Solar MPPT - Freewheel diodes, buck regulators etc

[RayJones]

I've been playing around with a PIC controlled buck regulator for the purposes of a Max Peak Power tracker for solar panels.

The original design I started with was based upon Tim Nolan's where he used a synchronous rectifier (Mosfet) in place of the back swing diode. (www.timnolan.com)

For sure this improves efficiency tremendously due to the much reduced voltage drop, BUT.

The BUT comes about when you have a day where the sun quickly goes behind clouds and the panel's output drops quickly.
I've determined what happens is the synchronous buck regulator suddenly becomes a synchronous boost regulator in the reverse direction
ie the battery can supply current in the reverse direction through the inductor to ground via the synchronous lower mosfet.
When that switches off, the top mosfet transfers the voltage kick in the reverse direction back to the panel's input (albeit with a blocking diode) - Ala the classic boost regulator topology.

After blowing yet another set of mosfets with god knows what current reversals and voltage spikes going on today, I'm now considering reliability over efficiency and using a true back swing diode in the usual location.
A true diode will simply not allow the reversed inductor currents to flow.

So there then becomes two obvious choices:

Fast recovery, or Schottky?

It looks like the Fast Recovery diodes have a much higher Vf, but are the switching losses more pronounced if you use a Schottky?

I'm interested to hear from others with perhaps more experience in switching regulator design, and even more so a discussion on MPPTs would be great....

This power electronics is spectacular at times 

[Simon]

yes I've been wanting to design one of these for a while but am still messing about with another project.

I'm no expert on diodes and switching regulators, I started by tring to use a switcher on my panels but figured i was getting no increase in efficiency due to the regs dropout but i was using a single chip reg. I think the proper IC controllers with an external mosfet (my chip used a BJT hence the losses ?) is going to be more efficient

You may be able to get arouind your issues by putting some sort of control in the MCU conntrol system ? I can see the sort of trouble your having as I'm doing a project to control a car dynamo and need to know if the dynamo is outputting enough to charge the battery instead of the dynamo becoming a motor and being run by the battery, in my case I sense the current and cut off power under a certain current value, in your case you could actually sense the direction of the current and refuse to let it to work if the current starts to go the wrong way ? or have an ADC input to the MCU so it can check the voltage and switchoff when it goes low, you could also have a battery and panel ADC input and check that the difference bettwen the two is great enough for it to work ?

[mikeselectricstuff]

I wonder if a compromise would be to have a schottky diode, and a MOSFET that only turns on when you know it's OK...

Another thing to bear in mind is that sync rectification only has major efficiency benefits when the Vf of a diode is significant compared to the output voltage - for something like a 12V supply, a big chunky schottky diode is only going to be a few % less efficient.
Choice of diode type will depend on frequency, which again depends on lots of other factors, but it generaly boils down to switching losses versus static losses and inductor/cap size. if you are not pushed for space, there is probably little need to have the frequency particularly high, so a Schottky is probably going to give the best efficiency. Shottkys aren't exactly slow.

There are some more complex buck-boost architectures like sepic which can deal with inputs higher or lower then the output, and as the source can potentially be floated, an inverting topology may be worth looking at to deal with the input range issue.

[Zero999]

You could connect two MOSFETs back-to-back, it would mean a higher voltage drop and gate capacitance but still might work out more efficient than a Schottky diode.

[RayJones]

Yeah that's a good idea to check the current direction, but I'm pretty sure the high side current sensor I'm currently using only works in one direction.

I first tried low side sensing of panel and battery, but found I was having all sorts of trouble getting consistent readings with increasing currents due to the "ground" wandering about.
The new design using high side sensors works a treat, and I now have a dirty big fat single ground across the bottom of the power control stage.

There is already a schottky diode across the lower mosfet, this takes care of the initial back swing as the FET drive is deliberately non overlapping - for obvious punch through reasons (which is perhaps what is taking out my mosfets anyway when the sun rapidly goes away...)

My initial check was for low solar panel volts to stop the switcher, but that was when I found the boost converter action never allowed the "panel" volts to fall!

I'll have a closer look at the battery current sense, and see about making it bipolar somehow.
Actually even just a comparator looking for -ve flow would work as a yes/no control into the micro or even, shock horror, direct into the FET driver.

One thing you must be careful of with mosfets is their parasitic diode. I'm not sure what is meant by back to back mosfets, but the didoe may become another issue.

[Simon]

well I looked at the diagram on your link and the current sensor chip is a MAXIM one with a very similar part number to the one I'm using on my dynamo controller and that is one way only.
I think the issue with bidirectional is you need a dual supply or a biased chip or  the chip has to have a H bridge output that is indipendant from ground which makes feeding the reading into a micro very hard.
My get out of jail clause was to put a diode in paralele with the Relay that cuts the dynamo in and out so that when rurrent starts to flow through the diode I turn on the relay and then turn it off again when the current drops, the diode really allows the relay to cut in in the first place with no chance of current going the wrong way.

What I suggest you do is use a comparator with an input either side of the current sensing resistor so that if a reverse current condition occurs the ouput will go whichever way you want and signal something to cut off, you can either use a relay like i am or if you are using a SMPS controller with an enable pin you can just turn the chip off. If using a relay i suggest a latch relay for efficiency saving they are like 3.5 pounds from RS (nice bits of kit, I'm using one to eliminate heat dissipation as a problem in a confined space)

[Pyr0Beast]

To prevent energy going back to the panel use ordinary schottky diode.

In buck-boost converter use schottky diodes as well. Schottky diodes also don't have reverse recovery time.

With a drop of 0.3V (steal them from a broken psu) efficiency at 12V suffers less than 5%.

[Simon]

thats not quite his problem, the problem is to stop it acting as a booster alltogether, this project is more than juist a SMPS and needs the appropriate control

[RayJones]

Hi Simon,

actually the Tim Nolan design has since changed since I last was there (the hazards of blindly posting links!)
His original design used an op amp over a low side current sense, and used a PIC, not an Arduino.

I had two issues with the low side sensing - working close to the "zero" volt rail was not good without a -ve supply, and it didn't provide a solid rail through the unit on either the +ve or -ve side of the panel, hence my "zero" volt rail was broken in several places, and could quite easily be circumvented externally by accident.

And yes, I want to prevent the switcher reversing roles when the panel suddenly loses output due to quick cloud cover.

Now with schottky diodes, I'm looking at around a 10A output current and all the Schottky diodes on Farnell have Vf more like 0.7 to 1V
That's 10W of dissipation during the backswing.
I certainly have not seen any specs that approach 0.3V at these current levels....

The fast recovery have an even worse Vf.

The mosfets however are only a few milliohm when on and the Vds is thus very low, BUT role reversal happens oh so easily with a nice low impedance battery willing and ready for action  .

I think it is time to abort the PCB I've been using and go back to the old trusty matrix board for the power stage, and try some of the ideas suggested.
The current sense I'm using is a the LT6100 and looks something like a excrement of a fly, so I'll probably need to chop that bit of board out to haywire it in.

An an SPI based current sense would be brilliant for so many reasons

[RayJones]

Here's the original Tim Nolan design if you are interested:

homepower.com/files/webextras/NolanPPT102-94.pdf

[Simon]

well it uses a differential op amp sensing circuit anyhow so why not just more it to the positive rail ? I think you would just have the reverse the inputs and it would work fine. My reason for choosing a dedicated current sensing circuit that cost 4 times my micro was because I wanted to work 20+A so chose a 0.001 ohm resistor which made using the diff amp tricky due to offset

[Zero999]

Most op-amps' common mode range does not include the positive supply.

[Simon]

you mean the voltage on the current resistor will be too close the the supply voltage to be read correctly ? personally the MAX4172 does it all for me

[RayJones]

Yep, running near, or even *above* the opamps nominal supply will be even more troublesome.

That's why I used the LT6100, the sense must be well over the chip's supply.

I'm not sure why we are discussing the current sense as that is the part that does work correctly
We probably got there because what I currently have cannot check for negative battery flow into the inductor, but all are tarred with the same brush it seems, or you need at least a *true* bipolar ADC and most likely dual supplies (There's another field of mis-leading data sheets).

My main issue is the reversal situation is what most likely took out the mosfets last time. I observed the sun quickly vanish, then re-appear. Bam and pop - the magic smoke was released.

The use of a conventional diode as the free wheeler would automatically prevent the reversal, but at the cost of efficiency.
I'll just need to try a few hijacked devices from dead PC SMPS's and see how they travel.

Thanks for your thoughts so far anyway.

[Simon]

well you could use a mosfet as a "gate" to separate the battery from the circuit when the sun goes in, you can switch it off when the current gets very low (ie about to reverse), if your trying to pull all you can put a diode in parallel with the mosfet so it will pass small currents, I'm afraid you will never hit ideal efficiency you have to loose something somewhere

[jahonen]

The current sense circuit on Nolan's original circuit can be made bidirectional by shifting the output zero point to mid-scale (or other convenient voltage), by connecting R22 and C25 not to ground but to voltage where you want the zero level. This should be low-impedance source since otherwise it will affect the current sense amplifier CMRR. Of course, doing this will change the measurement range, so you might want to change the amplifier gain also.

That will make it possible to detect if current starts to run backwards.

Regards,
Janne

[Simon]

yep that would do it, if your running 10 bit resolution then 512 will be "0", you could also bias the "ground" by less than half as you don't need a full negative scale and can get more resolution out of the positive scale which is what you want to be accurate, say 1/20 so it will be 0-51 is negative and 52-1024 is positive

[RayJones]

The current sense circuit on Nolan's original circuit can be made bidirectional by shifting the output zero point to mid-scale (or other convenient voltage), by connecting R22 and C25 not to ground but to voltage where you want the zero level. This should be low-impedance source since otherwise it will affect the current sense amplifier CMRR. Of course, doing this will change the measurement range, so you might want to change the amplifier gain also.

That will make it possible to detect if current starts to run backwards.

Regards,
Janne

Hi Janne,

yeah, that's one way to do it.
My earlier problems with this original design were most likely to do with operating right down near the 0V rail. Just the simple act of biasing upwards would help put the operating range in the more linear region between rails, never mind being able to sense reverse flow.
Nevertheless, my preference is to do the current sense on the high side so I maintain a solid continuous ground through the controller from panel to battery.

Biasing like you suggest is actually what you also need to do with a lot of so called "bipolar" ADCs I alluded to earlier.
I don't know where chip manufacturers get off calling an ADC bipolar when the input needs to be biased midscale 
If you check the analog ADC input spec, there is usually a -0.2Vish absolute limit. What a load of crap for a "bipolar" unit!

Yeah OK, perhaps they give 2's complement outputs with 0 @ half scale, but it ain't that hard to do in software

Back to my design, it is becoming obvious I need to take far more care with the sense inputs to allow a rapid shutoff for those days when the panels ouput rapidly is productive then non productive due to patchy cloud cover.

The switcher itself is working well when it is meant to, so I'm happy with my inductor sizing etc.

[TechGuy]

1. Google for Trench Schottky diodes since they have the lowest forward voltage drop. However for these diodes to get ultralow Vf, they need to run hot. Be sure to look at the datasheet graphs that plot Vf versus Current and temperature. The headline Vf usually is cherry picked and may not be realistic for your application.
Here is an example Trench Schottky http://www.onsemi.com/pub_link/Collateral/NTST30100S-D.PDF

2. Low voltage reverse voltage protection can be done with a pair of "Back to Back Mosfets". A google search will turn on a lot of pages on this approach. You need to use back to back mosfet in order to avoid the reverse protection diode that is built in with mosfet devices.
Note: that you need a high-side driver when using N-Channel Mosfets. Ideally N-Channel mosfets are preferred because they are more efficient than P-Channel mosfets.

 Here is an example of using back to back mosfets for reverse voltage protection (using N-Channel):
http://datasheets.maxim-ic.com/en/ds/MAX1614.pdf
http://www.cdistore.com/Linear/datasheets/linear/4356f.pdf (See Page 15 for the example)

[RayJones]

Just found on the Linear Technology web pages a thing called an "Ideal Diode"

It uses an N channel Mosfet along with a LTC4357 controller chip (gate boost included)
http://www.linear.com/pc/productDetail.jsp?navId=H0,C1,C1003,C1142,C1079,P38529

If the forward voltage across the FET exceeds 25mV the FET is turned on, likewise turned off for the reverse situation.

If I place one of these at the battery side of the buck converter I will have an efficient diode function out to the battery, and can continue to use the synchronous low side FET in the buck regulator without worrying about the reverse feed.

8 bucks at Farnell, but it is one of those horrid MSOP8 packages, not easy for quick prototype lashups.
I need to make up some MSOP8 to DIP8 PCBs methinks.

[Simon]

Basically it does what I described with discrete parts a few posts back and is what i'm doing with a relay instead of the mosfet (but i can afford 30ms lee way) although i may consider mesfets later, Having the built in boost circuit is good as you can use N channel mosfets with lower RSDon resistance as opposed to P channel mosfets which is what I had considered

[jeremy84olsen]

You mean to say parasitic diodes should be avoided ?

[Simon]

what do you mean by parasitic ? the object here is to replace the diode because it has a fixed voltage of 0.5+ volts, a mosfet or relay will have a very small resistance in the range of milli ohms so will make less heat and be more efficient at low currents