Re: System Monitor Virtex-5 behavior question

07-15-2011 03:31 AM

When looking at the temperature measured via the System Monitor or XADC it is possible that there are discrepancies between the package temperature (T-top / Tc) and the Die temperature or SysMon (XADC) readings (Tj).

The reason for the differences is not usually an issue with the System Monitor / XADC readings but rather errors in how the temperature is measured.

In a customer system the following factors can affect the accuracy of temperature measurements:

1.

Measuring T-top (temperature of the package) in an accurate and repeatable way is always a challenge, when measured it will usually be slightly lower than the actual T-top. The factors that influence the accuracy of these measurements are the location and size of the thermocouple both of which are crucial.

• • Location - placing the thermocouple dead centre on the package is key, even once this is done there can be approx a 2 deg C error even when the correct size thermocouple is used.

• • Thermocouple size – A 36 gauge thermocouple should be used. If a larger one is used it will only act like a heat sink and sucks heat from the product, and cools the top, leading to a larger difference between the die and package temperatures.

Even when the location and size of thermocouple are correct a difference for between 2 -5 C can be seen between T-top measured and the actual T-top. There are also additional factors that need to be correct in order to get the T-top measurements as accurate as possible:

• • Contact surface integrity - i.e. media used to connected thermocouple to package. Using a bare measurement has gap issues and is subject to convection and contact pressure issues. The thermocouple should be attached using a compound in the right bead amount to minimize issues from this.

• • The pressure on top of the package and thermocouple also has an effect. The Thermocouple (bead) has to rest on the surface within the paste, so that the resistance of the past does not affect the measurement. Adding pressure will help, but this will affect the thermal characteristics of the package and again effect the accuracy of the reading obtained.

Some customers use an infrared sensor to measure the temperature, while this removes the variable of the contact media required and pressure required it brings its own measurement error. They average the area around the “dead centre” so an error of 2 – 3 C can be seen using these. Improving the area resolution and surface compatibility can improve this.

2.

The Tj measured through the System Monitor (XADC) is done through a directly probed passive diode, this can result is approx 3 C difference with diode if the reference voltage is properly stabilized. However if the reference voltage is not regulated correctly this can cause an increase in the measurement error of the System Monitor in excess of 3 C and therefore further exacerbating the delta between T-Top and the SysMon Tj.

Given these potential source of measurement error in both T-Top and Tj from the system monitor it can be seen why some customers measure a larger than expected delta between T-top and Tj.