It might make sense to do some thermal cycle tests with a few temperature sensor breakout boards like this one:š https://www.sparkfun.com/products/9418š--On Sat, Jul 28, 2012 at 11:17 AM, Andrey Sviyazov <andreysviyaz@gmail.com> wrote:
Hi Srdjan.Can we measure of all important for us instabilities of calibrations vs temperature and then write it to EEPROM for mass production?I mean, do you have information about repeatability of curves (laws) of deg C instabilities for chips with same and different date codes of manufacture?Also, how near should be temperature sensor to the LSM chip?I mean, can be used averaged relative temperature of the housing heatsink, with known gradient, of course?Please recommend us measure method, even if we need personal temperature sensor for each onboard LMS's.Best regards,Andrey Sviyazov.
2012/7/28 Srdjan Milenkovic <s.milenkovic@limemicro.com>Hi Alexander,
I understand. It is the same with WCDMA and other FDD modulation standards, BTS can not stop transmitting/receiving. However, we can not expect temperature to change from -40 to +85 deg C without doing anything about it.
LMS is quite robust regarding temperature change. For example, PLL will stay locked within +/-20 deg C from the tuning point.
The graph below shows DC/LO leakage calibration stability vs temperature. Applications based on OFDM (LTE, WiMAX for example) do not need to wary about DC/LO leakage recalibration as DC OFDM carrier is not used. Other modulation schemes will have to take care of this. If we define the spec of -60dBc LO leakage, the graph below shows that LMS IC is OK within +/-10 deg C offset from the calibration point.
BTW, you do not need UmTRXv3 at the moment. You can build a small board with temperature sensor alone and interface it back to BB/FPGA.
Best regards, Srdjan
On 28/07/2012 03:15, Alexander Chemeris wrote:
Hi Srdjan, Thank you for the comment. Unfortunately, we can't perform regular re-calibration, because a GSM base station can't stop transmitting. This means we need a full blow temperature sensor and this is a bad news, as we have to create UmTRXv3 to add it. This is something you should definitely consider for your next-gen chip - an embedded temperature sensor. Do you know how sensitive those parameters are to a temperature? I.e. should we calibrate adjust them every centigrade or every five centigrade? On Fri, Jul 27, 2012 at 6:28 PM, Srdjan Milenkovic <s.milenkovic@limemicro.com> wrote:Hi Alexander, As I mentioned before, IQ phase error is quite stable over temperature in 900/1800MHz region so let us put it aside. This can be tuned and fixed in production. On the other hand, DC offset and LO leakage are temperature dependent as one would expect. However, they can be calibrated in the product as well not just in the lab. You need just to trigger built in DC cal blocks and optionally use RF loop back to further improve LO leakage. If you want to avoid temperature sensor you have an option to do these calibrations regularly, every 30min-1h for example. This should track the temperature change without having the sensor on board. Our other customers are using both approaches, with and without temperature sensor, depending on application and spec. If they have on board temperature sensor they are using it for multiple purpose such as monitoring BB, PA, etc. Best regards, Srdjan On 26/07/2012 16:48, Alexander Chemeris wrote: On Thu, Jul 26, 2012 at 4:37 PM, Alexander Chemeris <alexander.chemeris@gmail.com> wrote: On Thu, Jul 26, 2012 at 3:31 PM, Srdjan Milenkovic <s.milenkovic@limemicro.com> wrote: Combining Thomas's idea of DC recalibration and, if necessary, making look up table for IQ phase error correction vs temperature should put us in good position to meet the specs. Well, AFAIK, LMS6002D doesn't have a temperature sensor built in and this means we have to install an external one, which again increase BoM cost. Is that's how this issue is solved by your other customers? I wonder is there a way to solve this without introducing a temperature compensation loop. Let me clarify - DC offset re-calibration is (more or less) fine in lab setup, but it's expensive to manufacture if we have to add temperature sensors and put every unit into a temperature camera for calibration. -- Regards, Alexander Chemeris. CEO, Fairwaves LLC / ïïï õÍòÁÄÉÏ http://fairwaves.ru
Robin Coxe š| šClose-Haul Communications, Inc. š| šBoston, MA šš
+1-617-470-8825