DIY SCPI programmable dual channel bench PSU 0-50V/3A (now EEZ H24005)

[prasimix]

Issues #207 and #210 are fixed and ready for testing in master branch. If everything works fine we can release that as hotfix v1.1.2.

[Pjoms]

Issues #207 and #210 are fixed and ready for testing in master branch. If everything works fine we can release that as hotfix v1.1.2.

Well done! I hope I can test #210 tonight.

[Pjoms]

I have made a quick test of the latest master branch firmware regarding #210 and it looks promising. No x100 jump after one hour, but I have only tried once yet.

[prasimix]

I'd like to inform you that thanks to LJJS's effort, project is now available also in KiCAD format. I'm not familiar with it (and have to install it once again ), but you're more then welcome to check that and eventually report what you find here or on the GitHub under #37.

[Rerouter]

I am assuming these are meant to match the originals as closely as possible?

Currently they throw some DRC warning in KiCad, Any issues if I go through and clean them up?

[prasimix]

You are welcome to do that, and if possible create new pull request.

[Rerouter]

Could I ask, for IC10, your footprint doesnt seem to be made to the normal 0.65mm pitch, is this how its done on your eagle client?, or an import error?

Equally a first glance shows the relays footprints missing, I'll pack this into a push request.

[prasimix]

IC10 (DAC8552) package is imported from TI.com site and converted with Ultra Librarian and it seems that it's fine at least in Eagle.

[Rerouter]

Ok, Pull request made.

[prasimix]

Ok, Pull request made.

Accepted, thanks!

[s8548a]

Issues #207 and #210 are fixed and ready for testing in master branch. If everything works fine we can release that as hotfix v1.1.2.

Trigger system is working fine, tested the Manual and Pin1 with Step mode and List mode 

the Bus mode (need to confirm), when INIT the output is OFF and when *TRG, directly the Step value is updated to the channel. (when init from the touch screen and *TRG via BUS it works as it should)

[prasimix]

Huh, it was happened again: ADuM3160 USB isolator failed again: this is a second time in less then a year and I cannot see a pattern of what was happened. My EEZ H24005 PSU is continuously powered on or in standby state (i.e. the AUX PS board where the ADuM3160 is located is always powered on). USB cable is also continuously connected to my desktop PC which is also most of the time powered on or in standby state. Today, when I tried to make connection with PSU, USB port is not recognized any more, and ADuM3160 is very hot. If I disconnect the USB cable, it's start to cool down. But I had the same issue with other USB isolator and PCB board, like on that is shown on picture below. It was continuously connected to the same PC and after few months it just mysteriously stops to respond and when PC side is connected USB isolator IC (ADuM4160 in that case) also is running extremely hot.
Does anyone experienced similar issue with ADuM's isolators?

[Rerouter]

Where would the design files be for him? I can have a look over if there are any datasheet Gotcha's If its available.

[prasimix]

EEZ H24005 USB isolation section can be found on sheet 7 here (png) or here (pdf).
Mentioned stand alone isolator board I bought some time ago on the ebay, and cannot find any schematics.

[Rerouter]

At a glance, your USB input has no AC shunting between the USB ground and the PE ground, So at a guess some noise got in, and exceeded common mode.

The other thing of note is your PE ground trace has very little separation to the traces of the isolated side, If you had a bad ground connection and it floated to half mains there is the possibility that something jumped to a data pin.

[prasimix]

Thanks for that. Another thing that crossed my mind and could make a situation even worser is possibility that PC and PSU are on different AC phases (I didn't check that yet). Anyway, instead of using ferrite bead that can be often seen when USB isolator is not in place, perhaps I should use a capacitor with hi-value resistor in parallel, or you'd like to suggest something else?

[T3sl4co1l]

No TVS diodes?  May be hot plugging transient.

Tim

[prasimix]

No discrete TVS but IC16 is in place that is 2-channel TVS. Perhaps I should use 4-channel IC to add protection to VBUS and Gnd lines, too?

[T3sl4co1l]

FYI, referring to
https://github.com/eez-open/psu-hw/blob/master/Consolidated/EEZ%20PSU%20consolidated%20r5B13a.pdf
ZD4 and 5 are wrong, if one fails shorted, mains becomes connected to protective earth.  SAR1 is all that's necessary.

The TPD2E001 (IC16) covers VBUS so ZD17 isn't needed, and it should be fine with hotplugging.  Do eliminate FB6 with great prejudice, however, and tie shell/ground to XUSB_GND in as many points as possible.  Pour a ground plane in the area (stitched, both layers; or inner layer if multilayer).

Leave a generous isolation gap around the XUSB_GND pour, say 1.6mm or more to main circuit ground.  You can add an R+C across the isolation gap to dampen EMI on the cable (as Rerouter hinted).  Typical values might be 220pF + 100R.  Doesn't need to be Y-type (it's ground to ground, if everything is behaved..), but can be.

Tim

[prasimix]

FYI, referring to
https://github.com/eez-open/psu-hw/blob/master/Consolidated/EEZ%20PSU%20consolidated%20r5B13a.pdf
ZD4 and 5 are wrong, if one fails shorted, mains becomes connected to protective earth.  SAR1 is all that's necessary.

Thanks, but in that case one or both fuses (housed in AC inlet) should blow (or rather explode ). My idea was to have two level of protection: TVS and MOV+SAR. So far, such failure didn't happened to me, neither other users reported it.

The TPD2E001 (IC16) covers VBUS so ZD17 isn't needed, and it should be fine with hotplugging.  Do eliminate FB6 with great prejudice, however, and tie shell/ground to XUSB_GND in as many points as possible.  Pour a ground plane in the area (stitched, both layers; or inner layer if multilayer).

ZD17 is optional, not mounted, added just to make place on the PCB if IC16 is not used.

[T3sl4co1l]

No, I mean it, you're just making it needlessly vulnerable and dangerous.  Those poor TVS diodes will explode as soon as a mains surge hits the thing.  They're frail at those voltages.

It will fail.  Maybe not today, maybe not in ten years, but you're cursing some poor bastard who builds/buys it and plugs it into their dirty power, and it blows up while they're out at work and burns down their apartment and all their neighbors have to suffer their mistake (or something hyperbolic like that).

(It takes a big fucking TVS to handle mains transients, typically over "30kW" rated (versus the "1.5kW" parts shown).  Those suckers are expensive, too.  Upside, they don't wear like MOVs do, as long as you never exceed their ratings.  Which isn't the easiest thing to guarantee, given the random nature of real surge events, so you may wind up using many together (even more $$) to ensure that.)

And anyway, you've got transformers!  What are you worrying about the common mode voltage?  That GDT + MOVs will keep things well within the typically 2.5kV rated range of anything safe to plug into the wall!  Just the MOVs would be fine, even kind of overkill as the transformer itself is capable of limiting transients to a fair degree...

FYI,
Tim

[prasimix]

Thanks Tim for additional information about TVS. Huh, it sounds dangerous. You are suggesting that it will catch the fire before fuse blow. At least backers from US is in the better position since their units also comes with TVS rated for 400V. Therefore their AC mains input has to be very dirty .

Regarding transformer, two AC/DC modules (Mean well LRS-150-48) are used in this model instead of bulky mains frequency transformer. I presume that SMPS transformer is not so capable of limiting transients.

[T3sl4co1l]

Ahh, all the more reason, the Meanwell units are already rated for it!  (Notably, Safety & EMC spec, IEC 61000-4-5.  Industrial, too.)  I wouldn't even bother with the MOVs, but you can if you like.

Tim

[morris6]

Huh, it was happened again: ADuM3160 USB isolator failed again:

Sorry for this somewhat late reaction..

Thinking about the failing ADuM 3160 isolator:

Normally the whole control section of the H24005, the arduino and so on, is totally isolated from the rest of the system. It is isolated by the ADuM 3160 USB isolator IC17 and control signal isolators IC19 - IC23. Power, 5V, comes from the TR1 supply module, the GND net of the control section is not connected to anything outside this section. So it looks like there is no path to prevent any buildup of a static charge.

In fact, when measuring voltage between PE and digital GND on point 8 of the left-hand auxiliary connector there is some voltage building up slowly. This happens while the H24005 is in standby as well as when switched on. Highest I have seen, after +/- 15 minutes was more than plus 10 Volts on a high impedance voltmeter. What is the cause? The meter, with its input open and nothing connected, has the habit of going to a negative indication. So it looks as if something else is loading the GND net positive referenced to PE. When switching the meter out of 'high-Z' to 10 Meg input the voltage drops to about zero.

I think there is nothing against connecting the control section GND net direct to PE with a piece of wire from point 6 to point 8 on the left-hand auxiliary connector. Just to provide a discharge path. Or a resistor 1 Meg or so? This will keep the 'inner' side of the isolator under control.

Another thing.. Also both power regulator sections are fully isolated and not protected against static. Shouldn't there be a (high) resistive discharge path as well? Or do the MeanWell supplies take care of that?

--------

PS. It looks like there is a typo in schematic 7/12 'Ethernet and USB PCB connectors' : near IC17 the GND symbols are wrongly identified as 'PGND', actually they are on the 'GND' net.

[prasimix]

Another thing.. Also both power regulator sections are fully isolated and not protected against static. Shouldn't there be a (high) resistive discharge path as well? Or do the MeanWell supplies take care of that?

I think that Mean Well has output coupled to PE with capactitor. Also both Power modules binding posts are coupled to the PE. Don't know if that is enough or even relevant as protection against building of static.