EEVblog #1266 - PSU Probing Screw Up

[EEVblog]

How to measure switching noise on a power supply properly, because Dave completely goofed it in the Riden RD6006 PSU review video.

EEVblog on LBRY: https://lbry.tv/@eevblog:7

[johnlsenchak]

You can find those  Meanwell  power  supplies  pretty cheap  on Ebay.com


Following on  https://lbry.tv/   

[thinkfat]

The mismatched piece of coax behaves like a tank circuit. The burst you see on the scope is the resonant frequency of that tank. The switching pulse puts energy into the tank. If you put a resistor in series, it will reduce the Q of the tank and this will dampen the oscillation. Probably the waveform you see is not true due to the reflection on the scope side but I think you'll still be fine for a Vpp measurement.

How long is that piece of coax, 1 meter? Then the frequency burst should be around 70-80 MHz or so (4m wavelength).

EDIT: judging from what is visible in some of the video frames, the frequency is more like 20MHz-ish. I guess I didn't account for some extra capacitive load from the scope input and extra inductance from the probing leads etc...

[EEVblog]

The mismatched piece of coax behaves like a tank circuit. The burst you see on the scope is the resonant frequency of that tank. The switching pulse puts energy into the tank. If you put a resistor in series, it will reduce the Q of the tank and this will dampen the oscillation. Probably the waveform you see is not true due to the reflection on the scope side but I think you'll still be fine for a Vpp measurement.

It's pretty darn close to the x10 scope probe measurement, as I showed in the video. Single input side termination is fine for this application. You can double side terminate but then you have to X2 gain your scope, and not all scopes have that option.

[Fran_PSR]

Good Morning.
Although I think it is not the right channel for this doubt I will expose it equally. I have a discussion with the members of my company about the difference between earth and ground (negative). The situation is the following:
We have to connect an RS485 cable with shield and we do not agree if the shield must be connected to ground (negative) or earth. Most of the information I find from other manufacturers suggest that the shield must go to earth but a team brand technician tells us that it must be ground (negative). These are security central that communicate with keyboards and other modules through RS485. I do not want to say the brand of the plant but as a reference I can say that it is Australian ... ;-)

Thank you so much

[Raymund Hofmann]

To evaluate this power supply the "screwed up measurement" is quite of some worth.

The high frequency noise above 10Mhz is too high for a good lab supply. Simple filtering could improve it a lot.

[thinkfat]

To evaluate this power supply the "screwed up measurement" is quite of some worth.

The high frequency noise above 10Mhz is too high for a good lab supply. Simple filtering could improve it a lot.

Not sure if even the improved probing technique is good for >20MHz. The switching pulses have a lot of energy in harmonics. You'd need to make sure that your probing setup has a flat response over the whole bandwidth you're interested in. Otherwise you'll just hit another resonance somewhere higher up and be fooled again.

[BravoV]

How much difference does this method compared to the global standard for measuring PC PSU noise as Formfactor Organization's specification like this below ?

[VK3DRB]

Depending upon why you are doing the measurement and what it is you are measuring, sometimes you want to see the >20MHz noise especially if it is for a design that has to go through EMC compliance testing. A spectrum analyser and a near field test set and an LPDA in an anechoic chamber are very useful to have around, albeit spectrum analysers are damned expensive. It used to be the world was only interested in emissions above 30MHz, but these days, it is almost DC to daylight. Which is good because of the amount of HF EMI in Melbourne cities is now beyond the joke - a lot of it due to illegal imports from China.

[Kleinstein]

To evaluate this power supply the "screwed up measurement" is quite of some worth.

The high frequency noise above 10Mhz is too high for a good lab supply. Simple filtering could improve it a lot.

Not sure if even the improved probing technique is good for >20MHz. The switching pulses have a lot of energy in harmonics. You'd need to make sure that your probing setup has a flat response over the whole bandwidth you're interested in. Otherwise you'll just hit another resonance somewhere higher up and be fooled again.

The 50 ohms termination at the input side is not that bad, it is at least better than using one of the normal times 1 probes. There is an upper limit to this, as the oscilloscope side of the coax is no longer high impedance when the frequency goes up. So higher frequencies (e.g. beyond some 10-20 MHz) get gradually attenuated. When the scope input approaches 50 Ohms this would be 50% attenuation. The problem is less with seeing additional resonances, more with reduced (up to 50%) amplitude at higher frequencies.