I have an application which will receive power from two redundant DC inputs. The input voltage is nominally 24V but can vary from 6V to around 130V from various transient conditions. I also need the unit to survive (but not operate during) the case of reversed power connections.

There is a DC-DC converter that converts the main input voltage to 5V / 30W to power the internal circuits in the unit. To support the redundant inputs, I want to OR the input power connections. But using a regular diode gives me very little headroom in the case where the input is 6V. Therefore, I would like to use an ideal diode.

Looking at commercially available ideal diode controllers, I can't find any that operate over 100V. I did find some workarounds by adding some extra circuitry to either the LTC4359 or LM5050. I am also aware that one can use an op-amp to construct an ideal diode.

I want to explore the possibility of constructing a solution based only on discrete components.
Are there any discrete ideal diode circuits that can meet my requirements?

  • operate at 6V
  • operate at 130V (preferably with a good safety margin)
  • Pass 36W of power (6A @ 5V, less current at higher voltages).
  • Have better input-output drop than a regular diode (so like less than 0.5V @ 6A).

I came across this circuit online for creating an ideal diode from discrete components.

enter image description here

This circuit has some problems.

  • When the diode is off, the voltage VOUT - 0.7V - VIN is put across the base emitter junction of Q1. This basically prevents its use at voltages higher than around 5V.
  • The gate of Q1 is not protected when operating past VGS max of M1.
  • To prevent excessive power dissipation at high input voltages, R1/R2 need to be fairly large values. But this makes the turn off time of M1 fairly slow (a good fraction of a ms), especially at lower input voltages. In the case where one redundant input suddenly shorts to ground, this could cause a significant glitch in the ORed input power.
  • If Q1/Q2 are not perfectly matched then it's actually possible for M1 to conduct some current in the reverse direction (from VOUT to VIN) when VIN is slightly less than VOUT. This could be a good fraction of an Amp given the right combination of MOSFET on resistance, VIN, and VOUT. Most commercial ideal diode controllers incorporate a small voltage offset (around 50mV) to account for this.

enter image description here

  • I can protect the gate by adding a zener diode.
  • I can minimize offsets by using a matched transistor pair such as the DMMT5401.
  • I can make the circuit work at high voltage and reverse input by adding some diodes to protect the transistors. But the diodes are in the VBE path, so they need to be matched. I couldn't find any diode pairs that are sold as "matched", but something like the BAV23,215 that contains two diodes on the same die should be pretty closely matched (although not specified how close).

The remaining problems to be solved are...

  • There can still be some reverse current. I need to find a good way create a small offset (say 20mV~50mV) in the Q1/Q2 base voltages.
  • Turn off is still slow. That might be solved by adding another transistor that connects to the collector of Q2 to increase the drive strength on the gate. Also, using current sources in place of R1/R2 might improve performance at low input voltages (but they would need to be matched, unless of course that's how I create the offset also).
  • It might be possible to just replace Q1 with a 200V diode that has a forward voltage that is lower than VBE of Q2. This gives the protection afforded by D1-D4 and solves the offset problem. But I would need some way to guarantee the relationship between the diode and Q2 over temperature and manufacturing tolerances.
  • \$\begingroup\$ Ah, your spec is too ideal and too expensive to make it practical. \$\endgroup\$
    – tlfong01
    Feb 2, 2022 at 8:09
  • \$\begingroup\$ Making your 100K resisters not equal will give a small offset .Remember the diode equation .Your mosfet is a P channel which gives a simple circuit but it needs low on resistance . \$\endgroup\$
    – Autistic
    Feb 2, 2022 at 9:08
  • \$\begingroup\$ @Autistic I agree. Making R1 go from 100K to 33K would provide around 59mV of offset (if the diodes behave ideally). It's a little extra power, but not much, so it might be worth it. \$\endgroup\$
    – user4574
    Feb 2, 2022 at 16:03


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.