DCM MPPT rev3 preliminary efficiency results

This is pretty rough territory for trackers to operate in – low input voltage, high input current, huge boost ratios. Datasheets for commercially available trackers show extremely low boost ratios that are unlikely in silicon solar car arrays.

We’ll continue to try and improve upon the high boost ratio design, but what we have today is already very good. It provides quite acceptable efficiency at a very low mass. Each board has two channels and weighs 347g without its case. That makes it 3.75 times the tracker-density of Drivetek trackers. There are still some losses to wring out of the inductor and the output diode.

11 Comments
  • Avatar
    Sam Lenius
    Posted at 19:26h, 24 September

    Hey Sasha,

    I was curious what your operating conditions were. Specifically output voltage and input current.

    Thanks
    -Sam

    • Avatar
      Sasha Zbrozek
      Posted at 00:08h, 27 September

      The code on the trackers was self regulating input current to 5.44A. Output voltage was 150.1v in all cases.

      I am preparing a more informative contour map of efficiencies. One team was interested in a specific data point at higher input power, resulting in:

      6.03A @ 51.455v on the input = 310.27w
      2.539A @ 118.1v on the output = 299.86w

      96.64%

  • Avatar
    Sam Lenius
    Posted at 18:28h, 27 September

    Thanks for the info Sasha. I took some data today for comparison to the new rev I built up this past weekend.

    I made some plots from the data you gave me:
    http://imgur.com/0b3hD.png

    I think your design is superior for very high boost ratios, however across the board our efficiencies are right in line. My constant frequency powdered inductor design may have the edge at low boost ratios however.

    For reference/comparison,
    6.13A @ 50.5378v on the input = 310.077w
    2.577A @ 118.115v on the output = 304.469w

    98.2%

    I find it really interesting how perfectly linear the efficiency falloff is versus boost ratio. Makes it easy to predict tracker efficiency for a given array/battery design.

    I’m going to try to produce some of these plots with the drivetek and AERL trackers that I have on hand so we can see how we really stack up.

  • Avatar
    Sam Lenius
    Posted at 18:29h, 27 September

    Forgot to include this plot:
    http://imgur.com/diDtL.png

  • Avatar
    Sasha Zbrozek
    Posted at 21:49h, 27 September

    I’m curious how you manage to avoid having very large reverse recovery losses in your output rectifier at extreme duty cycles. After my GaAs oriented SEPIC design, I moved to a CCM boost with a powdered MPP core – very similar to your design. I too saw very high efficiencies at low boost ratios, but couldn’t get as good a performance as you are at extreme ratios. In particular, at ratios >12, efficiency dropped to about 85% and the trackers were breaking quiet on my ham radio from half a mile away at 440MHz.

    The DCM design definitely has an advantage for very high boost ratios – that was the whole premise behind its construction. Boosting from 10v to 150v at 6A is about 94.1%, noticeably above the CCM curve in that operating region. DCM’s big disadvantage is the additional loss in the inductor. They’re critical conduction mode trackers, so the peak flux in the inductor is twice the average and minimum flux is zero. Increasing input voltage and increasing output voltage both increase the frequency (and thus core losses) but also push up the power. The result is an extremely flat efficiency vs. boost ratio curve. In the CCM converter inductor losses are more or less fixed by the input current.

    I suspect that I’ve oversized my FET a little bit and that I can reduce overall losses a little bit by moving to something with a lower C at the expense of R. I have a game of musical semiconductors to play before I’ve settled on the “finished” arrangement for revision 3.

  • Avatar
    Sam Lenius
    Posted at 23:24h, 27 September

    The answer is model based design. I have a very powerful and empirically confirmed to be accurate model of my tracker. I’ve accounted for every theoretical loss that I can, and I’ve used direct measurement to tune it. My first revisions did have low efficiency at high boost ratios, but by careful selection of components, I was able to significantly improve it.

    Additionally, I had significant conducted EMI issues like you mention. Mine would swamp out the nearby FM receiver. It took a helluva lot of work to fix. Since it’s hard switched, EMI is a huge concern. This board is the 5th generation of this design, and the 3rd complete relayout. Minimizing stray inductance in the switch node and adding additional conducted EMI filtering has helped a tremendous amount. The conducted noise is just barely triggerable now – around FCC Class A levels.

    I was inspired by your mention of a contour map, so I made my own. Check it out:
    (X’s are the measured datapoints)
    http://i.imgur.com/yy4QC.png

    I tried a 3D graph as well, but it’s pretty much impossible to interpret.
    http://i.imgur.com/wwyoL.png

    I might be able to tune my inductor better….it looks like my current design may be optimized for 4A. Theres always more fiddling to do.

  • Avatar
    Tim Gamber
    Posted at 11:38h, 03 October

    Hey Sam and Sasha,

    Fantastic work on the trackers!

    I have also started work on our very own tracker and it’s a simple diode rectified design like Sasha’s. It looks like so far I’ve been getting 93% conversion efficiency at ~12V in and ~110V out (at ~60W input) so approximately a boost factor of 9, which is definitely pretty respectable considering I built it three days ago hehe.

    I will definitely post up some efficiency plots and more info when I think I have some usable and accurate results.

    Also at the moment the tracker is powered off a separate 12V supply. I was just wondering what you guys think about this. I thought it made sense because then you can easily control when the tracker is on/off and you can also have one main DC-DC somewhere on your car instead of many.

    Well anyway it sounds like you guys have some solid trackers.

    Keep up the good work!

    See Ya

    • Avatar
      Sasha Zbrozek
      Posted at 23:31h, 07 October

      My first version tracker used the car’s CAN supply, but I found it was a bit annoying during testing. I didn’t have a reliable CAN power + data test platform and so I was constantly powering my trackers with soldered-on leads going to a bench supply. Further, it makes them less useful for doing things other than solar cars. My dad, for example, has expressed some interest in leaving one connected to a bank of lead-acid batteries for household backup.

      Further, you have a much larger stackup of efficiencies. Converting once, going to a pack, converting once or twice more, going to a tracker over a long wire, yadda yadda… it’s a bit circuitous.

      One caveat of the self-powered tracker design: be careful about the Isc checks. There are plenty of ways to make a tracker work under such circumstances. I diode-OR the cells from two channels into a local switching supply and am careful to do IV sweeps on one channel at a time. Should there not be a second supply, sweeps only sweep until they hit a voltage threshold that’s above what it takes to keep the switcher happy.

      Are you going with a CCM or a DCM optimized topology?

  • Avatar
    Tim Gamber
    Posted at 00:51h, 08 October

    Hey Sasha,

    You make a good point about the self powered tracker being useful for other things. For now though I have been powering the tracker from a few 9V batts we have laying around so its reasonably convenient. I guess the overall efficiency will be slightly higher on the car then now because of the supply voltage difference but I tested it out and found it is a fairly small change in eff.

    I ran some calcs the other day and after swapping out a component or two and changing the switching frequency I should be able to bump the tracker up to a little over 96% eff under the conditions I mentioned earlier. I guess I’ll just have to wait and see until I actually change out the components and test it out for real to see if my calcs were correct. I don’t want to go around making any false claims. I’ll let you know how it goes.

    For now it’s running CCM but I was thinking that maybe I could have it actively switch between CCM and DCM depending on the output load and/or boost factor. This isn’t a definite thing but more so an interesting idea I had.

    Just out of curiosity can your trackers actively change the switching frequency? On my tracker the switching frequency is programmable and can actively change during run time as well. Virtually any switching frequency between 1Hz and a few MHz is possible. And of course the duty cycle changes during run time as well…

    BTW how is the car build going? Ours is going pretty decent as we are prepping our top shell mold again to get it ready for the new top shell we are doing and we are also working on weight reductions on the rest of the car.

  • Avatar
    Sasha Zbrozek
    Posted at 11:17h, 08 October

    It gets pretty hard to do efficiency calculations as efficiency starts getting close to 100% – small measurement errors start to become large fractions of your losses.

    I just bought some new higher accuracy test equipment (come on UPS!!) and am writing some software to partially automate efficiency mapping.

    The DCM tracker would more accurately be called a critical conduction mode tracker – it varies its frequency to ensure minimal dead time on the inductor. That limits the peak current and thus the peak flux and the losses while maintaining the benefit for the output rectifier diode. It varies between 10KHz and 180KHz, depending on input I/V and output V.

    I’ve considered switching between DCM and CCM, but I would need a significantly higher inductance to make CCM efficient. One possibility is to have a second FET and another tap on my inductor, so I can choose how many windings I want to use. Of course, that’s more board area and more capacitive load.

    As for the car, it’s going pretty well. Our new molds are being cut as I type this and should be ready for shell making by Thanksgiving. Electronics-wise we’re in pretty good shape, everything is just refinements from the last car and we have a great deal of designer continuity. Our biggest stumbling block at the moment is sourcing a CSIRO motor. We’ve finished the new design under the assumption that we could get one and now we’re having some trouble with that.

  • Avatar
    Tim Gamber
    Posted at 13:32h, 08 October

    Hi Sasha,

    Sounds like you still have tons of ideas for different tracker setups. Switching in between CCM and DCM would be pretty crazy. For now though I think I’ll work on the tracking algorithm and the power conversion efficiency in CCM.

    If you guys can get the CSIRO motor running definitely go for it. The nicest thing of all is being able to talk to your driver over the driver radio without hearing that ridiculous rumble from the NGM.

    We got our CSIRO motors way back in 2004 I think, and we didn’t get them running until this last race. I’m actually not sure why the team before us didn’t try to get it running. Maybe it was just due to the usual time crunch. Anyway I’m not sure who or how we sourced the kits but I can look into it if you like.

    Also if you do get a CSIRO don’t trust the data sheet. If you have already designed your motor casing for it based on the data sheet I can pretty much guarantee you it won’t work. Both of the stators we have are a fair bit different then the dimensions and tolerances specified in the data sheet. So much so in fact that our first motor casing caused the stator and rotor to rub together (we trusted the data sheet).