It has occurred to me that using 3/4 of a quad 256-step digital pot (like the AD5263, only 6$) set up as a Kelvin-Varley divider might be an alternative to a DAC. Use 2 sections both voltage driven but set 2 values apart, such that the "wiper" arms are just a fraction of the reference apart. That's a coarse adjust, effectively 8-bit. Use a 3rd section between the wiper arms of the first 2 for a fine adjust, now effectively 7 bits since the coarse are 2 steps apart. Generally monotonic since all the divider resistors are integrated on one chip, and with about 30 ppm resolution. Seems a 5 ppm/C reference would be a good match, the LTC6652 is also about $6. (Possibly even use the last digipot section with a filter amplifier to trim the result +/- 50 ppm or so.) With the AD5263 you get 5 ppm/C stability, so only 6C temp swing eats up one LSB. I'd think about using only 25 ppm/C or better resistors and putting the whole thing in an oven...
Then again, a small micro with two 10-bit PWM outputs is a lot cheaper, and they can be combined and filtered to effect a 16-bit converter with only a few parts. ________________________________ From: Hal Murray <[email protected]> To: Discussion of precise time and frequency measurement <[email protected]> Sent: Fri, December 30, 2011 5:35:56 PM Subject: Re: [time-nuts] Thunderbolt? (re simple gpsdo.) > Yes, the DAC+reference is challenging and one way to go may be the > coarse+fine approach to avoid large (18bit and beyond) DAC. My last GPSDO > has an 18bit DAC but now I'm thinking to try the 8bit digital pot + 16bit > DAC op-amp combined. The reference can't be overlooked anyway. Be careful, there is no free lunch. If you use a pair of DACs, the stability of the coarse DAC needs to be evaluated relative to the bottom bits of the fine DAC. _______________________________________________ time-nuts mailing list -- [email protected] To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
