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

How about it? Experiments of the third kind , take 999999. Update Assembling the hardware, part 2

Now that the glue has dried, assembling the circuit in the little space available, Put it in a little vice, can rotate it because of the screw:

First added some power wiring, the crystal connections, so the PIC can actually do something:

About halfway I estimate:

To the right the 2 photo cells, both looking at the tritium light. One still needs to be blacked out, could not do it yet because the lenses are so beautifully coated, will be the last thing. To the right of the photo cells the TLC274 opamp, note the small 120M SMDs (top one is clearly visible). Below this opamp is the MCP1525 2.5V reference, also in thermal contact with the alu hotplate, and the extra filter 1u 10k 1u to its - input is already fitted. More to the left the PIC 18F14K22, note the programming header, consisting of 2 parts (3 and 1) made from IC sockets. More to the left the crystal, and above that hardly visible, a 1N4184 diode glued to the hot plate as temperature sensor, that needs some more components (resistor + capacitor). All the way to the right the 2 24LC256 EEPROMS, one on sub address 0, the other wired for sub address 1, i2c connections to PIC already in place... There is a 100k pull down to Vpp pin 4 of the PIC. The box still closes, 3D design here it is :-)

vailable,

tually do something:

s (top one is clearly visible).

ith the alu hotplate,

de glued

resistor + capacitor).

other wired for

Hi Jan, My guess is your biggest headache will be optical alingment. What happens when things wiggle just a bit? How does gravity move the PD's around? I'd want the optics really tight. I'm not sure the best way to do that.

George H.

On a sunny day (Mon, 9 Apr 2012 07:51:23 -0700 (PDT)) it happened George Herold wrote in :

The 'Velpon' glue is largely transparent, and I will use some to secure the position of the photo cells against the tritium tube.

Even now, it is pretty rigid with those short wires. The glue should make it one big solid piece.

I have put all parts in the box, except for the diff pair and LED in the power supply regulator. Have to sleep on that to see if I want the series PNP on the main board, that way no unregulated voltages in to the black box, cool I think I just answered my own question. It is 8 in the evening now here, enough soldering for today, it also needs some more resistors and caps to measure reference and battery.

position

Hi Jan,

It looks great, don't put glue between the photodiode and light source, that is one more variable, even if it starts transparent, a lot of glues will become cloudy over time from what I've noticed anyway.

cheers, Jamie

he position

Ewww, What's Velpon glue? Is it hydroscopic? What if your glue expands and contracts by a few microns with humidity?

Keep the glue as thin as possible!

George H.

ry.

On a sunny day (Mon, 9 Apr 2012 17:37:17 -0700 (PDT)) it happened George Herold wrote in :

Its a common glue here, been available since the fifties IIRC. Dunno whats in it, but if you ask me it is simply plastic dissolved in acetone Once the acetone evaporates the plastic remains. You can make your own plastic glue by scraping some plastic of say a toothbrush (pick your color), and dissolving it in acetone. Sure it may expand and contract, but was it not in a temperature controlled box? Not much humidity will be left at 40 °C in that box, and plastic is not hygroscopic as far as I know.

mmm The tritium light is glued with it, can you see it in the picture? And it used a lot! Velpon does stay a bit flexible if used in a large blob. That will make it shock proof. The idea is not to launch this experiment to orbit or an other planet, it is going to sit in some cabinet in my house where both humidity will be about constant, temperature will be in a reasonable range, and vibration will be minimum. Large accelerations will hopefully not happen either.

On a sunny day (Mon, 09 Apr 2012 17:05:51 -0700) it happened Jamie M wrote in :

Thanks, yes that the glue can become cloudy is a good point I think. I spoiled some glue over the tube of glue cap, and it sort of hardened over the period of a year, had to use huge pliers to get the cap of, it had formed a clear thick seal, not as good as glass, but not cloudy except for sporadic bubbles. I will follow your advice and apply the glue as little in the optical path as possible. The tritium quartz light tube is glued with a lot of it to the alu plate, can you see the glue in the picture?

its almost like this: _ /O\_ ==========

How about it? Experiments of the third kind , take 999999. Update Get ready for programming:

I finished the hardware in the black box, made a small design change for safety: put a 1 k resistor in series with SDA, and use a 12k pull up. This is to avoid any i2c bus conflicts during software tinkering. A good practice.

Had the box powered up, fixed the usual wiring errors, and everything behaves as in the initial tests, of course the photo cells both give a huge signal with the box open.

Added wires for the various signals to the outside world, these exit the box at the bottom:

that hole will have to be made light proof too.

Programming, and that includes code changes, will be done at 5V, for that I made a little connector that fits into my PIC programmer [1]:

Here is the programmer in view on the left, you can see how small the black box really is:

The yellow and brown wire connect to the unregulated 4 to 7.8V supply for the heater of the hot plate, the grey wire on the left is the control voltage for the 3.5 V supply, it will go way high as now that supply is forced to 5V by the programmer, but is not used during programming.

I made a design change so reduce the number of wires, and the number of components in the box: I will use a second PIC located on the main board (yet to be made) listening on logic level RS232 to the one in the box to handle user interface things like thermostat alarm LED, if that feature turns out to be needed can of course be done automatically, but for example a beeper would be cool, still must have some small piezo beepers somewhere. It would be disappointing to find out after one year that it already failed the first week... You gotta be realistic with these things :-)

[1] my PIC programmer and PIC programming software: No I do not use MPlab, I only use gpasm and a scope.

On a sunny day (Mon, 9 Apr 2012 17:37:17 -0700 (PDT)) it happened George Herold wrote in :

George was it you who asked about PID without the 'D'? Here is a nice paper from Microchip:

I am not doing it that way, but it does show that leaving out the 'D' in PID only makes things a bit slower perhaps. I want to avoid D as oscillations is the last thing I need in this experiment. And lock in time is not so important. In a way you could say I left out I too, only use the integral part. Will see if it does as I hope it will, else I will change the soft. That application note uses unsigned 24 bit math too I think, and I use signed 32 bit for everything. On top of that I only have a 1 second timer interrupt where I do things.

As you can see I am now thinking about software again... now that there is hardware to use it with.

How about it? Experiments of the third kind , take 999999. Update Programming, and it works!

The avid reader and PIC expert would have noticed in the previous picture:

The 18F14K22 has 20 pins, and I used the wrong IC socket as connector. Put it in the programmer and id not even notice,,, Vpp goes to the wrong pin..... I expected a dead PIC, but those PICs are strong, it survived, it programmed OK after replacing that connector with a 20 pin one... The idea of breaking out the chip and wiring it again ... not something for people with sensitive nerves and little patience.

Here the programming setup:

The USB stick holds the software sources, I had to move to a different location as the programmer needs a PC with parallel printer port. The little black box (I have loads of these little black boxes they are great for small PIC projects) holds a RS232 to logic level interface. Below that a supply of 18F14K22 Microchip PICs.... Luckily they were not needed for replacement... That pink thing top right has nothing to do with electronics, but is normally used under a cup of coffee.

Most amazing of all, after programming the initial test works! Now that is a rare thing for 5000 lines of code that was not checked on any hardware, I deserve an award for that, Of course I used old libraries I wrote, but even then it amazes me. Connected it to the RS232, and it immediately said hello at the correct baudrate:

I pressed 'D" for debug, and it started outputting the ADC values one per second. The first one is 779 or 780 for measuring the 2.5 V MCP1525 reference against the PIC internal 1024mV reference, there is a 1k versus 2k2 divider, lets see: ( (1000 / (1000 + 2200) ) * 2.5) = .78125 V. If 1.024 V is 1024 steps of the ADC then we should see 781, there is tolerances in both PIC internal reference and the MCP1525 reference, close enough. The 1 step change (one bit jitter) is to be expected, no averaging done in debug mode, just directs the ADC. The second ADC channel is not connected as moved the battery voltage measurement to the PIC on the main board, but it is still in this software version. The next channel is the voltage over the 1N4148 diode temperature sensor, it has 12k to +5 here, a few uA, and this matches what I measured over that diode.

The last 2 channels are the light and dark channel and over-range as the box is open and light floods in.

Wow!

Enough for today I think.

I am so amazed it works first time... :-)

possible.

Its a lovely looking circuit! (don't forget to do ADC dithering maybe)

cheers, Jamie

the first week...

Surely you aren't going to run this for exactly one year? That would be a poor choice indeed if it's a signal with a one-year period you're trying to detect.

Jeroen Belleman

On a sunny day (Wed, 11 Apr 2012 08:46:49 +0200) it happened Jeroen Belleman wrote in :

the first week...

I disagree and fail to see your argument. Is that CERN thinking? Oh me oh my For a periodic signal (think simply a mains 50Hz wave) the best time is one period, as then you end up at the same level you started (after subtracting other known events). With 8760 samples of the 'wave' (if it exists) (265 x 24 for a sample each hour), you get a very nice oversampled representation of an annual wave. But in fact there is a sample taken each second, and stored as 3600 samples averaged each hour.

CERN should be scrapped and the tunnels used for bomb shelters.

IF YOU CANNOT DO IT WITH THOSE TINY PARTICLES ON THE LAB DESKTOP THEN YOU CANNOT DO IT IN A MACHINE THE SIZE OF THE UNIVERSE.

!!!!!!!!!!!!!

the first week...

I'll ignore your snide remarks.

The signal you're looking for is necessarily small, hidden in noise, or this would have been a well-known and commonly acknowledged effect. You're not exactly the first to look for this, you know. I've seen publications claiming yearly decay rate variations in the 1e-3 ballpark.

So, to bring out a signal with a one-year period, you're going to have to apply a Fourier transform to your recording, and see if a line at the 1/year frequency stands out from the background. If your record length is exactly one year, it *will* have a line at that frequency. How will you know it's a true signal you are looking at then?

Besides, with a record that contains only one period of the signal you are seeking, if it exists at all, you certainly won't have enough data to draw any useful conclusions.

Jeroen Belleman

The first sanity check will be whether or not it exhibits diurnal variation in recorded intensity. Until all known systematic errors from ambient thermal fluctuations have been eliminated anything else is a complete waste of time. He should probably look at the initial data daily, weekly then monthly and quarterly as the experiment progresses to get a feel for how it behaves and refine the signal to noise.

In a domestic environment I'd expect to see a 5+2 weekly thermal signal in addition to the diurnal variation.

It might be good enough to use a low pass filter and MkI eyeball if there is enough data. He should be able to measure an approx halflife in under a month if the kit is anything remotely like up to snuff.

OTOH he knows a priori that he is looking for a signal at exactly 1 solar year and with a functional form governed by distance from the sun so he can fit that to his corrected data and get a direct estimate. My money is on an answer that is not distinguishable from the average of measurement noise over the period if the experiment is done correctly.

And that any observed variations will be strongly correlated with the ambient room temperature and/or the external one. This applies to all the purported detections so far published in the literature too. It would be easy enough to check this by running a similar experiment in Australia or South America where seasonal temperatures and solar distance are in the opposite phase to in the northern hemisphere.

Iff both northern and southern groups show the same behaviour then I might accept there is some solar influence on halflife, but *NOT* until.

--
Regards,
Martin Brown

On a sunny day (Wed, 11 Apr 2012 11:49:28 +0200) it happened Jeroen Belleman wrote in :

the first week...

Yes I have read the publications, several.

No that is not correct 'have to use a F transform',

That is not correct either, if a periodic effect is present then it is present in that year.

To continue with th text you did not seem to like: I am not into looking for a 'maximum' like LIGO, just ever increasing sensitivity just to probe Einstein right with gravity waves, or an idiotic project like ITER that is a tokamak that will never produce any energy and costs billions.

I want to see a big signal, and if I do not see this signal after one year then as far as I am concerned there is no such signal, as then it would be of no use to me. JUST like (imagine all caps here) it is no use to drive 2 Mercedeses SL at top speed into each other as a way to get a deeper understanding of the combustion engine and car electronics from the parts that fly about LIKE CERN. Sure they need a F transform just to see the day and night interval in the sunlight Like I said: All it is good for is bomb shelters. :-)

the first week...

suppose I provide you a signal which you decide is going to be a 50Hz periodic signal and so measure 20ms worth of it and decide it's a 50Hz sawtooth.

but the signal was actually a 1Hz sine wave and you only sampled part of it...

--
?? 100% natural

--- Posted via news://freenews.netfront.net/ - Complaints to news@netfront.net

there are several other ways to look for a perodic signal but they are not significantly different from the fourier transform. converting time domain data to frequency domain data is at best a fourier tranform, this is by definition.

it's present, but it is indistinguishable from lower frequency signals,

you'll get that, it's called noise.

--
?? 100% natural

--- Posted via news://freenews.netfront.net/ - Complaints to news@netfront.net

the first week...

But to be fair there is good reason here to believe that there is a possible annual variation and armed with that a priori knowledge you can test the hypothesis that the periodic signal is above the noise.

I know what result I expect to see. But that doesn't invalidate the method. You would obviously get a null or ambiguous result if it actually depends on the position of Jupiter instead.

In practice at first he is almost certainly going to see a 24 hour period roughly truncated sine wave modulated with a 7 day period 5:2 mark space ratio rectangular wave and whatever seasonal temperature variations do to his experiment. Worth monitoring mains voltage, atmospheric pressure, relative humidity, ambient and external temperature to check for systematic errors introduced by the environment.

Until the kit is demonstrably stable at measuring intensity over a weekly cycle there isn't any point in going any further. The interesting question is how long a run of data will it take to be able to determine the half life of tritium to within 0.05% or better.

--
Regards,
Martin Brown