Open Source MPPT

[whiskthecat]

Okay I have changed out some parts and am now sitting at about $50 of components for specifications of 60v @ 2x30Amps or 60v @ 60Amps. What do you guys think is a fair market price for a finished unit of those specs so I have a general idea how much money is left to work with?

[thomasow]

whiskthecat / All:

Just stumbled across this posting, and turns out I too am working on an open-source MPPT charge controller. Settled on a distributed system where one can use controllers as needed to support larger installs. A couple of the basic design parameters are: 85V Panel max, 25A output max. And have been targeting 12v .. 48v batteries.

Though I like PIC, I selected the Arduino IDE as it is simple for folks to use, and am currently debugging the hardware / porting of the IDE.

Smart MPPT Solar Controller

BOM cost are around $75, and will come down as I pull out redundant 'debug stuff' and/or depending on stuffing options one selects. (e.g., integrated USB, CAN ports..)

btw: This is the a continuation of a series of open source charging projects I have mapped out. Already have an external Alternator Regulator, and DC Generator controller. After this MPPT will look to integration via the CAN for coordinating charging between multiple sources, as well as installation cabling reduction.


Perhaps we can team up?

-al-

[whiskthecat]

That'd be awesome Thomas. I'm kind of dreading writing the code as the software side of things has never been my favorite part. I've still been looking at different designs this past week and an even better deal has come up locally for the 20% efficiency 100v 435watt Sunpower panels and I don't think I'll be able to pass them up.

Being stuck with such a high open circuit voltage I've started on yet another BOM focusing not on costs savings but absolute performance. In doing so I've come across quite a few cool ideas. I'm looking into using Gallium Nitride FETs to increase efficiency. They have much quicker switching times and lower on resistance than traditional silicon fets. Still using a dual phase design to distribute power dissipation and keep current down, the higher frequency allowed by the GaN fets can likely make up for their increased costs by reducing the size of the inductors and capacitors. I'd also like to implement zero voltage switching to eliminate turn on losses and any possible reverse recovery. I'm not aware of any commercial controller that has any of these 3 advanced features and it should be a respectable bump in overall efficiency allowing small size and minimal heatsinking.

If anyone has a link to a teardown of a higher end controller please post it.

[nimblemotors]

What I've been working on requires no software, and is 'open source', because it is all done by hardware and the chip maker provides all the circuit layouts, etc. The cost of the board is probably $20-$30 in qty.
I'm talking about using the SPV1020. It seems you just didn't get it.

Here is the app note for it. http://www.st.com/st-web-ui/static/a...DM00026751.pdf

The circuit board with the SPV1020 is put right at the solar panel.
It takes the panel input, does a MPPT boost to a fix voltage.
There are no 'home runs'. All panels will output the same voltage, say 30 volts,
and their outputs connect to a single 30v two wire BUS.
The amount of CURRENT varies. If one panel is shaded, it just contributes less current onto the BUS.

The BUS wires go into the interior, where you can a BATTERY CHARGER that will take the voltage and current from the solar panels and adjust them to charge a battery. It isn't a MPPT controller, it would be a simple DC-DC Buck converter to convert down the 30 V to say 14.5v.
This unit needs to handle the full current of ALL the panels.
The SPV1020 board only handle the current of ONE panel, which is limited to 9amps. 9amps * 17v panel = 150 watt panel. So the SPV1020 board would work with up to a 150 watt solar panel. I think the smaller the better in terms of shading, so my view is use 50 watt panels, get as many as you want, just need a SPV1020 board for each one.

Here is the diagram, to make it crystal clear. The ISV009v1 is the SPV1020 board.

[seandepagnier]

This really takes me back to 6 years ago when I actually built prototype open source mppt.

It worked but I needed to respin the board a few more times to make it useful.

Basically my parts cost was below $10 and could handle 30 amps 30 volts (was intended for 12v panels) It could work in buck or boost mode (by flipping input and output wires) so you could charge single cell lithium batteries off a 3v flexible panel, or it could act as a 5v power supply if you needed that, or whatever...

Now I have more experience as I have lived on a boat for over 5 years and maybe I can help with this project.

To give some background, I kept it as simple as possible with only one phase. Maybe it's noisier power but it really doesn't matter in this application. I used a small avr microprocessor and two fairchild mosfets (in some stupid package that is a pain to solder but they were the best parts to use) The idea is minimum on resistance of the mosfet (needed for efficiency at low voltages), and very low gate capacitance (needed for higher switching frequency) Then I found using a fairly large toroid inductor gave good efficiency. I didn't use electrolytic caps because you can get big enough ceramic capacitors for cheap and they are more efficient as well.

I experimented using hall-effect current sensors but found they actually used a lot of power (30mW or so) so I ended up measuring the voltage drop on the mosfet (which has a low on resistance so works like a shunt) If you install a separate shunt it will waste some power.

I thought it would be possible with just voltage feedback from the battery. You basically vary the duty cycle in 3 different places, and perform a parabolic curve fit, and you can find which way to shift and how far really easily. If you are right on the mark, you can sleep for a longer time before waking up to try again. Indeed this works if there is not a varying load on the battery, but not well when the load is changing rapidly if you want to constantly produce maximum power.

Instead I believe it is best to measure current as well as panel voltage, battery voltage, and possibly temperature as well. Over time you will learn the correct duty cycle and frequencies to use (you can get away with lower frequencies when the current is lower to reduce switching losses) and there are also times when it's just simply more efficient to stay on 100% of the time without switching.

I was using geda and pcb for schematics and layout, but maybe kicad is good now? Can you post a link to the repository?

Please do not consider ideas like 2.4ghz wireless etc... it would completely defeat the purpose of mppt which is to have maximum efficiency and not waste power. If you really get it working and are bored, consider pwm output to a winch servo (like for remote control sailboats) that is wired up to rotate the panel.

Maybe to save on cost, instead of an lcd screen, it could just have a usb cable which can relay statistics but more importantly configure battery chemistry and number of cells as well as control of equalization. Otherwise maybe 2 buttons and 3 leds is enough.

Finally, if you want to cater to the boat market, you should make the controller completely waterproof and submersible. This will allow it to last many years as corrosion will not get into it and destroy the components.

A lot of the cheap chinese mppt controllers actually put less amps into the battery than directly connecting the solar panel because the controller itself consumes so much, and doesn't properly track peak power anyway.

Originally Posted by travellerw

Many cruisers run panels in series due to the layout of their boats..

A bad on most sailboats because of partial shade. Lots of small panels in parallel is best. It is also best to have several small mppt controllers which are all very efficient so each panel can run at it's peak power point (which is a different voltage from other panels if it is at a different angle to the sun or partially shaded)

[whiskthecat]

I still don't understand nimblemotors. You are taking panels that are 18+v, boosting that to 30v, and then stepping that down to battery voltage. Why add an additional conversion when you can just step the initial panel voltage down? It can't be more efficient to have an additional conversion and considering that chip costs 10$ it likely won't be cheaper either once you outfit it with all the components it needs and then you still have to provide the buck stage that could have done the entire job by itself. I assume the 30v DC bus provided by the chips must sag down once the battery charger starts pulling on it and the IC's throttle back when doing MPPT to each panel. It's certainly a very interesting setup and has me thinking. What happens in low light conditions when the panels are supplying much less than the charger would like? Conventional MPPT charge controllers work by varying the apparent load of the battery to the panels. I see no way for your proposed setup to do this so it would seem like the boost section would be effectively bypassed.

Interesting stuff alexandra. I've only ever used Eagle for laying out PCBs as it seemed to have the most resources available as far as online documentation goes. Modern 2.4ghz wireless uses very little power. The last application I used it in had an update frequency of 60hz moving 64bits of data each transmission with full ACK and the battery life of a single cr2032 watch cell was approaching 300 days. For this application you'd likely have an update rate closer to 1hz and only moving data when you had to. The addition to the overall losses of the system would be less than negligible.

If I can find some help with the programming I see no reason not to include fancy features such as data logging and fully modifiable configuration via USB, buttons, or 2.4ghz. As for surviving in a marine environment I haven't put a whole lot of thought into that yet but with the levels of power dissipation I am expecting it will likely be possible to encase the entire device in thermally conductive epoxy with just the heatsink exposed. This sucks for serviceability of course but it might eliminate the need to anyways.

Yea I really like the idea of multiple small panels but the dollars per watt price and the surface area efficiency of the larger panels is quite attractive if you can mount them somewhere with minimal shading.

[seandepagnier]

Originally Posted by whiskthecat

Modern 2.4ghz wireless uses very little power. The last application I used it in had an update frequency of 60hz moving 64bits of data each transmission with full ACK and the battery life of a single cr2032 watch cell was approaching 300 days.

If you can keep the power consumption low enough then ok! I would say if it's less than 10% of what the controller is using.

Basically my avr was using 30mW but was sleeping 95% of the time using < 1mW, but driving the gates of the mosfets used quite a bit of power maybe 50mW or so, I cannot remember exact figures it's been a while.

Basically it would need at least a 3 watt solar panel before the mppt gain could cover the controler power consumption, and then only in good conditions.

If I can find some help with the programming I see no reason not to include fancy features such as data logging and fully modifiable configuration via USB, buttons, or 2.4ghz.

Except to do this you need higher power and higher cost micro. To make it easy, sure run something with linux then it's no sweat but the power consumption and cost is considerable. To do it on a lower power micro is more work, but you can probably find some small arms with drivers already written.

As for surviving in a marine environment I haven't put a whole lot of thought into that yet but with the levels of power dissipation I am expecting it will likely be possible to encase the entire device in thermally conductive epoxy with just the heatsink exposed. This sucks for serviceability of course but it might eliminate the need to anyways.

Maybe it can go in an annodized aluminum box

[whiskthecat]

The power usage of the radio would be in the single digit microwatt range I assure you. It uses 1uA of current when sleeping which is what it will be doing the majority of the time. If you look at my earlier post in this thread you will see my micro controller of preference. It's a $1 ARM chip that is quite beefy. I certainly don't want to overcomplicate things to the point of having full on linux being used.

[thomasow]

Hi,

Lots of conversations these past couple of days - some random thoughts and comments:

I get nimblemotors approach - and like that it places a focused MPPT engine at each (or small group of) solar panels to optimize to those conditions. We all know a major differences between marina install vs. land based is the probability of partial shading; and that that condition changes over time. - Hence the one controller per panel seems a promising approach. But I do not see the SPV1020 as a solution directly to the battery - it provides an intermittent voltage for later conversion. Perhaps makes sense for grid-tied offerings, but I think for our DC charging solution this would be an extra step - as well as requiring perhaps a very large (high amp) 2nd stage which I think is perhaps beyond the scope of these types of efforts.

Adding a simple LCD display would not involve lots of code - just matters what one wants. And of course makes the case / waterproofing a bit more challenging. But from a coding stand point - should be a SMOP (And Wiskercat - am happy to lend a hand out on that front).

Wiskercat: A question on the GaN parts (Which is nice to see a growing field) - true they allow higher operating frequencies, with smaller caps and inductors, but from what I have seen higher frequencies also tends to reduce overall efficiency a few % points. Any insight?

And back to programming: I have a core of ASCII data status output, as well as ASCII based configuration options already coded up for my alternator regulator. This would not be too hard to port to an MPPT controller (and what I am looking to do with my MPPT project). Though I noted the uC linked to previously lacks any EEPROM, so would need some additional hardware if one wanted to be able to retain configuration changes.

For IDE's, I have been focusing on the Arduino IDE - mostly because it is very simple for folks to install and use. Even though other (and arguable better) IDEs can be had for low cost - needing to purchase even a $40 programming tool, as well as the added complexity of even simple IDEs can in my experience be a barrier. Does restrict uC selections a bit (e.g., the whole PIC family) – but for these type of open-sources projects I think the software access is important. Now: what is interesting is the ability to add additional uC support into the Arduino IDE. And even more interesting is the talk about an as of yet unreleased Arduino Zero - which uses an ARM Core M0 based uC... So if you want to use the EFM32ZG108 uC, it might be possible to port it to the Arduino IDE…

FWIW: I like the approach of paralleled low cost controllers – mostly to address the changing shading issue with boats. Support the multi-phase converter approach, to reduce cap size, as well as allow for off-the shelve inductors to be used – do wonder about the option to configure as two single converters if that will end up driving up cap size needs (being that each side would then be a single phase). Think it would be best to keep the MPPT controller focused on one battery output – with a different device addressing multi battery bank needs (as that solution will be able to also provide the same benefit with other charging sources – shore, alternator, etc..). And think there will be an opportunity to utilize module communications – though wonder if RF is the correct approach as it opens up not only association / configuration issues, but also issues around isolation and security (Be a bummer of your slip mate had a like system but used 24v batteries )

-al-

[OceanSeaSpray]

Interesting. Some time ago I actually bought a bunch of components to build a MPPT buck converter, but I have been busy with other things since.

High-efficiency requires:
1/ Very low intrinsic consumption or the result will be worse than a simple switching regulator in low light with the circuit eating up most of the output...
2/ Low frequency switching
3/ Ultra-low resistance

...and this all leads to large inductors, MOSFETs and larger capacitive filtering and higher cost. With losses going up with RxI^2, it also means that making high-current single-phase units becomes more costly and more difficult quite quickly.
Multiphase is an optimised way of operating multiple cheaper smaller single-phase units in parallel. Splitting the input into several sources and using smaller simple single-phase converters instead comes to mind as some already pointed out

I had initially sketched a board with the power stage and sensing and I was thinking about driving it off something like an Arduino. In terms of cost, it makes far more sense to directly embed a MCU, but then it is harder to play with for common hobbyists. Hobbyists, however, are a small subset in itself and I think it would be going down the wrong pathway.

Nothing done and since I started thinking about this, MPPT has become cheaper, and often too cheap.
If you want to have a peek into a pretty solid unit, have a look at: http://www.roguepowertech.com/docume...Difference.pdf. It is not even that expensive from memory.

I also had apart a little Genasun 10A MPPT unit not long ago, overpriced for what it is, but looking quite decent and essentially producing no heat at all. Making a really affordable top-quality small controller with no bells and whistles whatsoever often feels like the most sensible pathway to me. Simple, cheap and scalable, just add more little controllers as needed, they could even be made waterproof and installed almost at the panels.

Basically I still like the idea of knocking up my own MPPT converter, but justifying it is not always obvious and I certainly wouldn't consider a large, heavy-current or high-voltage design. A high-voltage panel is no better than connecting two or more lower voltage ones in series and the only sensible configuration on a boat is parallel as far as I am concerned. Too much shading no matter what for series-connected/HV systems.
The first concern should always be trying to almost match the array voltage at Pmax to the system bulk charging voltage and then most of the MPPT justification/hype goes out the window anyway.

My panels have Pmax around 17.2V and I bulk charge from 13.2V up to 14.2V max, so I have a little bit to reclaim there as long as waste is kept to an absolute minimum. With MPPT, there is a point in low light when it is no longer worth switching; just turn the upper MOSFET on and flow the DC through the inductor with everything else in sleep mode most of the time. A lot of crappy designs can't do that because they need the PWM switching to generate the higher gate voltage to keep the top N-channel FET on. They are focused on best efficiency at full output. Efficiency at low output is often much more essential when living aboard. It is a lot more practical to make a converter that can keep switching and gaining in low light if it is not designed for heavy current in the first place.

[whiskthecat]

I do not see the SPV1020 as a solution directly to the battery - it provides an intermittent voltage for later conversion. Perhaps makes sense for grid-tied offerings, but I think for our DC charging solution this would be an extra step

Exactly what I was getting at. I'm all for individual MPPT for every panel, but I don't think that's the chip to do it with.

A question on the GaN parts (Which is nice to see a growing field) - true they allow higher operating frequencies, with smaller caps and inductors, but from what I have seen higher frequencies also tends to reduce overall efficiency a few % points. Any insight?

Okay so the way these GaN fets work better is by having a smaller input capacitance so that they turn on and off faster than a traditional mosfet would. This turn on and off time is where your switching losses come from. In the nanoseconds it takes for the FET to go from being very high resistance to very low resistance there is current flowing through a non zero resistance and therefore wasting power. The more often you switch the more often you are in this transitional power wasting state. So the GaN fets allow you to operate at a higher frequency with equal or less switching losses than a traditional mosfet at a lower frequency because they switch faster at all frequencies. Of course the trade off here is that GaN fets are much more expensive, but once we account for the reduction in inductance and capacitance they provide it's not so bad.

hough I noted the uC linked to previously lacks any EEPROM, so would need some additional hardware if one wanted to be able to retain configuration changes.

These uC's allow you to write to the flash memory using the CPU itself. So however much flash you have minus the size of your program is your data storage space. We shouldn't need much to hold a few config values.

1/ Very low intrinsic consumption or the result will be worse than a simple switching regulator in low light with the circuit eating up most of the output...
2/ Low frequency switching
3/ Ultra-low resistance

...and this all leads to large inductors, MOSFETs and larger capacitive filtering and higher cost.

Back to GaN fets again. They solve multiple problems at once here. With traditional mosfets you have #2 and #1+3 competing against each other. If you want a fast switching fet, it won't be the lowest on resistance. If you want low on resistance, it won't switch that fast. The compromise case is often to choose the low resistance fet and drive it hard, which leads us to giving in to #1 and wasting power to drive the fets with high current. GaN fets have lower on resistance and faster switching speed than traditional mosfets, allowing you to operate at a higher frequency while still having ultra low resistance. In fact Google is running a competition right now to design solar charge controllers for large arrays using these new GaN devices to shrink the overall size of the controller.

Hobbyists, however, are a small subset in itself and I think it would be going down the wrong pathway.

This. Just because I want the controller to be open source doesn't mean that I want to limit it's design in any way just to try and increase its access to tinkering about.

Nothing done and since I started thinking about this, MPPT has become cheaper, and often too cheap.
If you want to have a peek into a pretty solid unit, have a look at: http://www.roguepowertech.com/docume...Difference.pdf. It is not even that expensive from memory.

That controller is single phase and the price I've found for it is $400. I think it can be beaten by quite a margin.

A high-voltage panel is no better than connecting two or more lower voltage ones in series and the only sensible configuration on a boat is parallel as far as I am concerned. Too much shading no matter what for series-connected/HV systems.

I was under this impression as well. But in my neverending quest to try and convince myself to buy/not-buy these 100v sunpower panels

I stumbled across this video

. Therefore I'd like to make the controller support up to 100v. It'll keep the people who insist on running panels in series for whatever reason happy as well.

The first concern should always be trying to almost match the array voltage at Pmax to the system bulk charging voltage and then most of the MPPT justification/hype goes out the window anyway.

The problem with this is in tropical climates (where the nice sailing is IMO) your panels are going to heat up and drop voltage so the margin you need over bank voltage is going to be large enough to justify the MPPT. I have given up on using the 17v renogy panels because after running all the numbers on them their performance is quite upsetting above 85 degrees F. You would need to run them in series (elminates the shading resistance I wanted from the small panels in the first place) or get 24v panels (not as cheap). Interestingly enough, for a dual phase design it would actually be optimal to have 2x the panel voltage as the bank voltage since you are switching at a 50% duty cycle then basically eliminating all ripple current.

[OceanSeaSpray]

In retrospect, I do indeed remember seeing a price like $369 for that Rogue controller now, spot on. I also suspect it to have too high an intrinsic consumption for my liking due to the number of "features" available. There is a lot of stuff on that board.

Most useful panels for 12VDC systems have 33 or 36 cells in series and while heating always degrades output voltage, that heat is associated with intense irradiation and therefore good power anyway. 36-cell panels usually have VOC=22V and Vmp=17-18V, so they do have margin for heating from there down to 14V. As you said, it just further negates the usefulness of MPPT in higher temperatures, but MPPT doesn't offset the heating effects regardless of panel specs: power not produced cannot be recovered afterwards of course. There is no better situation than getting maximum power without MPPT after all.
This is why ultra-low intrinsic consumption is really important in my view. The gain from MPPT is most valuable when power is precisely not plentiful.

Interesting info about these 100V panels. They basically have a bypass diode for each cell instead of a few only across "strings". This does speak for supporting higher input voltage for sure and the panel wiring would be smaller in section.

GaN FETs are a great find indeed. rDSon goes up quite quickly as rated VDS increases, so there will be a trade-off somewhere. The best ones seem to be relatively low voltage.

At the time I was thinking 15-20A would be a good figure for a simple black-box converter. The idea was building something that would always run cool without any external heatsinking.
I didn't look too closely inside the little Genasun unit, but it had three D2PAK MOSFETs. Two for the synchronous switch obviously, third one might have been some kind of ideal diode arrangement. At 10A everything is easy.

[whiskthecat]

Originally Posted by OceanSeaSpray

I also suspect it to have too high an intrinsic consumption for my liking due to the number of "features" available. There is a lot of stuff on that board.

Agreed. I have no idea why there are so many IC's. You pretty much just need an MCU, a logic supply and some mosfet drivers.

Originally Posted by OceanSeaSpray

There is no better situation than getting maximum power without MPPT after all.

You'll never get maximum power consistently without MPPT. Even with small panels, dropping a single volt from the panels to the battery bank is going to cost you upwards of 5% power. More likely you will drop 3v and lose 15%. If you're MPPT is losing anywhere near 5% then there is something bad wrong with it's design (As most chinese controllers likely do considering the size of their heatsinks). The Renogy panels I ran the numbers on would actually dip below 14v in tropical conditions, this is a huge problem because once you start getting that close to the battery voltage, current can begin to drop from the internal resistance of the battery and if you do dip under bank voltage you will get a whopping 100% loss from your panels as no current will flow. Considering this I don't plan on installing anything with a Vmp less than 24v and this might be why many designs don't account for 100% duty cycle as you pointed out earlier, I likely won't either.

Originally Posted by OceanSeaSpray

GaN FETs are a great find indeed. rDSon goes up quite quickly as rated VDS increases, so there will be a trade-off somewhere. The best ones seem to be relatively low voltage.

Yea at 100v the GaN fets were more than 20x faster at switching with comparable rDSon. These faster switching times are also accompanied with lower gate driver power consumption. If built this would be without a doubt the most efficient controller as far as intrinsic consumption goes and very likely overall efficiency as well. The next model of GaN fet has already been announced with even lower rDSon. They may not even benefit from ZVS.

(eGaN FET at 375 kHz and MOSFET at 250kHz) Even at a higher frequency the GaN fet wins across the board.

Also in reply to Thomas earlier I forgot

do wonder about the option to configure as two single converters if that will end up driving up cap size needs (being that each side would then be a single phase). Think it would be best to keep the MPPT controller focused on one battery output – with a different device addressing multi battery bank needs

The idea of going from 1 dual phase to 2 single phases is not specifically to support 2 separate batteries (although it would be capable and this could be seen as an added benefit). It's more to just allow for different panel configurations from one controller. The same controller could support 1 large high power panel or 2 lesser ones each being MPPT controlled instead of just paralleled, this would give better shaded performance. The capacitance is not an issue since the current in each phase is still the same.

[OceanSeaSpray]

Originally Posted by whiskthecat

Agreed. I have no idea why there are so many IC's. You pretty much just need an MCU, a logic supply and some mosfet drivers.

There is a need for a little more for instrumentation, but that's all I had put on my board indeed. Isolated COM ports and the like chew through more juice and make no sense to me on a boat.

Originally Posted by whiskthecat

...if you do dip under bank voltage you will get a whopping 100% loss from your panels as no current will flow. Considering this I don't plan on installing anything with a Vmp less than 24v and this might be why many designs don't account for 100% duty cycle as you pointed out earlier, I likely won't either.

VOC would need to dip below the bank voltage to get 100% loss, not very likely. A panel where VMP would drop below charging voltage under normal conditions would be a poor choice. Now mine (I just went out to check) give VOC=21.3V and VMP=17.75V @ 25degC. They get warm in intense sun, but just warm and there is no observable reduction in current. You can also leave your hand on them. They just have good air flow underneath…

An easy way of taking care of the 100% duty cycle is bypassing the top N-channel FET and inductor with a large P-channel FET, some have really low rDSon values now. At a minimum, it would be sensible to reduce the switching frequency down to the minimum needed by the gate driver when full conduction is best.
It is not just low light operation, as soon as you don’t take more than the panel current, switching makes no sense anymore and it also takes care of the problems of switching into no load etc.

Originally Posted by whiskthecat

Yea at 100v the GaN fets were more than 20x faster at switching with comparable rDSon. These faster switching times are also accompanied with lower gate driver power consumption. ... eGaN FET at 375 kHz and MOSFET at 250kHz Even at a higher frequency the GaN fet wins across the board.

You need to give full credit to the standard FETs too and they come with at least 3-4 times lower rDSon than the GaN FETs I looked up earlier. A 100V GaN FET had rDSon=4mOhm and I wouldn't want to hook up a 100V panel to it because a good lightning flash nearby could take it out. It would take 150V FETs to accept a 100V input safely still with surge suppression and then rDSon was climbing up further again.
GaN FETs are a trade-off between full load efficiency (the only one ever quoted and least meaningful) and really useful efficiency when available power is reduced. They will be especially better at low power where intrinsic consumption is more important. That is fine by me, by the way. Lastly, the ones I was looking at weren't cheap and didn't really have provisions for any heatsinking. 1W was something like 50degC package temperature rise. That is 15A @4mOhm and this goes back towards a modestly sized unit unless you found much better devices.

The frequencies you are quoting would keep everything very small, but they would also be extremely high for an application like this. High-efficiency converters are running below 50kHz and sometimes as low as half of this with serious inductors. The toroid core I had sourced for this is 2’’ diameter. You can’t beat switching less when it comes to switching losses. With GaN FETs, the optimum would likely be a bit higher in terms of efficiency vs cost.


Separating the converter from the charge regulation is what I had in mind. It is the best way of handling multiple batteries properly. I have built the isolating switch that controls the charging side and the buck converter is next, some day!

[whiskthecat]

Originally Posted by OceanSeaSpray

An easy way of taking care of the 100% duty cycle is bypassing the top N-channel FET and inductor with a large P-channel FET, some have really low rDSon values now. At a minimum, it would be sensible to reduce the switching frequency down to the minimum needed by the gate driver when full conduction is best.
It is not just low light operation, as soon as you don’t take more than the panel current, switching makes no sense anymore and it also takes care of the problems of switching into no load etc.

Good points. I hadn't thought of using the Pfet as I was too hung up on trying to get the higher voltage gate driver for the top fet to be always on. Your way seems simple and worth implementing. What had you came up with for a startup power supply to run your logic before the main buck had started?

Originally Posted by OceanSeaSpray

You need to give full credit to the standard FETs too and they come with at least 3-4 times lower rDSon than the GaN FETs I looked up earlier. A 100V GaN FET had rDSon=4mOhm and I wouldn't want to hook up a 100V panel to it because a good lightning flash nearby could take it out. It would take 150V FETs to accept a 100V input safely still with surge suppression and then rDSon was climbing up further again.
GaN FETs are a trade-off between full load efficiency (the only one ever quoted and least meaningful) and really useful efficiency when available power is reduced. They will be especially better at low power where intrinsic consumption is more important. That is fine by me, by the way. Lastly, the ones I was looking at weren't cheap and didn't really have provisions for any heatsinking. 1W was something like 50degC package temperature rise. That is 15A @4mOhm and this goes back towards a modestly sized unit unless you found much better devices.

Yes I had noticed some of the popular brand name MPPT's quoting 99% efficiency, what a joke. Maybe when the thing isn't switching and the panel just happens to be matched to the battery on it's own. The GaN quoted thermal characteristic are with no top package heatsink on a single layer 2oz board. This is far from what I would implement. I was looking into using proper power PCBs with copper substrate. Even still a regular board with top heatsink provided and vias to the bottom with a back whole board heatsink would allow huge improvements in the quoted thermal performance. Also remember that we are using dual phase so that current rating is about where we need to be at already, so we likely could get away with no heatsinking and these devices are especially tolerant to high temperatures (300C) but I wouldn't want to push them if not necessary. As for the lightning scenario I would think it hardly matters. If a device is rated at 100V there is already surge overhead room of about 20v built into that. And if lightning hits you its not gonna be just a 20v bump anwyays, nothing is going to protect from that. Additionally I have been rounding up on describing these "100V" panels as they are actually 86v open circuit. I just picked 100v as there are far fewer component options at 90v if any. Since they are the only panels I know of with bypass diodes for every cell and the only reason I am designing to such a high voltage in the first place, 100V is plenty of headroom for all panels.

Here is the GaN I was looking at by the way. They are $6 per chip which isn't so bad when you look at the price of comparable FETs. To keep the size of the thing and the cost of the passive components down, before I ever even thought to use GaN I was looking to keep the frequency above 100khz (dual phase equivalent 50khz) I haven't put a lot of thought into this yet but it might be possible to run both GaN fet and traditional FET in parallel to try and achieve the benefits of both. The controller would switch the GaN FET on first, which comes on very fast with little gate losses. You'd then be running at about 7mOhm and then you switch on the mosfet at your leisure (the switching doesn't have to be fast because there is already close to 0v across the mosfet since the GaN is turned on next to it) and once it comes on you now have the lower 2mOhm path in parallel. No idea if that's practical yet it might be more efficient (and expensive) to just run 2 GaNs in parallel to reduce the rDSon.