Jeremy Hise posted an update in the group Datalogger developers 6 years, 4 months ago
Hi all, I’m pretty new to data logger development but have certainly gotten my feet wet. A project I am currently involved in requires measuring tiny changes in tree stem growth. So tiny, in fact, I believe I have to deal with signal noise from my power source/14 bit ADC. I know I can apply some statistical methods (ie. averaging with oversampling) to deal with noise but is there any circuitry that can be used as well, such as a low-pass filter? I would love to hear about general approaches to this problem. Thanks!
Hey thanks for posting. Interesting issue.
Noise in circuits is well studied and a deeeeeep issue. In circuits there are two types of noise to manage, faster than 1hz and then probably your main concern and more challenging less that 1Hz . Filtering out noise is useful at higher frequencies where low frequency noise is not an issue.
Managing noise in measurements is considered a magic art – the first step is to reduce it with selection of low noise components, and then good circuit design. To do that you need to look for the noise specification of components – and its good to understand the different types of noise. You are probably interested in shot noise, thermal noise,
However component drift may also be an issue with measurement – which is not strictly noise, but is variations in a sampled signal.
The blogs I’ve been posting on measuring low water levels are all essentially about characterizing the low frequency noise sources in the sensors+amplifiers+digitization circuitry. The manufacturer’s have put out product specifications, and the characterization is partly about determining the way they are trading off different solutions to the front end sensors they have chosen.
I’d be interested in the equipment you are thinking of using.
See this link here to my blog…..
Wow some great info. Thank you very much and I will check out the your blog as it seems this career path will be heavily involved in these kinds of problems. The design is straight forward, a Li-Polymer battery, arduino pro mini with the ADS1115, xbee series 2 and a Midori Precisions linear sensor…essential a variable resistor up to 1K. Trees are constantly soaking up water but the transpiration during the day causes their main stem to shrink. This stem size change is what I’m trying to measure which, depending on the species, can be up to about 6 microns. I’ve done some rough measurements but need to get it under the oscilloscope to measure the Hz. The noise is obfuscating the tree stem size change but I’m hoping there will be a way to differentiate between the slow change in voltage produced by stem size changes and the much higher frequency noise.
One requirement of the project is battery longevity, hence the pro mini.
Interesting project there – I’ve wondered about the effect on the trees when they are really transpiring in a heat wave,
and also what the delay was between heat peak for transpiration, and pulling the water from the ground, and stream.
So looking to measure 6microns, or 0.006mm across the LP-10F travel of 10.0 mm
I’m guess maybe you are looking for some resolution within this possibly 0.006/16 mm?
Those are some great questions. Tree water/soil/atmospheric relations is where a lot of this work is pointed. Some species, like the Picea abies, can swell up to 0.006 mm in a 24 hour period, others less so, so I’m looking to detect changes within the 0.000-0.006mm range.
Generally speaking, would one start with identifying noise at the power source and tracing additional noise along the “chain” of components/sensors? Is it possible that a linear sensor would introduce new signals?
I think you have to look at all the noise sources with the level of measreument that you are looking at. So broadly for a 0.006mm range, you want some resolution in it, which defines what exactly is your 0. That in turn will translate to the noise floor that you need to exceed. So so you could use a scale of 16 – then you need to be able to measure 0.006mm/16 which is 0.000375 mm – which is very small.
Possibly a challenge for the LP-10F which isn’t specifying anything to suggest it can be used for such “micromovements”. Its suggesting an Output Smoothness of 0.1% – which for 10mm would suggest 0.01mm – so I wonder what Output Smoothness is. Also the TCR (temperature coefficient of resistance) of the LP-10F seems very high. What daily fluctuation of temperature do you think the measurement apparatus will be subject to.?
So just checking and wondering where the suggestion for LP-10F came from – wondering if it has been used before for such an operation – just wondering if there is something that is known about it not stated on the data sheet.
Neil this has been such a great conversation. Thank you.
I don’t understand the scale of 16. Could you elaborate or point me to a reference. I’m having a hard time googling that one. This sensor is a “cheaper” version of one that is in the field being used exactly for this purpose. That sensor has a temp. coefficient of +-400 ppm/K. This seems to be specifically relevant to the resistance applied to the voltage of the device. This will be deployed in the field during spring/summer/fall so a temp range of 15F wouldn’t be surprising. The sensor was more or less supplied/suggested by the project manager who works on the science side. Again, the LP-10F is cheaper than the LP-20F and was assumed they where more or less the same, though the temp coefficient is obviously better on the 20F. Also to note is there is a known voltage sensitivity of 1mv=4.0micrometers. Would this mean that for each mv the variation would be mv*4.0 basically?
My reference to 16 was that for a total change of 0.006mm its going to be broken down to a scale, and if you choose 1/16 (as its digital) then for 0.006mm/16 you would have a target of 0.000375 mm – which is very small.
Now looking at the LP-10F spec – which with a total travel of 10mm is better than 20mm for measuring something very small.
However looking again at LP-10F spec – not knowing anything about the device – is it mechanically capable of accurately detecting 0.000375mm change of distance.
So I’m just asking the question – since I don’t know anything about the LP-10F devices – is it capable mechanically of measuring a change of 0.000375mm – when considering wiper action, the everyday cycle of changing direction while the tree trunk expands/contracts – also known as hystersis.
So just an initial “dumb question”, before looking at the electronic noise and effects of temperature change of 15C.
Of course it is possible to measure very small changes of distance, across very large distances – and the latest most well known (with a hefty price tag..
is the LIGO – they talk about measuring a change of 10E-18m across 4000m.!!!
Just came across a devices that maybe could make small distance measurements
Piezoelectric film – the problem might be calibrating it.
That is it might be able to show a change in trunk diameter based on stress applied to the film, but how easy would it be to translate to metric mm?
Also it doesn’t contain a temperature dependency specification, so seems unlikely – so just share it as an idea.
eg TE DT2-028K/L – which is 0.64″ wide and 2.86″ long. I think it would take some experimentation to see if it would wrap round a trunk, and if the film can be securely attached to ribbon.