Figure 4: Lab setup during temperature tests. Note the shields to reduce the airflow
Figure 5: detail of temperature tests. Note the plastic cover.
The pot of glue puts pressure on the temperature sensor. |
With the PCB placed horizontally and shields around it to prevent any airflow from the side (the top is still open), in an ambient temperature of 22 °C, the temperature of the G-LINK case got upto 60 ±2 °C, while the PCB around the G-LINK was 37 ±2 °C . With a very slight airflow from a small fan placed 30 cm away, the temperatures got immediately down to 47 ±2 °C and 32 ±2 °C respectively. As the G-LINK is specified with a case temperature Tc up to 85 °C, I don't believe it is necessary to put a cooling tower on the G-LINK. During those measurements the G-LINK received FF/00 data running at 50 MHz (around 1 Gbps).
In another measurement I had put a plastic cover on top of the PCB, where the PCB was placed horizontally. Shields where placed 10 cm from the PCB to prevent any forced airflow. The bottom of the PCB was 10 mm away from the motherboard PCB. With this, the temperature of the G-LINK chip did not rise above 76 °C. This means that with the little free airflow at the bottom of the card was already enough cooling to stay within temperature. Normally of course one will have air flowing on both sides of the card. The board even works with a Vcc of 4.0 V and up to 5.7 V it has been tested OK! In those cases (still with the cover on top of it) the case temperature was 57 °C and 93 (!) °C respectively. Both AA/55 and FF/00 patterns were send and checked, while also control words were sent. |
Table 1: Case temperature of G-LINK (data FF/00 or AA/55 and checked to be correct)
|
Case temperature [°C]
(Vcc = 4.0 V)
|
Case temperature [°C]
(Vcc = 5.0 V)
|
Case temperature [°C]
Vcc = 5.7 V
|
| Slow forced airflow |
|
47
|
|
| No forced airflow |
|
60
|
|
| G-LINK and top of PCB covered |
57
|
76
|
93
|
The transmitter board (G-LSC)
A similar layout with the local ground plane has been used for the transmitter board, using the HDMP-1022 chip. The only difference is that 1206 size capacitors have been used, which are a factor of two larger in size than the 0603 size used on the receiver board. Also those have been placed as close as possible to the G-LINK.
The measurement results are similar to the ones described above: the board works well with voltages in the range of 4.00 Volt to 6.00 Volt. With the G-LINK and top of the PCB covered, the HDMP-1022 runs less hot than the receiver: 50, 62 and 74 ±2°C for a Vcc of 4.0, 5.0 and 6.0 respectively.
Conclusions
There are rumours that the G-LINK is very difficult to get to work reliably. I've made a simple system which gives a robust power supply to the G-LINK chip which also helps in cooling the chip. With this system the G-LINK worked very reliably under varying conditions.
The precautions taken are probably overkill, but it has been shown that with those simple measures that don't add any extra cost (and possibly reduce the cost as no cooling fin is needed), one can do a design which is state of the art. It is also nice to see that the local ground plane helps to solve many problems at the same time. The design is shown to be very robust with respect to functioning with high temperatures and voltages working between 4.0 and 5.7 Volt.
A possible improvement is that the decoupling capacitors may be of a smaller value (47 nF or 22 nF) to get a resonance frequency which is higher; 100 nF is not needed as the current surges that the G-LINK needs are not that high.
Another improvement may be to have as Layer 2 (the layer under the local Gnd plane) a Vcc plane. This will form a local capacitor with a value in the order of 100 pF which will bypass very high frequencies. In the current design Layer 2 is the Gnd plane.
The above described technique of a local surface with a local ground plane directly under a chip can be used for other IC's such as PLLs and Gigabit Ethernet transceiver chips that run at a high frequency or at a high temperature. |
