Thermal Diode Reader
 

I was reading up on pH's article on amd's thermal diode calibration and would like to do something similar and I was wondering where I coudl go about purchasing the Maxim 6655EVSYS diode evaluation system. It is the one that pH uses and it seems to be very accurate and have a very useful UI. Maxim has the item on their site listed as 100 dollars, but i dont think i can stomache 100 dollars for an eval board.

If anyone has any suggestions I would greatly appreciate it. Or a low cost alternative that is as accurate would be great too.

http://www.procooling.com/articles/a...x6655evsys.jpg

It only takes three components (chip and two capacitors) and some miniature shielded cable to make one yourself. You do have to be able to read a DMM and solder. You can request samples from MAXIM, so the chip is free. Solder the components right on the back side of one of your DIMM slots where you gain access to the SMBus, assuming you motherboard does not have an SMBus header. The hardest part, regardless of who's board you use is isolating the CPUs thermal diode pins from circuitry already on your motherboard, so that they don't interfere with your reading. That takes some investigating, but if worse comes to worse, you can do like I did and solder right to the CPU itself. Articles are around detailing how to do it.

Hoot

I already did the soldering to the cpu with an old dummy p3 866. its increidbly easy.. I will ask my father (works in a music studio so they have tons of wires and bread bord and caps) for some of the other parts so that it wont cost me anything. I suppose that shielded cables are a very big part of it as any interfearence wil screw up the readings. but do they have 26 gauge shielded cable? kinda small eh..

Im pretty sure the test board has an smbus header. its hte nf7-s. When i pull it out of this rig I will have to take a look.

One thing, I was looking at maxims website and saw that they have other readers for temps that have .05+- acuracy compared to the 6655E which is 1C+- accuracy. seeems logical this would be a better choice. Am I wrong?

And to anyone who wants to solder directly to the cpu, if you have a 70watt iron like i do, TURN IT DOWN... lol... 70 watts on the back of the cpu isnt good. it starts to melt things when u get about 3mm from the surface.

70w iron? jeez i use a 25w.

Anyway if you want to go cheap, theres always the 6658 (IIRC) which you can just request a free sample. I doubt they'll give you a free sample of the 6655EVSYS. The downside is that its an SMB part, so it plugs directly into a motherboard, and not to a serial port. On the upside, it'll read down to -50C, so you can use it with phase change systems too.

For shielded cable, use either coax cable (like for TVs) or just use networking cable. Either will reduce noise a good bit. If you want even clearer readings, you can hook into labview like pH does, and add a 60Hz notch filter there to remove all the power supply noise from the system.

MaxxxRacer: Did you find a flowmeter yet?

For small work I use a 17W pen. The NF7S has components connected to the thermal diode to drive the Attansic ATTP1 thermal protection IC, so forget leeching off of the motherboard. I used short lengths (~1 inch) of 30ga enameled wire to get the two pins away from the socket, then transitioned them to 26ga shielded pair. Keep the run to the MAXIM or whoevers chip as short as possible and with the 2200pf cap across the input pins, you won't have trouble with noise. Originally, I ran two 6in lengths of the enameled wire with a light twist and the 2200pf cap kept the noise on them down just fine. I actually soldered a 2200pf chip cap right across the two input leads on the MAXIM chip. Do not route the thermal diode to the Winbond W83627THF chip. It is a bad choice for reading thermal diodes. It even uses a method not recommended by AMD for reading the diode, but it's a cheap solution so many motherboard manufacturers use it.

Hoot

Thanks for the info gus..

My iron is adjustale (edsy loner baby!) so the wattage is not an issue.

hoot, im gonna just use some short 26ga wire that is ultra low resistance to low the capacitance in the wire and to help with noise. My dad did have some teflon coated silver core wire but they used it all up and it costs an arm and a leg.

All 26 gauge wire has very nearly the same resistance, so theres no need to worry about that. At 43.6mOhms per foot, its essentially zero for you needs. Likewise, since you're adding a capacitor to the circuit, worring about a few pF of capacitance in the wire (which will actually reduce noise) is not productive.

Also, all the noise in the system will be inductively coupled into the circuit by motherboard and power supply. Thats why you have the cap to filter out as much as possible. Beyond that, all you can do is use sheilded or twisted pair wire and try to make your soldering and wiring as clean as possible.

Read through the Maxim data sheets. They explain this stuff clearly.

will do but which part should i use? The maxim that is.

The DS1631 seems to have a better accuracy than the other maxim sensors at .5C+- at 0 to 70C and at 30C has ~0 error.

Looking at its pdf file there seems to be a parabola of sorts that im sure has an equation to it. If i could get maxim to do a equation fit to that curve I could level out the results so that they have even less error...

The DS1631 has better accuracy, but it does not accept input from an external thermal diode. It uses one that is inside the chip. PM on the way.

Hoot

hoot, i wasnt aware. I was trying to find some info in the pdf but it wasnt anything i could decipher.

Did you look at the 665x specs? Its probably not a bad choice. The sheilding stuff I mentioned is in there as well.

http://pdfserv.maxim-ic.com/en/ds/MAX6657-MAX6659.pdf

I will be using the 6658 i think. I might take hoot up on his offer. btw sorry if i sounded rude in the pm hoot, i thought i would be able to use the other chip hence i declined. a bit to hasty.

EDIT: I got the board done except for the chip.. dont have the chip yet but i was bored so i made the board.

One question though. do the ground ad 5 volt have to come from the smbus header. i wouldr rather it not, becuase its in a really awkward place. i would rather hook up directly to the psu or something.

You can get the voltage from anywhere you please. I have 5 pin IDC connectors with pigtails to run to your board. It plugs into the SMBus header with no soldering required. You're welcome to one under the same offer from the PM I sent you.

One thing you are going to have trouble with in your board design is the fact that these chips, regardless of who makes them are not .100 center stick-through packages like the board needs. They are extremely tiny surface mount devices. The .pdf file has the dimensions at the end. They can be, with a small soldering pen and for my old eyes a stereo microscope, soldered to a mini-dip component carrier and then mounted in a .100 center stick-through board. Again, same offer as in the PM..

Good luck with whatever you choose to do. The effort is well worth it.

Hoot

ohhh... hoot i think I will have to take your offer on all of those items. having the socket would do wonders. My dad was mentioning the fact that when making bread board components (he's made some insanely intricate stuff with them), solding the chips directly to the board is kinda crazy as you cant remove them easily and its harder to solder.

btw, would you mind if i just paypaled you like 7 dollars. of you u still want me to send you the manila envelope?

At the forums where I moderate, it is not appropriate to discuss such matters in an open forum. I'll PM you. I'm an "Elmer" kind of guy. Some who read this will know what I mean. Some will not. I like to encourage people to do things themselves, such as what you're doing. Just my way of giving back all that I have enjoyed in Electronics these past 35 years.

Hoot

thats quite a number of years. lol.. though i think after soldering krytec and stripping it with a standard wire stripper i can solder just about anything... stripping teflon off a 30ga wire is not fun at all.. luckily i got smart and grabbed all the tools for this out of the toolbox. It helps to have a toolbox with every known hand wiring tool.. lol..

Elmer? hmm.. got a cat named elmer...