MPPT buck converter that isn't optimized for battery recharging?

[mberg2007]

Yes. However, your Raspberry/DCDC converter combo is NOT like a resistor. You can turn that knob until your resistor takes 20 watts, but you can never get your Pi to take 20 watts.

You're losing me a bit here. No I cant force the pi to take 20 watts since the system running flat out will draw a maximum of 10 watts with the units I have attached to it. But I can use a resistor in parallel with the pi which will present the panel with an arbitrary load, can I not?

If I cannot do this manually, how does an MPPT charge controller manage to do this with a battery, which has a very low impedance as far as I know?

The DC/DC will keep decreasing its impedance to try to pull more power from its source. If the source was a voltage source that would work. But it's not - once you start moving to the right side of the VI plot for a panel it becomes more like a current source. So the power goes DOWN as you decrease the impedance.

When you say "Power", are you referring to watts? If so, why does it matter if the input is a voltage or a current source? Power is the product of those two.

Can you show me a few simple math examples with volts and current and power, and what happens when cloudy conditions start to limit the power coming out of the panel?

In practice this means the system would work fine until a cloud passed in front of the sun, at which point you'd see a rapid drop to close to zero power, with no recovery until you were in bright sun.

This would be okay. A system operating without a battery would have to suffer such conditions. With a 40 watt panel I expect the system to keep running in bright and semi-bright conditions. As long as the panel is producing 10 watts and sending that into the buck converter, the pi will not "overload" the input power will it?

That would take care of power conversion. You would also need an UVLO circuit to reduce power output when the panel drops below its lowest possible operating point.

I have bought an LVD for different reasons but I had not heard of an UVLO until now. Are those two things the same? As I am reading the specs on such a thing, it cuts off output when the input power drops below a certain point. Which is also what an LVD does.

How would this change things? My pi would still die when a cloud passed by. Just as it would without an UVLO... right?

Or if you really, really want to spend the money, you could get something like a Midnite Kid. But there are three problems there:

I actually have a Tracer 1210RN which is an MPPT capable solar charge controller. Is the Midnite Kid the same type of device?

Having a fully fledged charge controller defeats the purpose of this exercise as I am then back to needing a battery of some kind. This is the setup I have now. I want a setup which is much cheaper, to see if simply buying a lot more panel area is more cost effective than setting up a system that requires a battery. Imagine that the pi is producing something valuable, like bitcoins. So every hour of on-time has a certain yield. Will the yield over a year be the same, lower or higher given a $100 system that is based on small panels, an MPPT solar charge controller and a lead acid battery, OR a system which has a huge panel and a cheap buck converter. Also factor in replacing the lead acid battery.

(The bitcoin thing is just an example, my pi won't be generating fortunes for me I understand that

-Michael

[ButchDeal]

The cheapest thing you can do is grid tie. We have a customer with a large bit coin farm and large grid tie array. Anything else and the computer will never be able to make up the cost of the array.

[jflorey2]

You're losing me a bit here. No I cant force the pi to take 20 watts since the system running flat out will draw a maximum of 10 watts with the units I have attached to it. But I can use a resistor in parallel with the pi which will present the panel with an arbitrary load, can I not?

Sure. But that will give you no more performance than a system without that arbitrary load. If you have a use for that occasional extra 10 watts of heat it might make sense; otherwise it's a lot of work for no improvement.

If I cannot do this manually, how does an MPPT charge controller manage to do this with a battery, which has a very low impedance as far as I know?

By changing the PWM of the power switch, which changes the impedance transformation ratio.

When you say "Power", are you referring to watts? If so, why does it matter if the input is a voltage or a current source? Power is the product of those two.

It matters because when you decrease the impedance of a load connected to a voltage source, power goes UP. When you decrease the impedance of a load connected to a current source, power goes DOWN. That makes them very different.

Power is not the product of a current source or a voltage source. A current source is a source that always provides a fixed amount of current; a voltage source is a source that always provides a fixed amount of voltage. They are not the same as current or voltage. There are also no ideal current or voltage sources; ideal sources are abstractions that make electrical design easier, and many devices _approximate_ one or the other. For example, a fluorescent ballast approximates a current source; a battery approximates a voltage source. A solar panel has characteristics of both depending on what operating point you are at.

Can you show me a few simple math examples with volts and current and power, and what happens when cloudy conditions start to limit the power coming out of the panel?

OK. 20 watt panel, with max pwr point at 20 volts, 1 amp.
Ideal impedance is 20 ohms. Then you get 20 watts. (V^2/R or I^2R; at the maximum power point both work.)

Let's say it goes up to 40 ohms. Now the panel is a voltage source. V^2/R is now 10 watts. You can back-calculate the current; it is now .5 amps. The higher the resistance gets, the lower the power gets.

Let's say it goes down to 10 ohms. Now the panel is a current source. I^2/R is now 10 watts. You can back-calculate the voltage; it is now 10 volts. The lower the resistance gets, the lower the power gets.

From the above you can conclude that there is one ideal impedance that gives you the most power.

Now let's say you are operating close to the limit with a DC/DC converter. The converter is drawing 19 watts. Everything is OK. The Pi wants more. The converter decreases its impedance. Decreasing impedance on a voltage source means more power, so it does so and now the power is 20 watts. Still OK. Now the Pi wants more. The converter decreases its impedance - but now it has moved into the current source part of the curve. Decreasing impedance on a voltage source means LESS power, so now the power is 18 watts. The converter keeps decreasing its impedance trying to maintain its output voltage. Now power is 16, 14, 12, 10 watts etc and the panel power very quickly collapses to near zero.

If there's a cloud and you are only getting 10 watts on the panel, this happens at the 10 watt point.

This would be okay. A system operating without a battery would have to suffer such conditions. With a 40 watt panel I expect the system to keep running in bright and semi-bright conditions. As long as the panel is producing 10 watts and sending that into the buck converter, the pi will not "overload" the input power will it?

No. It will keep working as long as you stay away from that max power point. Unfortunately a cloud may bring you right up to the max power point, at which point the voltage will collapse.

I have bought an LVD for different reasons but I had not heard of an UVLO until now. Are those two things the same?

Sort of. In DC/DC converter language it's an undervoltage lockout. It means that the converter will not allow the input voltage to drop below a certain voltage; in the above example that would happen around 18 volts or so.

How would this change things? My pi would still die when a cloud passed by. Just as it would without an UVLO... right?

Yes. But the system would recover once the cloud passed.

Imagine that the pi is producing something valuable, like bitcoins. So every hour of on-time has a certain yield. Will the yield over a year be the same, lower or higher given a $100 system that is based on small panels, an MPPT solar charge controller and a lead acid battery, OR a system which has a huge panel and a cheap buck converter. Also factor in replacing the lead acid battery.

If you want reliable power you need a battery (or large capacitor.)
If you want no storage at all and want to maximize on time, use a sufficiently large panel and a DC/DC converter with an undervoltage lockout set appropriately.

[mberg2007]

Tanks a lot for your reply, I appreciate it a lot. There's a lot to digest

On ebay I found this LVD "module"; which is very easy to set up:

http://www.ebay.com/itm/201567740462?_trksid=p2060353.m2749.l2649&var=500774696360&ssPageName=STRK%3AMEBIDX%3AIT

I tried to look for an UVLO module but I had some trouble finding something simple and inexpensive. Do you know of such a module which I might use?

Right now I have the charge controller turning the load on and off as it wants, and it seems like it often turns off the load because it wants to send all the power to the battery. I'm going to change things around a bit so my load is connected directly to the battery, and protect the battery from complete discharge using an LVD. This way the controller only needs to worry about sending as much power as possible to the battery, and I'm going to leech a bit of that power without it knowing about it.

But I wonder how much power I will be able to get this way. If the impedance in the battery is very low, I suppose most of the power will prefer to flow that way. So if the panel is producing 10 watts, perhaps my pi will be able to draw only 4 of those. I don't know precisely what the impedance of my buck+pi is and I guess with what you said the buck will vary that impedance. Do you have any ideas on this type of setup?

Is it possible to have a super cap charged by a solar charge controller? That way I would not have to worry about LVD or maintaining the battery, it would just start at 0 every day.

-Michael

[mberg2007]

The cheapest thing you can do is grid tie. We have a customer with a large bit coin farm and large grid tie array. Anything else and the computer will never be able to make up the cost of the array.

I realise that. The goal was never to make this a cost effective setup, i.e. a system that would earn back its own cost. Consider this a hobby experiment where yield isn't important at all. What is important is what is the most efficient way to have a system like this running completely isolated. Is it a configuration with a solar charger and a battery, or is it a configuration with a buck converter and an oversized solar panel.

-Michael

[jflorey2]

On ebay I found this LVD "module"; which is very easy to set up:
http://www.ebay.com/itm/201567740462...%3AMEBIDX%3AIT
I tried to look for an UVLO module but I had some trouble finding something simple and inexpensive. Do you know of such a module which I might use?

No, sorry. I have built them for use with supplies with remote sense terminals but I know of nothing off-the-shelf. Sometimes you get lucky with a DC/DC converter; for example, a DC/DC that is rated for 18-36 volt input will sometimes refuse to convert below about 16 volts, which serves the same purpose.

Right now I have the charge controller turning the load on and off as it wants, and it seems like it often turns off the load because it wants to send all the power to the battery. I'm going to change things around a bit so my load is connected directly to the battery, and protect the battery from complete discharge using an LVD. This way the controller only needs to worry about sending as much power as possible to the battery, and I'm going to leech a bit of that power without it knowing about it.

But I wonder how much power I will be able to get this way. If the impedance in the battery is very low, I suppose most of the power will prefer to flow that way. So if the panel is producing 10 watts, perhaps my pi will be able to draw only 4 of those. I don't know precisely what the impedance of my buck+pi is and I guess with what you said the buck will vary that impedance. Do you have any ideas on this type of setup?

There is something wrong with your setup if it is "turning off the load because it wants to send all the power to the battery." That should never happen, unless you have an LVD that is shutting off the load to protect the battery from overdischarge.

Is it possible to have a super cap charged by a solar charge controller? That way I would not have to worry about LVD or maintaining the battery, it would just start at 0 every day.

Unfortunately no, because the controller needs power on the battery terminals to start up. However, if you keep the supercap above about 10 volts at all times (via an LVD) then it should work.

[mberg2007]

There is something wrong with your setup if it is "turning off the load because it wants to send all the power to the battery." That should never happen, unless you have an LVD that is shutting off the load to protect the battery from overdischarge.

The battery I have is underdimensioned for the task. It is there to carry the computer through passing clouds and because the controller needs power (and as you point out this also means that using a supercap without an LVD is not possible, I had not considered that).

The system is designed to run on an opportunistic basis when conditions allow, and then a bit of buffering for quick transitions of clouds or short overcast periods. No night time running, no running during extended overcast periods. A system designed to keep running all day would be excessively expensive (10 watts/hour = 240 wH per day, figure twice that for the battery size, 12v*x=580w -> x=48 ah if this was a car battery for example. Also I would have something like 120 watts of panel power, 2 hours avg per day year round at full rate = 240 wH per day). Would be nice but this is out of scope for the time being.

I seem to remember that charge controllers can't directly charge supercaps for some reason. But it's something I had planned to experiment with later on. I guess I would need quite a bit of capacity for this type of application but the panels should be able to put a lot more power into the caps and recharge them much faster than a battery. Do you have any experience with super caps in relation to solar panels and systems in general? Would I need a special controller with a "supercap battery charging profile"? I don't think that the controller trying to recharge a supercap as if it was a lead acid battery is necessarily a good idea.

-Michael

[mberg2007]

No, sorry. I have built them for use with supplies with remote sense terminals but I know of nothing off-the-shelf. Sometimes you get lucky with a DC/DC converter; for example, a DC/DC that is rated for 18-36 volt input will sometimes refuse to convert below about 16 volts, which serves the same purpose.

Suppose I were to connect my panels to an LVD input, and pass the LVD output into my buck converter. The LVD is set to whatever is a good cutoff, like 16V or something. Assuming the LVD doesn't interfere with the operation of the buck converter at the other end, wouldn't it function as an UVLO in this configuration?

-Michael

[jflorey2]

Suppose I were to connect my panels to an LVD input, and pass the LVD output into my buck converter. The LVD is set to whatever is a good cutoff, like 16V or something. Assuming the LVD doesn't interfere with the operation of the buck converter at the other end, wouldn't it function as an UVLO in this configuration?

Yes - as long as you provided enough capacitance on the solar panel side of the switch so that the the sudden load doesn't collapse the panel instantly. Most DC-DC converters have some input capacitance so make sure you add significantly more (3 to 10 times) capacitance on the panel side.

I seem to remember that charge controllers can't directly charge supercaps for some reason. But it's something I had planned to experiment with later on. I guess I would need quite a bit of capacity for this type of application but the panels should be able to put a lot more power into the caps and recharge them much faster than a battery. Do you have any experience with super caps in relation to solar panels and systems in general? Would I need a special controller with a "supercap battery charging profile"? I don't think that the controller trying to recharge a supercap as if it was a lead acid battery is necessarily a good idea.

Controllers don't care whether it's a supercap or a lead acid battery as long as you keep the supercap above about 10 volts (so the charge controller can start up) and below its maximum voltage (to prevent damage to the supercap.) For example, the Maxwell BMOD0058 is good to 16 volts, so as long as you didn't let it discharge below 10 volts or so it would likely work with most charge controllers intended for lead acid batteries. Beware of charge controllers that do both equalization charging and temperature compensation; in extreme conditions the combination of those two could drive voltages up above 16 volts.

[mberg2007]

Yes - as long as you provided enough capacitance on the solar panel side of the switch so that the the sudden load doesn't collapse the panel instantly. Most DC-DC converters have some input capacitance so make sure you add significantly more (3 to 10 times) capacitance on the panel side.

Good point. As the LVD switches on, the buck powers up the raspberry pi and connected devices, causing a large power draw which will cause the buck to lower its impedance to find more power, which may cause the panel voltage to drop (crash) until the LVD switches off. Not a good scenario.

In my case I have some panels connected in series, with a nominal Voc of about 44 volts. Pretty high for my buck (40V max) but it could probably take it. When I measure the voltage across the panel I can see the controller typically keeps the voltage around 16-18V, so I guess that's where it thinks the MPP is for my load. The LVD I have is simple and will cut off at a configurable voltage interval from 10-12V. You think 10V cutoff and 12V reconnect is enough to prevent a crash and allow the panel to recover?

Controllers don't care whether it's a supercap or a lead acid battery as long as you keep the supercap above about 10 volts (so the charge controller can start up) and below its maximum voltage (to prevent damage to the supercap.)

Well yes and no. On my controller I have to select the type of battery connected (lead acid/gel etc), and it has different charging profiles for each type. For example, for a lead acid battery it will trickle charge at one voltage, whereas with a GEL battery it will trickle charge at a different voltage. It also tries to charge to 14+V in order to balance the cells, if it is a lead acid battery. So it does matter to a certain degree. Just not sure if this smartness will get in the way of charging the supercap as fast as possible.

Turns out there are plenty of supercap configurations for sale on eBay. Some as large as 100+ F. They aren't that expensive, but I'm a little in the dark about capacitors and in particular how fast they discharge. My pi operates at 5V and pulls up to a maximum of 2A (10 Watts) with the various usb attachments I have plugged in. If the controller will charge the caps to around 14V and I have to maintain around 10V of power in them, that means the caps can be drained from 14V to 10V but I have no idea what that translates to in minutes. Is that 1 minute or 1 hour?

-Michael

[jflorey2]

In my case I have some panels connected in series, with a nominal Voc of about 44 volts. Pretty high for my buck (40V max) but it could probably take it. When I measure the voltage across the panel I can see the controller typically keeps the voltage around 16-18V, so I guess that's where it thinks the MPP is for my load.

If your Voc is 44 volts and your buck is 40 volts you're going to blow it. Maybe not today but on the first cold full-sun day. And if your Voc is 44 volts and your system is operating at 16 volts you are missing most of the power from the panel. (Although that would not be unusual with a PWM controller.)

The LVD I have is simple and will cut off at a configurable voltage interval from 10-12V. You think 10V cutoff and 12V reconnect is enough to prevent a crash and allow the panel to recover?

Probably - will depend on the light level.

Well yes and no. On my controller I have to select the type of battery connected (lead acid/gel etc), and it has different charging profiles for each type. For example, for a lead acid battery it will trickle charge at one voltage, whereas with a GEL battery it will trickle charge at a different voltage. It also tries to charge to 14+V in order to balance the cells, if it is a lead acid battery. So it does matter to a certain degree. Just not sure if this smartness will get in the way of charging the supercap as fast as possible.

The supercap doesn't care what voltage it is. Only issue in terms of voltage is how much energy you store (which is equal to 1/2CV^2.)

If the controller will charge the caps to around 14V and I have to maintain around 10V of power in them, that means the caps can be drained from 14V to 10V but I have no idea what that translates to in minutes. Is that 1 minute or 1 hour?

Calculate energy at 14 volts and energy at 10 volts; subtract them. The result will be in joules (watt-seconds.) Divide joules by watts and you will have the time in seconds that you will be able to run.

[mberg2007]

If your Voc is 44 volts and your buck is 40 volts you're going to blow it. Maybe not today but on the first cold full-sun day. And if your Voc is 44 volts and your system is operating at 16 volts you are missing most of the power from the panel. (Although that would not be unusual with a PWM controller.)

Ok so I need a buck with a slightly higher voltage specification.

I'm a bit surprised that you say I am missing most of the power from the panels. Whether I use my MPPT charge controller or a buck converter, both will transform the DC into AC and back into DC at the output voltage precisely to avoid wasting power. Did I completely misunderstand that? If MPPT controllers are like PWM controllers this way, then what's the point of using an MPPT controller?

The supercap doesn't care what voltage it is. Only issue in terms of voltage is how much energy you store (which is equal to 1/2CV^2.)

Thanks for that formula. From 14-10 volts at 100F this would give me 8 minutes of running time at 10 watts (total darkness). This would be fine I think.

-Michael

[jflorey2]

I'm a bit surprised that you say I am missing most of the power from the panels. Whether I use my MPPT charge controller or a buck converter, both will transform the DC into AC and back into DC at the output voltage precisely to avoid wasting power.

You are correct. However, if your PANEL voltage is 16-18 volts during operating then the MPPT controller isn't doing its job. If your open circuit panel voltage is around 44 volts then I'd expect to see an Vmp around 36 volts, not 16. By keeping voltage at 16 volts you are not harvesting most of the panel's power.

[sensij]

You are correct. However, if your PANEL voltage is 16-18 volts during operating then the MPPT controller isn't doing its job. If your open circuit panel voltage is around 44 volts then I'd expect to see an Vmp around 36 volts, not 16. By keeping voltage at 16 volts you are not harvesting most of the panel's power.

If the controller is delivering all of the power the load requires, it is doing its job. That does not necessarily mean it will be operating at the maximum power point 100% of the time, if there is no place for that extra power to go.

If the panel is operating at 16 V, but the Vmp is 36 volts, it suggests that you could increase your load under those conditions.

[jflorey2]

If the controller is delivering all of the power the load requires, it is doing its job.

A reasonable point. However, most MPPT's let the voltage float _up_ when the output is not delivering maximum power. I would be very suspicious of an MPPT charge controller that pulled the voltage down when it became underloaded. I have never seen one behave like that.

[sensij]

In my case I have some panels connected in series, with a nominal Voc of about 44 volts. Pretty high for my buck (40V max) but it could probably take it. When I measure the voltage across the panel I can see the controller typically keeps the voltage around 16-18V, so I guess that's where it thinks the MPP is for my load. The LVD I have is simple and will cut off at a configurable voltage interval from 10-12V. You think 10V cutoff and 12V reconnect is enough to prevent a crash and allow the panel to recover?

You know, maybe we misunderstood what was written. If the voltage measured across a single panel (not the entire series string) is 16-18 V, that would be around the typical maximum power point for a 36 cell panel.