Cheap DC-DC converters have been a boon on the hobbyist bench for a while now, but they can wreak havoc with sensitive circuits if you’re not careful. The problem: noise generated by the switch-mode supply buried within them. Is there anything you can do about the noise?
As it turns out, yes there is, and [Shahriar] at The Signal Path walks us through a basic circuit to reduce noise from DC-DC converters. The module under the knife is a popular buck-boost converter with a wide input range, 0-32 VDC output at up to 5 amps, and a fancy controller with an LCD display. But putting the stock $32 supply on a scope reveals tons of harmonics across a 1 MHz band and overall ripple of about 66 mV. But a simple voltage follower built from a power op-amp and a Zener diode does a great job of reducing the spikes and halving the ripple. The circuit is just a prototype and is meant more as a proof of principle and launching point for further development, and as such it’s far from perfect. The main downside is the four-volt offset from the input voltage; there’s also a broad smear of noise at the high end of the spectrum that persists even with the circuit in place. Centered around 900 MHz as it is, we suspect a cell signal of some sort is getting in. 900 kHz.
If you haven’t checked out the videos at The Signal Path, you really should. [Shahriar] really has a knack for explaining advanced topics in RF engineering, and has a bench to die for. We’ve covered quite a few of his projects before, from salvaging a $2700 spectrum analyzer to multiplexing fiber optic transmissions.
Orrrrrrrrrrrrrrrrrrrrrrrr you could just use a Pi filter for $3 and go home in 10 minutes.
…and just to be clear… that has nothing to do with a Raspberry Pi.
Can never be too sure these days ;)
Oh, it must involve a Banana Pi, then? ;-P
pi filters cost at least $3.14
come on, now.
Buy ten for $31.41, or a hundred for #314.15 (they get more expensive per unit the more you get).
Cool, can you write up the performance of one?
And the theory?
Gee, Wikipedia sure has a decent write-up.
That’ll make it easy then!
Aren’t pi filters only particularly suited for low current applications. If you are using the power supply at 5 amp, it might not be suitable.
Only the coil needs to stand the current, but 5A is still ok with 3.14$, or do it yourself, it’s just wire
Or if you are that picky about the quality (or lack thereof), you could always spend a bit more to get something a bit better.
The price gap between these and something half decent is pretty big though. A $10 filter would still be quite cost effective.
A lot of interesting concepts in this.
Filtering a noisy switching supply by using the power supply noise rejection of the power op-amp.
A spectacular scope system to demonstrate the measurements in the discussion.
A brief demo of the use of a low cost buck-boost power supply and it’s control panel.
By the way,
What is the remaining noise caused by? Anybody figure it out?
The oscilloscope screen shows the noise hump at 900 kHz, not 900 MHz, so I don’t think it’s mobile phone interference. I think it might be that the hump is above the frequency based on the slew rate of the op amp he’s using, so his circuit can’t respond fast enough to filter it out.
I don’t know the source of the 900 kHz noise, but that would explain why his circuit didn’t get rid of it.
looks a lot like am radio to me….
The oscilloscope screen shows the noise hump at 900 kHz, not 900 MHz, so I don’t think it’s mobile phone interference. I think it might be that the noise is above the frequency cutoff based on the slew rate of the op amp he’s using.
An op-amp and a zener diode???
Ain’t that just a voltage regulator but in discrete form???
I searched for a 5A voltage regulator and found LM338
chuck an opamp feedback in between the feedback pin and have that track the requested voltage… will always be requested voltage minus 1.2v.
Though, surely there are low-voltage-drop versions of said device.
Ah, it awaits approval… anyway without links here is the content of said comment:
An op-amp and a zener diode???
Ain’t that just a voltage regulator but in discrete form???
I searched for a 5A voltage regulator and found LM338
chuck an opamp feedback in between the feedback pin and have that track the requested voltage… will always be requested voltage minus 1.2v.
Though, surely there are low-voltage-drop versions of said device.
Not a very good solution. The op-amp used needs at least 20V supply, so it becomes quite useless for a power supply with less than 15V output. Plus, I don’t see any decoupling capacitors for the op-amp. Also, breadboard.
Def AM radio. Look up his 20 and radio finder it. Could be Purdue 920 AM, though I don’t see the HD garbage.
And now you end up with a power supply that is always about 4 volts lower than what the display shows…
The problem here is that there is no panacea unless the additional filtering is designed into the supply itself. Any filter must be placed before the voltage is sensed by the error amplifier loop. Unless this is done, the output won’t be at the correct voltage and/or have poor transient response due to the dynamics of the filter.
Or you could just use a small capacitance multiplier running with a complementary pair and send the output of that through a regulator, thus removing both high and low frequency noise. Maybe throw in a pi or T filter for good measure. Depends on your requirements.