EEVblog Electronics Community Forum
Electronics => Beginners => Topic started by: berwick53 on April 19, 2017, 04:52:51 pm
-
Hi, I'm taking some power measurements from an AMD graphics card using a PR30 current clamp on a +12V PCIe 6-pin power connector and It appears that I have current going in the other direction occasionally when not at load..
I've 'zeroed' out the clamp meter when the PSU was turned off and have removed the remaining bias from my data but I'm still seeing values below 0 with the largest (negative) reading being -1.68A. I'm measuring the result on a Rigol DS2702A with it recording data at 100kS/s which is far more than I need so it this just high frequency noise ? As this appears to be rather large?
The attached graph is Power (W) vs Time (s) and the blue line is a 1000 point moving average.
Current clamp datasheet: http://www2.elo.utfsm.cl/~elo382/wp-includes/images/PR30.pdf (http://www2.elo.utfsm.cl/~elo382/wp-includes/images/PR30.pdf)
-
100kS/s is certainly not enough, and will cause aliasing. On top of that
1.68A in this case is only about 1.5 bits of scope ADC resolution, basically below the noise floor of oscilloscope.
EDIT: That said according to the scale on the picture attached. Dunno if 1.68 is according to those numbers as 100+ amps current is an order of magnitude too high.
EDIT2: Well, those are watts obviously. What goes negative should be noise + likely aliasing. Average is not that negative at all as is about those 1.5 bits of resolution mentioned.
-
100kS/s is certainly not enough, and will cause aliasing. On top of that 1.68A in this case is only about 1.5 bits of scope ADC resolution, basically below the noise floor of oscilloscope.
EDIT: That said according to the scale on the picture attached. Dunno if 1.68 is according to those numbers as 100+ amps current is an order of magnitude too high.
EDIT2: Well, those are watts obviously. What goes negative should be noise + likely aliasing. Average is not that negative at all as is about those 1.5 bits of resolution mentioned.
Thanks for answering, I've done some more digging and have taken some pictures of the system when idling at a much higher sample rate (500.0 MS/s & 2.0 GS/s) and with setting the BW limit to 20 MHz vs 300 MHz. I can see from this when applying the BW limit that some of the noise disappears but the negative part of the waveform still appears. Do you still think this could be due to aliasing or something else?
I'm thinking that I may need to measure via a current shunt if I can't explain why.