User talk:CTho

From Wikipedia, the free encyclopedia

Contents

[edit] Re: Overclocking

Resistors. Deviation from nominal resistance is usually 15-20% at best on typical CMOS anyway, so a large enough temperature swing can change resistance values enough to cause timing problems. -- mattb @ 2006-09-11T03:13Z
In a CPU? Very few. In a DSP or GPU? There will be many in the DAC portion. Also remember that many interconnect resistances are not negligible in cutting-edge CMOS processes. Anyway, temperature will affect the MOSFET thermal voltage (look at your SPICE model equations), but the S-D current is a pretty weak function of thermal voltage. Of course, the lower the threshold of inversion voltage, the more temperature is likely to have a detrimental effect on the circuitry (and this is the effect you're probably seeing in your simulation). Actually what I took issue with is the spurious claim that MOSFETs slow down with temperature. Parasitic capacitance is the largest factor that affects MOSFET switching speed, and it is a VERY weak function of temperature. One reason for this is that, within a certain temperature range, carrier mobility in the channel with an increase in temperature.
So it's not so much that MOSFETs slow down with temperature, but they do become more susceptible to noise because of the increased thermal voltage. The instability exhibited by, for example, an overheated microprocessor is probably due to some gate(s) not having enough gate voltage applied to drive it into inversion at the current temperature. If you want to re-add the text with that correction, it's fine by me. -- mattb @ 2006-09-12T05:53Z
I don't mean to insult you, either, but it sounds to me like you've never actually designed semiconductor devices and are looking at SPICE simulations at a circuit level and guessing as to the cause of the effects you observed. You did not bother to explain WHY you think a MOSFET would slow down with increasing temperature, just that you don't believe my explanation. I don't expect you to take what I say as gospel, but if you're going to accuse me of misunderstanding then I at least ask that you attempt to offer an explanation of the appropriate effect rather than insisting that the transistor is going "slower" for unknown reasons. Failing that, I encourage you to review your device physics and basic quantum mechanics.
Let me explain again. The factor that most affects a gate's switching speed is obviously capacitance. However, the temperature dependance of capacitance in a MOS system is VERY weak (if you don't believe me, look up the mathematical models SPICE uses for temperature variation effects on capacitance). As a result, parameters like cutoff frequency and yes, gate switching speed, are not affected significantly by these small temperature swings. As I explained before, threshold voltage is probably the culprit here. To drive a MOS gate into strong inversion, the surface band bending of the semiconductor needs to be many times the thermal voltage. Taking the 100 K temperature swing you mentioned, it's not at all unfathomable that the threshold voltage of a MOSFET could decrease by a large factor. A quick calculation with some typical Si CMOS process numbers indicates that your temperature swing of 100 C could easily change a nMOSFET's threshold voltage by 35%, which could be non-trivial. I don't know for sure if this is a contributor to the instability seen in a mildly overheated VLSI system, but I suspect it plays a big role. It certainly plays a very significant role if you're comparing it to capacitance variation with temperature. Keep in mind also that a Si system begins to approach intrinsic carrier concentration at the temperatures you're talking about. The upper limit of Si device operation is usually put at around 450 K. Beyond that, thermally excited carriers begin to dominate doping-generated carriers and the device will simply stop functioning.
If you can think of or find a better physical explanation of the instability in an overheated microprocessor, please feel free to correct me. I don't claim to know much about digital VLSI, but understand that it's slightly offensive when you claim that I'm "misunderstanding" the device but you yourself cannot offer a better physical explanation of the effects you're observing. -- mattb @ 2006-09-13T05:01Z

[edit] Welcome and such

I wanted to drop you a line to welcome you to Wikipedia and encourage you to stick around. I hope that my reverting your edit and making you engage in this rather long dialogue won't sour your experience; I certainly mean no personal slight against you. Anyways, your experience is much needed on Wikipedia. The general state of computer architecture and digital electronics related articles is often pretty sad (just look at where CPU was not so long ago!), and the attention of a CmpE is just what many of them need. Even many of the larger articles need some serious attention; DRAM, Computer storage, Clock signal/Clock rate, etc come to mind. Right now I'm slowly working with some other people on revising the Computer article at Talk:Computer/Temp. Your input there would be much appreciated, since the current state of that article is medicore at best.

Anyway, welcome again, and thanks for being patient and helpful! -- mattb @ 2006-09-14T04:25Z

[edit] Biology Class

Hey, I thought you might find it interesting that my professor found your diffusion animation and used it within his lecture. Cheers! J.Steinbock 20:56, 17 November 2006 (UTC)

Cool, thanks for letting me know! --CTho 23:23, 17 November 2006 (UTC)

[edit] My bad

Sorry about mislabeling Image:Viking1Launch.jpg as a candidate for SVG conversion, I've got some macros for pasting various groups of templates and selected the wrong move-to-commons group without noticing it in this case. I'm usually a bit less of a klutz about that sort of thing. :) Bryan Derksen 06:00, 13 August 2007 (UTC)

No problem. --CTho 12:49, 13 August 2007 (UTC)