Blog Post Actions
Group Actions
Engagement
  • Author Author: jc2048
  • Date Created: Date Created
  • Views 426 views
  • Likes 4 likes
  • Comments 5 comments
Related
Recommended

TPS54A20EVM evaluation - part 5

jc2048

This is part five of evaluating a TPS54A20 on an evaluation board. The TPS54A20 is a buck converter with

an interesting topology - it has two phases merged with a switched-capacitor. This isn't a formal road

test (I was given the board by , nice person that he is) and I'm just doing what interests me and

blogging about it. Disclaimer: some of what I do here is quite sloppy and, to be fair to TI, it shouldn't

be taken as a proper review.

 

This time I'm looking at the operation of the converter and specifically the role of the capacitor. See

this link for the discussion I had with Jan about the operation of the part - I won't repeat the details

here.

 

https://www.element14.com/community/groups/power-management/blog/2016/10/23/low-voltage-step-down-converter-tps54a20-series-capacitor

 

Here are the waveforms at each end of the capacitor with a load of about 9.5A. The low side (blue trace)

switches between 0 and 6 volts and the high side (yellow trace) between 6 and 12V. The input voltage is

12V. These traces were taken with the probes using the flying-lead ground and the pieces of wire that Jan

had soldered onto the ends of the capacitor.

 

image

 

The voltage across the capacitor is the difference between the two and you can see that on average it's

around 6 volts. After the trigger point, where both waveforms are high, you can see that the capacitor is

charging - the traces are diverging so the capacitor voltage is increasing. There is then a section where

the traces remain horizontal, followed by a third section where the traces close up - the capacitor is

discharging here and powering the other phase of the converter.

 

Here is the high part in more detail. (An obvious question is where does all the ringing come from?)

 

image

 

And here is the discharge section in more detail.

 

image

 

Here's the high part again but with spring clips for the ground rather than the flying lead (but Jan's

wires are still attached). The ringing is still there, but it's much more regular now. There are about 7

cycles in 40nS, so that's a frequency of about 175MHz. What I'm now struggling with is, how much is that

down to my probes and how much down to the circuit?

 

image

 

At this point I got diverted a bit and took a capture with the persistance set to infinity (though, needless to say, I didn't

wait for an infinite amount of time before stopping it - this is about 5 seconds worth, which will be a

few thousand captures). This shows us the jitter on the width which looks quite bad until you look at the

timebase and realise that it's not much more than 10nS.

 

image

 

This final trace is with Jan's wires removed, so is the best I can do with my passive probes.

 

image

 

The spring-clip grounds don't eliminate the lead inductance, they just reduce it compared to the flying

leads and push up the frequency at which we'll see ringing if the probe is hit by a fast edge [the

frequency is largely defined by the inductance and the various capacitances at the probe tip which form a

tuned circuit]. But what I can see here is far in excess of what I'd see probing fast logic, so it has to be

real. Unfortunately, it's also a prime EMC hazard - 175MHz is just nice for finding its way around

and radiating off of wiring and given the kinds of currents involved there's plenty of power.

 

Here's what I pick up with an informal and rather crude near-field probe (it consists of loop made from

the flying earth lead shorted to the probe tip)

 

 

image

 

image

 

It doesn't tell us anything about the levels (no calibration and you can't simply extrapolate from a

near-field measurement of the magnetic field to what happens further out anyway) but it does show that

there is something there. Shame I haven't saved up enough money for a spectrum analyser yet.

  • in reply to Instructorman

    Thank you. Transistors next; I don't know a great deal about designing with transistors, outside of switching loads, so I thought I'd learn and blog about it.

  • Jon,

     

    I've just ready through all 5 of your blog posts on the TPS54A20EVM evaluation board.  Well done!

    I like your clear conversational style.  Your presentations are easy to digest, you provide good and useful detail, and you infuse admirable levels of honesty and humility.

     

    Keep up the great work.

    I look forward to reading your future contributions to the E14 community.

     

    Mark A

  • in reply to jc2048

    After further thought, maybe building antennas doesn't work. Wideband aerials don't have a flat response, so you need the calibration data. If you built one yourself you'd have to get someone to calibrate it for you and I imagine that would probably cost as much as buying a new one with its calibration already done. But I'm not very clear about this - maybe you could use the calibration for the antenna being copied and it would be good enough. If anyone out there is an antenna expert, I'd be interested to know how close a match you get between antennas that are mechanically identical (ie matching dimensions and matching materials).

  • in reply to Jan Cumps

    Me too. The RSA306B looks superb, but the PC required would need to be far superior to anything I currently own and just to complete the other half of a test instrument. [I can't quite remember, but I'm not sure it had all the bandwidths and the quasi-peak detector you'd want for EMC.] I prefer standalone instruments - they are complete and do the job [not sure that's really true anymore, though, as my Tektronix scope is awful: it takes longer to boot up than an old CRT scope did to warm up the tube, the UI is quite frustrating (pan control that works backwards, anyone?), the menus are a muddle, it's so slow it's difficult to position the traces, and the marketing department cut down the features so much that it's quite mean-spirited and rather than encouraging me to buy a better one from them it will just encourage me to look at the competition]. It seems to me that a PC is a real liability; most PC hardware is designed to have a lifetime of only a few years, you have the problem of the OS steadily morphing into something else underneath, and there are almost always problems with licensing at some point (usually when the hard disk dies and you somehow need to get it all working again).

     

    If you have time and practical skills you could easily build antennas (just simple metalwork - being able to drill and tap holes and that kind of thing). Just get someone here who has access to one to give you the measurements and a photo so you can see the construction details. The other possibility, if you just want to look at the emissions from fairly small boards is a TEM cell [effectively, it's a giant piece of 50 ohm co-ax with open sides]. Working with near-field probes is useful too [essential really]to pin-point sources and how the signals run around the wiring and escape from apertures, so you can do a lot without antennas.

     

    The important thing here is to keep the noise away from anything that's of a dimension that it can act as a aerial. A quarter wave at 175Mhz is a bit under half a metre. A tenth of that will still radiate, though obviously not as well as the quarter-wave, and that's only 5 or 6 cm. If the board layout is good (and it had better be, with a 10A converter switching at 2MHz), you won't have current loops of that scale, so most of the worry is radiation off of longer traces and wiring, particularly the power distribution to the board. If you have a tin can it only works if you filter everything going in and out, otherwise the emissions just run out via the connections. In this case the filtering might be enough as long as it's close in to the converter. Either that or investigate ways to reduce the noise in the first place, which is usually better if you can do it - though here it may be inherent in the topology with the capcitor and the inductors (designing out emissions is normally better than trying to contain them).

  • There as a nice roadtest for an EMI measure device.

    RoadTest the TEKTRONIX  RSA306B

    I didn't enroll for that one for two reasons: I don't have a PC that meets the requirements and I'd have to invest significant money in antennas.

    It would be great to check these switching designs though.

     

    I guess the buck is expected to be mounted within a metal can, like many switching power parts in small devices.