DIY DC Current Probe?

[0xdeadbeef]

I think it's going to topple over when you attach a BNC. It'll be a pita.

Well, maybe. Probably not in the actual use case where the scopes are placed quite a bit above the loads. And the scope will not even be connected all the time. That's what the LEDs are for.
Besides, I could just define one of the sides as bottom. Or use a larger footprint off-the-shelf case. So IMHO this isn't really a problem.

But, today I think made a power supply circuit with a TPS6513x.
Now on to sourcing all the components, which isn't easy if you are not a business. Then create test board, since I never get these right first try.

Plus it's QFN which is something I try to avoid. Plus DC/DC stuff is much more critical regarding component selection and layout compared to a charge pump.
Side note: AFAIK a USB device is only allowed to draw 100mA until it negotiated its current draw with the host.

[Jeroen3]

Ah, yes. You have a workbench with equipment shelf. The LED's are a nice touch though! I'm might steal that.

I am aware of the 100 mA uninitialized limit, but I have yet to encounter a host that implemented it strictly. But I could search for a USB charge controller nonetheless.
The QFN doesn't scare me, the DC/DC frequency does. I was able to approach to the recommended layout reasonably well with 0805.

The 2.7 to 5.5 Vin of the TPS65131 should be suitable for 3xAA or one 18650. I'll have parts and boards over ~2 weeks. The Chinese have holidays...

[Leo Bodnar]

Side note: AFAIK a USB device is only allowed to draw 100mA until it negotiated its current draw with the host.

Bad news, bad news! USB "device" is only allowed to draw 100mA for 100ms and then possibly 2.5mA if you convince the judges that it is really a device. Even this is very Zen because how the device with no data pins would know that there is no activity on the bus data pins?

Good news is that nobody knows this (let alone follows the rules.)

USB Lava lamps are allowed to draw 600mA if they are pretty.

Leo

edit: evidence is attached

[Jeroen3]

Just use this cable and you get twice the power 

[Jeroen3]

So. I took out Fusion 360 and did my first 3D to see if it would fit in a box.
Finding a box was hard, I wanted to use a profile extruded enclosure, but these would become large if they could house the 23mm of the CMS3005 sensor.
Later I found this Hammond box, with mounting stubs in the lid. It looks like it would fit.

The 3D was even harder. Never touched such software before. But it's great, I like it.

[0xdeadbeef]

Some parts of this design look strangely familiar...
But isn't this going to topple over when you attach a BNC. Won't it be a PITA?
Just kidding 

[Jeroen3]

Yes, strangely familiar because I couldn't find an extruded profile enclosure small enough.
And yes, it will probably topple. I could prevent this not by mounting the banana's on the lid. Maybe I will.
But I'm afraid any cables will offset the sensor.

[mk_]

But I'm afraid any cables will offset the sensor.

You should talk about your concerns with Sensitecs tech-support... very helpfull and friendly people...

[IDEngineer]

Idea: ACS chips were mentioned earlier in this thread. Some members of the ACS current sensor family sense differentially, to reduce/eliminate sensitivity to external fields. We've used these extensively and they work really well. It would be quite straightforward to throw together a demo of one meant as a scope probe... pick the part with the lowest range so you get the greatest sensitivity. We're able to discern at least 10's of mA with them in our applications, if that's enough for you. Very low insertion resistance so minimal impact on the DUT. And you get DC+AC in the deal, so no weird rolloff as you approach DC.

I personally prefer the clamp-around style so you're not "in the circuit" at all, but most of the examples in this thread seem to be of the insertion type.

Just a thought....

[Jeroen3]

Allegro Microsystems isn't that high bandwidth or accurate ans Sensitec. But they are also cheaper and smaller.
I also prefer the clamp around style, but those are expensive. And, if you want to measure on a PCB, you are still out of luck with a clamp.

[capt bullshot]

AFAIK, sensitec senses the field gradient from the u-shaped conductor. So in theory, they are insensitive to more homogenous fields from distant conductors or e.g. earth magnetic field. What matters is the distance of the conductor, the more the better. Can easily be evaluated by placing a (AC) current conducting wire next to the sensor and moving this wire around while there is zero current in the sense conductor.

[0xdeadbeef]

For the protocol: finally finished my first box...

[Jeroen3]

It looks great  . But does it work?

I'm still at renderings, I've been having way too much fun learning 3D cad. 
But the parts are ordered. It's roughly the same size, but a standard enclosure instead of 3D printed. Since I will have to make a few for at work if it proves a success.

[0xdeadbeef]

The tests I did up to now were successful. Like I tested the SMD part without the sensor and now a quick DC current test with the sensor.
I.e. the first LED lights up somewhere slightly above 1A. If the current->voltage conversion fully works as expected with high bandwidth is something I can only test later this week at work.

About the case: yeah, I wasted hours and hours looking for a fitting off-the-shelf box. I finally gave up for the moment as the best one I found was much too large in X/Y dimension (note that my box needs to be pretty high - my printed one has an inner (!) height of 45mm).

BTW: what about some air vents?

[Jeroen3]

I'll try without vents first. Since I noticed most of the heat came from poor contact in the test breadboard.

[0xdeadbeef]

BTW: found the time to quickly test my two boxes at work with a 10kHz PWM at ~12V signal driving a current through a ~3.3Ohm load. Somewhat simplified, but I didn't have the time to set up the test I wanted to do (more complex current profile, inductive load and much higher peak current).

Anyway, the signal shape looked almost identical to that of a Fluke i30s letting aside the very low current range. At the end of the active phase, the signal of the i30s dropped to a flat line while the one of my boxes showed a certain (slow) undershoot/overshoot. I'm unsure though if this was an issue of my boxes or one of the Fluke i30s. Didn't look like a bandwidth thing, probably more one of offset or sensitivity.
Besides, the signal of the i30s looked much noisier (which also made the low current range difficult to compare) but in hindsight I guess this might have been due to  different filtering settings on the scope.
Anything above the 10mA range (or so) overlapped more or less completely, so the amplitude is OK and there is no general flaw in my circuit.
Also the 1st of the current LEDs was lit or blinking depending on the PWM frequency and this worked in both directions. So I'm confident that the LED gimmick also works as planned.

I still want to perform a test with a more complex load and signal shape (+higher currents) and compare the boxes to a Fluke i30s (100kHz) and Fluke i50s (50MHz) with screenshots and all.
Still, it's already clear that the boxes work and fulfill my expectations. Even if further tests should show that there is some kind of undershoot/offset issue in the low current range, this wouldn't affect the usability for the planned purpose (to check the amplitude and shape/timing of a current profile).

[0xdeadbeef]

Started testing a bit with an Electronic Load and was confused at first about getting ~0V at 100mA. Further investigation shows there is a 12mV offset on the output without any current flowing which means 120mA or ~0.5% of the 25A current range. This is a bit higher than the "typical" value of +/-0.3% given in the datasheet but still within the maximum of +/-0.8% (not taking into account an additional temperature offset).
This is a bit higher than I expected given that the datasheet defines these values for +/-15V while I'm using +/-12V. Then again, in the 100mV/div (or 1A/div) range this is not really visible and given that the currents to me measured are in the 10-20A range, this is not really an issue for me.

[leonerd]

I found one device on Tindie that looked like what I would want. Except not quite suitable for high side.

That's mine

You say "not quite suitable for high side use" - what do you mean there? The design is entirely built around having the sensing shunt resistor on high-side. This can be anywhere from 0V to 60V above the reference ground. Is that not high enough? What limits are you wanting to go to?

Simply put, I want to make a uCurrent with a BNC plug on it that can have some common mode voltage.

That was the exact intention of my design.

[Jeroen3]

It looks to be intended to measure a devices power, and it goes in series with the power supply. You don't have high side problems there.
I would like to measure current anywhere, also in-circuit.

[leonerd]

OK, so just pass the measured current between the V[in] and V[out] terminals. It will cope just fine with any input voltage from 0V up to about 60V - so it can do high-side, low-side, anything inbetween. It's really just a current shunt sensor, but arranged on a board convenient that it could measure a device's power supply as that was what I wanted it for. But it can measure whatever current really.

[Jeroen3]

Alright. That did not become clear on the Tindie page. However, I've gotten into the rabbit hole now. 

update:
I'm currently looking into putting two 14500 lithium cells inside.
They should fit, since two normal AA's do. I also considered two eneloops, but that would be only 2.4 Volts, which is too low and would require an additional booster.
A protection IC would become unnecessary with nimh though. Both 2x1900mAh nimh and 800mAh li-on carry equal Wh's.

[fuzzoli]

For folks in the US and who might be interested in the Sensitec sensors mentioned in Reply #2, I contacted and Sensitec and they pointed me to https://buy.fwbell.com/current-sensors.html

They only carry the 4000 series sensors, but this series requires just a single 5v supply (as opposed to the +/- 15v supply of the other sensors).

...free shipping too! 

-Frank

[Jeroen3]

But, today I think made a power supply circuit with a TPS6513x.
Now on to sourcing all the components, which isn't easy if you are not a business. Then create test board, since I never get these right first try.

Got the board in. Very unstable switching, thing burns 250 mW without load.
Replaced chip, thing works for a while, then starts burning 100 mW... still without load. 

1: V+
2: V-
4: Vin


Sch/Brd:

Now, I got to admit I didn't have 3.3uH, so I used 4.7uH. It works great at 2.5Vin, but at 5V it's awful.

[Jeroen3]

Progress! Replaced the inductors for 2.2 uH. More stable. Seems I really need to get close to the 3.3uH, which isn't in my lab kit.

I now got ripple at around 200 mV, at around 1 KHz. Still significant ringing though.
When I test only the positive I get under 60 mV ripple, so I have to fix some Vin issues I guess.

Also spotted one mistake already, Vneg feedback is taken from before the capacitors.

[Jeroen3]

As I said, switchmode power supply topology near or above 1 MHz is a bit voodoo I'm not master yet.
I replaced the output capacitors 2x10uF ceramic 1206 for 4x 4.7uF X5R 0805.
Massive improvement.
I also disabled power-saving mode on the TPS65130 to make it behave more predictably.

Ripple is now low and high speed that even the location of the probe has significant effect.
There is still a small peak, I think this is the diode.

On below image:
1: V+
2: V-
4: Vout (0A)