Trying to measure attributes of DC/DC SMPS

Discussion in 'General Electronics Chat' started by ssb2245, Jul 11, 2011.

  1. ssb2245

    Thread Starter New Member

    Jul 11, 2011
    Hey guys, I am new to these forums. Forgive me if i am a bit rusty on some things. I have basic EE knowledge, and i want to try something.

    I have PC motherboards laying around, and well I overclock systems for personal use. What makes a motherboard good for overclocking is how well its power supply is, usually a 12v in and 1-2v out PSU fed by the main AC/DC PSU, which needs to deal with load transients of 100A steps. Sometimes max current pull might be over 200. of course these power supplies are actually built to supply a lot more than that, but how well their output voltage is regulated can tell a lot about how well they are doing. Switching frequency is going to be from 250khz to 1mhz max.

    Until now I didn't have a scope at my house to do this testing, i just used them in labs. I have one now, and i want to measure, Peak to peak (ripple), overshoot, undershoot, and whatever else you guys think this basic scope can measure. It only cost me 200 but I read a lot of good reviews about it, its a DSO 2090.

    Anyways I have 1x/10x probes, and I was going to use a coaxial cable that is shielded soldered to MLCC caps on the underside of the CPU socket. This is how we measure the voltages with a multimeter.

    Anyways i was wondering if i need to compensate the probe or if it shouldn't matter?

    Any tips to get the best readings?
  2. billbehen

    Active Member

    May 10, 2006
    Soldered coax sounds like a good idea. In theory, you need to match the impedance of the cable to that of the oscilloscope. Sometimes scopes have a switch for low impedance measurements with a current probe, so you could use that. Obviously, it depends on the Zo (characterisitc impedance) of the cable, which you can learn from the type of cable, which is usually printed on its insulation.

    If you have to do it yourself, you need a low value resistor (equal to the cable's impedance) from the scopes input to GND. This needs to be right at the face of the scope, like not an inch of component lead allowed around 1MHz and above!

    Remeber that this forms a voltage divider with the cable's AC/DC impedance so you may read lower voltages than are on your board....

    Ripple & noise are good to measure; also step load variations. Line rejection is also important, as AC mains are not really very clean! Testing this is not so easy, as mains transients are not easy to simulate, but most power supplies don't have much rejection capability.

    Compensating a probe, if you use it instead/in addition to coax, is also a good idea! I think the instruction manual tells you how to do this....
  3. ssb2245

    Thread Starter New Member

    Jul 11, 2011
    hey thanks for you reply. So i figured out that this scope has compensation testing through a ground and 1khz 2vpp signal out the back of it. So when I hook up a probes leads to the back leads I should get a 2Vpp and 1khz signal exactly and that is exactly what i get.

    So i added the coax cable (about 3 inches as short as I could make it) Then i made sure not to create excess cable length.I ended up using a shielded USb cable, that has about 5 smaller wires, 3 are together, and 2 are outside the central line. So i use the 3 small ones and soldered them together and use teh two on the outside of the central line and use them as ground. So far this has worked better than using a normal coaxial cable, I don't know why, maybe the shielding is better.

    Other than that i tried an HDMI cable since I am thinking HDMI tech requires very well shielded wire especially for a long cable like the one I used and cut up. i felt as though its internal wires were too thin though.

    I wish I had a current probe, but i don't, and i really wanted to measure transient loads. i have a current monitor which measures current from the main PSU's 12v outputs to the SMPS on the motherboard, so I got to look at that maybe i can hook one probe to measure the voltage drop across the current sensing resistor, and use the math function and compare it to the vout of the SMPS. of course its not measuring the current output of the SMPS, so I don't know how helpful hat will be.

    Can i use this FFT test function at all for analyzing harmonics of an SMPS? or would it be worthless?
  4. ssb2245

    Thread Starter New Member

    Jul 11, 2011
    Hey so i did initial testing of ripple, 20mv/div, 1ms/div and left the scope in 1x mode, i didn't do any compensation of the coaxial cable, as its like 1-2inches in length.

    Is there anything wrong with this?

    Also I am in the correct mode right? you measure ripple in AC mode, even in a DC/DC SMPS? b/c a friend told me i was in the wrong mode. Probe is at 1x..

    Uploaded with