-
Welcome back Guest! Did you know you can mentor other members here at H-M? If not, please check out our Relaunch of Hobby Machinist Mentoring Program!
Needing more than a spark test?
- Thread starter graham-xrf
- Start date
I'm curious regarding what JFET you have decided to use for the TIA. Also, are you going to use a cascode topology to minimize the Miller capacitance, or is that a non-issue in this application?
Your physical arrangement of americium capsules and lead is pretty close to what I've been thinking about. I don't think the Am241 capsules have to point directly at the object under test because the 60Kev x-rays are uniformly emitted in all directions. Ones that are emitted into the body of the capsule (or lead shielding) are absorbed. It probably is more important to get them as close as possible to the sample as possible, so your idea of removing the small bit that actually has the Am241 would be optimal in this regard (but see my last paragraph below). Drill or punch a 14 mm diameter hole in your lead sheet (the smallest diameter a 10 x 10 mm square would go through) and arrange the emitters in a circle around the hole, but not overhanging it. That arrangement will provide a relatively high gamma-ray flux for exciting XRF without the need of making or forming a hemispherical mounting surface. Placing the detector as close as possible to the opposite side of the hole will give us the least attenuation due to inverse-square considerations, and none of the primary 60Kev gammas will make it to the detector.
My one concern regarding messing around with the Am241 capsules is inadvertently releasing some americium particles. Ingesting an alpha and gamma-ray emitter doesn't sound like a good idea to me. The extraction would have to be surrounded by all sorts of caveats to avoid possible legal problems down the road. "Don't really do this, it is presented for educational purposes only" and so on.
Your physical arrangement of americium capsules and lead is pretty close to what I've been thinking about. I don't think the Am241 capsules have to point directly at the object under test because the 60Kev x-rays are uniformly emitted in all directions. Ones that are emitted into the body of the capsule (or lead shielding) are absorbed. It probably is more important to get them as close as possible to the sample as possible, so your idea of removing the small bit that actually has the Am241 would be optimal in this regard (but see my last paragraph below). Drill or punch a 14 mm diameter hole in your lead sheet (the smallest diameter a 10 x 10 mm square would go through) and arrange the emitters in a circle around the hole, but not overhanging it. That arrangement will provide a relatively high gamma-ray flux for exciting XRF without the need of making or forming a hemispherical mounting surface. Placing the detector as close as possible to the opposite side of the hole will give us the least attenuation due to inverse-square considerations, and none of the primary 60Kev gammas will make it to the detector.
My one concern regarding messing around with the Am241 capsules is inadvertently releasing some americium particles. Ingesting an alpha and gamma-ray emitter doesn't sound like a good idea to me. The extraction would have to be surrounded by all sorts of caveats to avoid possible legal problems down the road. "Don't really do this, it is presented for educational purposes only" and so on.
I powered up my signal conditioning board today to see how it works. No problems, just the expected relatively-poor offset drift in the 100X gain mode. I also connected it up to my pocketgeiger and confirmed that the signal path is OK.
I went back to Graham's teardown of the Am241 capsule and belatedly realized that my previous "live" test of the pocketgeiger had the capsule turned backwards so it's not surprising that the count rate was rather low. My capsules are fairly easy to disassemble so I did that to one of them and checked what the count rate was. Much higher. This was with the copper shield still in place, so I expect the count rate to almost double when I replace it with aluminum foil. I bought a metal HF "ammo box" awhile back to put noise-sensitive circuits inside it so that's available if need be.
I'd be happier if the pocketgeiger's noise level was lower. I'm thinking about replacing the LMC662 with an LTC6269. At 25C its bias current is as low as the LMC662 but it's quieter.
I went back to Graham's teardown of the Am241 capsule and belatedly realized that my previous "live" test of the pocketgeiger had the capsule turned backwards so it's not surprising that the count rate was rather low. My capsules are fairly easy to disassemble so I did that to one of them and checked what the count rate was. Much higher. This was with the copper shield still in place, so I expect the count rate to almost double when I replace it with aluminum foil. I bought a metal HF "ammo box" awhile back to put noise-sensitive circuits inside it so that's available if need be.
I'd be happier if the pocketgeiger's noise level was lower. I'm thinking about replacing the LMC662 with an LTC6269. At 25C its bias current is as low as the LMC662 but it's quieter.
Hi Mark
Firstly - do tell what it is on your signal conditioning board?
Yes indeed, the LTC6269 is what I settled on also, and it can be used for a TIA directly. It works way better if the Rf is made 510K//2.2pf, and followed by another inverting stage of gain 180, using 180K and 1K. The noise is 4.3nV/√Hz.
If the FET circuit were adopted, so the other half of the LTC6269 is used for the FET bias loop, the noise goes somewhere very low, allegedly 900pV/√Hz, but I think lower. I was after something that would range to the noise floor. On a output signal (biggest) of 4V, then 90dB down is 126uV. On those numbers, one does not need the FET circuit at all. The LTC6269 in 2 stages should be enough.
Of course, the Pocket Geiger's 66Mohm TIA followed by a stage gain of 100 is just ridiculous, but since the pin-out of the LMC662 seems standard, can we simply "adapt the Pocket Geiger 5 circuit board, and duck the need to make another board altogether? I say this because a differential driver , which can also have some noise filtering functions, might go in the place of U3A and U3B, no longer as comparators, can easily drive a relatively short shielded twisted pair to a A/D converter mounted right at the computer, so making SPI at high speed somewhat easier.
LM662 cannot do this! It has a GBW of 1.4MHz, a noise of 22nV/√Hz, a Vos of 3mV. About the only thing that looks good is the 2fA input bias current.
About the FETs
The IFN 147 is the InterFET name for the old 2SK147 except they market it from Mouser under the code SMP147 for £5.86, or £4.72 each for 10.
The very old datasheet has the very same simulation parameters. The reverse transfer capacitance is a whole 15pF.
The whole bunch of FETs from Linear systems, LSK389 (dual), LSK170(single) etc are all lower capacitance Ciss (20pF-25pF) and Crss (5pF-5.5pF) Reverse Transfer Capacitance, which would make the higher frequency devices than IFN147. I left off checking out all Linear Systems stuff because of the purchase terms, and because of importation difficulties.
The gain-bandwidth product of the FET+ one LT1806 TIA opamp is 2.4GHz. I do not propose to put as much as 1Megohm in the first feedback loop. I settled for 270K to 510K, then followed by another op-amp stage with gain 180. The FETS are basically audio devices, and I realized just about any low noise FET would do, the better if one used a VHF or UHF low noise FET.
In extremes, even one of those low noise microwave FETs with 28K (NF=0.4dB) could be great, but expensive. The problem is how to use local feedback to knobble it's ability to have gain beyond UHF without having it become a superior loud audio square-wave oscillator!
In simulation, I used the LSK389, only because the model was conveniently already there in LTSpice, but I knew one could easily use any of a whole lot of others. I was looking at some costing less than £1. My simulation norm was a photodiode current pulse (Iph) delivered as a 100nS pause, then exponential function rise in 100nS, and a 200nS dwell as the charge goes over the hump, partly into 230pF diode capacitance, while being leaked by the 40Mohm diode bulk resistance, and pulsing into the amplifier. Then a 500nS exponential function fall with a tail extending to 10uS.
Iph could be from 500pA to 5nA, as a starting guess. If as high as 20nA, we need less gain.
Using these (audio-ish) FETs, the 4V output pulse is peak-delayed by about 2uS, and is "stretched" to about 14uS. It can do better. I just stopped tweaking. I may have to lower the gain if 5nA is too little. At the low end, I was using 500pA. At this stage, I ceased to worry about FETs . I would need a reminder where is the cascode circuit example if you think we need to go there. Crss less than 2pF is probably find-able in a FET. Nearly all discrete FETs around now seem hail from last century, even if they were recently made. I may have a little trawl for more low cost UHF FETS, but I will have to edit in my own .MODEL properties.
A little noise spreadsheet
It is a trivial calculation, but help yourself to the noise temperature spreadsheet. I expect you have LibreOffice Calc. Hmm wait up..
It seems .ODS, and .XLSX are not allowed extensions. I had to use .ZIP.
- - - - - - - - - - - -
The Illumination Geometry
About the photons being emitted in all directions is understood. The source only has to be sunk that tiny bit, or the sensor if you like, so that a sideways photon, unable to manage more than a right angle, cannot hit the sensor by refraction. A tiny "lip extension" on the tube around the sensor is enough, allowing then most of the hemisphere of radiation to play into the space, without wasting any being too far up a lead setting.
In my design, the axes of the radiation sources is not as in the image of the post #460. They are tilted to aim at the centre where the test metal is. The angle depends on the distance to the metal, and the radius of the ring of sources, but maximum effect is when that radius equals the height to the test stuff, and the angle is 45°. This choice is modified by how big a lump of lead we are willing to put up with. It is OK to make it all of something else, plastic even, but lined with lead, say 2mm or 3mm.
Re: Extraction of the Radioactive Sources
OK - getting down and dirty with it. I have already extracted one down to the stage of looking like those in the CAD image. The dimensions are not made up. That was real measured. We cannot rely on common sense, but we can easily find a one-liner, or a URL for a user to get some warnings. We can include with a warning not to eat the thing, nor let one fall into becoming playthings for children. So long as you don't ingest it, the risk is pretty much zero. Of course, we are not letting free the Am241 pellet. It always remains in it's little metal disc mounting.
I used a small file to take out the 3 outermost press-deform (peen?) fixings from the sides. For convenience, I post again the three most relevant images, one of them modified to show the file cut line. I think no matter what brand of smoke detector, the way these things are held down is always much the same. Everything about keeping hold of radioactive sources gets to very fundamental principles of shapes. Glues are never used.
To that end, we may have to pay attention to how we fix them in our gadget.
I put the narrow edge of the file down onto outer disc, and filed away the 3 peened over spots. I admit it is a bit disconcerting to be filing away on a radioactive thing plonked down onto a green cutting mat held down with the fingers. In the end it was easier to to hold it on left forefinger and keep filing. It does not take long.
From the side
I thought to take a side-cutter snips and cut across, and I nearly fouled up, not thinking that it would have to cut through the bigger disc of the setting, around the back. But then, I realized one could use the big cutters just to get a grab on it, and bend the outer metal curled around to just force it to let go. The radioactive pellet mounting dropped free.
From the back
After a minute or so, one's technique and options improve with practice. You could probably mangle the outer metal enough without filing, but if you don't want to have it ping away somewhere, then file away the squeeze-downs. It only takes a few minutes. I did it dirty, but one could put it over a hole drilled in some metal, large end down, and then punch or press it out. While I did not actually have my finger skin in direct contact with the Am241 thing in the middle, I got as close as 1mm or so.
Tip:
DO NOT stash the smoke detector pellets in the same bag as Thorium gas mantles, nestling right up against each other!
I don't know for sure what happens, but I think that gets close to making third substances. I cannot imagine we have anything close to breeder stuff here, so we have to ask @RJSakowski
Firstly - do tell what it is on your signal conditioning board?
Yes indeed, the LTC6269 is what I settled on also, and it can be used for a TIA directly. It works way better if the Rf is made 510K//2.2pf, and followed by another inverting stage of gain 180, using 180K and 1K. The noise is 4.3nV/√Hz.
If the FET circuit were adopted, so the other half of the LTC6269 is used for the FET bias loop, the noise goes somewhere very low, allegedly 900pV/√Hz, but I think lower. I was after something that would range to the noise floor. On a output signal (biggest) of 4V, then 90dB down is 126uV. On those numbers, one does not need the FET circuit at all. The LTC6269 in 2 stages should be enough.
Of course, the Pocket Geiger's 66Mohm TIA followed by a stage gain of 100 is just ridiculous, but since the pin-out of the LMC662 seems standard, can we simply "adapt the Pocket Geiger 5 circuit board, and duck the need to make another board altogether? I say this because a differential driver , which can also have some noise filtering functions, might go in the place of U3A and U3B, no longer as comparators, can easily drive a relatively short shielded twisted pair to a A/D converter mounted right at the computer, so making SPI at high speed somewhat easier.
LM662 cannot do this! It has a GBW of 1.4MHz, a noise of 22nV/√Hz, a Vos of 3mV. About the only thing that looks good is the 2fA input bias current.
About the FETs
The IFN 147 is the InterFET name for the old 2SK147 except they market it from Mouser under the code SMP147 for £5.86, or £4.72 each for 10.
The very old datasheet has the very same simulation parameters. The reverse transfer capacitance is a whole 15pF.
The whole bunch of FETs from Linear systems, LSK389 (dual), LSK170(single) etc are all lower capacitance Ciss (20pF-25pF) and Crss (5pF-5.5pF) Reverse Transfer Capacitance, which would make the higher frequency devices than IFN147. I left off checking out all Linear Systems stuff because of the purchase terms, and because of importation difficulties.
The gain-bandwidth product of the FET+ one LT1806 TIA opamp is 2.4GHz. I do not propose to put as much as 1Megohm in the first feedback loop. I settled for 270K to 510K, then followed by another op-amp stage with gain 180. The FETS are basically audio devices, and I realized just about any low noise FET would do, the better if one used a VHF or UHF low noise FET.
In extremes, even one of those low noise microwave FETs with 28K (NF=0.4dB) could be great, but expensive. The problem is how to use local feedback to knobble it's ability to have gain beyond UHF without having it become a superior loud audio square-wave oscillator!
In simulation, I used the LSK389, only because the model was conveniently already there in LTSpice, but I knew one could easily use any of a whole lot of others. I was looking at some costing less than £1. My simulation norm was a photodiode current pulse (Iph) delivered as a 100nS pause, then exponential function rise in 100nS, and a 200nS dwell as the charge goes over the hump, partly into 230pF diode capacitance, while being leaked by the 40Mohm diode bulk resistance, and pulsing into the amplifier. Then a 500nS exponential function fall with a tail extending to 10uS.
Iph could be from 500pA to 5nA, as a starting guess. If as high as 20nA, we need less gain.
Using these (audio-ish) FETs, the 4V output pulse is peak-delayed by about 2uS, and is "stretched" to about 14uS. It can do better. I just stopped tweaking. I may have to lower the gain if 5nA is too little. At the low end, I was using 500pA. At this stage, I ceased to worry about FETs . I would need a reminder where is the cascode circuit example if you think we need to go there. Crss less than 2pF is probably find-able in a FET. Nearly all discrete FETs around now seem hail from last century, even if they were recently made. I may have a little trawl for more low cost UHF FETS, but I will have to edit in my own .MODEL properties.
A little noise spreadsheet
It is a trivial calculation, but help yourself to the noise temperature spreadsheet. I expect you have LibreOffice Calc. Hmm wait up..
It seems .ODS, and .XLSX are not allowed extensions. I had to use .ZIP.
- - - - - - - - - - - -
The Illumination Geometry
About the photons being emitted in all directions is understood. The source only has to be sunk that tiny bit, or the sensor if you like, so that a sideways photon, unable to manage more than a right angle, cannot hit the sensor by refraction. A tiny "lip extension" on the tube around the sensor is enough, allowing then most of the hemisphere of radiation to play into the space, without wasting any being too far up a lead setting.
In my design, the axes of the radiation sources is not as in the image of the post #460. They are tilted to aim at the centre where the test metal is. The angle depends on the distance to the metal, and the radius of the ring of sources, but maximum effect is when that radius equals the height to the test stuff, and the angle is 45°. This choice is modified by how big a lump of lead we are willing to put up with. It is OK to make it all of something else, plastic even, but lined with lead, say 2mm or 3mm.
Re: Extraction of the Radioactive Sources
OK - getting down and dirty with it. I have already extracted one down to the stage of looking like those in the CAD image. The dimensions are not made up. That was real measured. We cannot rely on common sense, but we can easily find a one-liner, or a URL for a user to get some warnings. We can include with a warning not to eat the thing, nor let one fall into becoming playthings for children. So long as you don't ingest it, the risk is pretty much zero. Of course, we are not letting free the Am241 pellet. It always remains in it's little metal disc mounting.
I used a small file to take out the 3 outermost press-deform (peen?) fixings from the sides. For convenience, I post again the three most relevant images, one of them modified to show the file cut line. I think no matter what brand of smoke detector, the way these things are held down is always much the same. Everything about keeping hold of radioactive sources gets to very fundamental principles of shapes. Glues are never used.
To that end, we may have to pay attention to how we fix them in our gadget.
I put the narrow edge of the file down onto outer disc, and filed away the 3 peened over spots. I admit it is a bit disconcerting to be filing away on a radioactive thing plonked down onto a green cutting mat held down with the fingers. In the end it was easier to to hold it on left forefinger and keep filing. It does not take long.
From the side
I thought to take a side-cutter snips and cut across, and I nearly fouled up, not thinking that it would have to cut through the bigger disc of the setting, around the back. But then, I realized one could use the big cutters just to get a grab on it, and bend the outer metal curled around to just force it to let go. The radioactive pellet mounting dropped free.
From the back
After a minute or so, one's technique and options improve with practice. You could probably mangle the outer metal enough without filing, but if you don't want to have it ping away somewhere, then file away the squeeze-downs. It only takes a few minutes. I did it dirty, but one could put it over a hole drilled in some metal, large end down, and then punch or press it out. While I did not actually have my finger skin in direct contact with the Am241 thing in the middle, I got as close as 1mm or so.
Tip:
DO NOT stash the smoke detector pellets in the same bag as Thorium gas mantles, nestling right up against each other!
I don't know for sure what happens, but I think that gets close to making third substances. I cannot imagine we have anything close to breeder stuff here, so we have to ask @RJSakowski
Last edited:
Hi Mark
Did I, at any stage, gather you had a preference for parallel loading ADC?
I have been trawling mouser.co.uk
The very filtered-down choices are:
AD4000. In stock in 4 days. £21.67
Pros: It has pseudo differential input with fast settling drivers built-in, saving 2 external op-amps.
The price is sort of OK. You need to look to USA pricing. e.g. This one is $28.43 in a MSOP (RM10) package.
Cons: It speaks SPI only, needing 70MHz clocking, on 3 wires, and best lives right up close to computer kit.
It needs an external low noise Vref, and likely separate conditioned supply regulator (which might be a need for any)
I thought 4.096V, supplied by a low noise 5V LDO regulator that is used for the op-amps as well
- - - - - - - - - - - -
AD4001 is a bit lower cost. £18.65 or $24.46 in USA
It is just like the AD4000, except for one extra con, in that it needs a ADC differential driver - meaning 2 op-amps + noise filter parts.
AD4807-4 at £4.82 is OK for this role, and leaves spare op-amps for TIA signal conditioning.
- - - - - - - - - - - - -
LTC2370-16IMS at £35.99 It seems only the industrial temperature version is in stock. $47.21
Pros: It has pseudo-differential input with internal drivers, tempered by the recommendation to use a buffer anyway
It clocks conversion internally, with no pipeline latency. 76uV at LSB. 94dB SNR @2kHz impressive.
Cons: It requires an external low noise reference
It does SPI only, should one prefer parallel.
It is waaay too expensive (for me).
- - - - - - - - - - - - -
AD7622BSTZ costing £29.68 in UK, or for USA, it's $47.21 in singles.
Pros: It has onboard 2.048V buffered reference onboard, saving lots of low noise external regulator and bandgap reference.
It does both SPI and parallel 16-bit data ofload as a mode choice. One can play with either if high-speed serial is awkward in software dev.
I says SPI/ QSPI/MICROWIRE/DSP compatible. Except SPI, I don't have experience of these
It has 92.5dB SNR, and has no pipeline delay (I think it has a buffer register).
Cons: It still needs the external differential fast-settling driver.
It is still expensive, but this time the internal reference feature and parallel data may offset the extra circuit hassle.
- - - - - - - - - - - - - -
I was thinking the AD4000 would be an OK choice provided one can program to deal with 70MHz clocked serial data.
I am thinking to get a LTC2370-16, and also a AD4000, for development.
While I wait for this stuff to arrive, I may have another go at the physical build, and dig out the remaining Am241 sources.
- - - - - - - - - - - - - - -
Re: The home front:
The contractor visited yesterday afternoon. He knows exactly what I want, and made some suggestions. In this, there was a little "project creep".
Rules
The regulations, and boundary rules apply, being as I am in a National Park/ Site of Scenic Beauty / part Sites of Special Scientific Interest.
The outbuilding can be 5m to 5.5m long, and 3.5m to 3.8m wide. In USA-speak, that is about 18ft x 12ft.
The height to internal ceiling can be standard 2.5m or so, and total height not more than 4m. That allows a roof pitch of 35° to 40°.
It is to be on 150mm steel reinforced concrete, with the hardstand extended to put a surround of 1.2m of the same patio paving blocks.
It will have 50mm wall battens inside, with Celotex polyurethane insulation, and ply over (good to hang tools on).
Insulated
The floor is likely 35mm chipboard on 50mm bearers, and insulation under, and damp membrane. If I need machine hard points, I cut through for pucks with chemi-bolts.
He proposes 150 x 50 joists with 150mm insulation, and boarding for a storage up top, like a closed mezzanine if you like.
The whole lot will use up the remaining tiles I have that match the main house, but I have to get more.
Power
For 240V 1ph, I will be getting 50m x 6mm2 or maybe 10mm2 SWA armoured cable for direct burial, along with 25mm blue polythene (farm style) water supply.
Water
25mm supply from house. There will be a drain soakaway, but I cannot put any foul or poisoned stuff in that. Workshop poisons have to be taken away separately. We have gutters + water butts for garden use.
So, insulated, with window views to South and South-West. Out back there, it looks directly onto farm fields. It has mini patio-style surround, partly shrouded by flowers, shrubs, fruit trees, whatever, designed to "blend in". Then for inside, a mix of garage/garden tool overspill, + machine(s) & lab/hobby/man cave space. By many HM standards, I suppose it might be considered "a bit small", but it is bigger than the half a garage. Not really a "shed", but new build in a National Park just has to be made that bit classier, and not look like it was thrown together from a load of palettes and corrugated iron. It is waaay more expensive than my first idea, but if I am going to play with the stuff I do, there is little choice. The little silver lining is the possibility that I can offset some of this stuff, and the machinery in it, against tax. It's a "development lab" full of RF instruments and scientific development stuff for X-Ray Fluorescent materals identification, and "Nuclear Instrumentation Science".
Did I, at any stage, gather you had a preference for parallel loading ADC?
I have been trawling mouser.co.uk
The very filtered-down choices are:
AD4000. In stock in 4 days. £21.67
Pros: It has pseudo differential input with fast settling drivers built-in, saving 2 external op-amps.
The price is sort of OK. You need to look to USA pricing. e.g. This one is $28.43 in a MSOP (RM10) package.
Cons: It speaks SPI only, needing 70MHz clocking, on 3 wires, and best lives right up close to computer kit.
It needs an external low noise Vref, and likely separate conditioned supply regulator (which might be a need for any)
I thought 4.096V, supplied by a low noise 5V LDO regulator that is used for the op-amps as well
- - - - - - - - - - - -
AD4001 is a bit lower cost. £18.65 or $24.46 in USA
It is just like the AD4000, except for one extra con, in that it needs a ADC differential driver - meaning 2 op-amps + noise filter parts.
AD4807-4 at £4.82 is OK for this role, and leaves spare op-amps for TIA signal conditioning.
- - - - - - - - - - - - -
LTC2370-16IMS at £35.99 It seems only the industrial temperature version is in stock. $47.21
Pros: It has pseudo-differential input with internal drivers, tempered by the recommendation to use a buffer anyway
It clocks conversion internally, with no pipeline latency. 76uV at LSB. 94dB SNR @2kHz impressive.
Cons: It requires an external low noise reference
It does SPI only, should one prefer parallel.
It is waaay too expensive (for me).
- - - - - - - - - - - - -
AD7622BSTZ costing £29.68 in UK, or for USA, it's $47.21 in singles.
Pros: It has onboard 2.048V buffered reference onboard, saving lots of low noise external regulator and bandgap reference.
It does both SPI and parallel 16-bit data ofload as a mode choice. One can play with either if high-speed serial is awkward in software dev.
I says SPI/ QSPI/MICROWIRE/DSP compatible. Except SPI, I don't have experience of these
It has 92.5dB SNR, and has no pipeline delay (I think it has a buffer register).
Cons: It still needs the external differential fast-settling driver.
It is still expensive, but this time the internal reference feature and parallel data may offset the extra circuit hassle.
- - - - - - - - - - - - - -
I was thinking the AD4000 would be an OK choice provided one can program to deal with 70MHz clocked serial data.
I am thinking to get a LTC2370-16, and also a AD4000, for development.
While I wait for this stuff to arrive, I may have another go at the physical build, and dig out the remaining Am241 sources.
- - - - - - - - - - - - - - -
Re: The home front:
The contractor visited yesterday afternoon. He knows exactly what I want, and made some suggestions. In this, there was a little "project creep".
Rules
The regulations, and boundary rules apply, being as I am in a National Park/ Site of Scenic Beauty / part Sites of Special Scientific Interest.
The outbuilding can be 5m to 5.5m long, and 3.5m to 3.8m wide. In USA-speak, that is about 18ft x 12ft.
The height to internal ceiling can be standard 2.5m or so, and total height not more than 4m. That allows a roof pitch of 35° to 40°.
It is to be on 150mm steel reinforced concrete, with the hardstand extended to put a surround of 1.2m of the same patio paving blocks.
It will have 50mm wall battens inside, with Celotex polyurethane insulation, and ply over (good to hang tools on).
Insulated
The floor is likely 35mm chipboard on 50mm bearers, and insulation under, and damp membrane. If I need machine hard points, I cut through for pucks with chemi-bolts.
He proposes 150 x 50 joists with 150mm insulation, and boarding for a storage up top, like a closed mezzanine if you like.
The whole lot will use up the remaining tiles I have that match the main house, but I have to get more.
Power
For 240V 1ph, I will be getting 50m x 6mm2 or maybe 10mm2 SWA armoured cable for direct burial, along with 25mm blue polythene (farm style) water supply.
Water
25mm supply from house. There will be a drain soakaway, but I cannot put any foul or poisoned stuff in that. Workshop poisons have to be taken away separately. We have gutters + water butts for garden use.
So, insulated, with window views to South and South-West. Out back there, it looks directly onto farm fields. It has mini patio-style surround, partly shrouded by flowers, shrubs, fruit trees, whatever, designed to "blend in". Then for inside, a mix of garage/garden tool overspill, + machine(s) & lab/hobby/man cave space. By many HM standards, I suppose it might be considered "a bit small", but it is bigger than the half a garage. Not really a "shed", but new build in a National Park just has to be made that bit classier, and not look like it was thrown together from a load of palettes and corrugated iron. It is waaay more expensive than my first idea, but if I am going to play with the stuff I do, there is little choice. The little silver lining is the possibility that I can offset some of this stuff, and the machinery in it, against tax. It's a "development lab" full of RF instruments and scientific development stuff for X-Ray Fluorescent materals identification, and "Nuclear Instrumentation Science".
Last edited:
From an earlier question: my signal conditioning board uses an LMC662 and a TL072 to provide selectable gains of 1, 10 and 100. I really only need the LMC662 for the input amplifier, which has a capacitively-coupled input with a low pass corner around .1Hz. Probably overkill. It looks to me like most of the noise coming out of my board is what's coming out of the pocketgeiger board, so there probably isn't much to be gained by replacing those amplifiers with more expensive, lower noise types.
Regarding your question about serial or parallel-output ADCs, I really have no preference. It is mostly dictated by what kind of interface the processor board has got. A DSP chip with some built-in ADC interface features might be ideal for this, but some like the MSP430 launchpad don't directly support 16 bit I/O (at least, not by my superficial reading of the data sheet). I think the processor for the Teensy4 line can, but the board designer didn't lay it out to support a contiguous set of 16 bit I/O's.
So in general, the serial data approach seems to be more widely available, but more problematic due to the faster serial clock and data signals. While looking into this problem, I found a good article here that offers an elegant solution, but only if your microcontroller board can support two SPI channels, one acting as a master and the other as a slave. The master channel would be used to set up the ADC and the slave channel would receive the data. Clock and data delays in the slave channel remain synchronized because they travel the same distance back to the processor. I note that the master's SCK basically is turned around at the ADC end so there would be a few additional nS of delay if you need to buffer it. However, that could be matched by buffering the data coming out of the ADC as well. You could even go the LVDS route if you want to.
Your workshop sounds like it will turn out pretty nice. What national park will be your machining backdrop?
Regarding your question about serial or parallel-output ADCs, I really have no preference. It is mostly dictated by what kind of interface the processor board has got. A DSP chip with some built-in ADC interface features might be ideal for this, but some like the MSP430 launchpad don't directly support 16 bit I/O (at least, not by my superficial reading of the data sheet). I think the processor for the Teensy4 line can, but the board designer didn't lay it out to support a contiguous set of 16 bit I/O's.
So in general, the serial data approach seems to be more widely available, but more problematic due to the faster serial clock and data signals. While looking into this problem, I found a good article here that offers an elegant solution, but only if your microcontroller board can support two SPI channels, one acting as a master and the other as a slave. The master channel would be used to set up the ADC and the slave channel would receive the data. Clock and data delays in the slave channel remain synchronized because they travel the same distance back to the processor. I note that the master's SCK basically is turned around at the ADC end so there would be a few additional nS of delay if you need to buffer it. However, that could be matched by buffering the data coming out of the ADC as well. You could even go the LVDS route if you want to.
Your workshop sounds like it will turn out pretty nice. What national park will be your machining backdrop?
Is it that the noise coming out of the pocketgeiger board comes from stuff unrelated to the amplifiers? How?
If that is so, disconnect the noisy stuff from having a role.
Regarding the amplifier section.
It need not be blighted by anything else. Change out the LMC662 for almost anything with a higher GBW.
The capacitively coupled concept between the diode and the TIA does not have the RC roll-off you may be thinking about.
The capacitor is a near short-circuit to charge surges from the photodiode. No electrons cross, but their piled-up field does.
The shunt bulk resistance 40Meg, is not part of of an RC network pole. What arrives in the diode is new energy carriers, and the way this splurge of electron changes fields, and works as a charge amplifier is not the same as a low-pass filter. The feedback forces that a voltage change on the capacitor at the input is simply not allowed to happen. The million plus gain of the op-amp will swing the output so that, via the feedback resistor, all the electrons that deigned to move the voltage by charging the coupling capacitor are defeated.
Surely a low pass at 0.1Hz would not be able to show a pulse doing it's up 'n down in 10uS.
This is not like having an input resistor, and a feedback resistor, to get a voltage gain of Rf/Rin. There is no Rin.
The gain of that stage is equal to the value of the feedback resistor, The gain-bandwidth product of pocketgeiger would be fractional Hz, which is why I don't understand how it works, other than that some other component, like the capacitor across the Rf is acting as a sort of integrator.
Even with LMC662 granted the relief of Rf = 510K, the bandwidth is only 2.74Hz, unless it is doing something integration-like with the Cf
Changing out to LTC6269, and setting gain to about half a million, say 510K would give a bandwidth of 660KHz, and amplify what noise there is to a level where it cannot anymore be challenged by noise from other stuff. In effect, the S/N ratio becomes preserved, "locked in"
The best value for the coupling capacitor is about 1nF. Larger values can work OK ish, but the soak-up of the exchanging electrons starts to blunt the signal. Less than about 1nF gets a fairly close following of the pulse, but at lower amplitude as seen on the amplifier output. The voltage used for reverse bias had better be super-super clean noise free. That way, the only noise comes from the bias current carriers, and the little thermal noise.
The slew rate of the LMC662, and it's noise, is, for me, an outright deal-breaker. Even if you leave it in there, you will get a much better outcome if you take out the 22Meg resistors, and put 470K or 510K there. The noise from the reverse bias current lowers lots if you set the bias to 4V or so. The increase in diode capacitance is not too much of problem. Then, use the second stage to make up the gain going for about 180 to 200 or so. Then put a 1-pole RC filter between the output and the ADC.
Probably the biggest single thing you can do, still using the SparkFun PCB, is to change out the LMC662, alter the gain, take care of rail powers if you need to, and and altering the gain of the second stage.
That, anyway, is what I expended too much time with LTSpice to find out, and is why I will be using something else.
- - - - - - - - - - - - - - - - - -
For ADCs, the more expensive (£29) ADC AD7622 can offload 16-bit parallel chunks. If a Teensy4 can accept 16-bit chunks twice every microsecond, then you get full speed.
- - - - - - - - - - - - - - - - - -
This is the South Downs National Park, and pretty much similar to what I see. The crop just planted a few days ago is winter wheat.
--> South Downs National Park
If that is so, disconnect the noisy stuff from having a role.
Regarding the amplifier section.
It need not be blighted by anything else. Change out the LMC662 for almost anything with a higher GBW.
The capacitively coupled concept between the diode and the TIA does not have the RC roll-off you may be thinking about.
The capacitor is a near short-circuit to charge surges from the photodiode. No electrons cross, but their piled-up field does.
The shunt bulk resistance 40Meg, is not part of of an RC network pole. What arrives in the diode is new energy carriers, and the way this splurge of electron changes fields, and works as a charge amplifier is not the same as a low-pass filter. The feedback forces that a voltage change on the capacitor at the input is simply not allowed to happen. The million plus gain of the op-amp will swing the output so that, via the feedback resistor, all the electrons that deigned to move the voltage by charging the coupling capacitor are defeated.
Surely a low pass at 0.1Hz would not be able to show a pulse doing it's up 'n down in 10uS.
This is not like having an input resistor, and a feedback resistor, to get a voltage gain of Rf/Rin. There is no Rin.
The gain of that stage is equal to the value of the feedback resistor, The gain-bandwidth product of pocketgeiger would be fractional Hz, which is why I don't understand how it works, other than that some other component, like the capacitor across the Rf is acting as a sort of integrator.
Even with LMC662 granted the relief of Rf = 510K, the bandwidth is only 2.74Hz, unless it is doing something integration-like with the Cf
Changing out to LTC6269, and setting gain to about half a million, say 510K would give a bandwidth of 660KHz, and amplify what noise there is to a level where it cannot anymore be challenged by noise from other stuff. In effect, the S/N ratio becomes preserved, "locked in"
The best value for the coupling capacitor is about 1nF. Larger values can work OK ish, but the soak-up of the exchanging electrons starts to blunt the signal. Less than about 1nF gets a fairly close following of the pulse, but at lower amplitude as seen on the amplifier output. The voltage used for reverse bias had better be super-super clean noise free. That way, the only noise comes from the bias current carriers, and the little thermal noise.
The slew rate of the LMC662, and it's noise, is, for me, an outright deal-breaker. Even if you leave it in there, you will get a much better outcome if you take out the 22Meg resistors, and put 470K or 510K there. The noise from the reverse bias current lowers lots if you set the bias to 4V or so. The increase in diode capacitance is not too much of problem. Then, use the second stage to make up the gain going for about 180 to 200 or so. Then put a 1-pole RC filter between the output and the ADC.
Probably the biggest single thing you can do, still using the SparkFun PCB, is to change out the LMC662, alter the gain, take care of rail powers if you need to, and and altering the gain of the second stage.
That, anyway, is what I expended too much time with LTSpice to find out, and is why I will be using something else.
- - - - - - - - - - - - - - - - - -
For ADCs, the more expensive (£29) ADC AD7622 can offload 16-bit parallel chunks. If a Teensy4 can accept 16-bit chunks twice every microsecond, then you get full speed.
- - - - - - - - - - - - - - - - - -
This is the South Downs National Park, and pretty much similar to what I see. The crop just planted a few days ago is winter wheat.
--> South Downs National Park
Last edited:
Heh Heh.. The first little batch from mouser would hardly be heard shaking about in a small envelope, but it came to £62.33 before VAT. At least they waived the shipping charge. 
A passing question.. A gamma coming out of Am241, encountering one's thumb, would pass right through and out the other side, not likely encountering anything - right??
Where are our nuke experts? Gotta find one soon.
A passing question.. A gamma coming out of Am241, encountering one's thumb, would pass right through and out the other side, not likely encountering anything - right??
Where are our nuke experts? Gotta find one soon.
Last edited:
The .1Hz corner is my somewhat paranoid response to the Theremino design, which works real hard to minimize pulse undershoot. They claim the undershoot messes up the peak-voltage measurement. I debated over going to an entirely DC-coupled approach to avoid this, or go with a very low roll-on frequency to minimize the undershoot. I went with the latter just for simplicity's sake.
Simulating the pocket-geiger circuit with variations of the .1uf interstage capacitor really highlighted the undershoot. Dropping the capacitor to 1nF resulted in significant undershoot. Increasing it to 1uf greatly reduced it, but will drop the roll-on frequency. The pocketgeiger designers may have chosen .1uf in an attempt to reduce low frequency noise, but they should have done that by going with a better amplifier.
The landscape looks a lot like some areas in the dryer (eastern) part of Oregon. Very pretty. Not a bad view while you're making swarf
.I'd think that a 60Kev photon wouldn't find much in a thumb to impede it, but let's see. The most significant absorber would be the calcium phosphate A.K.A. bone. Calcium's u/p @60Kev is .66 and phosphorous' is half that. Calcium phosphate = Ca3(PO4)2 so it has a 3/5 concentration ratio. If the thumb bone is about 1cm thick, the calcium thickness equivalent would be .6cm. Calcium's density is given as 1.5g/cm^3 so for calcium (u/p)*p ~1. The transmitted intensity then is calculated as I = Izero * exp(-.6) = Izero* .55, so about half the gammas would make it through the thumb -- if the phosphorous wasn't present. Since it is, it absorbs some, too; but in addition to a lower absorptivity, its concentration is lower as well. Taking a guess based on these considerations, maybe about 45% of the incident gammas would exit a thumb.A passing question.. A gamma coming out of Am241, encountering one's thumb, would pass right through and out the other side, not likely encountering anything - right??
60 KeV is around the KV used for medical extremity radiography so I would expect significant absorption. 55% seems reasonable, maybe higher. The photon flux from your source is pretty low however.
In imaging the patient dose generally goes down as the KV goes up since more photons reach the image receptor. One other difference is the 60KeV represents a monochromatic beam. In imaging that would be the peak KV so the average energy is about 2/3rd.
R
In imaging the patient dose generally goes down as the KV goes up since more photons reach the image receptor. One other difference is the 60KeV represents a monochromatic beam. In imaging that would be the peak KV so the average energy is about 2/3rd.
R