Custom coil questions (TFR approximation, csv data limit etc)

[vapealone]

Repost from here:

I am just playing with the eScribe and have a few question regarding to custom coil settings I couldn't figure out without an actual DNA200 on hand:

1. Is it a must to normalise the TFR to 70°F or any other temp does the job?
2. Is it just the eScribe display which rounds TFR to 0.X or the sensitivity/precision of the DNA200 is 0.X?
3. Is there any limit of the input data pairs the board can handle?

Details:

1. I have tried and it is possible to upload csv with 68°F normalised TFR (which is 20°C metric). As all of my data is metric it would be awesome if I could simply convert the temp. But to do that the board should be able to handle csv where the TFR=1 belongs to 68°F
2. The graphic display of the custom settings rounds both temp and TFR to 0.X . Which means that e.g. for 316S/S the same TFR is shown from 160-350°F. I take it is just for demonstration only but again, w/o a board and 316 S/S on hand I can't tell how precise it is in the background.
3. It is also unknown to me how the algorithm behind uses the input data. The display shows simply linear approximation between steps. If this is the case I might want to load a tons of data (i.e. for every 5° or even more) depending on the material. However, it would be lovely to know if there is any particular data size the board is capable to handle. Again, eScribe has no problem with 40+ data pairs but the board can be an other thing.

P.S.:
If 68°F normalised TFR works and the board can handle 0.00x precision TFR wise and able to handle 40+ lines of data it is just minutes to convert all my data to F and make a new tab with tons of simply copyable csv materials. Just sayin'

[John]

It needs to be 70F, but realistically if you normalize around 68 you'll only have a few degrees worth of error. 

It is only Escribe. If you right click on the data point in the graph editor it will show you full precision. Internally it is 16 bits below the decimal point. That's increments of 0.0000152587890625

I'll let James answer about the internals. I know there is a limit just due to storage space (remember, we carry one for each profile) but I'm not sure what it is.

[vapealone]

John said:

It needs to be 70F, but realistically if you normalize around 68 you'll only have a few degrees worth of error. 

It is only Escribe. If you right click on the data point in the graph editor it will show you full precision. Internally it is 16 bits below the decimal point. That's increments of 0.0000152587890625

I'll let James answer about the internals. I know there is a limit just due to storage space (remember, we carry one for each profile) but I'm not sure what it is.

Thank you John and waiting for the info on data limit.

[James]

68F will be fine. You do not need to normalize to 70F with the more recent firmwares. It's a good idea to normalize to *some* temperature, though, as the EScribe curve editor only ranges from 0.1 to 10.

We can handle as many points as the internal curve storage allows. The storage space is shared between the curves of all profiles and the battery discharge curve. Our default Ni200 curve has 6 segments (7 points), and I was able to split to 22 before it ran out of space. So the practical limit is 8 segments (9 points) before you'll run out of space if you load the curve onto all 8 profiles. If you don't need it for all profiles, of course, you can do more.

There is a feature presently hidden that you may find useful: Go to Help->About and left click the Evolv logo seven times. That will unlock the "Reduce Number of Points" option in General->Material Profile->Special. When it removes points, it chooses segments that are closest to linear first, so that you don't end up losing meaningful detail.

[vapealone]

Thank you James, it was very informative.

[SSV]

James said:

There is a feature presently hidden that you may find useful: Go to Help->About and left click the Evolv logo seven times. That will unlock the "Reduce Number of Points" option in General->Material Profile->Special. When it removes points, it chooses segments that are closest to linear first, so that you don't end up losing meaningful detail.

points = metric 1 and the final metric correct?  Not splits? so starting metric and ending metric = 2 points or 1 total?

[John]

I don't know what you mean by "metric" in this context. A line is two points, two segments are three points, etc.

[SSV]

John said:

I don't know what you mean by "metric" in this context. A line is two points, two segments are three points, etc.

metric = set of parameters....I.E. 70,1 (temp, res)....

for example.....is this plot a total of 9 memory points, or 8?  I am assuming 9, according to your description....

[John]

9. (70,1) is a point. "Metric" has no mathematical meaning in your sentence. You're a sweet guy, but you are in my sandbox here. I'd be a lot more appreciative if you'd use everyday words and concepts in a technically correct manner, and as clearly as possible. If you can create and publish correct, repeatable TCR curves for the wire you sell that would be beneficial to both of us, but my patience for bamboozling is low and getting lower.

[vapealone]

A new question, hope you don't mind:

It seems that eScribe doesn't support TCR lower than 0.002 using the special wire feature.
However, I can add segment(s)/CSV with a resistance change representing materials with lower TCR (e.g. S/S316)
The question is if it works for real or not.
In terms of 316 S/S it has ~8 times higher resistance than Ni200 with ~1/4 of the TCR    resulting ~2 times more ohm change to detect for same length/gauge coils. (Eg. 0.09for Ni and 0.77 for S/S.)
So I take the hardware must be able to handle a well sized S/S coil (resistance at least 4 times of the default lower limit) which leaves the software.

I personally would be happy to know that I can load and use segments/CSV representing lower than 0.002* and ready to make sure that the initial coil resistance is high enough for safely detectable resistance change.
Thought, you may want to consider lowering the TCR limit  to e.g. 0.001 with a warning popup (the temp control can be inaccurate below a TCR of 0.002 and/or requires higher initial resistance) to satisfy other users as well

*: in my case one would be my 26ga.FeCrAl/32ga.Ni200 Hybrid Clapton with a TCR of ~0.0013-0.0011 and TFR of ~1.1-1.31 in the 200-600F range

[John]

You can load and use segments arbitrarily small with .csv and it will work... but the smaller your TCR the worse your temperature accuracy, obviously. We'll change the minimum to .001 in the next iteration of EScribe to accommodate. 

The initial resistance doesn't really enter into it unless you're doing something silly low, like a .02 ohm SS coil. With a low ohm coil you get more quantization and offset errors in the voltage sensing, but less with the current, so it more or less cancels out in the .05 to 1 ohm range. 

[vapealone]

John said:

You can load and use segments arbitrarily small with .csv and it will work... but the smaller your TCR the worse your temperature accuracy, obviously. We'll change the minimum to .001 in the next iteration of EScribe to accommodate. 

The initial resistance doesn't really enter into it unless you're doing something silly low, like a .02 ohm SS coil. With a low ohm coil you get more quantization and offset errors in the voltage sensing, but less with the current, so it more or less cancels out in the .05 to 1 ohm range. 

Thank you John again
You guys are awesome
I will spread the good news about the lower limit

P.S. I don't want to do any silly, neither enter the coil resistance. Just used as an example convincing mainly myself that the hardware must be potent enough