3
\$\begingroup\$

I'd like to build an audio amplifier using the TPA3004D2 IC (datasheet).

My supply input is 24 VDC and I'm currently using a switching regulator (datasheet) with a switching frequency of 500 kHz to get 17 V and was wondering if an additional LDO for smoothing the supply would provide any noticeable benefit or if I can just connect the 17 V to the amplifier IC.

My logic is that any switching noise is too high in frequency to be audible so an LDO would likely not have much of an effect, is this correct?

Also, Figure 18 in the TPA3004D2 shows -63 dB supply ripple rejection from 10 Hz to 1 kHz so is this already working like an LDO?

\$\endgroup\$
2
  • 4
    \$\begingroup\$ The Class D amp itself is a switching amplifier. So the load of the switching regulator will be switching currents, or in other words, 250 kHz current pulses. The transient responses in the datasheet are given for 1 kHz transients but I can't see what will happen at 250 kHz transients. \$\endgroup\$ Commented Sep 22, 2022 at 11:28
  • \$\begingroup\$ you wouldn't necessarily need a LDO - a standard linear regulator might also be suitable. \$\endgroup\$
    – Kartman
    Commented Sep 22, 2022 at 12:13

2 Answers 2

2
\$\begingroup\$

Like any regulated power supply, it is mainly there to provide low power supply impedance from DC up to perhaps ~1..10 kHz. This low impedance is provided by the error feedback of the PSU and has nothing to do with whether its output stage is linear or switched. The low frequency error feedback mechanism are similar in both switched and linear PSUs, so one cannot say that one performs better than the other.

Also consider, that you Class-D amp will also have error feedback, which automatically brings more low frequency power supply rejection (-63 dB in you case). Therefore, typically one doesn't need that much PSRR in the power supply itself, which means: Usually, any regulated power supply will be good enough.

Note, that you also need good power supply impedance at your switching frequency of perhaps a few 100 kHz and linear PSUs may have higher error feedback bandwidth. However, this is never tackled by the PSU feedback, because at high frequencies beyond a few 10s of kHz, bulk electrolytic caps are a much easier way to achieve very low power supply impedance.

\$\endgroup\$
2
  • \$\begingroup\$ So it should be enough to add enough capacitance (or the right capacitors) so that the output impedance of my regulator is small at the switching frequency? What about the switching frequency (250 kHz) of the class-D amp as mentioned by Rohat Kılıç? \$\endgroup\$ Commented Sep 23, 2022 at 11:34
  • \$\begingroup\$ @celeryperson Yes, just adding capacitance is a standard tactic for low power supply impedance beyond ~10kHz. I am also talking about the Class D switching frequency. The Al electrolytics also have a very good ratio of capacitance to ESR to work as snubbers and prevent any peaks in supply impedance and related supply ringing. However, they become increasingly useless about 1 MHz, so back up with ceramics to provide charge fast during the switch transients. \$\endgroup\$
    – tobalt
    Commented Sep 23, 2022 at 14:12
0
\$\begingroup\$

-63 dB supply ripple rejection isn't fantastic. If yours is specified from 10 Hz to 10 kHz, that's good. Shady ones may have and there may be additional details here, for instance if they specified it at 50/60 Hz but you have noise in the kHz audio range and the supply rejection is even lower there.

If you are ok with -63 dB, then I see no need for any linear regulator after it. If you have a large capacitor bank between your power supply and the class D module, that will add to your rejection as well. Simulate or calculate it with parasitics such as ESL and ESR to find out by how much.

Given the high dynamic range of music, your linear regulator needs to handle a lot of current, which complicates things.

I would first try passive filtering with both bulk (electrolytic) and fast/low ESR+ESL capacitors such as plastics in parallel.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.