2
\$\begingroup\$

EDN have an interesting (for novices like myself!) guide on testing a power supply's design, in which they suggest a 500MHz oscilloscope as one piece of equipment necessary for "proper" testing:

The test equipment needed for proper power supply testing varies with the type of power system being tested, as well as the financial budget for the equipment. Here is an example list of equipment:

  • ...
  • Oscilloscope 500 MHz BW or greater with probe for noise measurements.
  • ...

It seems a random choice, except that it is among the very highest spec'd oscilloscope I've seen on the market.

Are they trying to say you should go for the best your budget can stretch to, or is there some technical reason why 500MHz is the magic number?

I'm guessing it might be related to typical SMPS frequencies, PCB and other parasitics, and other characteristics of current technologies, or some artifact of the nature of EMC?

Firstly it bothers me it's seemingly been plucked out if thin air. And it's weird that the article seems to be written for informing readers of sub-professional level, yet recommending top-end equipment (totalling a near 6 figure sum).

Can someone make sense of this? If that's too vague a question, then simply, why 500MHz? Maybe it's just the one sitting in front of him.

\$\endgroup\$
  • 2
    \$\begingroup\$ Seems like it's the author's opinion that noise measurements should include 0-500MHz. IMHO, this doesn't seem to be an accurate generalization; if neither your circuit or supply has significant harmonics above 100MHz (which is true of many systems), than you may be OK with a 100MHz scope. However, be aware that when qualifying/certifying equipment (i.e. if you are selling it and you need to pass FCC stuff), a 500MHz may very well be necessary for checking radiated/conducted emissions and similar. \$\endgroup\$ – uint128_t May 13 '17 at 0:25
  • 4
    \$\begingroup\$ I'm not sure I'd say a 500 MHz scope is close to the highest-spec'd scope on the market. I know Keysight makes a 63 GHz model (that costs as much as a large house!), for example. \$\endgroup\$ – Hearth May 13 '17 at 0:42
  • 2
    \$\begingroup\$ Keysight, Tek, LeCroy all have real-time scopes in the 10's of GHz nowadays; and have had ~10 GHz repetitive sampling scopes for at least 15 years. \$\endgroup\$ – The Photon May 13 '17 at 2:06
4
\$\begingroup\$

The parasitics of 0.01uH (10nH) and 10pF will ring at 500MHz.

The parasitics of 0.001uH (1nH, or 1mm of wire) and 100pF will ring at 500Mhz.

SwitchRegs have several situations where 1mm wire and 100pF co-exist as parasitics. You need to see that, and be aware.

Why? The ability of OpAmps and other ICs to reject 500MHz on their VDD pins is, basically, ZERO.

You, the power-system and power-circuit designer, have to take responsibility for identifying and quantifying and managing the risks.

EDIT Some years back, checking out a 4-channel camera interface, I found the top 1/3 of the 4096 codes went to trash for the 6 LSBs. Bringing in the Tek P6201 FET probe with 900MHz bandwidth (1970 era technology), I eventually saw a 0.01volt PP oscillation at 600MHz on the -5volt power ---- the ADC was oscillating. Cure? Install 27ohm Rdampen in the (4) ADC input paths; we gently bent up that VIN lead on the IC, and soldered a vertically-oriented SMT resistor. Somehow my 8-layer PCB, with 0.1uF bypass caps diligently placed, with 4 ADCs sharing the front/back, perhaps with ferrite-beads to isolate the VDD, was oscillating. The IC vendor stated "Never seen that happen before."

Without that 900MHz probe and 500MHz Tek7904, I'd have had to punt. Touching the ADC Vin pin with fingers caused changes in the output code-blurring, so I had a hint of weirdities.

\$\endgroup\$
3
\$\begingroup\$

If you watch the video (5:58 min) on the second page, he measures a transient noise frequency of 141MHz, which justifies the 500MHz.

The text sets up the situation, but the videos are the key to understanding why.

Engineer It - How to test power supplies - Measuring Noise

\$\endgroup\$
  • \$\begingroup\$ Well spotted. I hate video tutorials! \$\endgroup\$ – CL22 May 13 '17 at 10:18
3
\$\begingroup\$

The basic principle is that you can't fix what you cant see, so get the best eyes you can afford.

Most switch mode power supplies operate at a few MHz, but in order to maximize efficiency they have extremely fast switching transitions. This means you get signal components at much higher frequencies, and parasitic inductances and capacitances can easily form relatively high Q resonances at those frequencies.

This information is also probably targeting/assuming commercial applications. It is generally possible to over-engineer your circuit and be confident that it is going to perform. For instance, you can slow down the edge rate of your switcher to reduce harmonics, use extra shielding and filters, and so on. But that shielding adds to the cost and size, and the lowered efficiency generates more heat, which you have to dissipate, and also hurts you in the market. And some times you are wrong -- a poorly chosen filter component can have parasitic element that totally eliminates its effectiveness.

Another big factor is whether your circuit needs to be pass EMI certification. If it does, and you run a pre-qualification test and find out the circuit is radiating at 300 MHz, you need an instrument that can see 300 MHz. So in that way of looking at it, the answer is "if you care about how this circuit operates at frequency X, you better have a scope that can measure X"

One last point is that all the kids these days are into high speed digital signals where the oscilloscope is king and they completely forget the awesomeness that is the spectrum analyzer. A used 2-4 GHz spectrum analyzer is pretty cheap and for noise measurements its dynamic range will blow the doors off any 500 MHz oscilloscope.

\$\endgroup\$
  • \$\begingroup\$ +1 especially for spectrum analyser. But don't almost all oscilloscopes have FFT these days? Tho obviously limited by its bandwidth. Besides that and lower cost of an older model, any reason why you'd suggest a dedicated SA? \$\endgroup\$ – CL22 May 13 '17 at 10:17
  • 2
    \$\begingroup\$ The only people who takes scope FFTs seriously as a substitute for a real spectrum analyser have never used a real spectrum analyser. The issue is dynamic range, even a cheap SA will see two signals 100kHz, and 70dB apart in level, you are not going to do that is reasonable time with an FFT driven by an 8 bit ADC... Most 'scopes have an 8 bit ADC which implies a -48dBFS broadband noise floor, my old SA can see a weak signal over 90dB down on a strong one at the same time and can do it without massive amounts of averaging. You need different tools for time and frequency domain measurements. \$\endgroup\$ – Dan Mills May 13 '17 at 13:38
  • \$\begingroup\$ I addition to @DanMills comment about dynamic range, the ADC of most scopes is not optimized for low distortion so it is hard to tell harmonics in your signal from those introduced by the scope, and since they operate at base band, they typically don't have enough memory (or FFT size) to get good spectral resolution for narrow peaks. \$\endgroup\$ – Evan May 13 '17 at 16:49

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.