How about it? Experiments of the third kind , take 999999.

past.

ET

ox.

to the back box too.

he larger alu box.

temperature

ritium

ature,

s

Neat, Not all plastic will block NIR light. You'll want to check how light tight the box is. (Just 'cause you can't see through it doesn't mean that the photodiode can't.) (sorry for that triple negative.)

George H.

I think the BPW34 has a slightly higher spektral sensitivity, as well as quantum yield. It is a PIN photodiode.

On a sunny day (Mon, 26 Mar 2012 08:24:45 -0700 (PDT)) it happened George Herold wrote in :

Thanks

Yes, I have been holding it against the sun etc. Of course it is the second barrier, the alu box is the first, so it is not going to see a lot of IR. But if measurements turn out to show some effect o sunlight on the setup, then I can add some black paper wraps, paint inside of alu box black, lots of possibilities.

On a sunny day (Mon, 26 Mar 2012 19:09:24 +0200) it happened Johann Klammer wrote in :

Yes, 0.9 electrons per 'photon' IIRC. OTOH the spectrum is wider (to 1100 nm, and we do not want any IR sensitivity. Actually the light from the tritium tube is not really that low in absolute sense, I got good signal with Nichia green LEDs as sensor. I hope the BPW21 will give a lot more signal than the LEDs. As to PIN, speed is no issue, in fact slower is better (integrating over individual flashes of the tritium tube, but I also use a triple RC filter as integrator. So the choice between the BPW34B and the BPM21 was a very conscious one.

larger alu box.

temperature

And one year from now, Jan discovers that the paint he used to obscure one photo-diode was mildly radioactive...

Keep it up and report back in a year. This is truly science.

Design changes:

Acquisition data will be stored in a 24LC256 i2c serial EEPROM (on order). I ran out of enough code space in the 18F14K22 to store 8760 x 2 bytes (24 hours x 365 days), as I have a lot of extra test stuff added. Now I have 32 kB.

As the 18F14K22 SPI is broken, and the i2c uses the same hardware, I wrote i2c software today to write and verify the data, just in case. I replaced the LM335 temp sensor with a normal Si diode, smaller thermal mass, and the about .7Vis right in the middle of the ADC range, the LM has 3.21 V at 45 C, so that would have to be divided, losing accuracy, or some opamp offset circuit, diodes are good temp sensors, used them before. Will calibrate using my thermocouples and EEPROM setpoint. Photocells should be here Wednesday, curious....

Whats broken with the SPI ??

. 4 hours x 365 days),

range,

acy,

ore.

Hey Jan, I was thinking about you design last night. Why the two photodiodes? If you want the lowest leakage then don't bias the photodiode. Just ground one side and run the other into a TIA. You don't need any speed. And with a grounded photodiode and no light the only offset will be from the opamp. (Vos and I-bias* R-feedback)

George H.

On a sunny day (Tue, 27 Mar 2012 13:50:41 -0600) it happened hamilton wrote in :

I would have to look it up, but it was reported here by me last year, google may still have it. Let me look >After waisting some hours to get SPI working with the PIC 18F14K22 connected to

an ENC28J60 Ethernet cotroller,

SPI.

On a sunny day (Tue, 27 Mar 2012 12:54:57 -0700 (PDT)) it happened George Herold wrote in :

The reason is the temperature drift of the dark current, keeping 2 photodiodes in your configuration, but connected the opposite way, one in the dark, eliminates any changes in dark current to a large extend. At least I hope so.

I have made 2 designs, one with TIA as you suggest, I will do some measurement as soon as those diodes arrive, R-feedback is very big... :-)

About measuring light and dark....

After some math, and remembering the great advice from J Larkin CEO of Highland Electronics regarding averaging of ADC values, it looks like I am going to use 2 completely identical gain channels, one with the photo cell (arrived today, man those are big) looking at the tritium light, and the other one with the photo cell painted black, but mounted on the same tritium light to get thermal coupling[1].

Samples are taken of both inputs every second, added to a total for each, and after 3600 seconds, one hour, the total is divided by 3600, and BOTH samples (light and dark channel) will be stored. Storing light an dark values makes it easier later, during data analysis, to find systematic errors. Averaging every time of 3600 samples gets at least rid of some noise. There is also RC filtering.

A third ADC channel measures the voltage drop over a si diode that is also thermally coupled to the tritium light, This temperature channel is also sampled once per second, and used to set a PWM value (PID controller) that drives a IRLZ34 power MOSFET that heats 2 resistors that are also thermal coupled to the tritium tube [2]. The 'D' in the PID is zero, I expect a worst case lock in for the temperature regulator of about 2 minutes, this startup time is not so important, acquisition will not start until the temperature is in range. A fourth ADC channel also sampled 1 x per second, measures the battery backup voltage.

I am thinking about a fifth channel to measure a second external to the PIC reference, (MCP1525, a 2.5 V reference), to be able to detect reference voltage changes of the PIC internal reference. That MCP1525 is there for supply stabilization anyways.

Timing: I have one PIC running on a 32 kHz watch crystal (lightning detector circuit), and that only wakes up when needed, but in this case as we run from a wallwart normally, I will use a 10 MHz crystal, no PLL multiplier, such a still low frequency only consumes a few mA. It is more accurate than the PIC internal oscillator (that one is 2%), so the timing intervals are more the same. In absolute sense it does not matter if the one hour is exactly one hour, because if you stop the experiment after exactly one year, and you find you have n samples, then you know the sample duration was 365 x 24 / n hours. And you can still draw the exact graph of the light variation, all of course if all the sample intervals had the same length, the crystal oscillator makes sure that is to better than 10^-4, no calibration needed.

Now to measure them photo diodes, weekend coming up, soldering iron next to me, most parts have arrived, even did draw a diagram...

[1] May all change after I measured the new photo diodes. [2] Will the PWM changes (because of external temperature changes) affect the input signals? Maybe I should filter and run the heating resistors with DC?

Photo cell test did a simple photo cell test today:

The signal is amplified by a very simple one stage amplifier, a TCL247 CMOS opamp, with a (YES!) 120 MOhm resistor in the feedback. About 6 V supply, no bias on the photo diode. This is what we get if dark (no tritium light):

that is 15 mV.

And this is what we get with the tritium light:

847 mV.

Nice within the ADC range of 1.024 V, and due to the radioactive decay it will only get lower over time.

As I did not screen anything, scoping revealed hum:

This is after connecting 470 pF parallel to the 120 MOhm, bigger capacitor values caused the opamp to oscillate at RF. Of course this is not the final setup and circuit, just needed to get an idea of the signal levels with these BPW21 photo diodes. All very nice, very sensitive photo diodes really, good dark performance too. Exceeds my expectations.

470p ---||--- | 120M | -----====----- | - | || ----------- |\ | || | + | >------------- multi meter -- || --- ---|/ TCL247 | || / \ BPW21 | | || --- /// /// || | || /// The TCL247 allows you to drive the inputs below Vss, common mode range is to -.2 V.

So, now we know some real values, and can proceed.

Jan, I recommend you change your circuit. You want to put the diode on the + terminal rather than have it part of the gain circuit.

Also I would recommend operating diode in reverse biased mode rather than in the non-linear voltage generation mode.

And of course the entire apparatus should be inside copper box to shield from hum.

On a sunny day (Fri, 30 Mar 2012 11:28:36 -0400) it happened " snipped-for-privacy@teranews.com" wrote in :

I think this is not correct. Look at the diode as a current source, after all electrickity is about ELECTRONS, the current is compensated by the current Vout / 120M (R-feedback). There is no 'voltage' at the - input, it is kept at zero by the opamp!

No, again, this is about electrons, all of electronics is about electrons, little charged balls, that move, no other way to understand electronics, everything else is illusion or in short bullshit.

The thing is in a light proof box within an alu box:

I did this test with the alu box closed. Copper is heavy, expensive, and has no advantage over alu here. The parts inside of the black box will be kept at a constant temperature too.

Um, yes, it is a current source. (more or less) The problem is you are running it unbiased and you are putting the diode in the gain-determining feed back. But in a way it doesn't matter be cause no matter how non-linear you've made your circuit, you are only looking for differences anyway. Op amp thermal drift could be a problem, however.

You have already agreed with me above, you just don't understand that to diode will work much better reverse biased (note 10 volts max)to insure that all electrons as you think they are are collected. Look around for some photo diode circuits and you may see what I mean. Running the diode at zero volts will not give any kind of linear stable output.

Yeah, aluminum should work fine. It's just harder to make good contact on seams with aluminum.

On a sunny day (Fri, 30 Mar 2012 14:24:24 -0400) it happened " snipped-for-privacy@teranews.com" wrote in :

No no, do not run in front of the wagon... Let's do some math (EEEEEEHHH), I measure .847 V. The feedback resistor is 120 M, so now we know the current from the diode .847 / 120.10^6, or about 7 nA. (nano ampere). Now lets look up the definition of Ampere:

6.241 × 10^18 electrons / second.

So 7 nA is 7 x 6.241 x 10^9 = 43.687 x 10^9 electrons per second.

The quantum efficiency of the BPW21 photo diode is specified as .8 electrons per photon[1]. So now we know we had .8 x 43.687 x 10^9 = 34.9496 x 10^9 photons per second hitting the photo diode. Nothing 'more or less' linear, just simply linear. I have assumed a perfect opamp here.

(Hope I did the math right, its late here, but just for the curious),

But in a way it doesn't matter be cause no

Right that is why the thing will be kept at 45 C. That includes the opamp, the tritium tube, and the photo diodes. I may go for +40 C, as the leakage of the opamp goes up very fast with temperature. When kept out of the sun it is never hotter than 35 C here.

See the math above.

Never had a problem with that with these boxes. The surface where the parts touch is very big.

Painted boxes *are* a problem.

[1] it is NOT specified as *volts* per photon, but as electrons per photon. I have played a bit with it in just plan daylight (no direct sunlight cloudy), I get 30 uA on a moving coil meter when connected to the BPW21, you could make a crystal radio amplifier with it. 1000 would drive a toy car.. :-)

per photon[1].

Knew it, late, math... (1 / .8) x 43.687 = 54,6 x 10^9 photons per second Or whatever else I did wrong.

per photon[1].

Hi Jan,

Crazy idea for you: to get more dynamic range of the light signal you could accelerate the emitted electrons in the tritium light tube, and target the electrons to a small spot on the surface of the tube to get a brighter light. From wikipedia:

the tritium beta/electron decay energy is about 18keV, so I guess you would need more than this to notice an effect on the electrons.

cheers, Jamie

To reduce drift due to temperature changes caused by external changes in temperature and by heat generated in the electronics, it is important to know what the thermal resistance in degrees centigrade per watt of the little black box with the critical temperature sensitive components is. So that we know how many watts the temperature controller needs to put in to keep temperature constant. A simple way to found that out, is put 1 Watt heat in, and measure the temperature rise. One should of course do that with a 1 watt capable resistor, not with an 1/8 or

1/4 as I did here, as else the consequences are as shown here.

Anyways I found the value. I took a 12 Ohm resistor, and my thermocouple sensor as described here

and put those in the small black box.

The small black box was put into the big alu box surrounded by thermal padding:

Finally the alu box was closed: you can see the heating cable and the thermocouple wires coming out, if only my ISP did not f*ck up my picture uploads, stay clear of godaddy [2]!

Then I fed it with 1 watt of electrickity:

using my lab supply described here: this thing saves you calculating current x voltage, so that makes things so much easier :-)

I waited and waited, and yes I did put up the power to 2.5W for a short while, should not have done that, went back to 1W, and did read out the temperatures after an hour or so.

The cold side was 23.9 and the box was 46.4 (channel 1), makes 22.5 C temperature rise for 1 W input [1] This is what I wanted to know, and opened the alu box to find this: Now should not have done the 2.5 W.... Further inspection after opening the black box showed this:

Anyways, now I can proceed, the neural net has some more ideas of power and reality, sometimes called 'experience'.

[1] The resistor melting into the box and the thermocouple wire touching the box does of course affect the measurement, but in a negative way I think, so the real thermal resistance will probably be higher,

It all does show quite a bit of power is needed to even maintain a delta t of

22.5C, for a 40 C working temperature that would mean 40 - 22.5 = 17.5 C minimum outside temperature for 1 W, this is not acceptable, need at least 2 Watt, makes -5 C minimum, will keep the experiment in house it should never freeze there. The 1 to 2 W power means the batteries will have to deliver this during a mains interruption, for a low voltage of 4 V for the batteries, 2 W is 500 mA, for 1500 mAh batteries this would cover only 3 hours.

Compromises compromises..

Anyways, it shows you can make most experimental stuff for very little money.

[2] If you think the pictures look like shit you are right, godaddy changes them! Comparing what I download back from them to my originals shows they change content, they perhaps, to rip you of some more, run their webservers in 8 or 16 bit mode?

# wget