Page 1 of 1

Exploring Ideal Diodes

Posted: Wed Sep 08, 2021 1:03 pm
by parkview
I have used some Ideal Diode circuits in a few projects now, but have found that they leak a reverse voltage back through the Diode (MOSFET). Scouring the Internet, I have found three different designs all based around the idea of turning off a MOSFET if there is no voltage on one side of the MOSFET.

I have tried making some of them up on a piece of prototype PCB, but it's hard to swap components out, and some of them are quite small - SOT-236. I have bit the bullet and designed up an Ideal Diode Test PCB. This board tests out all three of the designs, using a variety (3 different footprints) of MOSFETs and matched transistors.

To boot, JLCPCB now offer a Purple solder mask PCB, so I ordered with that colour, PLUS, I see they are now offering much cheaper shipping, as low as A$6, so in theory, you can have 5 10x10cm PCB's made for around A$10!
Ideal-Diode-Tester.jpg (47.7 KiB) Viewed 1394 times
The above pin headers give me a flexible input/output arrangements. I have included 5 or so test pads per test circuit. I have also included a MT9700 electronic switch IC, that apparently offers reverse voltage protection at 3A. We live in interesting times!

Re: Exploring Ideal Diodes

Posted: Sat Dec 18, 2021 10:32 pm
by parkview
As I mentioned above, I have found three different Ideal Diode circuit diagrams on line (note the resistor values in #1 and #2), these are:
ideal-diode-schematic.jpg (53.85 KiB) Viewed 1203 times
ideal-diode-schematic2.jpg (85.81 KiB) Viewed 1203 times
ideal-diode-schematic.3.jpg (12.46 KiB) Viewed 1203 times
These are all based on a transistor current mirror arrangement. I have previously unsuccessfully tested circuits 1 and 2. Maybe the problem was the MOSFET, or dual transistor package I was using and that was why it failed. Failed, in that I would always see a residual 2.6V at the input with a voltage present at the output. But this was meant to be a Ideal Diode, so I shouldn't be seeing no where near that high a voltage being passed through to the input. This is the reason I created the Ideal Diode test PCB:
ideal-diodeV1.0_sml.jpg (315.46 KiB) Viewed 1203 times
Note: in the above photo, I am feeding 12V in INPUT 2 through circuit 5 and then backwards through circuit #3. You can see the bright(er) LED in #5 and a very faint LED in #3.

I have marked each of the circuits with a texter. 1 & 3 are based on circuit 3 above (triple resistor), and have a AO3401 and RU30L30 MOSFET. Both of these never turned the MOSFET on as much as the other design.

PCB circuits marked 2, 4 and 5 are based on my schematic shown below (and circuits 1 & 2 shown above). I spent a lot of time experimenting with these.

I have laid out all three examples and placed some with variable resistors, so I can tinker with the values to see what happens. Here is main circuit:
ideal-diode_schematic.jpg (112.53 KiB) Viewed 1203 times
This is similar to diagrams 1 and 2 above. I have included pads for three different MOSFET footprints, so I can hot air different ones in and out of the PCB. Same with the fixed and variable resistors for the transistors. Each circuit can be isolated for the others using the colourful jumpers.

It's been a while since I designed up the PCB and I quickly reminded myself to review the MT9700 high side switch IC datasheet - AFTER I had let out the magic smoke by putting 12V into the 3-5V IC :-( Luckily I have a stash of them, so it was easy to hot air another one back into place.

So after 7 or so hours of playing around with the PCB, what have I found out? They all *kind* of work when inputting 12V, I say kind of, because I think ALL of them where mostly operating with a partially biased MOSFET and perhaps the body diode was helping to pass the current through to the output.

What worked best was diagram #1, with the 10K and 47K resistors transposed, ie: the 47K ohm resistor on the left side of the diagram. 90K and 20K values work as well, and the MOSFET gate was driven a bit stronger.

Ok, some data. I tested two MOSFETS: AO3401 (footprint=SOT23, 30V 4A, Vgs=12Vmax) and a RU30L30 (footprint=PDFN3333, 30V 30A, Vgs=20Vmax) both P-channel. Yes, the AO3401 Vgs was pushed to the max and survived, but not recommended. I tested both the DMMT5401 and BCM857B dual matched transistors, with no difference found between the two dual transistors packages.

A) Using my diagram, with RV6 set to 10K and RV5 set to 47K ohms, the MOSFET gate was at 8.9V with respect to GND, or -3.1V from the input voltage. I measured 49mV drop across the MOSFET with a load of 510mA at 12V input. This is a 24.99mW power loss across the MOSFET.

B) With RV6 set to 47K and RV5 set to 10K ohms, the MOSFET gate is driven harder at 1.93V with respect to GND or -10.07V from the 12V input voltage. The MOSFET had a 13mV drop across it with a 510mA load. This is a 6.63mW power loss across the MOSFET.

SO, thats was when the circuit was fed from the input. What happens if I back feed the 12V into the output? What do I see at the input side? Being an ideal diode, it should be 0V right? Wrong! This is the reason I built this PCB, as I always seemed to get a small residual voltage at the input. It was always in the 2.5-2.7 voltage range, which is enough to just light up the red LED off a 15K ohm resistor! After lots of trial and error and testing different combinations (even a back to back MOSFET arrangement), I think the voltage is due to the transistor/resistor arrangement and nothing to do with the MOSFET. When I place a 6 ohm load at the input, I only get a 52mA current flowing backwards through the Ideal Diode arrangement.

Even if I feed the voltage back in reversed at the output, I get a -2.7V, ie: reversed out at the input! Not that that will happen very often, but it was a surprise.

These do act as a good reverse voltage input protection, with only around -2mV to -38mV appearing at the output.

Re: Exploring Ideal Diodes

Posted: Sat Dec 18, 2021 10:52 pm
by parkview
That last post was all done with 12V. When I tested the circuits with 5V and RV6 set to 47K and RV5 set to 10K, I measured a 7mV across the RU30L30 MOSFET at a 190mA load current. The gate measured 0.88V to GND. When I fed 5V into the output, I measured -39mV at the input - nice!

When I put 3.3V into the Input side and RV6 set to 47K and RV5 set to 10K, I measured a 14mV across the RU30L30 MOSFET at a 100mA load current. The gate measured 0.42V to GND. When I fed 3.3V into the output, I measured -38mV at the input!

Just to clarify my results, I created this graph to show the relationship between feeding a 3 to 12V voltage into the output side of a AO3401 MOSFET circuit and measuring the corresponding voltage at the input side:
output-vs-input_voltage.jpg (13.84 KiB) Viewed 1188 times

The bottom X axis is the voltage being fed into the Output. The Y axis is what is being measured at the input. At around 7.7V placed on the Output side, the input starts going increasingly positive and from around 9V, it increases rapidly. Interestingly, the results are slightly different for the RU30L30 MOSFET when the voltage fed into the Output was below 7.7V. Voltages fluctuated +- 50mV on either side of 0V.

[Edit] after a bit of a break I thought it might be handy to redo the above graph, and this time include a small 20 ohm load on the Input, this makes for better voltages being measured at the input.

The MT9700 works as advertised. I should have placed a small 10K variable resistor on the current set pin 3, so then I could much easier test out the current limiting capabilities of the IC. I had placed the jumpers too close to the IC, so the hot air melted the plastic bases of the jumpers, especially after I replaced it the second time. I measured 260mV across the MT9700 with 1.84A flowing through it into a 1 ohm load. I was using a 3.6K ohm current limiting resistor, which equated to a current limit of 1.88A. Of course, dumping 5V across a 1 ohm load, I should have been passing 5A, so the IC was doing it's limiting job.

With 5V applied to the output of the MT9700 and the IC disabled, I measured 620uV at the input, with no input load. So it does effectively act as an ideal diode at 3V to 5V. This will make for a great little (SOT23-5), high side switch for controlling power to various circuits and it's cheap too, at around 12 cents each in lots of 50.

I have confirmed empirically that the MT9700 IC doesn't like a reversed 5V placed on the input, so it would need a input reverse protection diode/MOSFET in front of it if want that protection.

I have enjoyed designing, making and experimenting with this Ideal Diode project. Don't believe everything you read on the Internet, or at least test it out yourself!