Buck converter sensitivity to noise in the power line

[roleroz]

I know that it's normally a bad idea to power semiconductors with the same power line that you use to drive inductive loads because of the noise that those loads put on the power line, so I decided to separate the power domains for my project into 2. I'm using a 5V power rail for all things semiconductor (microcontrollers, sensors, leds and the like), and a 12V power rail for all things inductive (motors, steppers, servos)

This has to be battery powered, so I'm planning on using 4 18650 cells, and use 2 buck converters to create both power rails

My main question is, how good are buck converters on insulating the power lines from noise?, i.e., if I have stepper motors producing noise in the 12V rail, will that noise cross the 12V buck converter into the power line coming from the battery?, and if it does, will it cross the 5V buck converter into the 5V power rail? My other option is to put another set of 2 18650 batteries to power the 5V buck converter, but I'd prefer to not do that if at all possible

Open to other power designs as well

Both buck converters have to be reasonably beefy, as I need around 5A per rail (batteries should be ok with providing that amount of current)

Thanks

Joso

[m k]

My other option is to put another set of 2 18650 batteries to power the 5V buck converter, but I'd prefer to not do that if at all possible

Why?

[roleroz]

It would make the whole system too big and heavy

[JustMeHere]

Use a 4-layer board in your MC layout area.  Understand ground loops, and keep them small.

As long as your all digital, MCUs can be very tolerant to noise.

If this were a big problem, there would be no drones, electric cars, portable CD players, or  electric tooth brushes.

[Ian.M]

Apply some critical thinking - Apart from the control circuits, the input side of a buck converter during its on time looks like a big inductor.  Its fairly obvious that transients with a period at least an order of magnitude smaller than the buck's on time are going to have negligible impact, and the control loop should be able to handle anything with a risetime greater than several buck cycles, so at worst you only need input filtering good for a limited frequency range.   The fly in the ointment is the control circuits - consult its datasheet to see what PSRR vs frequency you can expect from it.  If its got an integrated switch and its PSRR isn't good enough, you are S.O.L. (unless it has separate Vin pins for the switch and the control circuits) as *good* high current filtering is invariably bulky, heavy and expensive. A separate buck controller with an external MOSFET may well be preferable, just because you can often run them from a *good* linear regulator, with a well filtered input, or simply filter the relatively low current Vin to the buck controller, to further improve its PSRR.

You'll also need to start putting some numbers on the problem - e.g. write test firmware to produce worst case noise on the 12V rail by rapidly and repeatedly reversing the steppers and other motors and scope the rail to measure the noise amplitude, then decide how much of that you can tolerate on your 5V rail to get a target PSRR to choose the 5V regulator.

However if you don't want to tell us which buck IC you will be using, don't want to consult datasheets, or want to use cheap far east buck modules with probably fake chips, the only possible answer is:

42

[redkitedesign]

My main question is, how good are buck converters on insulating the power lines from noise?,

Well, that kind of depends on your implementation...

The plethora of battery-powered, MCU controlled mechanical thingies like drones, RC cars, normal cars etc. that work properly shows that it can be done.

In general, know how the noise from your motors looks, and know how it can propagate (resistively, capacitively, inductively), and take appropriate countermeasures.

Most buck converters are quite good in suppressing any noise on the power line that's lower frequency than the bandwidth of the feedback loop. Check your design.

[james_s]

It shouldn't bother the buck converter any.

Driving inductive loads does require some care but it isn't particularly difficult. Just make sure you have plenty of decoupling on the power rails and put a freewheel diode or snubber across the inductor.

[roleroz]

Thanks for all the replies

I'm still designing the buck converter, so I haven't selected all my pieces, but I was planning on using https://datasheet.lcsc.com/lcsc/2005281834_Texas-Instruments-TPS54623RHLR_C560970.pdf for both of them

I guess I'll design the converter, and make a PCB just with it (I'll be embedding it into the main PCB for my final design) and test it out

Thanks again

[redkitedesign]

Use TI's Webbench. It's about as obvious as using the datasheet.