Post by thread:[PICLIST] [OT] test

>Also, how can I test the xtals? The easiest way to test them is to have a spectrum analyzer with tracking generator. One manufacturer (Marconi) had an app note with a circuit described as follows (my ascii art is hopeless) Put a 50 ohm load on the tracking generator, and a 50 ohm load resistor at the end of a coax lead to the analyzer input. connect the crystal between the active ends of the 50 ohm loads. run the analyzer and tracking generator through the frequency of the crystal, with a sweep of +/-5khz (probably for 18Mhz), and you should see a signal on the analyzer which looks a bit like the graphs for phase response. The lowest amplitude point is the parallel resonant frequency of the crystal, and the highest response peak is the series resonant frequency. The depth of the dip is a function of the ESR of the crystal, (remember at parallel resonance it will be seen as a parallel high value resistance) and I would bet that the crystals that do not work will no have a deep dip compared to ones that do work. This ESR figure will also show in the height of the series resonant peak, in that a high ESR will limit the height of the peak. If you have the time, you could do a similar arrangement with a signal generator and an RF voltmeter or oscilloscope, manually tuning the generator to look for the min and max response points. You will see several of each, and need to pick the highest and lowest points to get the relevant ones. If you really want to get fancy, it would be possible to calculate the ESR of the crystal by knowing the voltage division ratio into the 50 ohms on the output side of the network. From this I think you will see that the problem you have is the crystal manufacturer is not controlling the ESR of the crystals (probably because there is no specification) and Microchip are not controlling the spread of gain and phase characteristics of the oscillator, so sometimes you end up with a combination where the oscillator gain is insufficient to overcome the ESR loss. Your suggestion of adding a series resistor will only make this worse, and should only be used with crystals which have a low ESR or oscillator with very high gain, to stop the crystal being overdriven.

>If this were the case then using a "bad" crystal in >another board would likely work. >I don't believe that is the case, however. Which to me implies that the spread of characteristics for the crystals goes outside that allowable for the chips. Hence my comment that the ESR of the crystals is getting too high on some units because the manufacturer is not testing it, because it is not spec'd, because Microchip do not characterize the maximum allowable ESR, because they (apparently) do not spec the gain of the oscillator portion of the chip. (Whew - draws deep breath). for a bit of a treatise on how these parameters affect the oscillator have a look at this Maxim app note. dbserv.maxim-ic.com/appnotes.cfm?appnote_number=220 It is a little different in that the oscillator is a colpitts one, but the principle is the same.