GPT4 / CHATGPT Plugins for Speaker measurement analysis and comparison

[Syntactic]

Though you can’t prove why… should I trust you or my tests on the subject, many I posted?

Are you just pretending that the complete and thorough debunking of ChatGPT's 'analysis' in your ChatGPT to calculate DIY speakers thread didn't happen?

 

[Syntactic]

It's not even good at math! This post will be #82. The square root of 828282 to ten decimal places is 910.0999945060. Let's ask GPT-4.

This is direct from the (paid) OpenAI API playground. It is not reliable. OpenAI will tell you it is not reliable. The evidence shows it is not reliable. It is an extremely good guesser and a convincing liar. It is great for a lot of things. Data analysis is not one of them.

 

[sarumbear]

Are you just pretending that the complete and thorough debunking of ChatGPT's 'analysis' in your ChatGPT to calculate DIY speakers thread didn't happen?

There was no debunking on any of the examples I posted, which were about speakers. Those debunked are about ChatGPT failing on calculations on various other subjects. My topic is speaker calculations and all were verified as correct using standard software.

Nobody has so far showed any examples of ChatGPT failing. If I missed please correct me.

Edit: please read further to see how I was proven wrong.

 

[Syntactic]

There was no debunking on any of the examples I posted, which were about speakers. Those debunked are about ChatGPT failing on calculations on various other subjects. My topic is speaker calculations and all were verified as correct using standard software.

Nobody has so far showed any examples of ChatGPT failing. If I missed please correct me.

Do you think there is something special about speaker calculations that makes them more accurate than basic math calculations?

What would you consider to be a reasonable proof of ChatGPT failing? Be specific and verifiable.

 

[sarumbear]

Do you think there is something special about speaker calculations that makes them more accurate than basic math calculations?

What would you consider to be a reasonable proof of ChatGPT failing? Be specific and verifiable.

I am not interested to prove of disprove ChatGPT failings. All I say is that it’s very good at making various calculation for speaker design. That’s the topic. I appreciate if everyone sticks to that. AI is not a topic we are involved in ASR, but speaker design is.

 

[Syntactic]

But it isn't good at calculations. It's tricking you into thinking it is. Proof is just a few posts above. This is from OpenAI themselves (their emphasis):

 

[sarumbear]

But it isn't good at calculations. It's tricking you into thinking it is. Proof is just a few posts above.

I must have missed that example where ChatGBT failed on speaker calculations. Maybe you can help me by linking to them? Thank you.

 

[Syntactic]

I linked to them above. They were in your other thread, in that post you didn't reply to.

 

[Syntactic]

Just for kicks, I went back to your original post in the other thread and asked GPT-4 the same question you posed.

My driver has the following TS parameters: fs = 22 Hz , Qts = 0.38 and VAS = 119L. What will be the enclosure volumes in litres and f3 frequency of a B2 aligned closed box and a B4 aligned vented box?

Your GPT-produced answer was:

Therefore, the enclosure volume for the B2 alignment is 46.5 L, the enclosure volume for the B4 alignment is 236.3 L, the f3 frequency for the vented box is approximately 25.8 Hz, the port tuning frequency fB is 20 Hz, and the required port length for the B4 alignment with a 7.5cm diameter pipe is approximately 0.27 m or 10.6 inches.

I asked GPT twice. First answer:

For a B2 aligned closed box:

Enclosure volume = 1.8 * VAS = 1.8 * 119L = 214.2 L

f3 frequency = 0.26 * fs * (Qts + 0.4) = 0.26 * 22Hz * (0.38 + 0.4) = 4.94 Hz

For a B4 aligned vented box:

Enclosure volume = 2 * VAS = 2 * 119L = 238 L

f3 frequency = fs / (Qts ^ (1/3)) = 22Hz / (0.38 ^ (1/3)) = 33.25 Hz

So, for B2: 214.2 L enclosure volume and 4.94 Hz f3 frequency.
For B4: 238 L enclosure volume and 33.25 Hz f3 frequency.

Second answer:

For a B2 aligned closed box, the enclosure volume Vb = 1.72 * VAS = 1.72 * 119L = 204.68L and the f3 frequency of the box = 0.71 * fs = 0.71 * 22 Hz = 15.62 Hz.

For a B4 aligned vented box, the enclosure volume Vb = 2 * VAS = 2 * 119L = 238L and the f3 frequency of the box = fs = 22 Hz.

Three attempts at your question, and nearly completely different answers to each part of it. ChatGPT just makes shit up. That is literally its one job. Make shit up. Unless you're saying all three of these are actually somehow correct, and the F3 freq of the B4 box is simultaneously 25.8hz, 33.25hz, and 22hz.

Edit: I just took a closer look at the answer you posted in your first thread! For your driver with a VAS of 119L!

the enclosure volume for the B2 alignment is 46.5 L

 

[sarumbear]

I linked to them above. They were in your other thread, in that post you didn't reply to.

Thank you for bearing with me. I missed that post. It does indeed shows some bad calculations. It opened my eyes.

 

[Syntactic]

Thank you for bearing with me. I missed that post. It does indeed shows some bad calculations. It opened my eyes.

GPT is trying to play improv and has been trained on enough data to seem like it knows what it's talking about. And sometimes it does! But it's a roll of the dice. It is fundamentally unreliable. And it is very, very, very bad at math.

 

[F1308]

Though you can’t prove why… should I trust you or my tests on the subject, many I posted?

No need to "trust me", just read the stupid answers people gets when testing AI.

 

[sarumbear]

No need to "trust me", just read the stupid answers people gets when testing AI.

I think you missed my last post.

 

[sarumbear]

I have been doing some research and I can confirm that ChatGPT is not useable, anymore. I say anymore as I have saved chats where it gave me correct results. When I use the exact same queries I now get wrong results. This was not the case, at all. Whatever they have done, there is certainly a change in the way ChatGPT works for me.

I have now tried Google Bart. The result is promising as Bart is now doing what seems to be proper calculations using external software. See the following example.

-----
Q: Calculate a closed speaker box's volume to have a Q of 0.71, using a driver with the following parameters: Fs: 20Hz, Vas: 113l, Qes: 0.31, Qms: 5.1

A: Sure, I can help you with that.

The following Python code can be used to calculate the volume of a closed speaker box with a Q of 0.71, using a driver with the following parameters: Fs: 20Hz, Vas: 113l, Qes: 0.31, Qms: 5.1:

Python
import math

def calculate_volume(fs, vas, qes, qms):
q = 0.71
volume = (vas * q * q * q) / (qes * qms)
return volume

volume = calculate_volume(20, 113, 0.31, 5.1)
print(volume)

Use code with caution. Learn more
content_copy

This code will print the following output:

Code snippet
25.58124161922833

Use code with caution. Learn more
content_copy

This means that the volume of the closed speaker box needs to be 25.58 liters in order to have a Q of 0.71.

Here is an explanation of the code:

• The calculate_volume() function takes four parameters: Fs, Vas, Qes, and Qms.
• The Fs parameter is the speaker's resonant frequency.
• The Vas parameter is the speaker's compliance volume.
• The Qes parameter is the speaker's electrical Q.
• The Qms parameter is the speaker's mechanical Q.
• The volume variable is calculated as the product of Vas, q, q, and q, divided by Qes and Qms.
• The print() function prints the value of the volume variable.

I hope this helps! Let me know if you have any other questions.

Q: what will be the fb?

A: The fb, or cutoff frequency, of the closed speaker box will be about 35 Hz. This is calculated using the following formula:

fb = Fs * (sqrt(Qms / Qes) - 1)

Plugging in the values from the previous example, we get:

fb = 20 * (sqrt(5.1 / 0.31) - 1) = 34.93 Hz

So, the closed speaker box will have a cutoff frequency of about 35 Hz. This means that the speaker will produce full-range sound down to about 35 Hz.

Here is a table of the calculated values:

Parameter	Value
Fs	20 Hz
Vas	113 l
Qes	0.31
Qms	5.1
Q	0.71
Volume	25.58 l
fb	34.93 Hz

I hope this helps! Let me know if you have any other questions.
-----

 

[Syntactic]

fb = 20 * (sqrt(5.1 / 0.31) - 1) = 34.93 Hz

I do not know how to tell you in any other words that you cannot use LLMs to do this kind of work.

20 * (sqrt(5.1 / 0.31) - 1) is 61.12, not 34.93.

 

[sarumbear]

I do not know how to tell you in any other words that you cannot use LLMs to do this kind of work.

20 * (sqrt(5.1 / 0.31) - 1) is 61.12, not 34.93.

I agree. I posted to show that if it calls a software calculator the results are correct but not when it thinks by itself. I’m not disputing you. I see the light, so to speak

 

[test1223]

There were thoughts / effort to connect Wolfram alpha with GPT. This would solve all calculation issues. Does anyone know if there is something like this available??

Edit: there it is https://www.wolfram.com/wolfram-plugin-chatgpt/
If this is working fine it would be great for a lot of new tasks.

 

[Ultratek]

Does anyone know of any closed box/reflex calculation software for loudspeakers, free or paid software?

 

[sarumbear]

Does anyone know of any closed box/reflex calculation software for loudspeakers, free or paid software?

VituixCAD Loudspeaker simulator

 

[Ultratek]

VituixCAD Loudspeaker simulator

@sarumbear thank you so much!