USB noise / voltage fluctuations and how to condition for RF use

[cmorgan]

I'm building a usb powered device that, at present, is plugged into a raspberry Pi 3B that is powered by a digital bench supply (Rigol DP832A) at 5V with no current limit.

Measuring the voltage on my prototype device I see about 140vpp on vbus (noise), and periods of voltage dip. I believe these dips are when the PI is performing some power heavy operations, either cpu or wifi.

Attached image shows (traces from top to bottom):
- yellow = usb vbus
- teal = after FB
- red = near AD8313 pin, so after 10 ohm resistor and on the net with the filtering caps and AD9313 VPOS4 pin
- blue = 3.3v (also filtered with pi filter, generated from LDO on filtered 5V rail)

The prototype has a PI filter with FB on 5V, the signal looks basically the same on both sides of the filter. I've even tried adding ~10uF of capacitance on the output side and 4.7uF on the input side with little change. (I'm guessing this may not be enough capacitance).

If I power my prototype directly with the bench supply I don't see these dips.

I've got some RF circuits on the device, in particular an AD8313 rf detector. I'm not sure that these power fluctuations have any impact on the performance of the circuit. Nevertheless I was hoping to see if anyone had guidance on various approach of ensuring usb vbus was rock solid.

Lest readers think I'm just spam posting without thought, a few ideas came to mind.

- Use an LDO to regulate vbus down to 4.7V. FTDI usb-serial can operate with VCC at that value, AD8313 the same. Regulating down to 4.7V would provide a consistent voltage even when vbus drops.

- Look to add 30uF - 50uF of bulk capacitance to the filtered side of the ferrite bead. In-rush current would concern me.

- Select a better FB. The one at present is only 30 ohms @ 100MHz, I found a few others that are 120 ohms @ 100MHz. This could help reduce the higher frequency noise in the system, improving RF performance.

[mtwieg]

There's likely a combination of low and high frequency components to that measurement. Which are you concerned about? Ferrite beads won't do anything for low frequency dips. Even lots of capacitance probably won't help much for dips lasting milliseconds or longer. Using separate regulators is probably the only robust method for that.

If the RF component is what concerns you, using the scope's FFT is likely a much more useful way of observing it.

[cmorgan]

That makes sense. I do think there is a lot of low frequency noise (in the form of voltage ripple).

Would you know of any guides that you'd recommend related to power supplies for low noise RF circuits?

[mtwieg]

"RF circuits" is far to broad of a description to give any useful input. There are plenty of cases where RF systems operate fine from noisy/poorly regulated supply rails (like any USB bluetooth/wifi adaptor). In other cases you need to power things from a battery because even the best-performing voltage regulators are too noisy (like for powering a VCO which requires extremely low phase noise).