No, diodes won’t protect your circuit

Comments · 295

• BT

  @bttlab · 3 weeks ago · pinned

  I hope I managed to get my point across. The dual-diode clamp protection can and oftentimes should be used; it is pretty good, all things considered. In order for it to fail so spectacularly, many factors actually have to align. The R_in and the impedance of all devices connected to the 3.3V rail form a voltage divider. Just make sure the resulting voltage stays below 3.3V in the worst-case scenario, and there won't be any such problem.

  42

• TZ

  @tze-ven · 12 days ago

  That is a common misconception. Dual-diode clamps are designed to suppress fast, transient voltage spikes, such as those from ESD or transmission line reflections. They cannot survive the continuous, high-energy impact of an electrical overstress (EOS) event like the one you demonstrated, because they are not built for that purpose.

  308

• T_

  @T_Mo271 · 12 days ago

  Those protection diodes are intended for ESD, not for DC over-voltages.

  129

• FE

  @fernandoportal5422 · 2 hours ago

  This video was insightful

• D6

  @D614G_aka_Doug · 2 weeks ago

  It is quite easy and inexpensive to make a shunt-regulated &quot;clamping rail.&quot; One of the simplest ways is to use something like the TL431 adjustable &quot;precision&quot; shunt regulator. If you configure a TL431 to regulated at 2.5 V and then use ordinary silicon PN diodes from the input(s) to that rail, you&apos;ll get a clamping voltage starting at something around 2.9V at very low current to around 3.2V at moderately high clamping current, depending on the diodes you use. The TL431 can handle up to 100 mA (careful of dissipation!), but a PNP transitor can be used to increase that.<br>This isn&apos;t a &quot;micropower&quot; solution since you may need the voltage setting resistors for the 431 and you need to assure it gets at least a few hundred microamps of &quot;anode&quot; current (from, say, a 3.3 V supply via a resistor) to assure it stays in the active region. But if you are concerned with clamping 50 ohm sources, your system probably isn&apos;t micropower anyway. Pay attention to the amount of capacitance you need across the 431 for stability. This is laid out in the datasheet.<br><br>I noticed some of the clamping circuits in the video use Schottky diodes. They have their merits, but ordinary Schottky diodes are notorious for high reverse leakage current which can cause significant problems if high impedance circuits are involved. Low leakage Schottkys exist, but they are pretty expensive. Even ordinary switching diodes like the 1N4148 can be too leaky for some applications. The junctions of bipolar transistors can be very much better and cheaper than low-leakage diodes. Bob Pease wrote on use of BJT junctions as alternatives to diodes many years ago. JFET gate-body diodes exhibit extremely low reverse leakage, but they cost a good deal more than BJTs.<br><br>If you use conventional zener diodes, you&apos;ll generally get better characteristics (sharper &quot;knee&quot;) with lower power devices. 500 mW zeners are likely to be better than 1 W zeners. There used to be at least one family that was better than most, but I don&apos;t recall the part numbers. It can be tedious reviewing datasheets.<br><br>There are some interesting input protection networks available in SMT packages. There are also TVS diodes designed for very low capacitance.<br><br>If your clamping voltage is a bit higher than you would like, a small resistance between the clamp circuit and the input can be very helpful, often without introducing objectionable delay caused by the resistance and the input capacitance. Most internal input protection diodes can handle at least 2 mA and some are rated for 5 mA or more.<br><br>Any time you use any sort of active circuitry for clamping that needs to be fast, try to assure that the circuit is biased into its normal operating range when it isn&apos;t clamping. You want to avoid situations where slew rate limiting comes into play or where BJTs in the device are saturated, since it can take considerable time to get them out of saturation.

  53

• RI

  @rivimey · 10 days ago

  I found the video really interesting, and the comments add so much more. Please, please do another video pulling together the alternatives outlined in the comments into more accessible visual form. Would be really helpful!

  5

• SK

  @skveery · 11 days ago

  You can also put a clamp on the PSU. &nbsp;This will allow the traditional circuit to work on any pin.

  34

• OP

  @opros7 · 5 days ago

  I tend to avoid using schottky diodes for that because of their reverse leakage. Some can reach up to 1mA at 85°C which trashes any high impedance analog signal. So i usually use a large 10k - 1MOhm input resistor into low leakage diodes and then another 47-100k into the chip so the clamping diodes inside the chip carry a very low current. Also another 100pF to 1nF for clean ADC readings. If the input impedance is too high, the signal can be buffered again. ESP32s have the issue that if the ADC goes slightly over the supply rail, the entire input bank readings are off. <br>For anything lower than 1-10k i use bidirectional TVS diodes in SMA or SMB with a low value resistor. A low value resistor 10-470 Ohm is also great for shaving a bit of the high frequency edge off the signal which helps passing EMV. And for signals like USB there are dedicted clamping chips with diodes, a resistive element and a zehner. I managed to pass 4kV surge/burst tests with such designs where the data line is next to the mains or coupled but also had boards fail because i forgot the clamping on another pin that then had the supply bridged to. You need to be aware of the signal and what you are clamping to, correct.

  1

• MA

  @massimiliano9864 · 11 days ago

  Love the no AI statement in the banner of the profile, just subscribed

  11

• HL

  @HL65536 · 10 days ago

  What I usually use is 2 resistors: one before the diodes, one after the diodes. This also reduces the requirements on the diodes (no need for low Vf), which can then be small signal diodes that have very low capacitance. Also, the VCC clamp doesn&apos;t need to be on the I/O pin. One on the VCC rail for all pins at once will do the trick (also reducing capacitance) as the diodes conduct to there. With all that, it&apos;s possible to up the resistor value without impacting high-ish frequency signals too much (280 Ohms in total (which also protects against output short circuits) still allows 50MHz signals through).

  4

• DI

  @divyakumar8147 · 5 days ago

  thanks &nbsp;video was insightful!!

  1

• LO

  @louco2 · 10 days ago

  Thank you for taking the time, learned something new today!

  2