I have been given some advice in the past with regard to purchasing components.
Be wary of cheaper components. They are likely counterfeit, repurposed (second-hand), or out of specification.
It’s almost laughable how much cheaper some components are overseas. Up until this point, I hadn’t experienced any problems with cheaper components sourced from a country that will remain unnamed.
Up until this point…
So I'm working on a unit and I noticed the clock losing a few seconds every day. I wondered about the quality of the 32.768 kHz crystal I was using. I never had a frequency counter, and the resolution offered by my work oscilloscope wasn’t enough to give me a decent result. I applied for a RoadTest for the TBS 1202B-EDU a few months ago because I noticed it had a 6 digit frequency counter. I wasn’t successful in that application, but I was successful for the Ben Heck scope giveaway of the exact same oscilloscope. I completely forgot about the frequency counter until I saw the functionality on the scope when I booted it up. I decided to make a few measurements.
Here are the frequency measurements of the 32.768 kHz and 48.00 MHz crystals on a Silicon Labs starter kit:
So, let’s calculate the accuracy of the 32.768 kHz crystal:
(32,768 – 32,767.7)/32,768 = 9 ppm
And the 48.00 MHz crystal:
(48,000,000 – 47,996,900)/48,000,000 = 64 ppm
The specification on the NX5032GA-48.000M (48 MHz crystal) is 50 ppm, but that’s at 25 degrees Celsius, and it’s very rarely 25 degrees Celsius in Southern California. Because of the temperature tolerance, it’s understandable the frequency may be slightly incorrect. The 48 MHz clock is used for clocking the core of the microcontroller and for timing purposes the 32.768 kHz crystal is provided. Notice the accuracy of the 32.768 kHz crystal is much better than the 48 MHz one. The specification for the MS3V-T1R (32.768 kHz crystal) is tighter at 20 ppm.
Now, here is the measurement of my crystal:
32.772kHz! Let's calculate the accuracy:
(32,772.0 – 32,768)/32,768 = 122 ppm
122 ppm! Five times worse than the stated specification of the previous clock chip and almost 14 times worse through measurement. How many seconds in one day is that?
122 ppm * 60 * 60 * 24 = 10 seconds.
This clock loses 10 seconds a day!
Now, here comes the obligatory quote from the specifications of the scope. The scope can have a maximum error of 51 ppm.
