Avoid reverse leakage current in diode-OR circuit

[Spike101]

Hi,

I'm working on a little project where I need to support powering my circuit from three different voltage sources:

- an internal 12V-Li-Ion battery (always there)
- the charger for the internal battery (can be disconnected)
- an externally connected 12V battery (can also be disconnected)

To do this, I'm using an LTC 4417 power path controller, which allows me to specify the priority of the different voltage sources. To power the device on and off, I'm using the shutdown line of the LTC 4417. So far, this works very well.

But in addition to this switchable power, I also need a little bit of stand-by power (at most 100mA if all the LEDs are on), to power a battery gauging circuit and some logic things that should always work. To this end I used a simple diode-OR circuit using three LSM115J schottky-diodes connected directly to the three inputs. Now there seems to be a problem, because within a few seconds, the input capacitors on the inputs where nothing is connected get charged by (i assume) the current leaking in reverse through the diodes. Then, the potential on these inputs gets high enough that the LTC4417 (and my logic stuff) thinks, that there is some voltage source connected, even tough it's not.
Now, I noticed, that the max. reverse leakage current on the LSM115J is quite high (10mA). But even if i choose a different diode, this will only delay the build-up of the potential on non-connected inputs, but eventually they will still be recognized as present..

Are there better solutions than the OR circuit, that don't require a much lager footprint on the pcb than the 3 diodes in a DO-214BA package?

For now, I removed the diodes for all inputs except the internal battery, so the stand-by power is always provided by the internal battery. But i fear, that if the battery is completely empty, the logic stuff won't work any more, and then the charger-detection logic would not connect the charger thourgh to the battery...

[IanB]

Schottky diodes have very high reverse leakage current. If you use ordinary silicon diodes the reverse leakage will be magnitudes less. If you want to avoid those capacitors charging up with even such a tiny leakage current you could just put a high value resistor across them.

[Spike101]

Thanks for the reply!

I thought about using a resistor, but to counter the 10mA, i can't really use a very large-value resistor, so i would waste 10 more mA at all times just for the resistor..
But i just tried using another diode (SB330) and there it's much better.
I'll see if I can find some better smd-diode, and then use a high-value resistor as you suggested.

[IanB]

Thanks for the reply!

I thought about using a resistor, but to counter the 10mA, i can't really use a very large-value resistor, so i would waste 10 more mA at all times just for the resistor..

10 mA is crazy. If you use a proper diode the reverse leakage current will most likely be < 1 µA. Then if you place a 1 M resistor across the capacitor it will keep the leakage voltage below 1 V.

[Zero999]

Something doesn't sound right.

I thought one of the main selling points of ICs such as the LTC4417, is you don't need Schottky diodes?

I admit, I haven't read the datasheet in great detail but I presume you could use the IC, rather than OR diodes.

[Spike101]

10 mA is crazy. If you use a proper diode the reverse leakage current will most likely be < 1 µA. Then if you place a 1 M resistor across the capacitor it will keep the leakage voltage below 1 V.

I know it sounds crazy, but if i use the LSM115J, i can watch the voltage across the input capacitor climb on the scope even if there is a 10k resistor across. With 1.5k it won't go above 1V (the input is about 13.5V).. With the SB330s i bodged on the smd pads, there is no visible increase on the scope even if no resistor is there. Maybe I got fake LSM115Js, i can't imagine that the reverse current is really that bad... But I already found a few alternatives that I will try.

Something doesn't sound right.

I thought one of the main selling points of ICs such as the LTC4417, is you don't need Schottky diodes?

I admit, I haven't read the datasheet in great detail but I presume you could use the IC, rather than OR diodes.

Yes, you're right. But the plan was to use the LTC4417 just for the main power, so i can use its shutdown line to power everything down except the stand-by stuff, and use the inrush current limiting everytime it's turned on. The stand-by circuit doesn't need so much power, so i thought a diode-or should work ok for this purpose...

[wraper]

I know it sounds crazy, but if i use the LSM115J, i can watch the voltage across the input capacitor climb on the scope even if there is a 10k resistor across. With 1.5k it won't go above 1V (the input is about 13.5V).. With the SB330s i bodged on the smd pads, there is no visible increase on the scope even if no resistor is there. Maybe I got fake LSM115Js, i can't imagine that the reverse current is really that bad...

According to the datasheet it should be within spec. At 25^o and 12V typical (not max allowed) reverse current is 3mA, at 50^oC 10.5 mA, at 100^oC, you will get 200mA typical reverse current  . Using a diode at 1.5V below it's max voltage is not a god idea too.

[Spike101]

According to the datasheet it should be within spec. At 25^o and 12V typical (not max allowed) reverse current is 3mA, at 50^oC 10.5 mA, at 100^oC, you will get 200mA typical reverse current  .

I think i just looked at the nice Vf figure when i chose this diode  .. I assume in that regard it's a nice diode, and the tradeoff regarding the reverse current is just natural... Well at least i learned about the importance of another parameter..

Using a diode at 1.5V below it's max voltage is not a god idea too.

Right, that probably contributes to the bad performance...

[IanB]

I know it sounds crazy, but if i use the LSM115J...

I think you have misunderstood. There is nothing wrong with your observations. You have chosen a part with a certain behavior and the part is behaving to spec.

What is crazy is to assume, somehow, that 10 mA is a reasonable reverse leakage current for a diode that is supposed to isolate one part of a circuit from another. Your immediate reaction should be "Oh crap, I have chosen the wrong part. I need my leakage to be close to zero."

Design mistakes happen. It is normal. But as you iterate a design towards success you need to know what your required design parameters are so you can try to satisfy them.