What do you want from a GPSDO?

[rahandford]

Hello everyone,

I was looking on Ebay and Aliexpress for a GPSDO. Most of the units available (and information obtained on here) are all the same and many are based on references that are second hand and of an unknown age with  regard to ageing characteristics.

So I decided to try  to make my own based on the uBlox LEA-M8F time and Frequency Module.

What I would like to know is if I make this design open source and make it available as a product to others, what features do you want from a GPSDO for your bench?
So far I have 1PPS and 10MHz(x2) also a 30.72MHz outputs. Do people use the  RS232 outputs and if so what data do you extract? DO you want TTL or true RS232 levels?
Do you need an LCD and if so what information do you find most useful? Should I include a fractional PLL so you can generate a range of outputs?

All input would be most appreciated and the results will be shared here.


Thanks to all for your help.
Rupert

[rfspezi]

I would appreciate at least six (better 10) 10 MHz outputs to keep all my measurement equipment in sync.
A LCD display that shows the accuracy that can be expected at the moment would also be important to me.
Generation of different output frequencies is less important to me since i generate them with a better suited function generator anyway.
USB Interface would be nice.

[rahandford]

Thanks for  the feedback 

[timgiles]

I dont know how much the 6 or 10 outputs would add to the cost - but - prehaps consider the ability to add a daughter board that splits / filters.... the signal. Certainly 2 outputs (10Mhz) is useful to many, not sure 6 or 10 outputs is but I agree, an option would be excellent.

[rfspezi]

Do you already have an idea of the system design?
Which VCOCXO, DAC, uC...?
A block diagram to discuss would be great.

[ramon]

I would want a good 'holdover' (that module seems to have a vctcxo, so maybe it is good).
True RS232 better than TTL. But never USB.
Ability to reprogram/reconfigure/hack the Ublox module as much as possible.
I do not need LCD if we can check status via RS232. One 'power on' LED and another 'locked LED' (or TTL signal) is enough. But I don't mind if LCD option is included.
Fixed 10 MHz output + 1 PPS, and also configurable output (Because I need a good 2.048MHz reference output too, others might need different frequency).

The last one : a good case with universal power supply (100V/110V/220V), and good antenna kit.

[tkamiya]

No external power brick for power source.  I hate those things.  Built in power supply with built-in short term battery backup would be nice.  It would only need to keep it running for few hours.

Other than that, I'd like to have it simple.  2 PPS, 3 10Mhz, Power indicator, 1 PPS indicator, RS232C/USB, and big UNLOCK alarm LED, and a reset button.  I really don't need an LCD. 

I'm actually on a fence on RS232C or USB.  Most PCs today no longer come with RS232C.  It is also starting to lose compatibility.  DTE/DCE was bad enough.  Now 3.3V, 5V, and TRUE RS232C involving +/- 24V.  USB may be a reality, and since we don't use that for timing, I'm not sure if it's a bad thing.  If it was for NTP feeding pin9, I would be against it.

[David Hess]

I am not so hot these days on an internal off-line power supply if it is custom, undocumented, unrepairable, and unreliable which is often the case.  So an external power brick is fine as long as external 12 volt DC from any source is also supported.

At this point I think configuration via Ethernet is more practical than RS-232 but both would be useful.

Must have:

Power Indicator
Lock Indicator
Support for some type of redundant power or battery backup.

Should have:

Self charging of external battery if supported.
Some type of Ethernet connected real time monitoring.  This could be as simple as Telnet.

Would be nice:

NTP if an Ethernet port is included.
Built in distribution amplifier.

[Theboel]

1. ethernet
2. backup power conection
3. +12v DC input
4. Parameter can be adjusted with external correction (PPP etc)
5. Dual freq receiver

[tkamiya]

Ethernet to serial converters are already on market - except most are bulky and expensive.  Maybe something *like that* can be created and placed in where serial port would be?

[tkamiya]

I'm wrong.  Some quite small tcp/ip to serial modules are available.

Try doing some searches on Amazon.  They start around 20 dollars.

[beanflying]

Going down the path of battery backed GPSDO at present. Base unit I chose some time ago eBay auction: #253375324294 Multiple output options is a must IMO this many gives plenty for my usage. There is some shots of the internals here. https://www.eevblog.com/forum/testgear/bg7tbl-gpsdo-master-reference/msg1440787/#msg1440787 I can't see to many shortfalls with it as far as inputs and outputs go, sync's to Lady Heather, PPS, Square and Sine out.

8-10 hours GPSDO battery life from 10 D sized NiMh in the HP shell or circa 3 hours with the counter on as well using the GPSDO as the reference. Very worthwhile addition to the project or even a sub project. Reason for the NiMh over Lithium was ease of retrofitting the box and also less fussy for charging and discharge management.

[rfspezi]

I just thought about what happens if you don't have/want it powered on 24/7 and don't wanna use a battery.
To achieve lock and good stability as fast as possible after heating up of the VCOCXO, i guess the following would be useful.
* Every update of the DAC value needs to be stored in NVRAM
* Use frequency lock instead of phase lock (maybe optional in software?)

[rahandford]

Thanks for all the feedback so far - I will try to make all this into one device 

Datasheets for parts chosen so far:

LEA-M8F u-blox M8 time & frequency reference GNSS module: https://www.u-blox.com/sites/default/files/products/documents/LEA-M8F_DataSheet_%28UBX-14001772%29.pdf
DAC8571: http://www.ti.com/lit/ds/symlink/dac8571.pdf
Bliley TCVCXO: http://www.farnell.com/datasheets/2674668.pdf?_ga=2.168041506.2031284271.1552628280-648269207.1446653940&_gac=1.15460994.1552288765.EAIaIQobChMI6cXx7MX54AIVTZPtCh1MbgSdEAAYASAAEgL4pvD_BwE

Please let me know where I have gone wrong already! 

[beanflying]

I haven't looked at the M8F variant or specs but I have done some playing with the M8N. Depending on what frequencies you are going to try and generate other than a PPS the 8N suffers badly from Jitter. While my HP Counters measure 10MHz correctly against my GPSDO the Waveform is shocking at that point.

I just re did some testing here and it only works reasonably at integer divided frequencies of 24MHz (Max spec is 10Mhz but 12 is still reasonable) https://www.eevblog.com/forum/testgear/test-equipment-anonymous-(tea)-group-therapy-thread/msg2246022/#msg2246022

[dkozel]

I haven't seen the distribution of the 1PPS signal mentioned. Ideally there'd be several 1PPS and 10 MHz outputs all of which are length matched to keep the signals in alignment. The hardware for buffering and splitting the signals isn't that expensive, I'd say a minimum of four ports each makes for a useful system.

You'll find that square waves vs sine waves for the 10 MHz is a bit of a topic, I'd tend towards square waves by default, possibly with non-populated pads for a high order low pass filter to convert it to a sinewave.

For inspiration, the Octoclock has an open schematic of control and distribution. It uses a separate GPSDO, but maybe it will be of some use as a reference.
https://kb.ettus.com/OctoClock_CDA-2990

[texaspyro]

I think that you will find out that a M8F+TCXO will wind up having rather poor performance as a GPSDO.

[rahandford]

Why is that? What would you want to change? All advise much appreciated

[texaspyro]

Why is that? What would you want to change? All advise much appreciated

To start with, [eople that use GPSDOs want a stable, accurate output... something a TCXO is not known for.  Use a good OCXO.

[DimitriP]

The only features I need in my next GPSDO are:

A). Free
B)  4 months of time
C) room for one more bench to fill up with gear and projects
D) spare cash for mouser, digikey, and arrow orders

[rahandford]

Why is that? What would you want to change? All advise much appreciated

To start with, [eople that use GPSDOs want a stable, accurate output... something a TCXO is not known for.  Use a good OCXO.

Yes I appreciate that a OCXO is better but also (as far as I have found) much more expensive
Did you look at the specification for the Bliley TCVCXO? - its quoted at 2ppm.
I also believed that this is only really used in a holdover situation as during normal operation the pll will be locked to the GPS? 
Please feel free to educate me as I am no expert!!
Thanks for your feedback it's the only way to learn

[DimitriP]

Since you asked ....

I was looking on Ebay and Aliexpress for a GPSDO

That's the first wrong turn you took right out of the gate.

There is no such things as a "cheap" GPSDO. Perhaps "good enough for the price".
OXCOs are either expensive, small and consume very little power or are cheaper (by almost an order of magnitude) ,  old, bulky and a little on the power hungry (comparatively speaking ) side .

There are plenty (more than 2 ) published designs out there that can be built using 1PPS or 10KHz from a cheap GPS module ($10-$25), a surplus OCXO (like a 10811 for about $100) and a control board for about $20-$30  in parts.     

A search for GPSDO will return almost as many results as searching for kardashian

[texaspyro]

Yes I appreciate that a OCXO is better but also (as far as I have found) much more expensive
Did you look at the specification for the Bliley TCVCXO? - its quoted at 2ppm.
I also believed that this is only really used in a holdover situation as during normal operation the pll will be locked to the GPS? 

No correction mechanism is perfect or instantaneous.  You want to minimize the amount and number of corrections that needs to be done.   An OCXO might be in the 2 ppb range, a far cry from 2 ppm.

[texaspyro]

There is no such things as a "cheap" GPSDO. Perhaps "good enough for the price".
OXCOs are either expensive, small and consume very little power or are cheaper (by almost an order of magnitude) ,  old, bulky and a little on the power hungry (comparatively speaking ) side .

There are plenty (more than 2 ) published designs out there that can be built using 1PPS or 10KHz from a cheap GPS module ($10-$25), a surplus OCXO (like a 10811 for about $100) and a control board for about $20-$30  in parts.     

The Trueposition GPSDO was around $50  and the Samsung/Symmetricom/Trimble "UCCM" GPSDOs can be had for less than $100.  The Samsung, in particular,  is surprisingly good.

[awallin]

For inspiration, the Octoclock has an open schematic of control and distribution. It uses a separate GPSDO, but maybe it will be of some use as a reference.
https://kb.ettus.com/OctoClock_CDA-2990

This wiki page has a list of open designs for pulse/frequency distribution:
https://www.ohwr.org/project/pda-8ch-fda-8ch/wikis/Similar-Projects

My own and John Ackermanns phase-noise measurements are here:
https://www.febo.com/pages/amplifier_phase_noise/
http://www.anderswallin.net/2016/09/ettus-octoclock-distribution-amplifier/fda_2016-09-12_pn/

If you go with an OCXO and care about phase-noise and/or short-term ADEV then the distribution-circuit should be chosen so it doesn't degrade the phase-noise of the oscillator.