Small project with Dodgy Diode issues...

[Knut]

Hi,

I've been dabbling with a small sensor project with a battery backed power supply. In short I want 5V power from a USB supply on the regular with a single cell LiPo as the backup. Simple enough, a charger chip, a couple of diodes and a mosfet should do the trick. So, i sketched one up, built it up on a breadboard, tested for several minutes and felt rather confident with myself. A couple of weeks later and the PCB shows up in the mail, solder it up and things are no longer as rosy as on the breadboard...

My problem is that I get battery voltage on the USB port when USB is disconnected (no 5V input). To be presise, I measure 3.45V on the USB connector when the battery is connected (no USB power), which is about a diode drop less than the 4.1V charge voltage of the battery. There is a Schottky diode in circuit to prevent this so I started to do some digging into other possible reasons for how this "reverse" voltage manages to get to the 5V rail. Both diodes was soldered in correctly (not reversed). I suspected maybe the charger chip could have some strange quirks and leaking back to Vdd, but nope, it is not the culprit. Turns out the Schottky diodes I got from DigiKey must be dodgy.  More on why I believe that in a bit.

Attached is the schematic of the power circuit.


As is evident, D1 will prevent the battery voltage from reaching Vdd (5V rail) so the charge chip will be unpowered and no battery voltage on the USB connector when the battery is in use. One could say that there is a pull-down resistor missing on the gate of Q3, but that is taken care of by R1 and R2 (the voltage divider for measuring the USB voltage) so pulling down the gate to enable the mosfet is covered. D2 protects the battery from getting 5V through the integral diode of Q3. Q1 powers the charge indicator and Q2 is added to the voltage divider for measuring battery voltage so that it does not drain when not being measured.

For my breadboard tests I used a BAT42 as D1 and D2, it's only rated for 300mA, but that is more than sufficient for my load.
For the actual PCB I selected Toshiba CUHS20S30H3F as it has a very low forward voltage (250mV@500mA).

After eliminating all the reasons I could think of as causes for my issue, I started testing the CUHS20S30 diodes and sure enough, when applying 5V in reverse (5V on the cathode), I get 5V on the anode - not what I expected for a 30V reverse rated diode.
So, on to testing with some meters where things got interesting... My bench multimeter (Rigol DM3058) shows me a forward voltage of 65mV and OPEN in the reverse direction. As expected and the same was the story with my other multimeters, EXCEPT the "disposable" little Aneng AN8009 that I keep in a little mobile toolbox. That little sucker shows me a forward voltage of 80mV, and in the reverse direction it shows 2.9V And, I did test several diodes.

Now, this is supposed to be a 30V Schottky diode, not 3V so did I get the wrong part?
Nope, not according to the datasheet which states that the marking on the diode should be "86", which under the microscope is correctly marked on the diodes I received from DigiKey.

Links to DigiKey and the datasheet at Toshiba:
https://www.digikey.com/en/products/detail/toshiba-semiconductor-and-storage/CUHS20S30-H3F/9749634?s=N4IgTCBcDaIMIFUASBlMAGFBmdIC6AvkA
https://toshiba.semicon-storage.com/info/docget.jsp?did=63606&prodName=CUHS20S30

So, anything obvious I am missing or misinterpreting, or have I simply got parts from a dodgy batch of Toshiba diodes?

BR,
Knut

[harerod]

Does that voltage hold up, when you put load on it?
I am hinting at the reverse leakage of that chip, which, however, shouldn't be able to produce that voltage on the input voltage sense resistors. Is there any difference between your real setup and the schematic?

[Kleinstein]

Schottky diodes have a relatively high leakage current. The given diode gives up to 500 µA for 30 V at room temperature. With the currently relatively high temperatures in much of Europe one could get a leakage current higher than the DMMs test current.
The 80 mV in forward direction also points to a high leakage (maybe due to a high temperature).

[Knut]

Hi,

Did some more testing.

In circuit (on the actual PCB) I put a 68Ω resistor from the 5V rail (where I get 3.45V from the battery) and GND in series with a current meter and a voltmeter in parallell over the 5V rail and GND. Put 4V to the battery terminal from the bench psu and with the resistor in as load the potential on the 5V rail drops to 11mV so it does not hold up under that load at least. The 110kΩ load of the voltage divider is not enough to pull the voltage down (and that may screw with my gate on Q3...). The current meter shows 170uA through the 68Ω resistor.

I tested the Schottky diode with the same 68Ω resistor as a series load out of circuit:
5V forward voltage to the anode measures 70mA through the resistor and diode. (As expected, 5V/68Ω = 73mA)
5V reverse voltage to the cathode measures 170uA (0.170mA) through the resistor and diode. Same reverse current as I measured above.

The datasheet for the CUHS20S30 specifies a reverse current of 170uA @10V so I guess my results are as per specs, though at 5V and not at 10V (I measure 197uA @10V).

I did the same measurements to the BAT42 I used on the breadboard:
5V forward voltage to the anode measures 70mA through the diode and the 68Ω resistor.
5V reverse voltage to the cathode measures 0.07uA (0.00007mA) through the diode and the 68Ω resistor.
The BAT42 measures 82mV at the anode when applying 5V to the cathode - seems reasonable to me.

And, in regards to temperature we are below 25C up here so don't think that is a factor.

I guess the diode works as advertised, however I do not understand why I get a full potential of 5V on the anode when applying 5V reverse voltage to the cathode.
Probably just not the most suitable for my application so I will look for some alternatives to test out.

The implementation on the board is per the schematic, thought there is more to the board (sensor, mcu etc), but those parts of the PCB is not populated on the one I am testing with at the moment to keep them out of the equation.

BR,
Knut

(Edit: Fixed some typos)