A Little About Temperatures (Just A Tad)

**Reloadron** · October 26th, 2007

The subject of temperatures seems to be just one of those things. I wanted to conduct a few experiments on the subject, however, things did not quite go as planned. Thought I would share some of what I did come up with. I took a look at several temperatures, actually four temperatures and how they relate. I used my workstation as the victim, err test machine. It needed a cleaning anyway so this was a good opportunity to "instrument" it.

"Instrumenting" simply means placing sensors in a device to measure and record data. We could be placing vibration (accelerometers) on a motor to measure vibration at certain speeds or maybe during the manufacture of a motor placing small thermocouples within the windings of the armature to measure temperatures looking for "hot spots". I thought it would be of interest to place a few thermocouples inside (and outside) the case of a workstation. That is what I did. Don't expect to see Speed Fan or Core Temp here. This is about something else and besides to my dismay they don't work on this system.

The system consist of an Intel 7525 series motherboard running a pair of dual core 3.6 GHz Xeons. The GPU is a 3D Labs Realizm 800 commercial GPU. Audio is Sound Blaster Audigy 4 and there are 6 HDDs in a few RAID arrays under an Intel SRCS16 controller card. Overall a top of the line workstation built about 2.5 years ago. The machine generally runs 24/7 and serves as a furnace during the Winter months.

One picture is worth a thousand words so.... Here are a few pictures of the guts of this sucker:

The first is a general picture of the interiour:

The blue cables are the SATA cables from the controller to the HDD rack. The large black thing below the cables is the GPU card. There are fans below the cowling. The next picture better shows things...

I like the small fan assembly devoted just to the RAM. The system has 4 GB of RAM. The processors can be seen below the HDD cage. Each processor has a fan with cowling to draw fresh air in through the front panel. Below is a better picture of the processors...

So much for the guts of the system. My plan was to place 4 thermocouples into the system to measure some temperatures. I wanted to focus on Case Inlet versus Case Outlet as well as PSU Inlet versus PSU Outlet temperatures. The PSU is a 600 Watt commercial workstation PSU. Here are the thermocouples I fabricated. They are shown in 3 images. The overall thermocouples, the tips and the plugs.

Overall:

The sensor tips:

Last but not least the plugs:

Thermocouples are little more than 2 disimilar alloys joined at a junction (the tips). When heated they provide a small voltage. The voltage though non linear is proportional to the temperature.

I designed a simple system to measure the thermocouples using devices called Temperature Transmitters (TTs). The thermocouples are connected to the TTs which amplify and linearize the thermocouple outputs into a more usable signal. Here is a lookie at the TTs:

The transmitters are programmable and in this case are programmed for a Type J thermocouple and a range of 0 to 400 Degrees F. The TTs in turn provide a linear output of 4 to 20 mA (Milliamps of current) accross the range of 0 to 400 Deg. F. They are powered by their own PSU which provides about 28 Volts DC. Looks like this:

Since I really needed a voltage and not a current output from the TTs I placed some small resistors in the current path. They are the small brown things on top of the TTs. When a current passes through a resistor we get a subsequent voltage drop proportional to the current. In this case the current is 4 to 20 mA flowing through a 500 Ohm resistance resulting in 2 to 10 Volts. So what do we really now have? We have 0 to 400 degrees F. scaled as 2 to 10 Volts. Instruments don't read temperature, they read a voltage or current and scale it to temperature. The next trick is to display or read this stuff. The picture now looks like this:

Notice we have an added gizmo?

That "Gizmo" is a small A to D (Analog to Digital) converter. The unit pictured measures 4 analog channels of voltage and converts it to a 12 bit binary signal passed through a RS 232 port.

I wrote a simple routine to display the actual channel data. When all is said and done the picture looks like this:

The system is at idle. The channels are defined as:

Channel 1 is Case Temp In (ambient room temperature)
Channel 2 is Case Exhaust Temperature.
Channel 3 is the PSU Inlet Temperature.
Channel 4 is the PSU Exhaust Temperature.

Now we will load all 4 cores of the Xeons:

After about 5 min. of all four cores running 100% we see the following:

Notice the increases. The only thing that has changed is the load on the CPUs. All of that added heat is a function of the loads on the processors.

The bummer here is I don't have a program to monitor or measure the actual CPU core temps. Because the motherboard chipset is different, programs like CoreTemp and SpeedFan can't read the SMBus. That sucks. You know those core temps are cooking!

The point of this exercise in futility was merely to show what goes on with temperatures inside the case. The data you see could easily be charted & recorded onto and into a spreadsheet.

Questions? Ideas? Thoughts?

Ron

Yellowhello · October 26th, 2007

Whoa, nice program! Wish I had those programming mad skillz...

Nice little write-up man! Great pics too...

**Reloadron** · October 27th, 2007

Quote:

Originally Posted by yellowhello

Whoa, nice program! Wish I had those programming mad skillz...

Nice little write-up man! Great pics too...

Thanks Yellow, but in an effort to be honest it is not much more than a few lines of Visual Basic 6.0 and some math to derive the degrees C. readings and the delta temperatures. I am not a programmer type. Not by any means! LOL.

The data acquisition device is a Dataq DI-154 which I bought a few years back. They sell for about $39.00. Not a bad little A/D unit but it does use a serial (RS232) port, which fewer and fewer machines have anymore. Nice unit in that it is a 12 bit converter so we do get very good resolution and accuracy. I did calibrate the sensors to the unit and the temperatures are extermely accurate (+ / - 0.5 Deg. F.). Actually the small A/D unit includes some charting and recording software and has fixed range input channels of 0 t0 10 Volts or actually -10 to 10 Volts.

Again, the exercise was just about looking at some temperatures inside the case we don't generally look at.

Ron

**Stormcrow** · October 27th, 2007

Woah, I didn't realize 3D Labs made such a brick. It could pass for an 8800.

**Reloadron** · October 27th, 2007

Quote:

Originally Posted by Stormcrow

Woah, I didn't realize 3D Labs made such a brick. It could pass for an 8800.

Yeah, that sucker is big. The bitch is that right after I bought that card 3D was that 3D was taken over by Creative Labs and they discontinued high end graphics for commercial applications. The support as to drivers for XP and XP 64 was good but ends there.

Ron

**PrOLifIC_onE** · October 27th, 2007

So, Ron, how can get one of your temperature gizmos?

See my PM.

polobunny · October 27th, 2007

A Wildcat Realizm 800? You had some money burning a hole in your pocket or what? Those things are freaking expensive. o_O
Nice writing though.

**Reloadron** · October 27th, 2007

Quote:

Originally Posted by polobunny

A Wildcat Realizm 800? You had some money burning a hole in your pocket or what? Those things are freaking expensive. o_O
Nice writing though.

That card was $2,000 new when I bought it. Now I will share my lesson in stupidity further. When I built the workststion it was to replace my older Dell 530 MT workststion. That unit had a perfectly good Wildcat II graphics card in it. I could have used that card but nope, I opted for the new card. The old card would have worked just fine. When I built I used the newest and latest and greatest of everything.

Ron