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

[IAtaman]

Yes, exactly.

Okay great, thank you.

Is it also fair to say that if we knew what is the output impedance of an amp, we would be able to estimate how it would behave driving a speaker with a known impedance / phase graph? Because to me it sounds like if the answer to that question is yes, then we don't need to do a variety of complex load tests with expensive test equipment, we can simply do the math or even use matlab to plot the expected FR. Is that right?

One few more questions if I am not giving you gentleman headache already; can we devise an accurate enough output impedance model for the amp using the sweep data? If yes, how many sweeps would be needed? What is the best way to measure/calculate the output impedance of an amp?

 

[SIY]

Okay great, thank you.

Is it also fair to say that if we knew what is the output impedance of an amp, we would be able to estimate how it would behave driving a speaker with a known impedance / phase graph? Because to me it sounds like if the answer to that question is yes, then we don't need to do a variety of complex load tests with expensive test equipment, we can simply do the math or even use matlab to plot the expected FR. Is that right?

One few more questions if I am not giving you gentleman headache already; can we devise an accurate enough output impedance model for the amp using the sweep data? If yes, how many sweeps would be needed? What is the best way to measure/calculate the output impedance of an amp?

1. Yes, exactly.

2. Yes, exactly. One sweep is enough. There is no one best way, but I do a load change and see what the amplitude and frequency response changes are. 99% of the time, a power amp's source impedance can be accurately modeled as a resistance in series with an inductance. Worst case, you have to do some curve fitting, but that's a couple minutes with Excel.

 

[Deleted member 48726]

Unfortunately, I’m not as I did controlled tests and found out. I have noticed also in just everyday use. I have thought sound stage was great while listening in mono. I have thought I was listening to one speaker when in fact the one next to it was on (different models) I have thought ARC was doing an amazing job when it wasn’t on.

This experience has been echoed in my actual field, visual art. I know that the amount of sleep, caffeine, and just my mood affects my color vision. I color balanced an entire edition to cyan one to me. Checked all the machinery for flaws before acknowledging I printed it on little sleep and it was my vision that was off. Super illuminating was seeing my work in a museum for the first time. Physically it was the same as in the studio, but it looked so much better in the museum. Why? Because we are conditioned to think things in museums are good.

Relying on your senses solely is very bad. Ask any pilot. I try to always use controls to verify what my senses are telling me when it comes to making art.

I'm sure your tests show that. But I'm still absolutely sure you underestimate memory.
We are not made of wax and we can rely a great deal on memories and experiences. Otherwise every day is like starting at a new job..

 

[pma]

Unfortunately, I’m not as I did controlled tests and found out. I have noticed also in just everyday use. I have thought sound stage was great while listening in mono. I have thought I was listening to one speaker when in fact the one next to it was on (different models) I have thought ARC was doing an amazing job when it wasn’t on.

That's normal, especially if we are not very concentrated to listening or trained to concentrate. I made many tests where audience was unable to tell that music used was in mono, that one speaker was phase-reversed or tweeter was disconnected. On the other hand, there were some listeners who were able to tell it quite immediately.

 

[KSTR]

As @SIY said, exactly. Both amplifier output impedance and speaker impedance are in general complex (resistive and reactive components).

Don't know if this has been mentioned yet but output impedance can be highly unstable and nonlinear, which is IMHO a much bigger issue than any non-resistive character (typically, output impedances are almost perfectly resistive in the audio range, then becoming inductive above). Together with even more non-linear speaker loads this creates a very complex dynamic scenario for distortions.

 

[IAtaman]

Don't know if this has been mentioned yet but output impedance can be highly unstable and nonlinear, which is IMHO a much bigger issue than any non-resistive character (typically, output impedances are almost perfectly resistive in the audio range, then becoming inductive above). Together with even more non-linear speaker loads this creates a very complex dynamic scenario for distortions.

That sounds interesting. Any reference reading you can link to?

 

[pma]

Don't know if this has been mentioned yet but output impedance can be highly unstable and nonlinear, which is IMHO a much bigger issue than any non-resistive character (typically, output impedances are almost perfectly resistive in the audio range, then becoming inductive above). Together with even more non-linear speaker loads this creates a very complex dynamic scenario for distortions.

Yes. X-Y force test at various frequencies is a good test to disclose such behaviour. Crossover area. Unfortunately, tests at ASR are quite limited, though it is understandable taking into account number of devices tested. However, if we do not know, who would have cared?
R + jwL only? As Stuart (@SIY ) suggests?

99% of the time, a power amp's source impedance can be accurately modeled as a resistance in series with an inductance.

Ha ha ha! And the smaller the output signal, the worse. But I am not the Expert, so maybe I just do not know.

 

[SIY]

Ha ha ha! And the smaller the output signal, the worse. But I am not the Expert, so maybe I just do not know.

Or you do know and deliberately chose a grossly misadjusted amplifier. I was very disappointed when I saw Benchmark do the same trick.

I've measured a LOT of AB class amplifiers and that sort of crossover spike just doesn't occur when idle current is set correctly. Curious folks should refer to Self's excellent analysis.

 

[IPunchCholla]

I'm sure your tests show that. But I'm still absolutely sure you underestimate memory.
We are not made of wax and we can rely a great deal on memories and experiences. Otherwise every day is like starting at a new job..

Most research shows that memory is highly unreliable and unstable. Yes, it is stable enough for us to function pretty damn well, but details are easily lost and manipulated. What’s more is every time we access a memory we change it. But that isn’t the type of memory I’m referring to.

I’m talking about sensory memories and in particular echoic memory and it’s implications for auditory memory overall.

That's normal, especially if we are not very concentrated to listening or trained to concentrate. I made many tests where audience was unable to tell that music used was in mono, that one speaker was phase-reversed or tweeter was disconnected. On the other hand, there were some listeners who were able to tell it quite immediately.

Yes. And I am in either of those groups depending on circumstances. Most of us are. We have all been mislead by our senses. We will all be mislead by them again.

 

[Sokel]

Most research shows that memory is highly unreliable and unstable. Yes, it is stable enough for us to function pretty damn well, but details are easily lost and manipulated. What’s more is every time we access a memory we change it. But that isn’t the type of memory I’m referring to.

Genuine question:
Which kind of auditory memory is unreliable and lost?

The reason I'm asking is cause I had a call from my girlfriend of 30 years ago,4 countries away and not only I remember instantly but I got emotional too
So,which part is the one we shouldn't rely on?

 

[IPunchCholla]

Genuine question:
Which kind of auditory memory is unreliable and lost?

The reason I'm asking is cause I had a call from my girlfriend of 30 years ago,4 countries away and not only I remember instantly but I got emotional too
So,which part is the one we shouldn't rely on?

Well I’m not an expert at all, but that would fall under long term memory. We seem to have the ability for some sounds, say your mother’s voice, to be able to recognize the sound using memory even though the sound is frequency wise very different over time. Which is kinda the opposite of what we are usually talking about in these circles ( comparing similar sounds to find differences over short durations). I started look into this by reading about echoic memory and have been filling in here and there as time permits. I have yet to find a text that looks at auditory memory from sensory stage to long term, but I’m sure that is because I haven’t looked very hard.

 

[IPunchCholla]

Well I’m not an expert at all, but that would fall under long term memory. We seem to have the ability for some sounds, say your mother’s voice, to be able to recognize the sound using memory even though the sound is frequency wise very different over time. Which is kinda the opposite of what we are usually talking about in these circles ( comparing similar sounds to find differences over short durations). I started look into this by reading about echoic memory and have been filling in here and there as time permits. I have yet to find a text that looks at auditory memory from sensory stage to long term, but I’m sure that is because I haven’t looked very hard.

Edit to add: the part that seems the most problematic is on the boundary of short term/long term and how those get translated. Think witness testimony to traumatic events, or where short term memories are manipulated to cause false long term memories. For sensory memory, think auditory or visual illusions and/or hallucinations.

 

[Sokel]

Well I’m not an expert at all, but that would fall under long term memory. We seem to have the ability for some sounds, say your mother’s voice, to be able to recognize the sound using memory even though the sound is frequency wise very different over time. Which is kinda the opposite of what we are usually talking about in these circles ( comparing similar sounds to find differences over short durations). I started look into this by reading about echoic memory and have been filling in here and there as time permits. I have yet to find a text that looks at auditory memory from sensory stage to long term, but I’m sure that is because I haven’t looked very hard.

(I know audiophiles that listen obsessionaly the same 5 songs for 40 years,they must know Lyn Stanley's voice better than their mother's )

 

[KSTR]

That sounds interesting. Any reference reading you can link to?

The output impedance of an emitter follower (or source follower) is roughly the inverse of the transconductance, Rout ~= 1/gm. But gm is a function of emitter (source) current, in case of a bipolar transistor gm = Ie/Vt (Vt is the thermal voltage which is temperature-dependent). This means output impedance is modulated by the load current. Emitter ballast resistors (required to get an AB push-pull stage thermaly stable) do help, as does global feedback, together with optimum choice of resistor values and bias current.

Another thing is non-linear resistors used as emitter/source ballast resistors. Usually one would think the typical choice of wire-wound or metal band types is safe, but sometimes it's not (I've got bitten by this big time in a current-drive amplifier where any non-linearity of the sense resistor outside the feedback loop).

 

[pma]

Real world measurement of output impedance HF nonlinearity:

Current is injected into the DUT amp output via 10 ohm resistor. Directly at output binding posts the residual voltage is measured by scope..

Injected current is 10kHz triangle with rounded corners (DAC). Ipeak is 1.2A approx. Vout is 15mV, so we have dynamic Zout = 0.0125 ohm. Yellow trace shows HF nonlinearity of the DUT output impedance. This is a good result, measuring with 10kHz sine would still give quite clean sine response.
This is the way, for circuit designer, to measure output impedance and to get some info about the DUT circuit behaviour.

with 1kHz sine. Current 100mA/div, voltage 5mV/div.

 

[pma]

Measuring output impedance at 10kHz

Zout = 0.02ohm approx. Inductive phase phase shift, output voltage before the current.

X-Y plot shows phase shift (inductive) and some nonlinearity of output impedance at 10kHz. Injected current to the output forced through 10 ohm resistor.

 

[theREALdotnet]

Injected current is 10kHz triangle with rounded corners (DAC). Ipeak is 1.2A approx. Vout is 15mV, so we have dynamic Zout = 0.0125 ohm.

That’s how we used to test regulated power supplies (we injected via a capacitor, of course).

While you have it there, can you measure at a few more injected frequencies between 20Hz and 20kHz and plot Zout (magnitude and phase angle)?

 

[SIY]

That’s how we used to test regulated power supplies (we injected via a capacitor, of course).

While you have it there, can you measure at a few more injected frequencies between 20Hz and 20kHz and plot Zout (magnitude and phase angle)?

These days, we can do this in one shot using a source impedance sweep utility (this is a standard function on the APx analyzer I use). There's a lot of methods to get source Z, but I think the easiest to do and interpret (as well as having fewer potential error sources) is doing a step load change.

 

[pma]

Hello all, after some cooling down period. This issue of Ncore complex load dependent distortion is keeping my curiosity. I have made another sets of measurements, now with much lower value of parallel load capacitance, namely 141nF and 282nF. So the load is 4.7ohm//141nF and 4.7ohm//282nF. The results are much worse than I have expected.

Let's start with THD and THD vs. power plots, at 1 kHz.

We can see that:
- usual SINAD numbers defined at 5W/1kHz/4ohm would not tell any important difference,
- the distortion result with 141nF (3 x 47nF) parallel capacitor is worse than with the 282nF (6 x 47nF) capacitor,
- the distortion starts to rise above 5W with the added capacitive load and will strongly depend on capacitance value.
This all indicates to the fact, that the Ncore high-order feedback loop compensation has been designed taking into account pure resistive load only.

THD spectrum at 1kHz 76W

Again and consistently, the distortion with 141nF capacitor is much higher than with 282nF. Unfortunately, UcD180HG is much more consistent regarding load than the Ncore.

 

[pma]

With the capacitive load (in parallel with 4.7ohm) in steps of 0nF, 47nF, 94nF, 141nF, 282nF there was no issue and no significant rise of distortion up to 47nF. The distortion started to rise from 94nF, as one can see in the cumulative set of plots below.