Sciencemadness Discussion Board - Thermogravimetry

Thermogravimetry

brubei - 19-1-2018 at 18:08

Thermogravimetric apparatus is a system recording both mass and temperature of a sample. By this way, chemist can follow the transformations occurring in a compound, mainly phase transition, weight loss (like dehydration), weight gain (oxidation) energy of reactions, etc ... these informations are always interesting in analytical investigations

typical analysis of a calcium oxalate sample, successive dehydration and decarbonation are quantified at various temperatures

The aim of this thread, is to easily build a cheap machine as accurate as possible with your participation (i hope)

First difficulties are :
- how to insert a heating source which don't disturb weight (touchless heating like induction should be a good answer).
- same for temperature probe (infrared thermometer sound good choice)
-how to record data (electronic skill needed)
-build a compact weight scale compatible with high temperatures.

See you soon after some investigations !

Fulmen - 19-1-2018 at 22:16

You will need to insulate the scale from any temperature variations. The simplest solution will probably be to hang the sample on the end of a long rod.

brubei - 21-1-2018 at 11:51

thanks for your attention, i'v found a vidéo on yt to build a weighting scale with high precision for 20$

https://www.youtube.com/watch?v=GjNk2_j021g

a Good start point

LearnedAmateur - 21-1-2018 at 23:00

I just drew up this quick and simple design, going off Fulmen’s tip, that enables heating even with a propane/blow torch for those decompositions unobtainable by an electric heater. Just make sure that your scales have a somewhat high upper limit, preferably where the stand and empty crucible weigh about the same as the midpoint (since this is where an electronic balance is most accurate).

78D5094C-55CD-4DBA-9B94-6C2B01914A22.jpeg - 128kB

Edit: you could make the entire stand out of thick gauge wire, preferably steel - something on the order of SWG 5-15 (~2 - ~5mm). Make a square about the size of the balance plate, bend one corner up at 90 degrees, have a vertical section about 30cm long, do another 90 degree bend towards the centre of the square, then bend in a hook. For the crucible, a small steel can should suffice (but check to see if it will react with the sample first), with three small holes drilled in at 120 degree angles. Thread in thin wire sections (2x100mm, 1x 150mm for example) and twist them together at the centre to create a cone. With the longer wire, loop around the protruding end and twist it tightly so it can be hung on the hook - a consideration is the size of the loop, which should be large enough so that it can be quickly removed if anything goes wrong.

[Edited on 22-1-2018 by LearnedAmateur]

aga - 22-1-2018 at 06:36

Neat design LearnedAmateur.

wg48 - 22-1-2018 at 10:30

Quote: Originally posted by brubei

thanks for your attention, i'v found a vidéo on yt to build a weighting scale with high precision for 20$

https://www.youtube.com/watch?v=GjNk2_j021g

a Good start point

That project uses a 24 bit ADC HX711 available on a break out board for less than £3.

That’s impressive. I could not find the accuracy or decode the accuracy from its spec. I did find a similar 24 bit ADC AD7780 chip that suggested its accuracy was 18 effective bits. Even if the HX711 is only 18 effective bits that still impressive 1 in 256,000 or 4uV per volt or 1mg in 256g. Its has dual inputs and a serial output but slow at 0.4s settling time though not a problem for a scales or for temperature sensing.

I should add the 18 bits does not necessarily include linearity errors or sensor errors.

aga - 22-1-2018 at 14:35

Generally 18-bit resolution is pretty much useless to an amateur.

12-bit is kinda high, given the lack of resolution in the other parts of the system, whatever it might be.

Shooting for the stars is admirable, however if you settle for a bit less, the First project becomes do-able.

If a home-made system works OK with 8-bit resolution, try it with 10-bit.

The system will have to evolve to cope, which is Progress.

DrP - 23-1-2018 at 07:54

What level of prescission do you require on the balance? Even the entry level TGA machines use 4dp balances... some maybe better still.

I guess I'm lucky... at work we are about to buy a TGA machine for my lab. Getting a pre used one as they are very silly money to buy brand new. I am looking forward to getting the kit.

Oh, yea, sorry - I have nothing to add to this thread at all.... other than to gloat and show off my new Perkin Elmer TGA! lol.

wg48 - 23-1-2018 at 09:39

Quote: Originally posted by DrP

What level of prescission do you require on the balance? Even the entry level TGA machines use 4dp balances... some maybe better still.

I guess I'm lucky... at work we are about to buy a TGA machine for my lab. Getting a pre used one as they are very silly money to buy brand new. I am looking forward to getting the kit.

Oh, yea, sorry - I have nothing to add to this thread at all.... other than to gloat and show off my new Perkin Elmer TGA! lol.

I am not impressed by 4 decimal places unless it measures kg. If you mean significant figures that’s not even 14 bits, my jewellery scales do that and they cost £10.

PS Thats not being nice to us to gloat it makes us jealous LOL(not a serious comment)

wg48 - 23-1-2018 at 10:12

Quote: Originally posted by aga

Generally 18-bit resolution is pretty much useless to an amateur.

12-bit is kinda high, given the lack of resolution in the other parts of the system, whatever it might be.

Shooting for the stars is admirable, however if you settle for a bit less, the First project becomes do-able.

If a home-made system works OK with 8-bit resolution, try it with 10-bit.

The system will have to evolve to cope, which is Progress.

Most of the hard design is done for you with the HX711 ADC. Just connect up the strain gauge/s with screened twisted pair and ground the metal work, screen the break out board and your done. I guess it would be difficult not to get 16 bits not counting drift and accuracy which will probably be dominated by the strain gauge.

I am going to order two boards I will report back on how reproducible and noisy the readings are. The Analog Devices spec suggests you can get two more bits by averaging four reading that still less than two seconds for a reading.

DrP - 24-1-2018 at 02:25

Quote: Originally posted by wg48

I am not impressed by 4 decimal places unless it measures kg. If you mean significant figures that’s not even 14 bits, my jewellery scales do that and they cost £10.

by 4dp I mean gram measurements to 4dp. so, accurate to 0.0001g. I seriously doubt your £10 scales will do this. Saying that - a few years back a 4 dp balance would set you back 10s of thousands... nowadays you can get a cheap one for a few hundred.

wg48 - 24-1-2018 at 06:53

Quote: Originally posted by DrP

Quote: Originally posted by wg48

I am not impressed by 4 decimal places unless it measures kg. If you mean significant figures that’s not even 14 bits, my jewellery scales do that and they cost £10.

I am certain my scales are not accurate or repeatable to 0.0001g. I just checked them with a 10g weight and it was 11mg out and that was with a previous calibration with the same weight. What is the max grams on your 0.0001g?

I also checked to the resolution it is 1mg to10.020g. ie 1 in 10,020

i also found scales on ebay 10mg in 500g thats 1 in 50,000 for £7. My scales are a pain as the max is only 10g and my other one I have is only about 0.5g to 1000g ie 1 in 2000. (its almost ten years old)

With the low cost of ICs these days getting resolution is cheap and easy. Accuracy even repeatability is still expensive.

I ordered three HX711. I found then on ebayuk for 99p each including postage

DrP - 24-1-2018 at 08:43

Quote: Originally posted by wg48

What is the max grams on your 0.0001g?

Goes up to 65 grams - but I don't need anywhere near that weight for weighing samples for calorimetry or TGA.... what we DO need is the 0.0001g accuracy though. :-)

.... so I was wondering, from looking at the OP, what accuracy his balance will need to be.

[Edited on 24-1-2018 by DrP]

brubei - 24-1-2018 at 13:16

4 dp is not too much, imagine a sample that decarboxylate and loss 10% in mass, then you want 99% accuracy on this value. It means that you have to measure a variation of 1mg for a sample of 1g (with 100mg loss).
Using 0.01g scale change the sample weight to 10g, that is really high for a destructive method. Keeping a homogeneous temperature will become much more dificult and analysis of energetic materials more risky too.

0.001mg to 10g scale are maybe cheap enough to be used

[Edited on 24-1-2018 by brubei]

DrP - 6-3-2018 at 07:14

OK - My new (old) Perkin Elmer TGA machine was installed today - getting my head around the functionality and trying to eliminate gas leaks... The in built balance? The in built balance is supposedly accurate to 7DP! :-D

DrP - 18-4-2018 at 05:29

The general sample size suggested for a run on my machine is 5 mg.... but with the nature of the samples I am running I need to use smaller sample sizes. about 1.5 to 2 or 3 mg seems about right for our samples. So the accuracy of the balance is needed if you are looking at small percentage weight changes in a sample that is only 0.0015g to start with anyway.