While investigating a rather unpleasant screen flicker issue (to be reported separately) I decided to observe waveforms of DC voltage present at the 3V3 and 5V buses. The setup is as follows:
Meanwell S-series PSU rated 12 Volt 350 Watt
↓
USB DP “Red board” 1.3
↓
Extender 1.3 ⇉ oscilloscope probes on 3.3 and 5 Volt lines
↓
Hydra 1.2 → SD 1.4 (empty)
↓↓↓↓
T35 display
What I am observing is rather astonishing, with ripple exceeding 10% of nominal value.
Is this ripple and noise behavior normal? Is anything to be done? (Greenwiring and SMD rework is ok)
Attached are the oscilloscope screenshots.
Yellow trace is 3.3 V bus, cyan trace is 5 V.
The wide wave ripples are about 24 ms apart (~41 Hz),
the spikes at about 260 kHz
with the underlying low-level waveform ondulating erratically around 4 MHz.
P.S.: I have two DP modules, both are nearly identical in this regard;
P.P.S.: the waveforms are even more horrible when running off the USB.
It’s there even on batteries, I’m almost sure it’s unrelated to external power, one last test pending.
It’s Taiwanese
The 26 mV/div trace is 13 V PSU as measured on the barrel connector.
It’s somewhat cleaner than the “filtered” voltage, and the onboard regulators ought to stabilize voltage, not exacerbate spikes.
I remember previously encountering issues related to noise on one of the power modules, either the SP or DP. It was bad enough that it had a very noticeable negative effect on attaining GPS fixes.
Have you tried adding a capacitor or two? That’s what the simple thinker in me suggests.
I looked into a schematic of USB Client DP (version unspecified).
There are almost none of those ubiqitious .1uF ceramic capactiors. A good engineering rule of thumb says “bypass every noise-producing or noise-sensitive device pin with a ceramic capacitor close to the device”.
But no! there are only two of them.
(a) C1, almost well-placed bypassing the output of IC9 (MIC4680 buck-mode 5 V regulator), however the ground trace is too long and too curved and may prevent it from doing its job properly.
(b) C3, could be reasonably well placed near IC1 IN1 (pin 8). However, it’s futile there: IC1, a TPS2115 power multiplexor* doesn’t seem to produce a lot of RF and spikes most of the time.
The proper places for the ceramic capacitors are, with every cap providing an RF bypass path to the ground at a power line it’s connected to:
(1) near the barrel connector before any diodes
(2) near the mini-USB connector
(3) at the output of IC9 critical (MIC4680, the 5V switching regulator, mistakenly labelled -ADJ in the sch)
(4) at input to IC9 important
(5) at the output from IC7 (LM1117, 3V3 linear regulator) required by manufacturer
(6)(7) at both inputs to IC1 (TPS2115, power mux) required by mfg
(8) at input to IC7
(9)(10) at the 2x5 connector.
Caps with notes in bold face MUST be on the board no matter what, caps without bold face SHOULD be there for us to live in a better world.
All caps should be placed near the associated pins. Copper artwork should feature good RF-quality polygons (both live and ground): short and wide, without sharp angles and excess bends.
I dunno why everyone skimps on the ceramics, they are really really cheap and the board’s aren’t quite a bargain.
Input power mux is a good thing to have, but it doesn’t work on me: when transitioning from external to USB power, the device crashes the moment I unplug the barrel.
Note: at least there are no reports on the regulators blowing up at random or after being touched
(like it was the case when Atmel screwed up with the AVR Dragon in-circuit debugger/programmer for their 8-bit lineup)
Note: I’m not talking about polar capacitors (aluminium and tantallum) here, they are an entirely different matter and serve a different purpose. And they do almost nothing when it comes to fighting noise.
Side note: data lines on the USB Client DP are good, nothing is left floating and generally doesn’t require bypassing.
I will provide additional “end user” advice on aftermarket patch-up somewhat later.
I’ve been using the Meanwell supplies for a few years now and found them to be reliable and inexpensive. As long as you have good ventilation, they handle heat quite well.
More than can be said for my Apple Airport Express which expired this week after just over 1 years use. After battling for hours to open the case, I found all the electrolytics were bulging on top.
Present are
C1 acting as (6) and (3) at the same time,
C3 as (7) and (2).
By adding 4 caps to the most important places I had only modest success in improving 5V ripple rejection though. 3V3 fared better but the change was not as drastic as I hoped it would be. Will post traces later.
Therefore I suppose adding more caps won’t do the trick without redesigning the PSU.
I consider any mods not necessary for the vast majority of users who aren’t doing precision analog things (they should have separate “analog grade” power rails anyway) or sensitive RF stuff.
And my LCD flicker issue is, unfortunately, a fundamental one unrelated to power. (temporal dithering).
They can’t do much, redoing DP artwork may be of too little marketing value for them
Nothing to be concerned of – operational stability is not compromised for nearly all users.
Wrong. The beauty of the hot brick PSU will be ruined by the inadequate ripple rejection of USBc-DP regulators and the power buses will be noisy despite your investment.
Your PSU needs to supply 3.3 and 5.0 V and you have to inject it into the Gadgeteer bus, e.g. by using Extender boards. If you do so, you don’t need the USB-DP or SP modules for power, and I don’t know whether it would be advisable to have one connected at all.
Besides, the SKU you mentioned is a switching mode PSU, and it’s costly because it is rated for medical applications.
