Teensy 4.0 killed by interfacing circuit to CGA card (TXS0108)

[inmbolmie]

Hi, if you want to fast-forward the story, the direct questions are at the end marked bold.

I have an issue with a report of a bricked Teensy 4.0 used with my EGA/CGA to VGA board described here.

The event was like this: The Teensy was working but after some cycles of programming/plugging/unplugging from a Tandy 1000EX went bricked (no life, no usb, 15 second restore doesn't do anything).

Me personally have treated my Teensies very roughly during development, even adding to this kind of operation some over voltages, overheating, reverse polarities, etc, for months, never having an issue. So it comes to my mind some kind of ESD event with the user setup as a possibility.

The complete schematics are here https://github.com/inmbolmie/T4_EGA_VGA#schematics. The computer side from the EGA port perspective is essentially a 74LS244 (it has very low current sourcing capabilities) driving the 8 CGA/EGA signals up to the TXS0108 inputs, plus a grounding shield that is NOT connected at the Teensy side. The VGA signals are only output so I think they can be ignored. The Teensy is powered from the USB connector.



My theory is like this: I simply ignored any ESD issue coming from the EGA connector because I have the TXS0108 in between, and it has in-built ESD protection. But what I didn't consider is that the 3,3V and 5V pins (VCCA, VCCB) to power the TXS, that are directly connected to the Teensy, can leak voltage back supposedly from the input pins protection diodes of the TXS inputs. In fact, now measuring the pins with the board unpowered and the EGA port plugged in, like 2-3 volts are coming back to the Teensy mainly from the VSYNC/HSYNC signals. So *maybe* ESD-protecting the input pins only had the effect of diverting the voltage peaks back to the Teensy via 3,3V and 5V pins through the protection diodes... The theory is, that fiddling with the EGA connector some sparks could have jumped from the computer grounding case to one of the pins (because inadequate grounding, ambient conditions etc), passed from the input pin to the 3,3V rail via ESD protection and proceed to fry the Teensy 4.0 internals.



I've been thinking about adding a layer of protection to the power pins like this, with a shottky to block reverse currents and a TVS device to divert voltage peaks out of the Teensy:



The questions are:

Does all of this weirdness of a voltage peak passing through a TXS0108 make any sense to you?

Would be enough to protect like described the 3,3V and 5V Teensy pins?

Is there any recommendation for adequate partnumbers for the TVS diodes or whatever parts you consider necessary?

I would be very thankful if anyone is willing to share its experiences here.

I also have read posts suggesting using 74LCX devices for high impedance while the circuit is powered-off, but I don't know if that would also apply to this problem of reverse voltages coming to Vcc from protection diodes.

Thanks and best regards.

 

[MarkT]

Clamping signal lines to the rails with Schottky diodes is standard really - for any logic signal going off-board.
I've seen BAT54S and QSBT40's and similar used for this. Some series resistance is good too, even on outputs,
as you get reflection-reduction thrown in with the spike-suppression - 68R is a common value for this, and
inputs can have a lot more (helps with voltage tolerance)

If you've got proper decoupling (high-speed and bulk) on the rails then this should provide good ESD protection.

And with all the Schottky clamps Vcc isn't going to be driven far below ground....

 

[PaulStoffregen]

Does all of this weirdness of a voltage peak passing through a TXS0108 make any sense to you?

Honestly, no, not really. Indeed the ESD protection diodes do tend to divert ESD charge to the power supplies, but those supplies have a lot of very low impedance (at high frequency) capacitance.

But these bidirectional level converters are notorious for getting "confused" about the intended data direction if there is noise or ESD or other signal quality problems. We've seen lots of that in the past when people tried to use them for driving WS2812 and other addressable LEDs.

In the signal is unidirectional, you should avoid these bidirectional chips and use a standard buffer chip.

 

[inmbolmie]

Thanks for your responses.

. Some series resistance is good too, even on outputs,
as you get reflection-reduction thrown in with the spike-suppression - 68R is a common value for this

I will test also this, but the idea is using the protection for the power pins, not I/O as I/O "should" be already protected by the TXS ESD protection, and adding 20+ components somehow kill the project purpose.

but those supplies have a lot of very low impedance (at high frequency) capacitance.

But in this case there is no "power supply" but the Teensy power pins. Do you mean that the ESD it will be diverted to the USB cable 5V line and not inside the Teensy? I maybe can imagine that for the 5V but I understood that 3,3v is unprotected.

But these bidirectional level converters are notorious for getting "confused" about the intended data direction

I feel direction is not a problem here as direction is always input, and ESD is also input so it's not a case of a signal going the wrong direction. I tested some interfacing chips and finally used the TXS because it was the alternative that gave by far the cleanest output for this particular application, and also supposedly gave good ESD protection. I plan to test also 74LCX245 but the problem may be still there if it has the same kind of ESD protection, and that kind of information isn't usually included in datasheets at all. I think all chip designs assume that you have an independent power supply capable of absorving the energy from any ESD event.

Thanks and best regards.