[Swlug] DIY Geiger counter

Mon Mar 17 11:50:31 UTC 2025

Hi Rhys, just seen this -- must have been delayed for moderation due to being more than 38KB, but glad to see it arrived this morning.

At Z+0000=2025-03-16Sun00:27:12, Rhys Sage via Swlug sent:
> I've put the circuit  together pretty much as in the diagram in the nuclear research document. And I'm getting a background ripple. It's not quite the 50-100 clicks a minute that I was expecting from what I'd read. I had to substitute the 10M resistor for 1M. I have tried other values. I tried 560K instead of the 1M and 5.6K instead of the 10K resistor just to see what would happen.

    Good to experiment, but trying random things without calculating anything will likely lead to safety issues and lots of smoking and failed components.  In general, take caution when trying smaller resistor values.  Whenever I do this, I calculate the current that would flow through the resistor at the upper-bound voltage that that resistor would see, the power dissipation of that, and then consider whether that power dissipation is within the power dissipation rating of the resistor's physical size, often quarter watt or so.  Also, algebra of V = IR and P = IV yield a shortcut formula for this:-

        R = V²/P

-:where R is a lower-bound on the safe resistance value of the resistor in question that we wish to calculate, V is the maximum voltage across said resistor (e.g. 3.9v, 4.7v, or 5v), and P is the resistor's power dissipation rating (e.g. 0.25W).

    E.g., for V = 5v and P = 0.25W, the minimum safe resistance R would be calculated (without dimensional analysis) as (5)²/(0.25) = 25/0.25 = 1/0.01 = 100 -- that's 100 ohms.  So in this case, your 5.6k resistor is fine -- at least it will be able to protect itself from the maximum voltage in your circuit.

    Note also though that resistors limit current for other components, so it is still useful to calculate the upper-bound current through each resistor, follow the path of that current, and check that each component that it is going through can cope with that level of current.  To do this, you will need to know the voltage drop across that component at that level of current.  Determine the worst-case scenario, calculate it, and check it against the component's power dissipation rating in the datasheet, usually given in terms of an ambient temperature or the case temperature, or both.  Some datasheets for some components give the current ratings instead or as well.

    In components with more than 2 leads, you can also have multiple currents with different voltage drops, and you consider the power dissipation rating as a budget, and look at how each current accumulates heat upto that budget total.  Also, for use in environments hotter than 25°C, factor-in the temperature derating charts usually found towards the end of a datasheet.

> The capacitor is spot on at 100nf. Nothing made enough of a difference. Though I can see the ripple on the oscilloscope, I doubt I will be able to read it on an Arduino. I reckon if I can see background radiation on the Arduino then I can count excess radiation.

    Yes.  It is also possible to calculate -- or programme your Arduino to calculate -- the length of time you need to detect a significance at a low level.  I watched a video about this a few weeks ago on a YT channel called Primer, and the video was called something like "How to catch a cheater using mathematics".  It's a very good video that made use of Binomial Theorem to set thresholds on false positives and false negatives, and then calculate the number of samples that you need to satisfy those thresholds.  Very interesting, and it is applicable here to be able to distinguish small readings from background radiation.  You could write it into your driver, perhaps.  It is not very computationally intensive at all, I don't think.

    Mhmm, then again, it contained factorials that can become difficult to calculate for large inputs.  They can be efficiently approximated (Zeita Function springs to mind), but the algorithm is a lot more complicated and I would have to look it up.

> The only differences are the voltage from the Buck converter is 4.7v so the circuit runs from 4.7v not 27v and I'm using a photodiode and not a GM tube.
> 
> Rhys Sage
> 
> [IMG_20250315_193410.jpg  image/jpeg (1986973 bytes)]

    I notice 5 tiny pulses that look the sort of shape I would be expecting -- sharp leading edge with decay thereafter.  But the time range appears to be 1µs per division, 12 divisions across, so 12µs in total.  5 events in 12µs is way too much for your 50--100 Becquerel of background radiation; it would be more like 416667 Becquerel.  So it can't be that.  It's also only -0.2v pulses after about 30k current amplification using an amplifier that should vary by upto about 3.9v, if my calculations are correct (4.7v minus about 0.8v of the Darlington Pair output).  So I wonder whether it is a tiny bit of avalanche noise, perhaps from the photodiode, the transistors, or inside the scope itself.  These are on the order of 5000 times too frequent to be from background radiation, and weaker than we're hoping for as well.  Not only weaker but also barely stronger than the 0.005v or so of the finer noise amplitude.

    I think we can safely ignore all of this as that there is no signal there whatsoever, only noise.  This is not a good place to be because it is difficult to know what to change next, but please see my earlier email for how to divide&conquer this difficult task into some smaller, easier tests.

> [IMG_20250315_162620.jpg  image/jpeg (1326123 bytes)]

    I already sent a screenshot of that page 78 schematic to the list.

    Ah, but no photo of your updated breadboard.  Please do send a photograph of it so that I can inspect for errors or other problems.

    I hope you manage to try some of the suggestions that you do not seem to have tried.  Particularly the control photodiode idea -- if you make/find an optoïsolator, then you can for example, try with and without a capacitor in parallel with it and see whether it provides the overall boost that I predict, as well as the stretch in time, both of which would be useful in your circuit, I think.  Getting loop area right down elsewhere in the circuit is going to help too, and may be necessary to get this kind of circuit working, given how RF-sensitive it is, given that you're trying to detect and count pulses as small as 20 nanoseconds.

Kind regards,
James.
-- 
Wealth doesn't bring happiness, but poverty brings sadness.
Sent from Debian with Claws Mail, using email subaddressing as an alternative to error-prone heuristical spam filtering.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 833 bytes
Desc: OpenPGP digital signature
URL: <http://mailman.lug.org.uk/pipermail/swlug/attachments/20250317/dab582f4/attachment.sig>