Microphone schematics by chat GPT

[shotfurer]

Hello!
I asked chat GPT to create a low noise condenser microphone schematics
That's what I got:

To design a low noise studio microphone circuit with a condenser capsule that is phantom powered, you can use the following schematic:

yaml
+48V DC Phantom Power
|
|
R1
|
|
+--------+--------+
| |
C1 R2
| |
| |
+------+-------+ +------+-------+
| | | |
R3 C2 R4 R5
| | | |
Audio Signal Output <-----+------+-------+ +-------+------+< Audio Signal Ground
| |
C3 R6
| |
| |
GND GND

Where:

• C1 is the condenser capsule microphone
• R1 is a 10 kOhm resistor that biases the capsule
• R2 is a 100 kOhm resistor that connects the capsule to ground
• C2 is a 4.7 nF capacitor that blocks DC voltage from passing through
• R3 is a 2.2 kOhm resistor that sets the impedance of the capsule
• R4 is a 2.7 kOhm resistor that sets the gain of the amplifier
• R5 is a 1 kOhm resistor that limits the output current of the amplifier
• C3 is a 47 uF capacitor that blocks any DC voltage on the output
• R6 is a 100 Ohm resistor that provides a DC path to ground for the amplifier

The microphone capsule is biased by the +48V DC phantom power through R1, which charges the backplate of the capsule. The front plate of the capsule is connected to ground through R2. The AC voltage across the capsule is amplified by the JFET amplifier consisting of R4, R5, and C2.

To reduce the noise of the circuit, the values of the components have been chosen carefully. The values of R3, R4, and R5 are chosen to provide a high gain, while keeping the noise levels low. C1 and C2 are chosen to provide a flat frequency response, while blocking any DC voltage. C3 is chosen to provide a low-frequency roll-off to reduce any rumble or handling noise.

R6 provides a DC path to ground for the amplifier, and the output impedance of the amplifier is set by R5, which is relatively low, so it can drive long cables without significant signal loss.

Overall, this circuit provides a high-quality, low-noise phantom-powered condenser microphone with a low output impedance and a flat frequency response, suitable for studio recording applications.


To connect the schematic to an audio interface with phantom power via XLR, follow these steps:

1. Locate the XLR input on the audio interface that provides phantom power.
2. Connect Pin 1 of the XLR connector to the ground symbol (GND) in the schematic.
3. Connect Pin 2 of the XLR connector to the output of the circuit, which is located on the right side of the schematic.
4. Connect Pin 3 of the XLR connector to the +48V DC Phantom Power symbol in the schematic.

Note: Some audio interfaces may have a switch to turn on or off the phantom power. Make sure the switch is turned on before connecting the circuit.

Here's a summary of the connections:

Pin 1 (GND) of XLR connector -> GND in schematic Pin 2 (Hot) of XLR connector -> Output of circuit in schematic Pin 3 (+48V) of XLR connector -> +48V DC Phantom Power in schematic

Once connected, the circuit should be powered by the phantom power from the audio interface and ready to use.


What do you think about it?

 

[bk]

Where's the JFET?

 

[gyraf]

..this is most probably just AI hallucination...

amplified by the JFET amplifier consisting of R4, R5, and C2

 

[komikuapples]

But why?

 

[soapfoot]

FYI, an acoustics forum I frequent had a very similar post from a new account about "asking ChatGPT to design a membrane absorber."

May not be worth engaging with this post until the OP comes back and assures us they're not a bot

 

[abbey road d enfer]

I'd be very worried if someone relied on AI to design and build nuclear plants.

 

[Grim Reaper]

I'd be very worried if someone relied on AI to design and build nuclear plants.

Chernobyl.

 

[komikuapples]

Chernobyl.

The irony here is that Chernoblyl is one of the most public and infamous resources for tested and failed nuclear power, it's not a huge hurdle to assume that a huge portion of chatGPT would be sampling from some of that dataset.

So it could very well/most likely give you a plan that enacts that exact failure vector.

 

[komikuapples]

I'd be very worried if someone relied on AI to design and build nuclear plants.

I wouldn't put it past the grifters who are riding the coattails of ML atm, I work for a major software vendor who is primarily a silicon producer, a solid 50% of our applications are junk about how 1337 the applicant is with prompts under SD or GPT.

 

[Grim Reaper]

The irony here is that Chernoblyl is one of the most public and infamous resources for tested and failed nuclear power, it's not a huge hurdle to assume that a huge portion of chatGPT would be sampling from some of that dataset.

So it could very well/most likely give you a plan that enacts that exact failure vector.

AI should be prevented from using erroneous, fake, harmful, unproven, etc., database content.

 

[Grim Reaper]

Hello!
I asked chat GPT to create a low noise condenser microphone schematics
That's what I got:

To design a low noise studio microphone circuit with a condenser capsule that is phantom powered, you can use the following schematic:

yaml
+48V DC Phantom Power
|
|
R1
|
|
+--------+--------+
| |
C1 R2
| |
| |
+------+-------+ +------+-------+
| | | |
R3 C2 R4 R5
| | | |
Audio Signal Output <-----+------+-------+ +-------+------+< Audio Signal Ground
| |
C3 R6
| |
| |
GND GND

Where:

• C1 is the condenser capsule microphone
• R1 is a 10 kOhm resistor that biases the capsule
• R2 is a 100 kOhm resistor that connects the capsule to ground
• C2 is a 4.7 nF capacitor that blocks DC voltage from passing through
• R3 is a 2.2 kOhm resistor that sets the impedance of the capsule
• R4 is a 2.7 kOhm resistor that sets the gain of the amplifier
• R5 is a 1 kOhm resistor that limits the output current of the amplifier
• C3 is a 47 uF capacitor that blocks any DC voltage on the output
• R6 is a 100 Ohm resistor that provides a DC path to ground for the amplifier

The microphone capsule is biased by the +48V DC phantom power through R1, which charges the backplate of the capsule. The front plate of the capsule is connected to ground through R2. The AC voltage across the capsule is amplified by the JFET amplifier consisting of R4, R5, and C2.

To reduce the noise of the circuit, the values of the components have been chosen carefully. The values of R3, R4, and R5 are chosen to provide a high gain, while keeping the noise levels low. C1 and C2 are chosen to provide a flat frequency response, while blocking any DC voltage. C3 is chosen to provide a low-frequency roll-off to reduce any rumble or handling noise.

R6 provides a DC path to ground for the amplifier, and the output impedance of the amplifier is set by R5, which is relatively low, so it can drive long cables without significant signal loss.

Overall, this circuit provides a high-quality, low-noise phantom-powered condenser microphone with a low output impedance and a flat frequency response, suitable for studio recording applications.


To connect the schematic to an audio interface with phantom power via XLR, follow these steps:

1. Locate the XLR input on the audio interface that provides phantom power.
2. Connect Pin 1 of the XLR connector to the ground symbol (GND) in the schematic.
3. Connect Pin 2 of the XLR connector to the output of the circuit, which is located on the right side of the schematic.
4. Connect Pin 3 of the XLR connector to the +48V DC Phantom Power symbol in the schematic.

Note: Some audio interfaces may have a switch to turn on or off the phantom power. Make sure the switch is turned on before connecting the circuit.

Here's a summary of the connections:

Pin 1 (GND) of XLR connector -> GND in schematic Pin 2 (Hot) of XLR connector -> Output of circuit in schematic Pin 3 (+48V) of XLR connector -> +48V DC Phantom Power in schematic

Once connected, the circuit should be powered by the phantom power from the audio interface and ready to use.


What do you think about it?

Too difficult for my aged brain to comprehend.

 

[komikuapples]

AI should be prevented from using erroneous, fake, harmful, unproven, etc., database content.

Easier said than done, sadly~ just the reality of the rapid development we're seeing. It's a matter of perceived quality of the output, not functional quality or ethics of the model. Though not completely a one-to-one comparison, AI translations of Asian languages are basically the meme example, all about the perceived quality, not accuracy and that's not even considering intricacies of casual conversation in say Japanese to its formally written usage. That's why so many models were shat out the door, most of them were projected to be released 3 to 4 years from now with a much smaller scope, at least that's what we were told behind closed doors. Now it's just a game of our marketing folk trying to figure out how to sell the junk while us engineers cry in the basement.

TLDR; This entire thread is junk and pointless.

 

[abbey road d enfer]

Too difficult for my aged brain to comprehend.

Actually too silly for anyone to comprehend. A preamp with only passive components, come on, AI...

AI should be prevented from using erroneous, fake, harmful, unproven, etc., database content.

Who/what would be the judge?

 

[Grim Reaper]

Actually too silly for anyone to comprehend. A preamp with only passive components, come on, AI...

Who/what would be the judge?

The problem with AI is that it is developed by humans who program in their fallability, prejudices, biases, etc.

 

[gyraf]

It's not so much about human fallibility and biases, as it is about what humans like. The GPT systems are ideally trained against human feedback - but it would be enormously expensive in time and energy to grant it feedback on a 1:1 basis while training its gigantic data set, so in stead it is trained against a model derived from human feedback, i.e. a sort of a first-order generalization of what humans would probably like. That makes likability the key word here, not correctness, precision or safety.

/Jakob E.

 

[Grim Reaper]

It's not so much about human fallibility and biases, as it is about what humans like. The GPT systems are ideally trained against human feedback - but it would be enormously expensive in time and energy to grant it feedback on a 1:1 basis while training its gigantic data set, so in stead it is trained against a model derived from human feedback, i.e. a sort of a first-order generalization of what humans would probably like. That makes likability the key word here, not correctness, precision or safety.

/Jakob E.

Models are only as good as the data used to create the model and the data used includes the biases, flaws, deficiencies, etc., just like the data used for AI and AI programming.

 

[JohnRoberts]

Perhaps we need a new sub forum for generative AI content so we can keep it all in one place to make it easier to ignore.

JR

 

[Grim Reaper]

Perhaps we need a new sub forum for generative AI content so we can keep it all in one place to make it easier to ignore.

JR

Won't be able to ignore it because new posts are brought to everyone's attention.

 

[Spino]

FYI, an acoustics forum I frequent had a very similar post from a new account about "asking ChatGPT to design a membrane absorber."

May not be worth engaging with this post until the OP comes back and assures us they're not a bot

So they're officially taking over. And they decided to start from audio forums.

 

[Phrazemaster]

It's not hard to imagine if they are trained on all the great designs from the last 70 years that they come up with some viable, interesting, and sonically great schematics.

It's only a matter of time. I thought this kind of AI explosion wouldn't happen for another 2-3 years; it's really been the last 2-3 months.

One phrase: exponential growth.