next topic »

[on:]

This question was posed in my "Managing and Maintaining a PC" class today by a student that is in the residential electrician field. I really couldn't think of a good response. Why do computer components operate on low voltages like 3.3v, 5v. and 12v? Wouldn't higher voltage mean than you could use smaller wires (lower current, like you see in high-voltage power lines)? The only thing I could think of is that maybe the higher voltage can cause problems when you're looking at CPUs built on the nano-meter scale. With all those traces so close together, would higher voltage mean more powerful EM field? I've got nothin'.

[Reply #1 on:]

In 2 years I migrated from 12v cmos techology, to briefly 5v logic, and now it seems 3v3 for the GPS / transceiver type modules available.

I think that the 24v , or 5 volts can flash over between tracks in the chip, with the spacings they need to use , so 3v3 works with the micron spacing for these chips.

[Reply #2 on:]

Well, at one time, they did operate at higher voltages -- hundreds of volts, even. Back when computers were made with vacuum tubes.
 

The problem with higher voltages is two-fold. First, higher voltages require thicker insulation. If you're trying to make a computer smaller, that's a problem.

Second, if you look at the specs for a transistor, you'll notice that many of the ratings are specified as CURRENTS. Although we often use voltages as a reference in digital circuits, deep down inside they're really ANALOG circuits, which require a certain amount of current to operate. If you design for a higher operating voltage, you still need to run a minimum amount of current through the transistors to get them to switch.

Do the math:

3.3 V x 0.010 A = 0.033 W

24 V x 0.010 A = 0.24 W


Multiply those Watts times thousands of transistors in an IC  and you can see where the power requirements (and heat) could quickly get out of hand.

Higher voltage logic is also slower for some reason. I think it's because of the inherent capacitance of the circuits and the charging time to reach the required switching voltage. But I could be wrong.

[Reply #3 on:]

The simple answer is that most PC manufactures don't manufaure the chips used in it's design. They rather select what chips the chip industry has avalible. So the question really morphs to why do chip manuafacures use the voltages the do offer rather then much higher voltages. The answer seems to be that as chips have become faster and more dense in transistor counts the physics favor moving to lower and lower chip voltages, where today we see that the internal Intel processor chips work at around 1 volt level internally to best deal with heat dissapation and clock speed used. So the answer to the electrician in your class is to tell him that wire or chip conductor size has not been a limit, but rather the internal heat dissapation by the transistors the make up the chip design, packing millions and millions of transistors onto a single small die creates very difficult heat management problems that can only be dealt with by using lower and lower switching voltages.

Lefty

[Reply #4 on:]

The problem is that as circuits get smaller and smaller and faster and faster the basic technology still remains the same. TRANSISTORS. ANd these transistors don't always switch on and off as cleanly as we might desire. So there is some time when 2 devices are conducting when one of them should be off. Kind of like an H-Bridge. By going to lower voltages the current in those shorted conditions is less and less heat is generated.

[Reply #5 on:]

Thanks for the clarification. The original assumption was that the higher voltage would give the same power output, with a lower current (and thus allow smaller wires/traces). I didn't even think about the current threshold of a transistor (although I should have), or the higher voltage making the transistors run hotter.

This class isn't supposed to delve into electronics this deeply, but its hard to avoid when we're troubleshooting hardware and trying to explain why its not working.

[Reply #6 on:]

There is another consideration for using higher voltage, ( as in CMOS at 12v that I designed around for 40 years until I saw the Arduino light )

With  12v CMOS logic there is a typical 0.45vcc noise immunity ,  so 4 volts of noise on an input can be tolerated without it corrupting an input.

I havn't yet built an embedded Arduino project for a car ( although I do have one in the pipeline ) but I would be wary about using 3v3 logic without a buffer and filter for that notoriously ( electrical ) environment.

[Reply #7 on:]

Look at it this way:  The physical limits of materials are based on the strength of the electrical fields that they're exposed to.  (Air arcs over about about 20,000V per cm, for instance.)  Electrical field strength is measured in "Volts per Meter."  You get stronger electric field strength by increasing the voltage or by making your distances smaller.  The reason that modern electronics are so cheap is that the size of individual transistors keeps getting smaller; to keep the electric field strength similar, the operating voltage has to go down.
A vacuum tube operates at hundred of volts and has distances of a few mm; about 20000V/m field strength.  The transistors in a modern CPU have layers than are about 5 nano-meters thick (that's "not very many" atoms thick, BTW.)   Even at 1V operating voltage, that's something like 200000000V/m field strength!

[Reply #8 on:]

Second, if you look at the specs for a transistor, you'll notice that many of the ratings are specified as CURRENTS. Although we often use voltages as a reference in digital circuits, deep down inside they're really ANALOG circuits, which require a certain amount of current to operate. If you design for a higher operating voltage, you still need to run a minimum amount of current through the transistors to get them to switch.

Do the math:

3.3 V x 0.010 A = 0.033 W

24 V x 0.010 A = 0.24 W


Multiply those Watts times thousands of transistors in an IC  and you can see where the power requirements (and heat) could quickly get out of hand.

Nicely! laid out

[Reply #9 on:]

Why don't CPUs and microcontrollers operate on 24v?

Slew rates?

[Reply #10 on:]

Every time you switch a transistor in a CPU, heat is released. The higher voltage you run, the more heat you generate. How fast can you run a 12V CPU without special cooling?

You want to run real fast highly complex CPU's then you gotta run the smallest transistors at low voltage, and still need heat sinks and fan for the hotrods.
 

[Reply #11 on:]

Why don't CPUs and microcontrollers operate on 24v?

Slew rates?

SLEW RATES!

That's the term I was struggling to remember. Thank you!

If the maximum slew rate of your circuit is 12 Volts / microsecond and you try to drive a 24 V logic signal, then it will take 2 microseconds to reach 24 V.
If you're running your logic at 3 volts with a similar circuit (and similar slew rates), a switching signal will take only 0.25 microsecond.

[Reply #12 on:]

Modern high speed processors have another problem to deal with as well, electromagnetic interference.

Every conductor has some inductance rating, even if it's in the range of nano or pico henries.  While that may not sound like much, it does result in the generation of electromagnetic fields around those conductors as high speed signals run through them, and as internal traces on processors get smaller and smaller (the latest Intel i processors are built using 32nanometer traces), these miniscule electromagnetic fields can generate currents in neighboring paths, thus requiring the processors to run on increasingly lower and lower voltages to keep these fields small enough to mitigate such interference.

[Reply #13 on:]

All the above, plus 24V lets the "magic smoke" out

[Reply #14 on:]

Go for smoke! God for smoke! Bzzzzzattt! Poof! Ahhhh!