Solar Freaks

The older light box, shown here, was not sufficient for my needs. I needed a constant voltage, zero ripple, lightbox for proper electronics unit testing, and the development of a new auto tracking maximiser. The AC lighbox produced a 100Hz light ripple, which was too fast to be noticed with normal eye, but still effected panel output current..

After a few modifications, and a new regulated power source, lightbox 2.0 was born . This is the same light box which I conduct all maximiser testing on, and also the same light box used for scrutineering the NSW event.

As can be seen, it is slightly larger in comparison to the standard AIMSCC light box. It consists of twenty globes in a 5 x 4 arrangement. Initially, I went with the larger/wider box to allow me to test larger panels. However, now, I have added diffused reflector tape, on the underside edges of the cover, decreasing the test area.

We know the edges of the light box provide a weaker light source compared to the center. The tape does two things, it blocks out these weaker edges, providing a much more uniform light source, and intensifies the overall light out of the box.

As can be seen, I also have reflector tape on the inside of the box. The tape is strategically placed within the box, to help provide a uniform light source on the box lid.

Each globe is rated at 50 watts, giving me a combined nominal power of 1KW. However, I am running a slightly higher voltage, to achieve 100% sun and more (up to 110%), giving a total power consumption of 1.15KW. I know, its huge.

During testing at 50% light, I just disconnect every second globe. This way I will have a light spectrum matching that of 100% sun. The light on the lid is still uniform.

The power source is a high end PC power supply (Thermaltake Toughpower 1500), usually used for fast gaming systems. It has 4 individual 12 volt rails, with a combined continuous current of 120 Amps. It has short circuit protection, and soft start. The power supply has been modified and the voltage can be tweaked, via an internal resistive voltage divider feedback circuit. It has extremely low ripple, only 50mV @ 60 KHz (tested with scope) at full load. It will always provide a 12.1V regulated DC voltage (or any other set voltage), regardless of the mains voltage, or load.

It is connected to the light box using 4 pairs of 6mm gold banana bullet connectors, and 4 pairs of 30 amp 12 gauge silicon cable. I used this method, to minimise voltage drop (due to high I²R), across the lamp wiring network.

If you look closely enough on the cables coming out of the supply, you will see two mini Dean’s plugs. These do two things. One is an input for a remote ‘power-on’ signal, and the other is an output, providing 5 volts, to power the Nanometer without batteries. This way, I can use the Nanometer to send the send an ‘ON’ signal, to switch on the 12 volt rails, just before I hit the test button.

This power supply is an extremely efficient switching power supply. It has been tested by the manufacturer and independent sources within the PC community, to achieve efficiencies up to 90%.

Very nice Tony. Good job.

I look forward to your seeing how your auto tracking maximiser turns out. I expect that you will do well with it once you get it on the market.

miseli

Marc, its going to be hard to design a unit more efficient than the Easymax. But as you've mentioned in another topic elsewhere, the main benefit would be the auto tracking, so you can cool panels before the race.

I have the basic prototype completed, and i even sent an early unit to Ian for testing. He confirmed its tracking, under various light levels. That unit i sent him was based on the larger FATMAX, which is designed for high power, like 100watts and more.

There are many changes i want to implement, to both the hardware and the firmware, but im struggling to find the time. Im hoping to have a production prototype completed before the nationals, so i can run a few tests, and maybe you can trial it in your test car.

Not a problem.

I'll try and bring up at least 2 cars (probably this and last year's test cars) along to the nationals so that the 2 can be tested against one another.

I think that they'll be setting up the track on the Thursday, so I was hoping I might rock up at the Science Works on Friday before the national committee meeting in the afternoon for some testing. Ian has sent me a Boxhill unit, so I'll bring that up as well.

I'm interested to see how much of a difference cooling down the panel will really make on the track. I doubt that I'll get around to it at our state event so it's something to look forward to in Melbourne.

You can work out how much you gain from cooling the panel with the formula used to standardise panel temperatures during scrutineering. You may also wish to investigate how much a panel's temperature changes during a race. During some testing I was doing earlier this year we discovered that the new scorpio panels, which are very lightweight, heat up incredibly quickly when on a light box and we were able to watch the power output drop rapidly. I'll see if I can find the numbers I got from that testing, I should have them somewhere.

unussapiens wrote:You can work out how much you gain from cooling the panel with the formula used to standardise panel temperatures during scrutineering. You may also wish to investigate how much a panel's temperature changes during a race. During some testing I was doing earlier this year we discovered that the new scorpio panels, which are very lightweight, heat up incredibly quickly when on a light box and we were able to watch the power output drop rapidly. I'll see if I can find the numbers I got from that testing, I should have them somewhere.

Indeed your right. Although the formula is not 100% accurate, itll still be within +/-5%.
Now input the extra power, which gained from cooling (as seen in the formula), into the simulator and you should see a fairly accurate estimate of the benefits.

Those scorpio panels are light weight, have high efficiency, and have high fill factors. The only downfall is the low thermal capacity, which is inherent with anything light weight, which will make them heat quickly. However, this can also be a benefit, as itll allow you to quickly cool your panel before a race. depends how you look at it.

Itll be good if we can see your results.

As Tony mentioned, you are correct there.

You will then however have to translate this extra power into your torque vs rpm curve/data points for use in Ian's simulator.

You can maybe approximate this by assuming the higher torque vs rpm curve of a slightly higher sunlight % (easily done with my simulator since it allows for sunlight % from anywhere between 5% and 100% to be entered).
However since the voltage is mainly varying with panel temperature, this would be slightly different to just varying the sunlight % which mainly varies the panel's current output.
A better way of doing it might be to hold your curve's torque values constant and vary your rpm values according to the change in power.

Because you are at Boxhill and have Ian's dynamometer test setup at hand, it might be best to just run some actual dynamometer tests.

Even through doing all of this, you're still only running a simulation and you will only ever be able to approximate the effects. That is why I would like to get some on-track testing done if possible.

Please do post some test results if you have any.

miseli

Its a shame we didnt get to test out panel heating/cooling effects . Im going to make this a priority in next years states.
I want to run three tests:

Code: Select all

1) Dont Cool Panel -> Set Maximiser -> Race and record time.
2) Cool Panel -> Dont set Maximiser (leave it set for normal temp) -> Race and record time.
3) Cool Panel -> Re-set Maximiser -> Race and record time.

If time and resources permit, i might build two identical test cars, based on my own rules, just for testing purposes. With two identical cars the difference would easily be seen.

Tony, a few comments on light boxes in general.
The AIMSC light boxes, and each one is different!, are FAR from ideal. They are too small, giving uneven light especially near the edges. The originals used a 3 x 6 array of 16 degree reflector lamps and an autotransformer to drop the voltage to 216vac. To put it mildly the light was not constant across the top of the box. Replacing the lamps with 60 degree versions and adding a prismatic diffuser helps no end but the mains voltage varies from state to state and with the use of extension cords the light level in definitely not constant. Depending on the size of the panel, different amounts of light are reflected back into the box, plus ambient light passes around the edges and into the box, all adding to the effect. If you add a small cell to the top of the box and introduce feedback you will be amazed at the variation obtained by simply placing your hand on top of the box or even just leaning over it. With your modified DC PSU it shouldn't be too hard to add feedback. The thermal inertia of the lamps if pretty high so the light ripple is really quite small. The current produced by the panel is directly proportional to the light so there will be a very small ripple in the current. However, the problem comes from the way some of the early power meters still used in some states operate. They sweep with a constant current loading. Near the maximum power point dV/dI is quite large, so, to supply the constant current there has to be a large voltage ripple. Hence there is a large power ripple as well, leading to errors. The power meter that I designed and that we have used for the last few Nationals uses constant voltage loading, as does yours. Since dI/dV is very small the ripple is now very small as is the power ripple. This leads to a much more accurate result. Also, the old system only took one reading, my box takes 8 readings and averages them so ripple is not an issue.

I remember seeing your light box as tassy last year, and i figured that cell on the top was for feedback. Its a great idea, and i was going to do something similar to mine but time didnt permit. Maybe ill work on it this holidays.

My light box was plagued by the non-uniform light issue around the edges as well, that's why, as you ca see in the pictures, Ive just added diffused reflector tape around the weaker edges. It decreased the overall testing area, but its still sufficient as its a larger light box.

The PSU im using for my light box is a real pain to work with because everything is so tightly packed inside, and its surface mount. I tried other various PSUs, and even 2 smaller 700watt PSUs, before this and they weren't up to the task, as my light box draws nearly 100 amps.

John how do you generate the load in your power meter? Im just using capacitors, is yours the same?

Tony,
My first attempt some years ago used a MOSFET as a constant current load driven by a DAC run by a PIC16F84 but it had problems. Then I tried a PIC16F88 using the PWM to ramp the MOSFET constant current, also with problems. Now I have a small capacitor charged by constant current to give a linear voltage ramp that drives the MOSFET via an op amp. This still has problems if you try to exceed 26V but that's not allowed in the rules anyway. It meant I didn't have to use a big cap and I can discharge it quickly to allow multiple scans. By changing the cap current I can control the scan speed as well.

Ian's box is a bit of a nightmare with different wattage bulbs and separate supplies for each. There is no way he can add feedback. I think you'll find that it's worthwhile if you can do it, even if you have to accept more ripple. There is no way the lamps can track much in the way of frequency.

My prototype unit involved a variable load using a MOSFET and DAC like your initial unit. I too had issues with it.
With the unit im using now, i just put a capacitor across the solar panel, and sample the charge current and voltage at very high speeds. A MOSFET is then used to short the capacitor.

Basic operation is like this:
1) Discharge the CAP using the FET.
2) Unload the CAP, and allow it to start charging from solar panel.
3) Take readings of current and voltage. Calculate power.. blah blah blah...
4) When current equals zero, caps have stopped charging, thus voltage across them is the Voc of the panel.

The Charge current will vary following the solar panels I/V curve. The problem here is you need a fast ADC as the capacitors charge up pretty quick.



The current sense circuit is using the LTC2050 precision current sense amplifier and a 0.01Ohm metal strip resistor.

That was my problem too. I didn't have access to a high speed ADC. I was just trying to make it out of the bits that I had. I had PIC16F88s left over from a job so I used one. By using a separate ramp generator I could make the ramp as slow as I needed and the charge rate was independent of the panel being tested or whether using 100% or 50% Sun. I can drive the FET directly as well to get Isc and Voc. Same thing, I read current across a low ohm resistor and read volts direct, then do the sums. I only had a 2 x 16 LCD display to hand and that restricted what I could show but I have added programming to allow the panel temperature to be input and the corrected power shown. I also added a 100/50% light switch and an automatic panel weight calculator. There is also a relay driver to operate the light box. One thing I hadn't gotten around to was was a 20% weight adder but that's not going to be used next year anyhow. I have a 4 x 20 LCD on my light correcting lap timer but I don't get the time to finish it either. I make my gadgets through hole because I have board loaders and a wave soldering machine. I make way too many to do by hand but I not enough boards to go to the huge expense of buying a surface mount loader and soldering machine.

John, the chip im using is an Atmel mega32. The ADC isn't too fast, its 15Ksps at 10 bit resolution. Im not to familiar with the PIC chips, but i do know that they are similar to the Atmel chips. I initially went for the Atmel chips because their architecture is designed to run efficiently with C, which is what i program in.

Having a controlled ramp is ideal, but as i said, there were to many complexities so i stayed away from that method. With a typical 10 watt panel, my unit will take around 1000 voltage and current samples which is more than enough, and then i do some digital filtering. My unit actually draws the curve, stores it in memory, and dumps it via USB to your computer.

The only issue it has is at low voltages, like 1.5 volts or less, itll only take around 100 samples as the caps reach full charge alot quicker, and if we take into consideration quantization error and noise, we could lose around 3% accuracy. I can overcome this by increasing the size of the capacitor bank.

All my prototyping is through hole, but assembly i do smd as its cheaper, have you seen this company: http://www.wavetronics.com.au/, they do heaps of SMD stuff.

John if you like and i can send you one of my meters to have a look at.

Hi Tony B,
I am interested in investigating the purchase of one of your light boxes for our school (Pacific Lutheran) and I haven't been able to contact you by other means. Can I get you email or phone number so I can get a quote from you for this item?

thanks
Mark _S