EEVblog Electronics Community Forum
General => General Technical Chat => Topic started by: MrOmnos on February 05, 2017, 04:08:49 pm
-
Hi! Most equipment these days use switching power supply because they are efficient. Today in embedded systems we were taught about powering an embedded system. Books say that you should not use switching power supplies with high frequency systems, which makes sense because switching PSU have 100-200khz switching frequency which can interfere with the other on board signals. But my question is how do you define high frequency systems? For example uC has a clock frequency of 16 Mhz, is that a high frequency system? A pc operates at 2.4GHz is that high frequency system? Or is this limited to radios and microwave systems? Why are uC and PCs not affected by SMPS?
-
They mean high frequency *analog* systems. But ironically, the high frequency stuff is less likely to be affected since switching PSUs don't have much harmonic content in the GHz range. What is more likely to be affected are lower frequency IF signals.
-
Technically HF is 3-30 MHz, so well within reach of the SMPS.
-
"High" is simply when you have to account for high frequency effects. Likewise, "mid" is where you don't, and "low" is where you have to account for LF effects instead.
In a very low power amplifier (or an ancient tube amplifier), "HF" is as low as ~kHz.
In a sense, all conventional op-amps operate in the high frequency regime. Recall that the textbook model of an op-amp is an infinite gain block, but that's physically impossible (this is one of those naughty lies they repeat). Rather, an integrator is a more realistic model. The DC gain is still infinite, but the AC gain is finite, and indeed, decreasing with frequency. Consequently, the phase shift is a constant 90 degrees. The decreasing gain, and nonzero phase shift, is a hallmark of the "high frequency" domain.
A "high frequency system", in regards to suitability of SMPS power, is anything that has poor rejection (PSRR) at the frequencies where the SMPS operates.
Example: most LDOs have poor PSRR above 10kHz. (Which, by the way, is a manifestation of the op-amp gain behavior I mentioned above. Most LDOs' internal amps roll off in the 100s of kHz range, hence the poor PSRR at nearby and greater frequencies.) It would be tempting to use an SMPS to get reasonable efficiency, then follow it with an LDO postreg for low ripple and precision stability, but this doesn't work out so well in practice, for this reason.
Indeed, it's often better to raise the SMPS switching frequency, so that its ripple can be more easily filtered.
Countless examples of precision and radio gear exist, that use switching converters, in all places. There is no single solution, nor is it reasonable to make such an ignorant and overzealous rule like "never combine SMPS with sensitive circuits". There is only good engineering!
Tim