this. The equivalent circuit of course accounts for “everything” including mechanical elements because they are attached to the crossover. Again however, equivalent passive circuits assume an inherently linear system, because the circuit is made up of ideal passive components. Is this valid? Yes, for most speakers, because when a speaker’s audio quality is checked at levels below its rating, the distortion is usually well below -10 dB, which means nonlinearities contribute very little to the behavior and can be ignored for purposes of modeling. if one wants to model the effects of distortion at high power handling, then the model has to include elements which distort, I.e., nonlinear elements. As an example, a variable capacitor (varactor, varicap, etc) whose capacitance varies with applied voltage is a nonlinear element. These are useful for modeling power-based nonlinearities in an otherwise linear system. Same idea for inductors - once their cores saturate, they can be modeled using similar nonlinear techniques.

Impedance / phase diagrams as measured in reviews is as real as it gets as far as I'm concerned. Equivalent circuits are almost as good if we are aware of it's characteristics as you've described.

Still I'm not convinced that @pma 's test setup that started this whole thing is actually reflecting real load. At least from what I've learned the last few days it doesn't look quite right.