Digital Potentiometers: How Small Can They Go

[HatMan117]

I have an idea that involves using something like digital potentiometers instead of transistors in microchips to carry out logic. I was wondering if anybody knows if it's even possible to shrink a digital potentiometer down to near the same scale as modern microchip transistors, and why/why not? Also, I'd like to know more about how they work to begin with.

[WaterCooledNoob]

yea its already been tried does not work on silicon due to its design. 

[HatMan117]

yea its already been tried does not work on silicon due to its design. 

 

Why doesn't it work on silicon? Was there any success in getting it that small, or was it a failure?

[WaterCooledNoob]

3 minutes ago, HatMan117 said:

 

Why doesn't it work on silicon? Was there any success in getting it that small, or was it a failure?

intel attempted to use this back in 2010 but they could not get the logic to perform in a controlled way . as well as being very unreliable and unstable . dont forget digital Potentiometers use transitors . it is just less effecient . it can be done but transistors work better. 

[mariushm]

Think of digital potentiometers as a lot of resistors (129 or 257 for example) and mosfets or some component that's shorting out some of those resistors

or connecting the wiper pin between 2 resistors.

a lot of components, of course more than a bunch of transistors.. including in some cases some memory to retain wiper position at power off and some logic to decode i2c or spi signals (quite a few transistors)

 

see for example block diagram in datasheet : http://ww1.microchip.com/downloads/en/DeviceDoc/22107B.pdf

 

[HatMan117]

16 hours ago, WaterCooledNoob said:

intel attempted to use this back in 2010 but they could not get the logic to perform in a controlled way . as well as being very unreliable and unstable . dont forget digital Potentiometers use transitors . it is just less effecient . it can be done but transistors work better. 

Maybe I should explain the full context of why I ask.

 

Recent research into plastic magnets suggests that it may be possible to make tunable plastic magnets that operate at room temperature in the near future. Basically, depending on the color of the laser you fire at it, you can change the strength of the magnet. Thus, it is theoretically possible to use plastic magnets in a hard drive to store data in more than two states (not just north and south, but also at certain strengths).

 

This would allow for far more compact storage of data, since you wouldn't have to store it in binary. You could theoretically store it in, who knows, maybe, for example, 8 different states depending on polarity and strength (a hypothetical "octinary" language)... However, I feel that the advantage of highly compact storage might be offset by the slower read and write speeds that would likely come with having to translate the data to and from binary on the computational side of things.

 

So, I figured, since such a computer would no longer be forced to store data in binary, perhaps it might be possible to apply the same base coding across the entire thing. If the computer computed in "octinary" as well as stored data that way, then the entire computer would have the benefit of a far more compact coding system (meaning potentially huge benefits in CPU, GPU, and HDD performance for the same sizes), without the disadvantage of translating between two different languages all the time.

 

I thought this could maybe be accomplished by making logic operate not in on/off states, but rather by assigning bits to the strength of a current... Which is where digital potentiometers come in.

 

So if Intel already tried this, was it less efficient because they were translating from binary because that's what the data was stored in? Or was it unreliable because temperature changes resulted in changes in current (which would in turn potentially change the value of a bit)? Or was it some other issue, like a loss in volume efficiency, since potentiometers are inherently larger than transistors? Do you perhaps have some reference materials about Intel's experiments that I could study?

 

Also, I understand that digital potentiometers contain many components, including transistors, and are as such, bound to be inherently larger than any transistor. However, given the exponential increase in computing power a bit with more than 2 values would allow, is it possible that doing things with potentiometers over transistors would be worth the loss in volume efficiency that would bring?

[WaterCooledNoob]

1 minute ago, HatMan117 said:

Maybe I should explain the full context of why I ask.

 

Recent research into plastic magnets suggests that it may be possible to make tunable plastic magnets that operate at room temperature in the near future. Basically, depending on the color of the laser you fire at it, you can change the strength of the magnet. Thus, it is theoretically possible to use plastic magnets in a hard drive to store data in more than two states (not just north and south, but also at certain strengths).

 

This would allow for far more compact storage of data, since you wouldn't have to store it in binary. You could theoretically store it in, who knows, maybe, for example, 8 different states depending on polarity and strength (a hypothetical "octinary" language)... However, I feel that the advantage of highly compact storage might be offset by the slower read and write speeds that would likely come with having to translate the data to and from binary on the computational side of things.

 

So, I figured, since such a computer would no longer be forced to store data in binary, perhaps it might be possible to apply the same base coding across the entire thing. If the computer computed in "octinary" as well as stored data that way, then the entire computer would have the benefit of a far more compact coding system (meaning potentially huge benefits in CPU, GPU, and HDD performance for the same sizes), without the disadvantage of translating between two different languages all the time.

 

I thought this could maybe be accomplished by making logic operate not in on/off states, but rather by assigning bits to the strength of a current... Which is where digital potentiometers come in.

 

So if Intel already tried this, was it less efficient because they were translating from binary because that's what the data was stored in? Or was it unreliable because temperature changes resulted in changes in current (which would in turn potentially change the value of a bit)? Or was it some other issue, like a loss in volume efficiency, since potentiometers are inherently larger than transistors? Do you perhaps have some reference materials about Intel's experiments that I could study?

 

Also, I understand that digital potentiometers contain many components, including transistors, and are as such, bound to be inherently larger than any transistor. However, given the exponential increase in computing power a bit with more than 2 values would allow, is it possible that doing things with potentiometers over transistors would be worth the loss in volume efficiency that would bring?

tbh i dont know i just read an article a while back about intel ettempting this method and failed . i dont know why they did not continue on but you have a very strong arugement . I wish i could help but its above me on this matter . good luck . 

[HatMan117]

17 hours ago, WaterCooledNoob said:

tbh i dont know i just read an article a while back about intel ettempting this method and failed . i dont know why they did not continue on but you have a very strong arugement . I wish i could help but its above me on this matter . good luck . 

Lemme know if you ever run into that article again. I'm having issues finding any such article, and I'd be very interested in reading it.

 

Thank you.

[WaterCooledNoob]

Just now, HatMan117 said:

Lemme know if you ever run into that article again. I'm having issues finding any such article, and I'd be very interested in reading it.

 

Thank you.

will do pal 

[Lolucoca]

Here are a couple MCP40151 digital 50k potentiometers and on the right there are the smallest SMD transistors I had on hand, just for reference.