Mastech HY3010E-3 Power Supply Noise

[tech_builder]

Hi All,

I've been a fan of the EEVblog for a while, but just registered for the forum now because I have a problem with a lab power supply I just got. The power supply is the Mastech HY3010E-3, which is a 3 output supply with 2x 0-30V 0-10A, and 1x 5V 3A. It looks like a nice bit of kit, unfortunately looks may be all that is good about it.

I fired it up on the bench and after testing the outputs with a multimeter (which checked out ok), I hooked it up to my oscilloscope (Rigol DS1052E). I was quite surprised by the noise I saw!

Over 200mVrms noise, which is much higher than the quoted 0.5mVrms noise they spec'd. Also the voltage spikes are over 2Vpp! Is this common?! I expect some noise because it is a switching supply and not linear, but this was quite a bit more than I thought.

I emailed Mastech and asked them about it and their reply was:
"use professional milivolt[sic] meter to measure the RMS ripple, using scope is not a correct way to measure the ripple."

What do I say to that? I'm hoping to get some help here from people who know more about this topic so that I can make an intelligent and informed response to them, and hopefully figure out if something is really wrong with the power supply or my method of measuring the output from it (which was to connect the leads of my oscilloscope to the power supply directly).

Thanks for your help!

[grumpydoc]

Was the PSU connected to a load?

How did you connect the PSU to the 'scope?

[amspire]

If you connect the scope probe and ground both to the negative terminal only, how much noise do you get?

The problem with a scope is it is connected to ground, and the power supply could make a mess of ground, but still have 0.5mV RMS on the output. Also 0.5v RMS could still have a high peak to peak if the pulses are short enough.

A simple way to see if there is a high peak to peak ripple on the output is to attach a 1N4148 type diode in series with a 0.1uF cap/1M resistor parallel combination across the output terminals. You can try the diode in both directions. If that pulse waveform is genuine, then with the diode in one direction, you will get a negative voltage on the cap. In the other direction, you will get a voltage above the set output voltage. Use a battery powered multimeter to measure the voltages.

[PA4TIM]

I meaure powersupplys and powerrails all the time with a scope and my SMPS labsupplys from Philips and Delta show no visible noise ( at least not without using a uV resolution scope plugin), my linear psus show no excessive noise too.

This powersupply or your scope/probe just are not good.
Or your probe attenuation is not set correct in the menu ;-)

I think it is the first, This is the third noise related mastech powersupply problem I read about the last month ( one in a TV, the adapter that kills their LCR meter and this one. So they seem to be not very good in psu design)

You could add a filter on the output, do some shielding ect to make a proper psu from it.

[grumpydoc]

Looking at the trace I suspect that the PSU was unloaded and that the 'scope was connected by a probe with a long ground lead, or even one of the "BNC to croc clip" type leads.

In that case the high impedance coupled with stray inductance in the leads causes large spikes on the trace which are not really present when the PSU is driving a load.

Use as short a ground connection as possible - watch the PSU review that Mike did (mikeselectricstuff) and watch how he connected the probe. It might also be worth making a 10:1 or 20:1 50 ohm probe to measure the ripple & noise.

Or use a differential probe.

[gibbled]

The specs for this supply say it weighs 25lbs.  I'd suspect it is a linear supply rather than a switcher.

[Telstar]

Yeah, but the ripple is too much even if he used a crappy cable/probe.

[ptricks]

Could be grounding issues between the scope and the supply.
If you look at the peaks in the waveform provided it is in the shape of a sine wave, possibly either an AC source or a switcher somewhere.
The peaks look like when a switching supply powers the inductor then cuts off power to the inductor causing the spikes seen. The way they are spaced at equal times looks like a switcher running at around 1Mhz.

[mzzj]

The way they are spaced at equal times looks like a switcher running at around 1Mhz.

More like 100 khz if timebase is 10usec per div.

[tech_builder]

Thanks for the replys!

grumpy doc:
The PSU was not connected to a load, it was connected via some banana plugs cables with gator clips on the end (as you suspected). I did a differential measurement using to probes and subtracting the signals. The excel sheet is attached, but the results show a 120mVrms ripple. I searched for the PSU review you mentioned and watched 2, but neither had a scope in them...I will continue looking as I would really like to know how to do this properly.

amspire:
I have not tried to connect the scope probe and ground both to ground, I will try that and report back.

PA4TIM:
The probe attenuation is correct:) I thought that might be the problem too, but alas it is not...hopefully I don't have to do anything to the supply to make it functional

I appreciate all of the help I am getting here. I will try a few more things that were suggested and see how things change, most notably get rid of the banana to gator wires...

BTW I came across this while I was looking for ways to measure the ripple:
http://www.micrel.com/_PDF/App-Notes/an-27.pdf

does anyone have any comments on the procedures used here?

Thanks!

[PA4TIM]

http://cds.linear.com/docs/en/application-note/an104f.pdf
I made a load based on this appnote. I use a piece of coax with bncs and the scope AC coupled.

I told I measure ripple often just using probes but I forgot to say I do this not on an unloaded powersupply. You can connect a 100 Ohm resistor between the termials and then hook a 10X compensated probe to the terminals.

The Rigols are rather noisy but not like this, so if you still have all those spikes this way it could be the scope/probe picking up noise from some switcher in your lab. The impedance of a powersupply is very low so the probe groundlead is, for RF allmost direct connected to the tip and forms a nice loop antenna. Connct probetip to groundclip and look at the scope if you still see the spikes. If so useit to find the source of the EMC, you see it when moving the probe around and watching the amplitude.

If you do not see spikes that way it probably is the psu.

You can use a ground spring like in your pdf, or attach the core of a piece of coax to the plus terminal of the PSU, and the coax shield to the min terminal. And make the connections as short as possible.
If you still see the spikes it is the psu. Maybe a capacitor dead, or just bad construction. But that can be improved by adding some filtering and shielding.
But in that case first do a dynamic load test like in the pdf from my link. If the psu behaves very bad there too better build or buy an other one.

[grumpydoc]

I searched for the PSU review you mentioned and watched 2, but neither had a scope in them...I will continue looking as I would really like to know how to do this properly.

starting at 7:25 to about 8:30 and you can see the effect of using a "long" ground lead on the noise measurement.

[tech_builder]

I did some further testing after watching Dave's teardown and testing of the Atten PPS3205T-3S Triple Output Power Supply Review (EEVblog #439). I turned the bandwidth limit on and set the probe to 1x for better resolution. I also constructed a shorter ground lead out of some copper strands (see picture). This had the effect of lowering the RMS voltage to about 3.5 mVrms. It's still not the 0.5 mVrms listed, but I don't have the equipment/technique to verify that. I'm happy to see how a little technique can make a lot of difference in the reading - for the better!
Thanks so much for the help, I really appreciate it.

[saturation]

To measure PSU ripple properly you must have a floating scope or use a differential probe, otherwise you'll always see common mode noise because the scope is grounded to earth.

[SeanB]

In my "to watch" queue..........

[tech_builder]

To close this story somewhat...I ended up being unable to use this power supply as the noise was too high. I bought a used Tektronix PS280 linear power supply and the difference is night and day between them. I was trying to make a simple low pass filter using an opamp and I couldn't see the signal through the noise with the Mastech PS, but when I connect the Tektronix to the same circuit clarity prevails. As a matter of fact, the same measurement techniques I used to measure the noise of the Mastech I used to measured the Tektronix. I could hardly see any noise taking the scope all the way to the highest setting! I suppose I can use the Mastech if I require high power, maybe something involving motors, but otherwise it will sit there...

[saturation]

If you want really low noise see the Power Designs Precision Power Supply thread; you can get a used one for less that 1/4 the price of the Mastech.

[ptricks]

If you are having problems with a circuit that you think might be power supply related don't forget you can use batteries. We don't think about battery power much when we have a bench supply available but it is a true 100% noise free source

If you have some old laptop batteries , take them apart, there are usually several 18650 lithium cells still usable and they make great power sources for the bench.

[saturation]

yes, batteries are simple solution for low noise,  and if your power draw is low, are quiet self 'regulated'.  What I used to do was use NiMH with a regulator on the front end, since they are rechargeable, the wasted power dissipation was not such an expense.  In even older days, I used a discarded lead acid car battery that are charged with a cruder and noiser constant current PSU.

If you are having problems with a circuit that you think might be power supply related don't forget you can use batteries. ..