It’s been a while since I’ve posted here, but I’ve had a bunch of irons in the fire. I’m working on adapting my GPSDO to the ECS ECOC-2522, which the manufacturer claims has a short term ADEV in the low -12s, but I haven’t gotten it doing that well yet. But one thing that is ready (well, electronically it is - I’m still working on the laser cut case for it) is my GPS Talking Clock. The story is that I called the USNO time number at midnight on New Year’s Day, but the wife noted that it was the wrong time zone. That got me thinking, and I wound up designing a GPS driven simulacrum. It’s an ATXmega32E5 with the usual Venus838 timing module and a µSD card slot. The card is loaded with audio samples that the 32E5 plays back through its DAC. I got double-buffered DMA to work to feed the DAC, so audio playback is a largely background task. I just have to fill the buffer with the next block from the file every ~30 ms or so. The ticks and beeps are generated from an on board 1 kHz source and are turned on by a PPS ISR, so they’re as accurate as possible. The whole thing is basically as accurate as an aural clock can be - the latency induced by the speed of sound has far more impact than anything else. While the audio is turned off, the clock can also do Westminster Quarters (or any other chime you wish to load in). The µSD card is FAT formatted and the audio sample files are easy to make with ‘sox’ (raw, 1 channel, 8 kHz, 16 bit little-endian, unsigned), so there’s no reason you can’t substitute my voice with your own, or make your own chimes. It’s available at https://www.tindie.com/products/nsayer/gps-talking-clock/

Hi > On Feb 17, 2018, at 11:01 PM, Nick Sayer via time-nuts <time-nuts@febo.com> wrote: > > It’s been a while since I’ve posted here, but I’ve had a bunch of irons in the fire. I’m working on adapting my GPSDO to the ECS ECOC-2522, which the manufacturer claims has a short term ADEV in the low -12s, but I haven’t gotten it doing that well yet. At least on this data sheet: https://www.ecsxtal.com/store/pdf/ECOC-2522.pdf <https://www.ecsxtal.com/store/pdf/ECOC-2522.pdf> They don’t say much of anything at all about ADEV. OCXO’s in the little packages are rarely super stars when it comes to ADEV. Bob > > But one thing that is ready (well, electronically it is - I’m still working on the laser cut case for it) is my GPS Talking Clock. > > The story is that I called the USNO time number at midnight on New Year’s Day, but the wife noted that it was the wrong time zone. That got me thinking, and I wound up designing a GPS driven simulacrum. > > It’s an ATXmega32E5 with the usual Venus838 timing module and a µSD card slot. The card is loaded with audio samples that the 32E5 plays back through its DAC. I got double-buffered DMA to work to feed the DAC, so audio playback is a largely background task. I just have to fill the buffer with the next block from the file every ~30 ms or so. The ticks and beeps are generated from an on board 1 kHz source and are turned on by a PPS ISR, so they’re as accurate as possible. The whole thing is basically as accurate as an aural clock can be - the latency induced by the speed of sound has far more impact than anything else. > > While the audio is turned off, the clock can also do Westminster Quarters (or any other chime you wish to load in). > > The µSD card is FAT formatted and the audio sample files are easy to make with ‘sox’ (raw, 1 channel, 8 kHz, 16 bit little-endian, unsigned), so there’s no reason you can’t substitute my voice with your own, or make your own chimes. > > It’s available at https://www.tindie.com/products/nsayer/gps-talking-clock/ > > _______________________________________________ > 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.

No they don’t. I wrote and asked them and they sent me back some sample data. They were a pretty pleasant surprise. > On Feb 18, 2018, at 6:04 AM, Bob kb8tq <kb8tq@n1k.org> wrote: > > Hi > > >> On Feb 17, 2018, at 11:01 PM, Nick Sayer via time-nuts <time-nuts@febo.com> wrote: >> >> It’s been a while since I’ve posted here, but I’ve had a bunch of irons in the fire. I’m working on adapting my GPSDO to the ECS ECOC-2522, which the manufacturer claims has a short term ADEV in the low -12s, but I haven’t gotten it doing that well yet. > > At least on this data sheet: > > https://www.ecsxtal.com/store/pdf/ECOC-2522.pdf > > They don’t say much of anything at all about ADEV. OCXO’s in the little packages are rarely super stars when it comes to ADEV. > > Bob > > > >> >> But one thing that is ready (well, electronically it is - I’m still working on the laser cut case for it) is my GPS Talking Clock. >> >> The story is that I called the USNO time number at midnight on New Year’s Day, but the wife noted that it was the wrong time zone. That got me thinking, and I wound up designing a GPS driven simulacrum. >> >> It’s an ATXmega32E5 with the usual Venus838 timing module and a µSD card slot. The card is loaded with audio samples that the 32E5 plays back through its DAC. I got double-buffered DMA to work to feed the DAC, so audio playback is a largely background task. I just have to fill the buffer with the next block from the file every ~30 ms or so. The ticks and beeps are generated from an on board 1 kHz source and are turned on by a PPS ISR, so they’re as accurate as possible. The whole thing is basically as accurate as an aural clock can be - the latency induced by the speed of sound has far more impact than anything else. >> >> While the audio is turned off, the clock can also do Westminster Quarters (or any other chime you wish to load in). >> >> The µSD card is FAT formatted and the audio sample files are easy to make with ‘sox’ (raw, 1 channel, 8 kHz, 16 bit little-endian, unsigned), so there’s no reason you can’t substitute my voice with your own, or make your own chimes. >> >> It’s available at https://www.tindie.com/products/nsayer/gps-talking-clock/ >> >> _______________________________________________ >> 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.

I’ve built two GPSDO units now with this OCXO. For the first one, I fed the DAC (AD5680) from the oscillator’s reference output. This resulted in a very poor (compared to expected) short term ADEV result (1-2 E-11). There was a lot of noise (something like 5 mV P-P) on the reference output. Also, rather oddly, the reference voltage was something like 2.8v. The datasheet doesn’t say what the expected voltage is, but if you were to use that as your DAC reference, you’d be throwing away the top 20% or so of the tuning range. For the second unit, I built it with the same circuit I use for the CW OH300, which lacks a reference output. There, instead, I use an NCP51460 precision regulator for a reference. There, I achieved a short term ADEV much closer to expectations. It’s just under 6E-12 at tau 1-5s. ECS claims it ought to be closer to 3, but it’s entirely possible that my reference (Thunderbolt) and/or counter (53220A) are contributing error, or that my design has some other noise contribution I haven’t yet found. Still, it’s at least in the ball park. And the oscillator hasn’t even yet been running for 24 hours, so it may get better with some time (I’m a little skeptical about wear-in helping low-tau ADEV though. Doesn’t that usually operate on longer term drift instead?). In both cases, the reference was bypassed with a 10 µF and 0.1 µF chip cap adjacent to the DAC. Both power supplies had similarly low levels of noise and ripple. The worst that you could say about either was the amount of blowback from the oscillator itself was far higher than the input noise. Of course, sample size here is 1 each, so it’s possible that the first one was just a dud. But the theory that the noise on the reference is FMing the output fits the observations. > On Feb 18, 2018, at 7:54 AM, Nick Sayer via time-nuts <time-nuts@febo.com> wrote: > > No they don’t. I wrote and asked them and they sent me back some sample data. They were a pretty pleasant surprise. > > <ECOC-2522-10.000-3-F-C Allan Deviation.pdf> > >> On Feb 18, 2018, at 6:04 AM, Bob kb8tq <kb8tq@n1k.org> wrote: >> >> Hi >> >> >>> On Feb 17, 2018, at 11:01 PM, Nick Sayer via time-nuts <time-nuts@febo.com> wrote: >>> >>> It’s been a while since I’ve posted here, but I’ve had a bunch of irons in the fire. I’m working on adapting my GPSDO to the ECS ECOC-2522, which the manufacturer claims has a short term ADEV in the low -12s, but I haven’t gotten it doing that well yet. >> >> At least on this data sheet: >> >> https://www.ecsxtal.com/store/pdf/ECOC-2522.pdf >> >> They don’t say much of anything at all about ADEV. OCXO’s in the little packages are rarely super stars when it comes to ADEV. >> >> Bob >> >> >> >>> >>> But one thing that is ready (well, electronically it is - I’m still working on the laser cut case for it) is my GPS Talking Clock. >>> >>> The story is that I called the USNO time number at midnight on New Year’s Day, but the wife noted that it was the wrong time zone. That got me thinking, and I wound up designing a GPS driven simulacrum. >>> >>> It’s an ATXmega32E5 with the usual Venus838 timing module and a µSD card slot. The card is loaded with audio samples that the 32E5 plays back through its DAC. I got double-buffered DMA to work to feed the DAC, so audio playback is a largely background task. I just have to fill the buffer with the next block from the file every ~30 ms or so. The ticks and beeps are generated from an on board 1 kHz source and are turned on by a PPS ISR, so they’re as accurate as possible. The whole thing is basically as accurate as an aural clock can be - the latency induced by the speed of sound has far more impact than anything else. >>> >>> While the audio is turned off, the clock can also do Westminster Quarters (or any other chime you wish to load in). >>> >>> The µSD card is FAT formatted and the audio sample files are easy to make with ‘sox’ (raw, 1 channel, 8 kHz, 16 bit little-endian, unsigned), so there’s no reason you can’t substitute my voice with your own, or make your own chimes. >>> >>> It’s available at https://www.tindie.com/products/nsayer/gps-talking-clock/ >>> >>> _______________________________________________ >>> 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.

		Attila Kinali

Hoi Nick, On Mon, 5 Mar 2018 09:36:01 -0800 Nick Sayer via time-nuts <time-nuts@febo.com> wrote: > I’ve built two GPSDO units now with this OCXO. For the first one, I fed the > DAC (AD5680) from the oscillator’s reference output. This resulted in a very > poor (compared to expected) short term ADEV result (1-2 E-11). There was a > lot of noise (something like 5 mV P-P) on the reference output. Also, rather > oddly, the reference voltage was something like 2.8v. The datasheet doesn’t > say what the expected voltage is, but if you were to use that as your DAC > reference, you’d be throwing away the top 20% or so of the tuning range. That sounds weird. Usually the reference voltage is the full scale of the EFC range. Als, it should be pretty quiet, as it's (again usually) derived from a internal Zener diode inside the oven. I could understand a few dozen µV of noise (in a BW of 0.1Hz to 10Hz), but it shouldn't be more than that. mV level of noise would mean that something is seriously wrong with the reference. > For the second unit, I built it with the same circuit I use for the CW > OH300, which lacks a reference output. There, instead, I use an NCP51460 > precision regulator for a reference. There, I achieved a short term ADEV > much closer to expectations. It’s just under 6E-12 at tau 1-5s. ECS claims > it ought to be closer to 3, but it’s entirely possible that my reference > (Thunderbolt) and/or counter (53220A) are contributing error, or that my > design has some other noise contribution I haven’t yet found. Still, it’s at > least in the ball park. And the oscillator hasn’t even yet been running for > 24 hours, so it may get better with some time (I’m a little skeptical about > wear-in helping low-tau ADEV though. Doesn’t that usually operate on longer > term drift instead?). For short tau measurements, It is advisable to set the Tbolt into holdover. If the drift of the Tblot is not negligible, then you can measure its drift and compensate for it manually after measurement. This will give you more accurate results for these short tau measurements. Also, keep in mind that precision OCXO take a couple of days to settle. I have a few measurements that a relative high frequency drift up to one to three weeks after power up, before the OCXO settles for something an order of magnitude lower. I don't know the OH300, but they might exhibit something similar, just at a different timescale. Look at the datasheet, it should state a number of how many hours/days the OCXO reaches the stability specs. Another aspect is that you should remove any apparent phase/frequency drift (and probably other deterministic effects) before calculating ADEV. Attila Kinali -- <JaberWorky> The bad part of Zurich is where the degenerates throw DARK chocolate at you.