Power supply drift battle: Keysight B2962A vs. MPJA 9623PS

[maxwell3e10]

On the left: Keysight B2962A 6.5 Digit Low Noise Power Source designed by Agilent
On the right: Generic MPJA 9623PS Analog triple power supply designed by unknown company

10 V output is recorded with a gain of 1000 and a bandwidth of 1 Hz for 200 seconds relative to a stable 10 V source (using AlphaLab LNA10 amplifier and SRS DC205 source)

And the winner is: MPJA 9623PS with less than 100 uV pp noise, while B2962A just meets its specs of 150 uVpp noise.

I'll have a more in-depth review of B2962A, this is just a teaser.

[Zucca]

mmmm...

[exe]

Is Keysight a switching-mode power supply? (probably with linear post-reg).

[maxwell3e10]

Yes, its a switching mode supply, the RMS noise is exactly 3 mVrms 10-20 MHz, as the specs say. They make a big deal of additional low-pass filters that one can purchase (for $1k-2$k !) to filter out the switching noise. But of course it doesn't reduce the low-frequency flicker noise.

[TimNJ]

That's interesting, though at first glance, I thought you were just showing the output ripple.

But then noticed the time-base. (Although, it says T: 24.00s..is this just the zoom?) The signal on the scope is the error voltage relative to the Stanford DC205, right?

It's interesting how quickly the error voltage can change on the Keysight, while the linear supply seems much more reluctant to change quickly. Keysight almost has a periodic nature to it. I wonder what reference/error amplifier the Keysight uses and if it's in a sub-optimal location (thermally related?).

[maxwell3e10]

First plots have 10sec/div scale and DC coupling with 10 V subtraction. Later plot has 2 usec/div scale and AC coupling.

B2962 has an LM399 reference, so the flicker noise must be from something else. It could be interference between multiple unsynchronized switching power supplies.

[Zucca]

Are you using Coax cables from PSU to the scope?

[maxwell3e10]

Yep, even got a nice double-shielded cable. One can also switch the B2962 from grounded to floating and add 50 Ohm load, which changes the shape of the switching ringing a bit. 

[Zucca]

Interesting. Next step is a shielded box or try to be in a quiet EMI place....

[maxwell3e10]

Here is a comparison of voltage spikes on MPJA 9623PS (similar to HY3003D), Rigol DP832 and Keysight B2962. All are taken with 5V output connected by a coaxial cable and loaded at the scope with 50 Ohm terminator. For MPJA supply similar spikes appear with the supply shut off, so that's a measure of interference (it helps to shut down my laptop while taking this data). For Rigol supply the spikes appear at some point as it starts up and are similar with the output is on or off, so it must be some clock. On B2962 the spikes also appear and change shape as it goes through the start-up sequence.

MPJA 9623

DP832

B2962