Hi So … how good is the “calibrate and go” (not the tone on second channel) approach likely to be?
If it’s a bare crystal or normal XO (not a TCXO) that is supplying the clock, the crystal will follow some fairly well known curves. Which one of the curves it follows will depend entirely on the individual sample you have. What I see on my sound card will not be what you see on a different sample of exactly the same sound card installed in a similar machine. It’s a pretty good bet that your device will change by 0.1 to 1 ppm per degree C. Yes it can be a bit better. It can also be a whole lot worse. Your room environment is likely to move 1 to 3 C per hour if you have the heat or air-conditioning turned on. If you are in an unheated garage … who knows. If you take the most likely 0.5 ppm/C tempco and the most likely 2C change you get a “wobble” of about 1 ppm. The period of this wobble probably will be in the 30 to 90 minute range. On top of the temperature induced effects there are also voltage issues and long term aging. Unless you are in a very unusual setting, both should be well below the temperature related effects until you get out to time periods of many months to several years. At 130 KHz, 1 ppm is 0.13 Hz. You will need to run a pretty good / long FFT on the sound card to get that sort of resolution. With some coding, you can do various things to speed the process up a bit. Depending on what you are doing, 1 ppm may indeed be “good enough”. A typical TBolt GPSDO when locked and running normally will supply an output that is good to better than 0.1 ppb 99% of the time. It will hit 0.01 ppb 90% of the time. With some care, it could be even better. Both assume you are using a counter gate in the 1 to 10 seconds range to check it. That moves you up 10,000 to 100,000 times more accurate than the 1ppm drift on the sound card time base. Getting *useful* accuracy at that level out of a straight sound card reading is quite a challenge. You also get into a lot of fancy math surrounding things like 6 hour long measurement times. My *guess* would be that the GPSDO is not really needed in this case ... Bob > On Oct 31, 2015, at 8:58 PM, Tom Van Baak <t...@leapsecond.com> wrote: > >> As an aside, I work low frequency RF transmissions on 136 Mhz, and >> very narrow bandwidth. Can a soundcard be locked to GPS instead of >> its own internal crystal for precise frequency output? > > Chris, > > It might first be interesting to see how far off the frequency is before you > worry about disciplining it with GPS. > > One trick that I use is to make the soundcard generate 1 Hz pulses and > compare that against GPS with a TI counter. The tool is 1hz.c 1hz.exe in my > tools directory (http://leapsecond.com/tools). This avoids having to open up > the computer and probe or buffer the soundcard oscillator. > > If you collect a long enough data set you will also get a feel for how stable > the oscillator is; something you will need to know to tune your GPSDO > algorithms. > > But wait, there's more. If you find that your soundcard oscillator is, say, > 12.34 ppm fast -- you don't actually have to tune or discipline or replace > the physical oscillator. Instead just change the software that's writing to > the soundcard and have it generate waveforms that it thinks are 12.34 ppm too > slow. In other words, instead of defining PCM_RATE 48000 as a constant, you > set pcm_rate = 48000 * (1 + 12.34e-6) as a variable. One line of code. > > It gets even better. If all you need is one channel of output, then generate > your waveform on the L channel and generate 1PPS on the R channel. Use a > TBolt and TIC to continuously measure R vs. 1PPS and send those readings back > to the PC for averaging. As the time interval grows (indicating a frequency > offset) beyond acceptable levels, then make corresponding changes your > pcm_rate variable. This essentially becomes a software GPSDO. No h/w changes > are required to the board; you don't even have to open the case. The output > waveform will always be spot on-frequency, regardless of soundcard oscillator > frequency offset and drift. > > /tvb > > _______________________________________________ > time-nuts mailing list -- time-nuts@febo.com > To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts > and follow the instructions there. _______________________________________________ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
