DC noise from Jetson makes LED flicker

I have a application with an embedded Jetson Nano and have a problem with DC noise.

The system is powered by 24V DC and I use a isolated DC/DC converter (CCG-30-24-05S TDK-Lambda - Board Mount Modules - Distributors, Price Comparison, and Datasheets | Octopart component search) to bring down the voltage.

The Jetson is also used to to control 24 V LED panels with. However when the Jetson is running under load doing some inference the in full power mode the LEDs start to flicker. Probably due to induced noise in the system. The problem is greater when I use a longer power cable but if I compensate the PSU to 26V to get 24V to the LEDs they still flicker so they don’t flicker due to low voltage. When the Jetson is powered on but not running inference utilizing high CPU and GPU there is no flickering.

My electronics knowledge is highly limited and I thought there might be some people in here that stumbled on similar problems? I’ve googled all day and I tried using small capacitors I had lying around but they don’t make a noticeable difference.

Would a EMC filter make a difference and should I put it on the 5V or 24V side of the circuit? I would also like to be able to dim the LEDS using PWM from the Jetson, then it would be better to put it on the 5V side?
Something like this: RSEG-2003 : Detailed Information | EMC Components - Power Line EMC Filters | TDK Product Center

Any ideas to reduce the noise would be greatly appreciated!

Can you share the schematic of how Jetson control LED panel? In general, the control signal would not cause LED flicker.

Here is a schematic of the simplified version with LEDS flickering under Jetson load. The LEDs is not controlled by the jetson at this moment. This is for isolating the noise issue.

This is essentially how I connect the LED-strip when controlling them from the Jetson through PWM

Seems the LED flicker caused by work load of Jetson and also depends on the output capability of 24V supply. You should check the character of 24V supply and add some capacitors to reduce flicker.

The power supply I use is a bench power supply with that can output 5A. Or a DIN-rail mounted 24VDC 5A AC/DC transformer. It should be able to provide sufficient current.

If I use a short cable (1-3 m) from the 24V PSU to the circuit with the 24/5V transformer and LEDs there is no flickering. If I use a longer (25 m) cable there is flickering. The cable I use is a CAT 7 23 AWG solid wire where I use one pair, 2 wires, for 24VDC and one pair, 2 wires, for GND (double 23 AWG wires). The resistance of the wire is 1.2 Ohms.

Since the problem is bigger with the longer cable I suspect there is problem with noise introduced from the longer cable. Magnetic inference or something like that? The cable should be able to handle the voltage drop and power loss I think so this is not the issue?

If I were to put in some capacitors. Could you give me a hint of which capacitors I should use and how to implement them in them to the circuit as a starting point? My knowledge in electronics is quite limited unfortunately…

Thanks a lot for helping out!

You should use a oscilloscope to observe the stability of 5V supply to Jetson, if it is stable, then it should not be a problem for Jetson. We don’t have suggestion on capacitors of your custom 24V supply, maybe you can get help from vendor of converter.

Thanks, I don’t have an oscilloscope. Maybe have to get a hold of one to be able to troubleshoot further.

But do you think a filter like this one could make a difference?

I would start by testing with a large capacitor right next to the input to the Nano. Maybe 2000uF to 5000uF. Place it as close as possible to the Nano’s power input. See if there is any change.

Thanks you for something concrete. I put three 470µF in parallel without any noticeable difference. I will order some larger capacitors such as this:

Would it be a problem putting several of these i parallel for testing? When do you reach a point where capacitors doesn’t make any further improvements?

That capacitor should be a good test case. I doubt more than one would be of use for testing this.

So far as whether it is actually possible to use several of those capacitors I suspect it depends on whether the power supply can handle the sudden charge requirements…if you have enough capacitors, then it will look like a “short” for a very brief moment when starting the power supply (I have no idea how much capacitance would be required, but I also have no doubt that a single capacitor of the spec you listed would be enough to test with). You’d likely have to have a lot of those capacitors before a power supply would care, but there really isn’t a need to try more than the one for a capacitor that is 8200uF.

Do make sure the capacitor is as close as possible to the Nano. Using heavier (larger diameter) wires is also better than using smaller wires (smaller diameter) which might introduce resistance.

Thanks you. I will order some capacitors from 2000uf to 8200uf to try with.

To be able to control the LED-strip from the Jetson I have to connect them to common ground. Would this have an effect on the capacitor preventing the “capacitance”? See picture below how I’m thinking to make the circuit.

Technically I think the two grounds would need to be connected if the GPIO is to be used from the Jetson to the strip, but it depends on the design of the DC/DC converter. If the converter is ok with connecting -Vout to Gnd, then this might be preferable. There are times when ground loops are an issue, but unless there is more going on than what I know about, then probably there is no issue with a ground loop, but there would be an issue if the GPIO ground signal is floating or isolated relative to the MOSFET and light strip. Does the DC/DC converter have any issue with tying -Vout to Gnd?