measuring class D efficiency

[el00]

Hello
I have designed a pcb board for class D evaluation. I am generating PWM modulated signal through a bench signal generator, and I have only the final stage of the class D: from the signal generator I go to an inverter (to generate complementary PWM for H and L), dead time generation, half bridge and low pass filter. PWM is modulated at ~300kHz with a rail of 100V.
Everything is working as expected, I get a very clean output on a resistive load, for now I am using 25ohm but I will go lower, later on.
But... problems arose when I tried to measure efficiency. I am using current probes (clamps), one on the power rail and the other on the output (this is redundant as I measure voltage). Probes are battery powered. However, the high frequency noise irradiated into the probes from the switching frequency is so high that I cannot even see a 1-2A signal on the scope coming from the probes. I have partially solved by putting a filter between the probe and the oscilloscope, but it is still not acceptable.
What is your experience in measuring class D efficiency? Is there any advice that I should be aware of?
Thank you

 

[FvM]

In a correctly designed class D power stage, input and output are filtered and don't expose so much high frequency noise. Digital oscilloscopes do a good job in processing the signals, e.g. by averaging. Power calculations shouldn't be a problem.

 

[BigBoss]

Why don't you measure with a standard DMM with rms function ??
Measure voltages and currents in rms value then multiply so you'll find the efficiency..

 

[KlausST]

Hi,

Why don't you measure with a standard DMM with rms function ??
Measure voltages and currents in rms value then multiply so you'll find the efficiency..

This won't work.
* If you measure on DC paths then an RMS meter gives the wrong value, it needs to be an averaging measurement tool.
* if you measure on AC paths, then there will be phase shift between voltage and current...thus multiplying voltage amd current RMS values give apparent power, not effective power. You need effective power for efficiency calculations.

Klaus

 

[schmitt trigger]

Back when I was very active with audio, I used non-inductive power resistors as the load.
If you employ a precision resistor or accurately measure one, then the output power is simply V^2/R.
Of course, your DMM must be accurate over the full audio range. Otherwise use a scope.
On the scope use averaging to remove the high frequency noise.

 

[BigBoss]

Hi,
This won't work.
* If you measure on DC paths then an RMS meter gives the wrong value, it needs to be an averaging measurement tool.
* if you measure on AC paths, then there will be phase shift between voltage and current...thus multiplying voltage amd current RMS values give apparent power, not effective power. You need effective power for efficiency calculations.
Klaus

I have meant DC current on DC path so the measurement result will be average value as usual..Even Simple analog Multimeter may also measure average voltage/current.

 

[el00]

In a correctly designed class D power stage, input and output are filtered and don't expose so much high frequency noise. Digital oscilloscopes do a good job in processing the signals, e.g. by averaging. Power calculations shouldn't be a problem.

What do you mean with "input and output are filtered"? Could you elaborate or provide a reference?
In my design the output of the half bridge is correctly filtered by a two stage LC, in fact the output signal measured across a resistor is very clean.

The problem is the current, measured directly on the output of the power supply, which is very noisy. However, being a clamp, I suspect this is irradiated and not conducted noise.
I could try to add a LC filter on the rails, but I am not sure it is a good idea. How do you filter the rails, then?
Would it help in case this is irradiated noise?

 

[FvM]

As for the input DC bus, standard bus capacitors should provide sufficient filtering to measure input power with multimeters, as suggested in previous comments. Typically, the voltage and current measurement of the used lab supply would do the job. If you have already 2 stage LC output filters, which isn't mentioned in your initial post, I don't see problems to measure the output power with an oscilloscope.

Voltage measurement over a precision and low inductance power resistor would avoid usage of your apparently problematic current probes.

 

[el00]

As for the input DC bus, standard bus capacitors should provide sufficient filtering to measure input power with multimeters, as suggested in previous comments. Typically, the voltage and current measurement of the used lab supply would do the job. If you have already 2 stage LC output filters, which isn't mentioned in your initial post, I don't see problems to measure the output power with an oscilloscope.

Voltage measurement over a precision and low inductance power resistor would avoid usage of your apparently problematic current probes.

Well.. I cannot use the current measurement from the power supply since at the moment I am sending short bursts, I do not have a CW. Therefore, the current provided by the power supply is not constant.
I will try to add a sense resistor as you suggested, maybe on the low side to avoid using insulated probes.

 

[KlausST]

Hi,

I'm a bit confused about "I cannot use the current measurement from the power supply".
What power do you want to measure? Average over ??? time. Or instantaneous...or what?

Klaus

 

[el00]

Hi,

I'm a bit confused about "I cannot use the current measurement from the power supply".
What power do you want to measure? Average over ??? time. Or instantaneous...or what?

Klaus

As it is a pulsed system, and the output is closed on a resistor, by recording the voltage on the scope I can obtain the current, so the output power is easy to obtain, until I am driving a resistive load.
For the input, since the pulse is very short, the power supply is not fast enough in showing the current: I see a small change on the current measurement, but I need to measure the instantaneous current and integrate it over time. Input voltage is constant so I can obtain input power. But I need instantaneous measurement.
This is what I meant. Of course if there was a way to have average current (or power) that would be acceptable, but I have no idea on how to get that.

 

[KlausST]

I need to measure the instantaneous current and integrate it over time

This usually is done by a low pass filter.

Klaus