Arb assist ask

My goal is to generate this 24 VDC arb:

Keysite suggests it's easy peasy with Excel:

My Signal Generator accepts CSV files as input:

It is indeed easy peasy for Excel to export the requisite CSV file.

Are there any available Excel arb gurus in-group to point me in the right direction?

If not, it's probably easier for me to use octave to create a MAT file for my Signal Generator.

Thank you in advance.

# # #

Curve fitting techniques found in Numerical Analysis can create arbs. On its surface, Cubic Spline Interpolation and other approaches look not so easy peasy.

It's time to move on to digitizing a freehand sketch.

Your picture is pretty small but to me that looks basically like positive gaussian pulses and then half of a positive gaussian pulse followed by a vertical negative spike followed by an exponential decay back to the x axis. Define your x axis point spacing from your arb waveform size, use Excel to model each shape in separate columns using the appropriate formula, copy and paste the models into one big column separated by sections of zeros for the baseline between pulses to get one cycle of your waveform, and output the column as csv. Carry the gaussian and exponential decay models out far enough for them to get close enough to zero so that you don't have a big discontinuity where you insert a run of actual zeros before the next pulse if that matters, or just insert a linear slope to zero long enough to smooth out the transition. If the positive pulses look more like some other standard pulse shape and it matters, use that formula instead of gaussian. That's how I'd approach it, anyway.

That's similar to a waveform that we're working with, a gear-tooth wheel with a variable-reluctance pickup. The gear has long or short or offset teeth to give a once-per-rev index.

One of my guy used some AI thing to simulate the geometry and make the waveform files.

If you draw on a 2000 x 1000 pixels screen (as example), you can scan left ot right x and y position of each dot. You could add log if you want, If the signal is periodic draw one period. Who needs big software packages? Bit of C coding.

Yes, it simultaneously occured to me too, that AI ought to be good for something. Give your guy a big fat raise. He earned it.

AI cooked up a GNU octave recipe. It ought to be enough to kickstart my arb:

Normalize the time axis between 0 and 1. Then assuming 24 V scale:

A (Amplitude): Peak voltage height (set to 15V u (Mean): Center location of the pulse on the time axis (placed at t = 0.15 and t = 0.35). s (sigma ) Standard Deviation/Width: Controls pulse sharpness (set to 0.03)

Vrise(t)=A*EXP(((t-0.55)/s)^2)

Vdecay(t)= -Aspike(t)*EXP(-lambda*(t-0.55))

-Aspike The maximum negative peak depth ( set to -20V) lambda (Decay Rate Constant): Controls how quickly the signal settles back to zero (set to 15) for a rapid recovery)

The CSV file contains a single column of 2,000 digitized voltage points representing a complete 24 VDC arbitrary waveform cycle. It models the components discussed in the forum: two distinct positive pulses, a rising half-pulse, a sharp negative vertical spike, and an exponential decay back to the baseline.

AI is kind of amazing. It still requires human oversight, but it's not much, and it saves time on the minutiae. It even prepares the 2000-point CSV file contents. Looks like one column for normalized time and a second for amplitude. Dazzlingly uncomplicated.

This situation requires the suggested 2000 duple CSV file. Where is it?

--

I can't tell if you're responding to me. But if you are, just link your waveform screenshot to AI and ask it for the CSV file. You need to tell it about the 24V and being a VR output, which gets the standard Gaussian pulse periodically interrupted by that negative spike and exponential decay, all very standard outputs. My guess is the AI digitization is good to 0.5%.

For Your Information your own followups make it difficult to tell if you're responding to me. That's why followups customarily include a few paragraphs from the thread's tail-end - it enables people to put things into context. My <snip> above is neither malicious nor spiteful. People customarily insert a line break between character columns 72 through 80 when they post. On the other hand, your most recent followup to this thread extends to column 370. A good rule of thumb is to break 370 character lines into about six separate lines by inserting line breaks. My own usenet server sometimes throws an error when your comments are posted "as is." And life's too short for me to clean your post up to appease my usenet server. Ergo, the above <snip>.

Now, to move on to the pertinent topic of AI digitizing my image. AI's able to bullscat rhetoric until the cows come home. It's perfect for political posers. On the other hand, in regards to technical answers, AI appears to always be needy. It always seems to need one more thing. It continually suggests that if you do one more thing for it you will finally receive The Answer (tm) in return as your reward. Early detection of AI stringing you along pays off. You want to avoid putting 90% of your effort into solving the last 10% of the problem. The recipe AI cooked up for me:

solves 90% of my problem. It's too time consuming to prod AI into answering the last 10%. It's easier for me to use my mind. Your Mileage May Vary.

The proclamation that you only need link your waveform screenshot and ask for a CSV file is a punchline:

AI response:

Assuming the image represents a repeating waveform with three sharp downward pulses, this Python script generates a CSV containing exactly 2,000 `(time_s, voltage_V)` tuples. The maximum voltage is limited to 24 VDC.

```python import csv import math

No thank you. GNU Octave is easiest for me. Your Mileage May Vary.

AI would have built those CSV files for you. All you have to do is download them. It's just an ASCII text file. The Arb app software should allow you to designate those files for upload to the generator. I don't see a problem here.

Several AI platforms refused to create a CSV file. They all tried to foist source code on me instead. And so, several struggle sessions ensued to correct syntactic errors embedded in the original source code. AI's silliest syntax errors include: intermittently capitalizing a variable name and using an em dash in place of a hyphen. The final nail in AI's coffin is AI's semantic error.

Another researcher posits the primary problem. And then submits a superior solution, in my opinion. Your mileage may vary.

Google has been systematically inserting itself between you and the original source, answering questions with AI- generated summaries before you ever reach the page those answers came from. The results you do see are filtered through an algorithm that weighs your search history, your location, and the billions of dollars advertisers have spent to appear for particular queries. Two people searching identical phrases on the same day can get meaningfully different results without either of them knowing it. And because Google controls roughly 90% of the world’s search traffic, most people have no frame of reference for what a less mediated search experience would even look like.

After thinking things through, it seems that creating a CSV file from an image ought to be commonly trivial. Using a plain vanilla search at a pertinent website opens a cornucopia of CSV conversion code:

Don snipped-for-privacy@CRComp.net wrote: |-----------------------------------------------------------------------| |" The final nail in AI's coffin is AI's semantic error. | |

:)

Yesterday I submitted a referee report about a LLM submission to a journal. That submission boasts 100% accuracy via GPT, but various persons (e.g. I) detect fundamental shortcomings in ChatGPT. In this referee report I challenge that / those (co-)author(ess(e)(s) to show what that LLM deduces from a test which I predict it to fail.

I uploaded that referee report to Manuscript Central, and I also uploaded it to

Funnily enough, the only HTTP or HTTPS uploads from this directory so far are to GPT! Maybe GPT is anxious about its reputation :)

|-----------------------------------------------------------------------| |"[. . .] | |Another researcher posits the primary problem. And then submits a | |superior solution, in my opinion. Your mileage may vary. | | | | Google [. . .] | |[. . .] | | [. . .] Two people | | searching identical phrases on the same day can get | | meaningfully different results without either of them | | knowing it. [. . .] | |[. . .]" | |-----------------------------------------------------------------------|

Truly, and remember that mathematicians distinguish between "Two" and "Two distinct" - i.e. a mathematician would say that a triangle has a million angles consisting of three distinct angles. I.e. "Two people searching" can be one distinct person repeating the same search query via Google seconds apart and getting different "AI" summaries. I did so a few times!

|-----------------------------------------------------------------------| |"After thinking things through, it seems that creating a CSV file from | |an image ought to be commonly trivial. [. . .] | |[. . .]" | |-----------------------------------------------------------------------|

Many years ago I downloaded and used well a program which is designed to convert a scan of a graph from an old scientific journal to a numerical text file (very possibly as CSV - and if not, its outputs would easily be convertible to CSV). I am sorry that I do not remember its name for you, and that I cannot access the computer onto which I installed it to check for you. I think that I downloaded it from SourceForge. (S.

There's definite benefits to avoiding AI altogether.

No worries about the lost program. Because, as expected, image to CSV is both common and trivial. A generalized (irrelevant top and bottom noise cancellation commented out) version of this script:

Creates these plots:

Some additional axial scaling will create an apropos arb CSV.