10
\$\begingroup\$

Probably a stupid question but I could not find an answer on the web.

I was looking for a diode to protect a 36V battery from a short circuit on the charging port. The best I could find (SB560-E3/54 and similar) have around 0.5V forward drop. That is a lot of heat at 2A charging current.

There are some OR-controllers (like LM5050) with reverse current protection function, but all of them require a GND connection and a few extra parts besides the FET.

Do 2-terminal ideal diodes exist and is it even possible to design such a device?

\$\endgroup\$
10
  • 3
    \$\begingroup\$ When the question is marked as "needs details or clarity" it is customary to at least mention what details you need or what is unclear... I will gladly provide additional details if I know what you need. \$\endgroup\$
    – Maple
    Commented Jun 15, 2022 at 16:58
  • 2
    \$\begingroup\$ This is a very nice question. The downvote and close flag are not warranted. Ideal is not absolute nor physically realizable. There are degrees. We can get pretty damn close to an ideal diode for specific applications, and this would be one well suited for that. \$\endgroup\$ Commented Jun 15, 2022 at 17:42
  • 2
    \$\begingroup\$ @SteKulov I was looking for "reverse current blocking" (TI definition), which is different from reverse polarity protection. Although when implemented properly it can certainly do both. However that was not my question, which was about plausibility of 2-terminal design \$\endgroup\$
    – Maple
    Commented Jun 15, 2022 at 22:12
  • 1
    \$\begingroup\$ @BruceAbbott "you didn't look very hard" Here are the requirements: 45V reverse, 2.5A average, axial through hole, in stock. You may try punching these at DK, Farnell, Mouser and Arrow before dismissing my efforts outright. Also, 1W in the airtight enclosure for 3 hours charging time is not exactly something to ignore. In any case, I only mentioned that to provide a background for the question. \$\endgroup\$
    – Maple
    Commented Jun 17, 2022 at 9:30
  • 1
    \$\begingroup\$ @BruceAbbott 2.5A is simply to have a safety margin for the 2A charger's rated current. When reduced to 2.1A the smallest available drop is still 420mV. Same with voltage - the charger outputs up to 42.4V, so we set 45V requirement. The enclosure contains the battery, motor controller and a whole mess of wiring. The closest similar device I can think of is the body of an electrical scooter. I don't need a two-terminal solution, I already found workable option with LM5050. The question is about possibility of one, from electrical engineering perspective \$\endgroup\$
    – Maple
    Commented Jun 17, 2022 at 16:06

6 Answers 6

13
\$\begingroup\$

For your application - it’s not possible to have such a diode without qualifications (unrestricted use), since they require power to operate, being active devices, and it’s not possible to get power out of a “resistor” that has 0V across its terminals.

It is possible to create approximations, though.

If the diode is expected to be periodically connected in reverse, you can use the reverse voltage and a boost converter to charge a low-leakage capacitor. The voltage on the capacitor would hold a mosfet turned on for a long time - perhaps days at room temperature. The mosfet’s body diode should act as a backup. A conducting low-RDS(ON) mosfet does a good job approximating a forward biased ideal diode.

Pursuing this idea further, another option opens if it’s possible for the diode to be non-ideal for a short time periodically, e.g. once in a few hours. These short periods of non-ideal behavior would be used to capture the diode drop voltage and use it to charge the gate capacitor using a step-up converter - a joule thief of sorts. There would need to be some reasonable forward current available to do this of course. Two forward diode drops will be plenty to run a boost converter to recharge a capacitor. One diode drop will make it less efficient but still possible. Once the capacitor is charged enough, the converter will turn off and the mosfet will turn back on.

Combining the two ideas, you can get a two-terminal diode that in certain operation regimes and certain load impedances will remain ideal either continuously or for the vast majority of the operating time. The nano power circuitry needed to switch between states would be an interesting challenge for sure. Low voltage CMOS gates (74 family) can be operated sub-threshold to manage that. With parts specified for 1.8V operation, 0.7V is not too hard, as the currents are still in the micro amps, so already a lot for this application.

In short: why do you care how many terminals you got? It’s not like you don’t have a power source available! It’ll make it miles easier. There are ideal diode controllers you can use directly then. So it’s an interesting challenge and a wonderful kludge to design into a one-off hobby project, but not suitable for production.

\$\endgroup\$
4
  • 1
    \$\begingroup\$ In this particular application diode connected in reverse means either short on charging port (easily done but rare) or charger with wrong pinout (unlikely). Stealing part of forward voltage is more plausible, as this corresponds to regular charging operation. So now the question comes down to this: is it possible to make so efficient circuit that for all practical purposes it will be close enough to ideal diode? You seem to be confident that it is. Now, is it suitable for production? \$\endgroup\$
    – Maple
    Commented Jun 15, 2022 at 20:57
  • \$\begingroup\$ If you look at literally thousands of diodes covering huge range of applications you'd notice very fine granularity of their parameters. But if you line them up by Vf you will eventually hit the limit in any Imax category. Wouldn't be nice to continue these categories with lower and lower Vf? Without the need to switch from 2-terminal schematics to much more complicated driver + FET + cap layout. \$\endgroup\$
    – Maple
    Commented Jun 15, 2022 at 21:03
  • \$\begingroup\$ An ideal diode of any kind will be a complex circuit. A practical implementation would be an analog ASIC with a couple hundred fets/transistors. There's no need for all that just to have two pins instead of four. Now, is it suitable for production? Yes. With four pins. \$\endgroup\$ Commented Jun 16, 2022 at 3:38
  • 3
    \$\begingroup\$ @Kubahasn'tforgottenMonica Or with two. \$\endgroup\$
    – Hearth
    Commented Jun 16, 2022 at 21:02
12
\$\begingroup\$

I would argue that an ideal two-terminal diode does not exist. I am also assuming that having an internal power source, such as a battery, is not allowed. My thinking is as follows:

  1. An ideal diode has zero volts across its terminals.

  2. No voltage difference can exist between any nodes inside the ideal diode.

  3. Without a voltage difference there can be no use of active circuitry, such as MOSFETs.

  4. Such a diode must rely on inherent characteristics of its internal components. In other words, an ideal diode can only be constructed if you already have an ideal diode to use inside it.

\$\endgroup\$
4
  • 1
    \$\begingroup\$ Very clear-cut logic! Indeed, one needs voltage difference to power internal circuits/gate driver, and there should not be difference if device is functioning as intended. Having said that, we both know that the term "ideal diode" does not really describe truly ideal part in real life applications. If we assume that some tiny Rds(on) is allowed in forward direction, and tiny leakage current in the reverse, can this be done then? \$\endgroup\$
    – Maple
    Commented Jun 15, 2022 at 16:50
  • \$\begingroup\$ Your answer was first and it is perfect from theoretical standpoint. I did accept Kuba's answer because it gives a [very] small hope we may see "almost ideal" devices one day \$\endgroup\$
    – Maple
    Commented Jun 16, 2022 at 19:47
  • 1
    \$\begingroup\$ I think this theoretical answer fails on theoretical grounds; it ignores time. First we notice that step 1 ignores the reverse operation: An ideal diode can have any reverse voltage across its terminals. This can be used to provide energy, which can be used later in time to power active circuitry. And active circuitry can be designed to consume electricity only while switching. The answer assumes that there is no energy storage over time. \$\endgroup\$
    – MSalters
    Commented Jun 17, 2022 at 14:11
  • 1
    \$\begingroup\$ @MSalters No, there is no concept of time for an ideal diode. And certainly there is no requirement that the diode spend part of its life in reverse bias. On the other hand, in reverse bias an ideal diode passes exactly zero current so your "active circuitry" causes the diode to be non-ideal. Ideal means ideal, not sometimes ideal. \$\endgroup\$ Commented Jun 20, 2022 at 10:39
9
\$\begingroup\$

While continuing a research on the topic, I've stumbled upon this chip: LM74610.

enter image description here

So, here is a simple answer: yes, it is absolutely possible to make 2-terminal "ideal" diode. Note the term in parenthesis denoting the usual approximation of the theoretical abstraction.

Not surprisingly it works by harvesting energy of the load current, just as @Kubahasn'tforgottenMonica suggested.

The chip requires only external MOSFET and a storage capacitor to work inline. The FETs are routinely embedded into PMICs (see LM66100 for example). I suspect capacitor can be embedded as well, or at least bonded to the die in production. This makes a single 2-terminal device perfectly plausible and not too expensive or hard to manufacture.

\$\endgroup\$
3
  • 1
    \$\begingroup\$ This circuit has a periodic >0.7V glitch to recharge the boost capacitor. Not quite ideal, but, it is a clever circuit. I'd rather stick with the circuit @Ste Kulov referenced in the comments of your original post due to simplicity and that it can be used above 45V reverse voltage limitation the LM74610 has. \$\endgroup\$
    – qrk
    Commented Jun 16, 2022 at 21:58
  • 1
    \$\begingroup\$ So what is the voltage from Vin to Vout during operation? Does the circuit require an ideal MOSFET in order to function as an ideal diode? \$\endgroup\$ Commented Jun 16, 2022 at 22:51
  • 2
    \$\begingroup\$ @ElliotAlderson Either ideal MOSFET or a pinch of fairy dust :) The way I understand datasheet, the gate is switched off with 2% duty cycle, which increases voltage drop to that of a body diode, allowing charge pump to steal some power. The rest 98% of time the voltage drop is defined by Rds(on) of the MOSFET. \$\endgroup\$
    – Maple
    Commented Jun 16, 2022 at 23:12
4
\$\begingroup\$

Consider a hypothetical ideal inline diode. Connect it in series with a resistor and a DC motor. The diode would rectify Johnson noise from the resistor, thus powering the motor, turning heat into work without a heat sink. This is a Maxwell's Demon for electrons, and violates the Second Law of Thermodynamics. Since the 2nd law is solidly verified physics, anyone who devises such a thing and proves that it works is guaranteed fame and fortune. Given that it has not happened, I expect it never will.

\$\endgroup\$
2
  • \$\begingroup\$ Ha! This is nice mental exercise. The problem is, it equally applies to 3-terminal ideal diodes, which do exist and widely used, thanks to generally accepted relaxed definition of "ideal" \$\endgroup\$
    – Maple
    Commented Jun 16, 2022 at 19:57
  • \$\begingroup\$ @Maple 3-terminal ideal diodes pass a small amount of current to the third terminal. That power consumption is sufficient to satisfy thermodynamics. \$\endgroup\$
    – John Doty
    Commented Jun 16, 2022 at 23:43
2
\$\begingroup\$

No, there are no ideal diodes, just as there are no other ideal elements but you can use a diode with a lower drop.

enter image description here Source: https://pediaa.com/difference-between-schottky-and-zener-diode/

Another option is to use a mosfet to do power blocking:
Equivalent ideal diode circuit analysis

Reverse polarity protection - PMOS vs Schottkey diode

\$\endgroup\$
2
\$\begingroup\$

No, for the reasons explained in other answers, but there are some things, like the TI SM74611, that get damn close.

The internal charge pump means that it injects a bit of switching noise, and it's probably quite slow, but they claim 26 mV forward voltage at 8 A, far better than any normal diode could achieve.

\$\endgroup\$
1
  • 1
    \$\begingroup\$ Please put the part number in your answer, right now it risks being deleted as a link-only answer. (The test is whether the answer remains useful if links are stripped out; yours is in a borderline category where a screenshot is useless but as long as the link URL remains readable, even if it cannot be followed, it still has some information. But better to indicate what part you're discussing directly in the text.) \$\endgroup\$
    – Ben Voigt
    Commented Jun 16, 2022 at 21:34

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.