Power Supply Off with another device in parallel

Dear all,

As mentioned in the forum according the power supply, https://devtalk.nvidia.com/default/topic/1048640/jetson-nano/power-supply-considerations-for-jetson-nano-developer-kit/, I am testing an external 4G modem with the Nano.

The Jetson and the Modem are connected to the same lab power supply (5A max) in parallel.
When the power is on, both device start but the Jetson shutdown few seconds after the start sequence (Power led off).

I tried various sequence with and without voltage regulator (3A max to see it it was due to a lower voltage than 5Vcc), it is always the same !
Also starting first the Jetson and powering the modem once the Nano is on…the Jetson shutdown immediately.

Please, do you have some feedback about this critical point ?

Thank you for your help and support.

Best regards,

Hook up a storage oscilloscope on the power rail of the Nano. Connect it to pins 4 and 6 of the 40-pin header.
Then, look at the 5V signal with a resolution of at least 0.1 ms/div or smaller, using a trigger when you turn on the modem (you might be able to find an appropriate pin on the modem’s power rail to trigger off of?)

My guess is that the inrush current of turning on the modem will make the voltage droop on the Jetson power, and the Jetson power supervisor is super sensitive; even a very brief power glitch will make it turn off.

Hi Snarky,

Agree with you, the monitoring of the Jetson power is too sensitive.

According to the datasheet of the TPS25944 (On the Dev. board side), the reverse blocking mode is active when Vin < (Vout-10mV) !!!
Question : Why so sensitive ? A reverse blocking set at 100mV must be sufficient for most applications.
This is why it is impossible, on the same power supply to add a parallel device.
It’s a shame because the TPS259444 can handle higher input voltages !

So, when I connect the modem, the voltage droop at Vin - 94mV, and its sufficient to block the Jetson.

I see 2 solutions to solve the problem:

1- Using 2 separate power supply (nasty solution)
2- Using a galvanic isolation with optocoupler and capacitor

What do you think ?

Best regards,

Ok, I found a solution, not the better, but it works.
I connect the modem first, and after the Jetson to avoid the droop voltage.

That’s it ! ;-)
I will build a small electronic interface to delayed the power supply to the jeston and let’s go !

Best regards,

A GPU is a very sensitive high-density piece of electronics. It’s similar to a CPU in that way. The difference is that, because Jetson is intended for embedded applications, the carrier board doesn’t do as much conditioning as, say, an Intel PC motherboard and power supply would do for a normal consumer. An Intel CPU (or an NVIDIA GPU in such a system) is just as sensitive, if not more! It’s just that those parts are made for end-user usage, and thus have more parts and cost added to them to be more robust to power variations.

It is not impossible at all. It doesn’t work for you, in your current power supply setup. That’s because your power supply setup doesn’t fulfill all the necessary requirements. You can fulfill those requirements in many different ways.

For example, you could use a split rail power supply, where there’s a different regulator rail for the modem than for the Jetson.
Or you could use an anti-sag circuit on the input. This is a diode, followed by a large capacitor, that you then hook up to the barrel input jack (or the 5V rail on the 40-pin header.) I’d recommend 1000 uF or more.
You’d need a diode with sufficient power handling capacity, and you’d need the input to the anti-sag circuit to be higher by the voltage drop of that diode (check the datasheet.)
Or you could use a higher-voltage input, and use a linear 5V regulator to keep the voltage consistent, although the drop-out times the amperage will dissipate a lot of heat so you’d need a component with very good cooling and power handling characteristics.

Hi Snarky,

Yes, you’re right but I solved the problem.

Best regards,