My 600MHz Prescaler (from the scrap box)

[Gyro]

Before I get all OCD about voltage references I thought it would be a good time to document my 600MHz divide by 10 prescaler. This is to boost the capabliity of my 80MHz Philips PM6622 Counter timer - it may not go very high but the beauty of its Panaplex plasma display more than makes up for it, compared to the higher frequency LED models.

As I indicated in the title, I built this one from bits I had in the scrap box, hence it uses an old Plessey SP8630 divide by 10 ECL 600MHz prescaler. Unlike more modern prescalers, it does have the advantage of having no particular low frequency cutoff (aside from slew rate) and its input can also be biased to give zero output when there is no input frequency (rather than a spurious one). In this circuit it happily goes down to 45MHz (sine wave) which is a more than ample overlap with the counter.

I wanted a decent input level range, but not having any PIN diodes available I went for a clamp approach, both on the input and after the gain stage to ensure that the max input spec of the SP8630 isn't exceeded. Not having any MMICs available either, I went for a single salvaged BFS55 transistor, this has an FT of 3.3GHz so should be up to the job. This combination achieves an input range of -24 to +23dBM (14mV to over 3V) at 180MHz (the highest source I have at the moment). I'm sure sensitivity will drop off a bit at higher frequncies. The upper limit is set by power handling of the 47R input resistor. The clamp diodes are HP 5082-2835 schottkys (1pF). Output is approx. 0.5V p-p.



Construction is combination of good old Manhattan style and air wired using a mixture of t/h and smd caps (hard to see in the photo). The main aim being to keep the input capacitance as low as possible. The input impedance is fairly indeterminate, I decided to have the input on the BNC male connector since it's most likely that I will use it close up to the target board using a PCB mounted BNC socket, so accurate termination of a coax isn't really an issue. Impedance won't drop below 50R anyway. One valuable lesson is that the center pins of BNCs move, which isn't good for smd caps!, so some strain relief was added.



Not shown in the photo are the 5V LDO, supply protection, caps and ferrite beads etc. mounted on the back of the board to give a 6-12V input supply range.

As indicated at the beginning, the design choices were based on what I had lying around, so some compromises, but it works well, with a decent input range and extends the frequency capabilities of the PM6622 quite a bit.



[Edit: Yes, the OCXO in the Adret has drifted off again  ]

[technix]

Is it possible to discipline the OCXO with something? GPS maybe?

You can get one of those second-hand Rubidium Standards as a replacement reference oscillator if the func gen takes 10MHz.

[Gyro]

Hi technix,

Yes it is possible, the Adret has a disciplining input on the back which will take either a DC level or a clock of several frequencies (pretty flexible). I guess that might be an excuse for its internal 5MHz OCXO being a quite drifty. I guess I'm spoiled by the PM6622 which came with their highest precision OCXO (PM9691) which never drifts.

I last calibrated both of them a few months ago with my ublox LEA-6T based GPSDO breadboard. Another project that I need to finish! 

www.eevblog.com/forum/projects/my-u-blox-lea-6t-based-gpsdo-(very-scruffy-initial-breadboard-stage)/

Chris

[technix]

Hi technix,

Yes it is possible, the Adret has a disciplining input on the back which will take either a DC level or a clock of several frequencies (pretty flexible). I guess that might be an excuse for its internal 5MHz OCXO being a quite drifty. I guess I'm spoiled by the PM6622 which came with their highest precision OCXO (PM9691) which never drifts.

I last calibrated both of them a few months ago with my ublox LEA-6T based GPSDO breadboard. Another project that I need to finish! 

www.eevblog.com/forum/projects/my-u-blox-lea-6t-based-gpsdo-(very-scruffy-initial-breadboard-stage)/

Chris

Or you can get a Rubidium standard that can be disciplined by GPS (your LEA-6T) and reference your entire lab off that? (reference clock or PPS for all frequency-sensitive equipment, and run a Raspberry Pi Stratum 1 NTP server off the GPS's NMEA and PPS off the Rubidium for all network-connected equipment?)

[Gyro]

Whoa there, you'll turn me into a Timenut! 

Yes, I did look at Rubidium oscillators but following several threads I got turned off by things like warm-up time and power consumption, power cycling etc. I really didn't want something that I needed to keep running for extended periods. With the Efratom OCXO disciplined by the ublox LEA-6T I can have the whole thing up and stable (enough) within 5 minutes (faster than the OCXOs in the PM6622 and Adret). My plan is to finish the project with several buffered outputs so that I can directly discipline the Adret and calibrate (more like check) the PM6622 occasionally.

BTW I noticed your post in another thread where you were talking about disciplining a 32kHz crystal. I just wanted to make sure you know that the ublox LEA range have two frequency programmable outputs - you're not restricted to 1pps, you could program one output to 32kHz and do a direct 1:1 pll if you wanted. The LEA-6T is relatively cheap to find on ebay etc. in China, fitted in some quadcopter gps modules (cheaper for you that is, they're getting more expensive to buy from here now!).

[technix]

Whoa there, you'll turn me into a Timenut! 

Yes, I did look at Rubidium oscillators but following several threads I got turned off by things like warm-up time and power consumption, power cycling etc. I really didn't want something that I needed to keep running for extended periods. With the Efratom OCXO disciplined by the ublox LEA-6T I can have the whole thing up and stable (enough) within 5 minutes (faster than the OCXOs in the PM6622 and Adret). My plan is to finish the project with several buffered outputs so that I can directly discipline the Adret and calibrate (more like check) the PM6622 occasionally.

BTW I noticed your post in another thread where you were talking about disciplining a 32kHz crystal. I just wanted to make sure you know that the ublox LEA range have two frequency programmable outputs - you're not restricted to 1pps, you could program one output to 32kHz and do a direct 1:1 pll if you wanted. The LEA-6T is relatively cheap to find on ebay etc. in China, fitted in some quadcopter gps modules (cheaper for you that is, they're getting more expensive to buy from here now!).

Currently I am running a straight GPS-disciplined Stratum 1 NTP on my Pi. I am not running traditional PLL as I am using DS3231 and using its built-in adjustment registers to discipline it digitally.

If possible, you should discipline all your frequency-sensitive equipment off the GPS. This will kill most the drifts.

[Gyro]

If possible, you should discipline all your frequency-sensitive equipment off the GPS. This will kill most the drifts.

Possibly true, but I have a feeling that the clock jitter of the PM6622 OCXO will always be lower than my GPSDO project and it is always nice to have a sanity check... The 'man with two clocks' may no longer know the right time, but at least he can tell if one of his clocks is broken! 

[technix]

If possible, you should discipline all your frequency-sensitive equipment off the GPS. This will kill most the drifts.

Possibly true, but I have a feeling that the clock jitter of the PM6622 OCXO will always be lower than my GPSDO project and it is always nice to have a sanity check... The 'man with two clocks' may no longer know the right time, but at least he can tell if one of his clocks is broken! 

Then just discipline the PM6622 with GPS, and then reference everything off the PM6622. This works the same way as a GPS-disciplined Rubidium standard (or my disciplining the DS3231 with GPS)

[Gyro]

Sadly it's not possible to discipline the OCXO in the PM6622, only select an external 10MHz source, so other things would then need to be manually adjusted. Discipling the Adret seems the most practical solution as that is the unit with the poor drift performance. The PM6622 is stable enough to just need checking every 6 months or so. Thanks for the suggestion though.

[technix]

Sadly it's not possible to discipline the OCXO in the PM6622, only select an external 10MHz source,

Rubidium?

so other things would then need to be manually adjusted. Discipling the Adret seems the most practical solution as that is the unit with the poor drift performance. The PM6622 is stable enough to just need checking every 6 months or so. Thanks for the suggestion though.

If you disciplined your Adret with GPS you may wound up getting a better reference than PM6622 out of it.

[Gyro]

Rubidium?

I refuse to become a Timenut!   I'm having a hard enough time resisting becoming a Voltnut. 

[technix]

Rubidium?

I refuse to become a Timenut!   I'm having a hard enough time resisting becoming a Voltnut. 

It costs way less to be a time nut and it is way less maintenance intensive. Used Rubidium standards are fairly cheap and you already have a timing-oriented GPS that can further discipline it. And besides those you need only a single RF distribution amp and a few BNC cables to reference your entire lab off GPS-disciplined Rubidium standard.

I don't know how energy intensive a Rubidium reference would be, but replacing several OCXO with a single atomic reference may increase timing accuracy and reduce lab energy consumption at the same time, as those OCXO can be kept off at all times.

[Gyro]

You're very persuasive! and you're right of course, it would be nice to have such a thing as an off-air primary voltage reference. Yes, the GPSDO will have plenty of buffered outputs and I'll probably extend its use over time.

Rubidium sources do seem to need fairly large heatsinks just to keep the internal electronics cool while the source is heated as far as I've seen, at least for the ebay ex-cell site ones. My ideal situation is still to go from all-off to reasonably fast startup where possible. The GPSDSO breadboard seems to achieve that and most of my stuff is well aged (old) and pretty stable during warm-up. I have two master switches in the lab, one for the general stuff and one for the equipment with OCXOs, voltage reference etc. Both are normally off overnight unless I'm monitoring something.

Chris (apprentice eco-nut and bill payer)

[technix]

You're very persuasive! and you're right of course, it would be nice to have such a thing as an off-air primary voltage reference. Yes, the GPSDO will have plenty of buffered outputs and I'll probably extend its use over time.

Rubidium sources do seem to need fairly large heatsinks just to keep the internal electronics cool while the source is heated as far as I've seen, at least for the ebay ex-cell site ones. My ideal situation is still to go from all-off to reasonably fast startup where possible. The GPSDSO breadboard seems to achieve that and most of my stuff is well aged (old) and pretty stable during warm-up. I have two master switches in the lab, one for the general stuff and one for the equipment with OCXOs, voltage reference etc. Both are normally off overnight unless I'm monitoring something.

Chris (apprentice eco-nut and bill payer)

How long does it take for the Rubidium to warm up?

[Gyro]

I think it varies depending on serveral variable (model, China salvage method, age etc), but seems to be anything up to an hour to stable lock but several more to stabilise to Timenut territory. Warm-up power looks to be in the 30-40W range and about half that running (there are several threads). I'm not sure how well they take to frequent power cycling either. Btw, Hydrogen Maser seems to be the current Timenut reference of choice (I know less than nothing about those!).

[technix]

I think it varies depending on serveral variable (model, China salvage method, age etc), but seems to be anything up to an hour to stable lock but several more to stabilise to Timenut territory. Warm-up power looks to be in the 30-40W range and about half that running (there are several threads). I'm not sure how well they take to frequent power cycling either. Btw, Hydrogen Maser seems to be the current Timenut reference of choice (I know less than nothing about those!).

If running those ovens for those OCXOs together take more power, you may be better off using an always-on Rubidium as an replacement. You can even disconnect those ovens and reference all your equipment off that single nutty Rubidium.

[Gyro]

Yes, from what I've read in other threads too, that pretty much sums up my reservations about any Rb source within my 'acceptable' price range (in comparison with GPSDO OCXO cost).

[Gyro]

Just a quick update....

Bypassing the 10R emitter resistor of the BFS55 with 1nF has improved input sensitivity by about 4dB at 180MHz (around -28dBm,  7.5mV) and levelled the frequency response vs lower frequencies. I can't think why I left it out in the first place.