No Ticket.
I recently came across someone using ChatGPT to get some information about a certain problem they were working on. I didn't go quite to plan. So, I thought I try to elicit a more correct answer from the AI. In other words, if we ask the question carefully enough, can it give a useful answer in this particular case. It made for a much more interesting discussion than I thought when I started and the experience gave me pause for thought on the nature of the tools in an engineer's toolbox and how the tools, the engineers and the folks who educate engineers might adapt. There are also some 'old faithful' rules of engagement to apply. Enjoy!
Hello ChatGPT.
Hello! How can I assist you today?
Could you tell me how to calculate the inductance of a bus bar which has a rectangular cross-section, please?
Sure, I can help you with that!
To calculate the inductance of a bus bar with a rectangular cross-section, you can use the following formula:
L = μ? * μ? * (b * h) / (6 * (b + 2 * h))
where:
Note that this formula assumes that the bus bar is infinitely long, which is a reasonable assumption for most practical applications.
To use this formula, you will need to know the values of b, h, and μ? for the specific bus bar you are interested in. Once you have those values, you can simply plug them into the formula and calculate the inductance.
I hope that helps!
Thanks. Could I ask you to demonstrate that the equation you provided is dimensionally consistent, please? By that I mean can you show that, if all the variables have their expected units, the answer will have the expected unit (henries)?
Sure! We can demonstrate that the equation is dimensionally consistent by verifying that the units of each term in the equation match the units of inductance (henries).
The units of each term in the equation are:
Substituting these units into the equation, we get:
L = (H/m) * dimensionless * (m * m) / (6 * (m + 2 * m))
Simplifying the equation, we get:
L = (H/m) * (m^2) / (6 * m)
Canceling out the units of meters, we get:
L = (H/m) * m / 6
L = (H/6)
Therefore, the units of the result of the equation are henries, which is the expected unit for inductance. Hence, the equation is dimensionally consistent.
Could you assume that the bus bar is a finite and known length, please?
Certainly! If the bus bar has a finite length and the length is known, then the inductance can be calculated using the following formula:
L = (μ? * μ? * A) / l
where:
To calculate the cross-sectional area of the bus bar, you can use the formula:
A = b * h
where b is the width of the bus bar in meters (m) and h is the height of the bus bar in meters (m).
Therefore, the formula for inductance of a finite bus bar becomes:
L = (μ? * μ? * b * h) / l
where all variables have the same meaning as before, but b and h represent the dimensions of the rectangular cross-section of the bus bar, and l is its length.
Note that this formula assumes that the bus bar is infinitely long in the direction of current flow, but of finite length in the other directions.
I hope that helps!
Could you assume that the bus bar is of finite length in the direction of current flow, please?
Certainly! If the bus bar is of finite length in the direction of current flow, then you can use the following formula to calculate its inductance:
L = μ? * μ? * (b * h) * [ln(2 * l / t) - 0.5]
where:
Note that this formula assumes that the bus bar has a rectangular cross-section and that the length of the bus bar is much greater than its thickness, i.e., l >> t.
The term [ln(2 * l / t) - 0.5] is known as the "end correction factor" and accounts for the fact that the inductance of a finite bus bar is greater than the inductance of an infinitely long bus bar with the same cross-sectional area.
I hope that helps!
Could you further explain the difference between the height of the bar and the thickness of the bar, please?
Sure! In the context of a bus bar, the terms "height" and "thickness" refer to two different dimensions of the rectangular cross-section of the bar.
The "height" of the bus bar refers to the distance between the top surface and the bottom surface of the bar, measured perpendicular to the length of the bar. This dimension is sometimes also referred to as the "depth" of the bar.
The "thickness" of the bus bar, on the other hand, refers to the distance between the front surface and the back surface of the bar, measured perpendicular to both the height and the length of the bar. This dimension is also sometimes referred to as the "width" of the bar.
So, in summary, the height of the bus bar is the distance between the top and bottom surfaces of the bar, while the thickness of the bar is the distance between the front and back surfaces of the bar.
It's worth noting that in some cases, the terms "height" and "thickness" may be used interchangeably or may have slightly different meanings depending on the context. However, in the context of a rectangular bus bar, the above definitions should generally apply.
Could you give me the most recent formula again but avoid using thickness if it is actually the same as width, it seems like there are too many physical dimensions for the bar, surely it can only have three, but the formula has height, length, width and thickness. It seems that two of these must refer to the same thing?
Something went wrong. If this issue persists please contact us through our help center at help.openai.com.
------------------------------------------------
So... it left... just when it was getting interesting [if you know the movie reference then you're officially awesome and can tell everyone that you're an honorary bear for the rest of the day. Another clue is in the title.]. I am sad, though, as it was just getting interesting. In my view this is still a phenomenal achievement. ChatGPT could probably get a 2:1 in ECE or EEE at undergraduate level as long as it doesn't have to do any coursework ;)
Seriously though it's demonstrative of the main thing that I think we should keep in mind with tools. It is up to us to make a judgement about the quality/utility of the answers/output we get from our tools. This is the same with MathCAD, SPICE, Simulink etc. what comes out is only as good as what goes in. If you ask SPICE for the distortion of a waveform in a small signal analysis of a circuit which includes active devices it would tell you there isn't any. But, we know that's not physically possible...
My only other tiny criticism is I wish ChatGPT would cite some sources so we could look back and see where it's "coming from" when it gives certain responses - like how it came to have four physical dimensions for something that exists in a three (macroscopic) dimensional universe. Naturally, it is trained on a large dataset and there is no particular paper or web-page that contains the text it displays verbatim - like a search engine - the text is a construction of the algorithm based on the training data. However, just like a research scientist writing a journal article, the words are written by the paper's authors but, their work, that they are discussing is built on the work of others... whom they cite... in the references. So, could ChatGPT effectively construct a set of references for us? Perhaps I'll ask it next time it's in the office.
Educators may well be furious with me for suggesting such a thing because it makes it easier to abuse the tool to write academic submissions in an educational setting but, my argument there is quite simple. If you want to test a would be engineer then ask them to do something from a higher level of Bloom's Taxonomy (https://cft.vanderbilt.edu/guides-sub-pages/blooms-taxonomy/) that the AI can not manage*. e.g. here are schematics for two discrete transistor front ends which perform the same primary function. Judge which is 'best' based on the following criteria, mass, likely pcb footprint, cost, volume, future proofing of component supply, power dissipation, robustness to overvoltage of power rails, internally generated noise, PSRR, feedback loop stability in the face of capacitive load on the output etc. etc. If the AI can make that holistic judgement and argue it convincingly then I will absolutely welcome my first synthetic employee at BI [film reference in there too for those still paying attention, clue is in the banner at the top.].
And if I did hire an AI, the first question then would be, "Hi, if you were a bear... what species would you be?"
For busbar inductance calculations see, for example [1].
* The example is purposely ridiculous for comic effect. No student could ever answer such a question :0
References
[1] https://pe.org.pl/articles/2012/8/73.pdf On-line [Accessed: 26/04/2023].
Senior Embedded Software Engineer
1 年I’ve recently found the Bing* AI useful in cutting through the dross. Putting my question in the ‘normal’ search engine, results in either products they want to sell me, or TV shows, …. With AI it delves deep and provides answers from forums it would take me hours to navigate otherwise. *other search engines are available