Apologies for resurrecting this old thread, but I thought it might be useful to discuss the pros and cons of TCXO with the DW1000. We have extensive experience with TCXOs, and there are a lot of subtle aspects that one would not expect which can affect how well they work in a Decawave system.
In one sentence: TCXOs are NOT an assured way to improve accuracy and performance of an UWB system, they have a lot of gotchas due to their internal architecture and other faults.
I think it is important to first discuss the various clock choices:
Crystal or standard crystal oscillator:
This provides a clock source which is not temperature compensated and thus can drift with temperature over some range. There is also an initial tolerance which usually can’t be any better than about +/- 10 ppm. If subjected to a temperature gradient, the timing can shift a lot, several ppm. If you are doing a ranging cycle, this will affect the result.
TCXO, old style:
Back in the old days, temperature compensated crystal oscillators were made with a complex analog network of thermistors which controlled the voltage on a varactor diode. If you set up the network just right, the change in resistance of the thermistors would change the capacitance of the varactor and pull the crystal into being mostly temperature compensated. Tuning this was complex and costly, and you couldn’t get all that accurate with it. But, and this is important, the frequency changed smoothly with temperature, no inflections or steps. This old style TCXO is no longer made, alas.
TCXO, modern style:
Modern TCXOs use a digital compensation circuit inside them. Periodically, they measure the temperature, index a table based on temperature, and then output a control voltage via a DAC to a varactor to pull the crystal. This circuit allows the manufacturer to precisely tune the crystal offset at every temperature, load the precise calibration table to the TCXO, and thus can supply devices as accurate as +/- 0.5 ppm.
Here is the important part: the modern TCXO updates its crystal pull relatively slowly, at some rate that is adequate enough to keep the crystal in tolerance for the fastest temperature gradient expected to see in service. This means that the adjustments do not occur smoothly all the time, but are really small piece wise linear segments. Every time the circuit loops through and adjusts, there is an inflection in the frequency.
If that inflection occurs in the middle of a ranging operation, then the time base has shifted ever so slightly and that can affect your ranging accuracy. One might think that shift would be so small as to not matter, but it does. Consider that a 20 ms ranging cycle will be off 10 cm if the crystal shifted 0.017 ppm (or 17 ppb). A TCXO can easily shift that much and still be within datasheet spec.
TCXOs vary in how much they bounce back and forth, and how often they adjust. We have found some makes/models which are terrible and don’t work well. We have found some that do.
Even with the ones that do work, we find some individual TCXOs which exhibit anomalous behavior that is know in the industry as “micro jumps”. These are not so much related to the temperature comp circuit as to certain intrinsic properties of the crystal itself. In one particular case, we found a TCXO in an anchor that jumped 4 Hz, or 0.10 ppm for 38.4 MHz, well within datasheet spec. However, that would trash our measurements making the data from that anchor bad. After a frustrating hunt for the cause, we eventually captured the event on a spectrum analyzer:
The two peaks represent two sweeps, the TXCO had jumped frequency by a few Hz (span is 100 Hz).
We added code to our system to specifically detect when a TCXO jumps and to map it out of the location solution during those times, then bring it back in again when stable.
Here’s a good background article on various TCXO issues and discussion of internal operation:
In closing, the clock source to the DW1000 is absolutely critical to get good results in an UWB system. You want it to not change temperature very fast, be stable, and be accurate. A TCXO may sound like the best way to do that, but it can introduce more problems than it solves. We do use TCXOs in our designs, but we only specify ones we have tested extensively and found to behave.
Mike Ciholas, President, Ciholas, Inc
3700 Bell Road, Newburgh, IN 47630 USA
mikec@ciholas.com
+1 812 962 9408
