Fig. 1 – The LimeSDR board. Courtesy of Lime Microsystems.

One thing that struck us as well as the community is the overall heat that is cumulating on the LimeSDR. Having worked with the Ettus B210, we were surprised of the 60°C shown in LimeSuite under no significant load. Some were concerned by the premature degradation of the board, but heat may also impact RF performances (chapter 4.4). As such, we decided to take action against what we considered could be a problem.

This SDR contains three main chips: a Cypress FX3 CPU, an Altera Cyclone IV FPGA and a LMS7002M RF chip. On Fig. 1, these can be found at the top, in the middle and at the bottom of the board, respectively. There are also some voltage regulators as well as a few inductors susceptible to heating on the top and on the left. The complete schematics and routing of the board are available on the LimeSDR’s Github page.

In order to have a clearer idea of what was going on, we used a Fluke Ti90 camera to inspect the thermal activity in details. For all the following measurements, we let the board run the gnuradio DVB-T project for 30 minutes before taking the pictures. Fig. 4 shows how the naked board performs out-of-the-box. The camera indicates an average board temperature of 64.0°C with a maxima on the center – the FPGA chip – around 67.0°C. Fig. 5 shows how the Ettus B210 performs in comparison: more than 25°C lower with an average temperature of 38.7°C!

Thermal view of a naked LimeSDR board

Fig. 4 – Thermal view of a bare LimeSDR board. From top-left to bottom-right: the board, the Cypress chip, the FPGA chip and the RF chip.

Thermal view of a naked Ettus B210 board.

Fig. 5 – Thermal view of a bare Ettus B210 board. From top-left to bottom-right: the board, the Cypress chip, the FPGA chip and the RF chip.


The first step in cooling down the Lime board is to heatsink every sensible chip. Some like the smaller ones are harder to find, but we ultimately found everything we need on rs-online.com (a complete BOM is available on our Github repository). As seen in Fig. 6, a reduction of nearly 10°C can be expected by using heatsinks alone.

Thermal view of a LimeSDR board with heatsinks mounted

Fig. 6 – Thermal view of a LimeSDR board with heatsinks on. From top-left to bottom-right: the board, the Cypress chip, the FPGA chip and the RF chip.

A top picture of the LimeSDR in its case with heatsinks.

Fig. 7 – Top view of the case. Heatsinks are mounted on 6 chips.


Heatsinking works best with an active fan and good air flow. Using Solidworks, we designed a plastic case to be 3D printed (also available on our Github project). The 12VDC fan is to be mounted on the top cover and is connected to the external supply on connector J21. A 9V to 12V PSU works best. Holes for SMA connectors allow access to every UF.l connector using SMA pigtails. A hole on the top cover permits access to the FPGA JTAG header. After 30 minutes of constant load, the board is still cool with an overall temperature of 31.5°C, a nearly 35°C improvement over the initial scenario! Fig. 8 shows the detailed picture.

Thermal view of a LimeSDR board with heatsinks and fan on.

Fig. 8 – Thermal view of a LimeSDR board with heatsinks and fan on. From top-left to bottom-right: the board, the Cypress chip, the FPGA chip and the RF chip.

A top picture of the LimeSDR in its case with fan, with top cover.

Fig. 9 – Top view of the case with fan and top cover. SMA connectors are dispatched on each side.


While cooling the LimeSDR is not mandatory, we believe that doing so may extend its lifetime and yield better performances. Either way, the 31.5°C average temperature is a good working point, much better than the 64°C measured out-of-the-box.

You will find all the files mentioned in this document on our Github repository.

Acknowledgement

This research work was supported by the Armasuisse Science and Technology Federal Office
as part of project WISE [grant 8003514629].