davidwhess@gmail.com said: > I think a PIC might be fast enough to DDS it. The output bandpass filter > will cure a lot of sin. Using a dedicated switched capacitor filter would > be fun but more expensive. There are two parts to a DDS like setup. One is the math for the DDS and then spining for the right number of cycles. The other is a PLL to measure the speed of the clock driving the CPU and tweaking the DDS "constants" so it tracks the PPS. It might be fun to do that with a fixed number of cycles. Or maybe you can use a counter/timer to count cycles. You don't need any filtering. The goal is not to make a pretty picture on a spectrum analyzer. All you have to do is get the long term timing right so the clock doesn't drift. -- These are my opinions. I hate spam.

David, Hal, When I did the PIC divider I first tried the normal phase accumulator (DDS) approach. But I couldn't fit it in 38 instructions. So that's why I went with the binary leap year-like approach instead. The code, and very detailed comments are at: http://www.leapsecond.com/pic/src/pd30.asm Note this chip, like most of my PIC dividers, is drop-in compatible with the TAPR TADD-2 mini board: https://www.tapr.org/kits_t2-mini.html /tvb ----- Original Message ----- From: "Hal Murray" <hmurray@megapathdsl.net> To: "Discussion of precise time and frequency measurement" <time-nuts@febo.com> Cc: <hmurray@megapathdsl.net> Sent: Wednesday, October 19, 2016 3:29 AM Subject: Re: [time-nuts] 1PPS to 32.768 khz > > davidwhess@gmail.com said: >> I think a PIC might be fast enough to DDS it. The output bandpass filter >> will cure a lot of sin. Using a dedicated switched capacitor filter would >> be fun but more expensive. > > There are two parts to a DDS like setup. One is the math for the DDS and > then spining for the right number of cycles. The other is a PLL to measure > the speed of the clock driving the CPU and tweaking the DDS "constants" so it > tracks the PPS. It might be fun to do that with a fixed number of cycles. > Or maybe you can use a counter/timer to count cycles. > > You don't need any filtering. The goal is not to make a pretty picture on a > spectrum analyzer. All you have to do is get the long term timing right so > the clock doesn't drift. > > -- > These are my opinions. I hate spam. >