Hardware improvement opportunity


I think there is a simple opportunity to save 0.4 Watt in the 10A design. Not a very big deal but still…

KR Stefan

Which is the simple opportunity?

For the tracking you do not need a calibrated current measurement. So instead of having the shunt you can take the difference of the lowpass filtered switch node signal and the output.

KR stefan

You mean current measurement via inductor resistance, e.g. as described in this application note?

I’m not sure how accurate it will get. And as I would like to get a fairly accurate calculation of the solar energy input, I chose a shunt-based current measurement.

No I was thinking simpler:

Resolution is compatible to a shunt measurement. I would say sufficient accurate for an indication of the solar power. A very accurate number is not very useful imho because illumination conditions are changing all the time so what is the point of a calibrated measurement. For the battery I agree that a calibrated measurement would be better to have an accurate energy storage measurement function.

Very interesting approach, thanks for sharing. I will include it in my LTSpice simulation to check how it works and which values for the filter RC are suitable.

Regarding calibrated measurement: Currently, the 2 shunts are located such that solar input current and load output current can be distinguished. The difference between solar and load is the battery current. I agree it would be possible to reduce to only one bi-directional shunt measuring battery input and output. However, you would not be able to “see” energy flow that goes directly from the solar input to the load. So you don’t get a proper production + consumption measurement (or only un-calibrated assumption).

Maybe with calibration of each PCB your method might get quite accurate. Would need some testing I guess…

Quick question wrt simulation. I was wondering why you are using an LM5107 as the MOSFET driver, I see it has no spice model. It is not a very ‘hard’ switch driver but I guess with frequencies under 100 kHz it is ok. The design looks quite nice.

The LM5107 is quite cheap and I found the driving current to be sufficient for the application. The faster you switch, the more difficult it gets to handle electromagnetic radiation.

In the simulation I used a similar driver (LTC4444) which exists in LTSpice library.

Yes, price is certainly an advantage. I agree radiated emission increases but also efficiency but since frequencies are low it may not be an issue. For radiated emission low L in the high dI/dt loop helps and small loop area. You have MOSFET packages that enable this. If you want I can take a look at the layout. Btw do you know if you have much ringing on the switch node? MOSFET drivers can have reliability issues with negative voltages on the switch node.

It seems like above method for current sensing is patented by Genasun: https://patents.google.com/patent/US9203300B2/en

The method described here also looks very similar: https://patents.google.com/patent/US9759750B2/en

So it’s probably not possible to use this method in an open source project before 2036. Very sad. The LTspice simulations looked kind of promising.

Wtf you got to be kidding such a simple thing. I can hardly imagine that there isn’t any prior art, there is a whole industry in power converters that need current measurements without extra losses . When I wrote down to explain I was thinking that in practice you probably have to filter also the battery voltage because the ripple voltage will be too high. A patent has only value if there is no prior art. I will look into it. I can also think of ways to go around it but lets not discuss here.
BTW very good of you that you checked IP, I didn’t think of that one.

Well amazed by you finding a patent for something so simple, I did some googling. It was not that difficult to find prior art:
https://www.eetimes.com/document.asp?doc_id=1273006# (see fig. 5)

https://pdfs.semanticscholar.org/dc10/496aa21b8322b6ccebc31df38f40993a34a0.pdf (fig. 2.11)
using an extra winding: fig. 2.10

So not a solid patent at all.
A shame that with a legal smoke screen they try to patent ohms law.