I picked up a couple of these 500W IRS2092 amplifier modules. They seem to do mostly what they claim, but there is a non-trivial amount of carrier on the output. With the default oscillation frequency of about 400 kHz, and a grounded input, there was about 1 V RMS on the output.
However, based on some discussion on this thread I adjusted the oscillation frequency down to about 200 kHz to improve the efficiency and reduce the load on the 2092.
Of course that allowed even more carrier through and I now measure about 3 V RMS with a 4 ohm load which works out to about 2 W. Given that this module delivers a lot of output that's not a big deal, but certainly higher than I had expected.
Should I look at more output capacitance on top of the existing 0.47 uF, or is there really no tangible benefit? Alternatively, should I jack up the frequency a bit instead? My speakers certainly won't notice 2 W, and I can't hear 200 kHz, so I am somewhat inclined not to worry about it.
Thoughts?
However, based on some discussion on this thread I adjusted the oscillation frequency down to about 200 kHz to improve the efficiency and reduce the load on the 2092.
Of course that allowed even more carrier through and I now measure about 3 V RMS with a 4 ohm load which works out to about 2 W. Given that this module delivers a lot of output that's not a big deal, but certainly higher than I had expected.
Should I look at more output capacitance on top of the existing 0.47 uF, or is there really no tangible benefit? Alternatively, should I jack up the frequency a bit instead? My speakers certainly won't notice 2 W, and I can't hear 200 kHz, so I am somewhat inclined not to worry about it.
Thoughts?
LC filters have a characteristic impedance, this needs to match the loading - either the speaker itself or the speaker in combination with an RC damping network. Simply adding more capacitance will reduce the characteristic impedance and hence result in an under-damped filter, calling for changes to the output RC network to maintain a flat FR.
I'd be inclined not to worry about it unless your speaker cables are exceedingly long.
I'd be inclined not to worry about it unless your speaker cables are exceedingly long.
One other thing to keep in mind is that speaker impedance is anywhere near flat. Due to voice coil inductance, the impedance in the hundreds of kHz will be in the hundreds of ohms, so the actual dissipated power, despite that 3Vrms carrier amplitude, should be near-insignificant (miliwatts range).
All good points. Thanks!
The class A purist side of me really cringes when I see all that fuzzy noise on a nice clean sine wave signal, but the class D engineer side says 500 W weighing 55 grams for $20: score!
The class A purist side of me really cringes when I see all that fuzzy noise on a nice clean sine wave signal, but the class D engineer side says 500 W weighing 55 grams for $20: score!
Remember that there is not such a thing as fully uniform electron flow. Inside semiconductors (and conductors) electrons flow irregularly, in chunks, randomly.
That truth of physics can be faced with two opposite approaches. One approach is to consider the absolute average of everything, the static model. The other approach is to characterize the average rate and magnitude of the chunks of electrons, the dynamic model.
A class AD amplifier has a monotonous transfer function between +/- few amps output, this is exactly like a class A amplifier. The best linearity happens around 0 volts.
Carrier residual is of little importance as long as it only contains the fundamental frequency and no substantial RF at dozens of Mhz. Carrier residual amplitude dissipates very low power due to the fact that load resistance at 400Khz can usually be 10 times nominal value in almost every speaker (except metal ribbon radiators, which need consideration). Audibility at 100s of Khz is null due to the fact that speaker response, and human body perception, can be 100s of dBs down or more w.r.t. 20khz at those frequencies (>20khz electrical or ultrasound would be perceived as heat rather than sound).
What matters in a class D amplifier is the content of audio frequencies at output. On the other hand, measuring the THD and noise floor of a class D amplifier will solve the doubt (is it measurably good?). That almost always requires a sharp LPF for removing content over 25Khz from the input of the measuring instrument, and suitable attenuator. Carrier residual can damage the input of the measuring instrument if a direct noise measurement is attempted, apart from disturbing the ADC in digital measurements.
That truth of physics can be faced with two opposite approaches. One approach is to consider the absolute average of everything, the static model. The other approach is to characterize the average rate and magnitude of the chunks of electrons, the dynamic model.
A class AD amplifier has a monotonous transfer function between +/- few amps output, this is exactly like a class A amplifier. The best linearity happens around 0 volts.
Carrier residual is of little importance as long as it only contains the fundamental frequency and no substantial RF at dozens of Mhz. Carrier residual amplitude dissipates very low power due to the fact that load resistance at 400Khz can usually be 10 times nominal value in almost every speaker (except metal ribbon radiators, which need consideration). Audibility at 100s of Khz is null due to the fact that speaker response, and human body perception, can be 100s of dBs down or more w.r.t. 20khz at those frequencies (>20khz electrical or ultrasound would be perceived as heat rather than sound).
What matters in a class D amplifier is the content of audio frequencies at output. On the other hand, measuring the THD and noise floor of a class D amplifier will solve the doubt (is it measurably good?). That almost always requires a sharp LPF for removing content over 25Khz from the input of the measuring instrument, and suitable attenuator. Carrier residual can damage the input of the measuring instrument if a direct noise measurement is attempted, apart from disturbing the ADC in digital measurements.
Last edited:
I reconfigured the test setup including dropping the supply rails to +/- 65 V, and raising the oscillation frequency to around 330 kHz. Re-testing the carrier noise showed it had dropped to about 1 V RMS, which is lower than expected. Not sure I can explain why the drop was that large but it was consistent on both bridged amplifiers.
Possibly the inductance of your chokes dropped due to core saturation at higher voltage and lower frequency - thus reducing its impedance.
Carrier residual (class AD) has a linear dependence on rail voltage, and a quadratic dependence on idle switching frequency.
- Status
- Not open for further replies.