Hi Labrat,
yesterday I found 2 hours and I reworked the OXCO on a strip board.
I put the TL431 direct on the can of the OCXO to keep at the same (stabilized) temperature and
covered the whole board with a lot of foam.
I use a low noise analog PSU with an LT1083 (dedicated for behind-SPSU regulation) that is powered by an wall Switching PSU 12V 3 A.
The LT1083 based PSU was on my bench for years
My OCXO seems to be an ISOTEMP (this 4 green glas stands on the bottom) but has a third pin layout that differs from the 2 of the data sheet.
But it works. After some days running and now in the foam coating I got 10mHz deviation against my GPSDO (1 ppb).
Not to bad!
On the top of the photo you see the display of my GPSDO with useful information.
The upper line shows the error of the 10 MHz against UTC, in this case 0.100 ppb.
Left from the display the foam coated board with the OCXO and the VCO control which is used as external reference for the HP 53131A.
I will use it as external reference for the HP counter and my SDG2042 untill the Trimble OXCOs 34310-T will have found their way
with the slow boat from China. I'm also awaiting a Aluminium enclosure to make it perfect.
By the way, has anyone a data sheet of the Trimble 34310-T?
Hi Volkmar,
That LT1083 based psu with its claimed ability to reduce switching noise as a post smpsu LDO regulator intrigued me enough to track down its datasheet to see whether it included ripple rejection figures for frequencies higher than the bare typical "at 120Hz" caveat given in most LDO datasheets since I have a rather jaundiced view of this technique to 'magically cure' the switching noise issue with smpsus and dc-dc converters.
I saw an actual graph going up to 100KHz with an impressive (for an LDO) rejection figure of 30dB where your standard LDO would start behaving more like a high pass filter, aggravating the problem rather than mitigating it. So, for once, this looks like it meets its claimed switching noise rejection performance and not be a waste of space as is so often the case with so many other LDOs that imply such a benefit when simply tacked directly onto the output of a DC-DC converter.
The standard LDO, lacking any such HF ripple rejection can still be used with an smpsu but in this case, its job is simply to restore voltage regulation lost by the use of an RC LPF to suppress the noise before it reaches the LDO which is all fine and dandy when you can afford to take the hit on efficiency by the 2 or 3 volts lost in such a filter'. I've download the datasheet for future reference so my thanks to you for leading me to an LDO that might actually reduce the noise and HF ripple of an smpsu.
Regarding that OCXO, I think you should be able to get it calibrated to better than 0.1ppb with a bit of care and perseverance. If you're using a trimpot to electrically tune it, the major problem when trying to get sub ppb accuracy lies with the quality of the trimpot and the need to really pad it out to limit the trimming range to make the adjustment less ticklish.
When I was trying to trim the 10MHz CQE OCXO I'd used in my FY6600 to get below the 10ppt mark for a few hours at a time, I landed up utilising the 2G tip over effect to achieve the very fine tuning required, without having to touch that damned trimpot which so often would make the calibration worse rather than better, by altering the tilt angle of the front panel using suitably sized items as shims (a coin, an india rubber pencil eraser or two, the odd expired credit card and a DIP pin straightening tool just to name four of the items used).
All of that messing around with makeshift shims to fine tune the FY6600's OCXO has become a thing of the past since I finally acquired a Rubidium Frequency Standard (RFS) and it's now used to represent the GPSDO's 10MHz reference frequency as a rather distinctive Sinc pulse via its external reference connection (the OCXO is now frequency locked to the GPSDO).
I can't help you with any information on that trimble I'm afraid. If it weren't for the fact that I can't track down a suitably sized extruded aluminium project box to house my RFS, I too would be awaitng delivery on a nice aluminium project case. It seems I'm destined to fabricate a "Special" to solve my RFS's lack of proper housing wherein I can create a thermostatically controlled environment to minimise frequency variations with changes of room ambient temperature.
As you may have surmised from my previous posting, I've been a little preoccupied with my latest "toy" (an Efratom LPRO-101) this past week along with other experimental activities involving my gpsdos (the MK I has been taken apart for spare parts) hence my absence from this topic thread these past few weeks.
I had hoped I could keep the MK I going for a few weeks longer but I'd had to disconnect so much of the wiring from the "Five Volt" 13MHz OCXO in order to run higher voltage (possibly destructive) tests, disturbing other connections to the point where it just didn't seem worth the effort of putting it all back together again anymore.
As it happened, that "Five Volt" ocxo proved to be a 12v part after all - it was rather handy that it happened to be able to pass through its activity dip transition temperature in rather short order from cold and start producing a stable 4V p-p square wave from a 5V supply with the only side effect of this low voltage operation being a protracted 7 to 8 minute warm up time.
When I'd bought this very first, one and only, OCXO module, I didn't have a variable voltage bench supply to test it out on progressively higher voltages and, since it was able to still output a stable 13MHz even on a supply as low as 4.8v, there seemed a real possibility that it was a 5 volt part that might well let out its 'magic smoke' if subjected to a 12V supply so I decided to treat it as a 5 volt part until I at least had some more 'spare' OCXOs to play with (and a variable bench supply to make any such potentially 'test to destruction' experiments a whole lot more informative). Anyway, I've now relabelled it as a 12V unit and added it to my parts bin (more as a keepsake than something I'm likely to find another use for).
As well as an SMA socket for monitoring both voltage and noise and ripple on the MK I's 5.17v Vcc rail (1uF ceramic cap with 1KR shunt in series with the connection to the Vcc rail), I'd added a couple of ugly banana sockets to connect a DMM to the ground and the buffered copy of the EFC voltage to monitor the output voltage from the PLL.
Initially, I could only see voltage changes in increments of ten mV with my existing DMM collection so I blew a whole 13 quid on a nice Mestek DM91A "9999 counts" DMM to improve the resolution by an order of magnitude to increments of 1mV. Even this didn't suffice to let me see voltage changes before I could predict them from the phase shifts shown by the scope traces so I used a 3.16v lithium coin cell in series opposing to provide a stable dc offset that would allow the DMM to autorange to the 1000mV scale, reading the difference between the 3.298 EFC voltage and this dc offset, allowing me to see changes in increments of 100uV.
Whilst a new unused CR2032 creates a remarkably stable offset voltage (better than you can get out of a TL431 imho), this was just a temporary bodge to prove the concept, hence my using a TL431 to provide a +3VDC offset to the black (formerly the ground) banana socket to allow my crappy (on account it suffers from a permanent, silent, 10 minute shut off time out battery life extender annoyance!) UNI-T 58A DMM to display the 290 odd millivolt difference in 1mV increments. The only reason for my resorting to such a low grade "Voltage Reference" was on account of the lack of a stable Vref from my under-volted 13MHz OCXO (I can now see a stable 5.1ish voltage once its supply rail tops the 11 volt mark).
For my MK II GPSDO, I use the 5.127v on the 12v 10MHz OCXO's Vref pin to generate dc offset voltages in 1v increments via a string of matched 1KR resistors fed via a 120ohm padding resistor with a 50K trimpot in series with a 20KR across the 5K's worth of resistors to trim the volt drops to within half a millivolt of the 1v drop across each of the lowermost 3 resistors (if I ever need to use a 4v dc offset, I'll have to apply a small adjustment to the shunt resistance of the trimpot otherwise - the 1, 2 and 3 volt taps are matched closely enough to avoid the need for such a recalibration).
The EFC, the (jumper selected) DC offset and the ground are wired to a 3.5mm stereo jack tip, ring, sleeve contacts respectively to provide access to these test points for an external voltmeter. Not only does this provide a neater connection point than the banana sockets I'd used on the MK I, it also allows me to measure between ground and the EFC to give a direct reading and also to measure the DC offset with respect to ground by way of sanity checks of the EFC minus DC offset voltage high resolution readings I normally monitor.
The Vref pin on an OCXO is the most stable voltage reference you're likely to be able to access outside of a professional calibration lab so it's the obvious no-brainer choice voltage reference available to the hobbyist. Normally, you wouldn't use it to supply tens of milliamps directly to other voltage sensitive parts of a DIY GPSDO. Indeed, the manufacturers' advice in this regard is typically not to load this pin beyond a maximum of 1mA.
In my case, this fixed 5.12K 1mA load reduced the open circuit voltage by 1mV from 5.128v to the 5.127v I now see across my tapped potentiometer circuit. This is of no consequence since this represents a tiny fixed additional loading on the oscillator's internal supply rail.
As far as the 1.216KR worst case DC offset's impedance effect on the EFC voltage measurement accuracy is concerned, at a worst case loading of just 10nA, the voltage 'error' is simply lost in the noise. In my case, absolute accuracy plays second fiddle to my primary requirement to monitor the delta V of the EFC voltage (in such a basic James Miller based GPSDO design, minor dc offsets and temperature effects are simply lost in the wash
).
Harking back to the original question of how best to divide the frequency output of a 65MHz OCXO down to the required 10MHz GPSDO reference output frequency (sharing similar issues to my own 13MHz to 10MHz efforts), it seems to me, as I originally suggested, that the best way to proceed would be to divide the 65MHz down to 32.5MHz and proceed from there (you can divide by 13 down to exactly 2.5MHz which is comfortably above the 2MHz minimum clock input for a 3N502 clock multiplier programmed to multiply by a factor of 4 to generate the required 10MHz).
One of the problems I discovered with all of the additional complexity (no doubt exacerbated by my using stripboard construction) was the storm of switching noise imposed by the logic gates on the supply rail polluting the 10MHz output which necessitated the addition of a series resonant 10MHz crystal between the LPF output and the BNC socket to filter the worst of the resulting jitter noise from the MK I's 10MHz sine wave output. The modest additional 1dB loss was neither here nor there (I already had a healthy +12.5dBm coming out of the LPF to start with).
The MK II is essentially a 're-spin' of the MK I but using an M8T in place of the M8N GPS receiver module with a 10MHz OCXO to avoid the divide by 13 jiggery pokery (four ICs and their additional 400mW energy requirement) making use of a single sided copper clad board to eliminate the self inflicted problems of logic gate induced noise on the supply rails and improve the effectiveness of the LPF suppression of switching noise on the outputs of the DC-DC buck converter (5.31v rail) and the 5 to 12 volt boost converter feeding the OCXO's 12v pin.
The end result of all these changes to the earlier design being a reduction of energy consumption once warmed up, down to just 1.4W from the MK I's already modest enough 1.8W at 12v in spite of the small (3 to 4% ) extra losses of an additional DC-DC boost converter to power the 12v OCXO.
I ran an energy consumption versus DC input voltage test over the 7 to 24Vdc range (none of this restriction of a 11.5v minimum to 15v maximum supply voltage nonsense for me, thank you very much!
) and, in summary, it ranged from a minimum of 1.318W at 7v to a maximum of 1.488W at the 24volt limit of the buck converter module (an increase of just 170mW for a 17 volt increase implying a fixed vampire loading of 10mA on the buck converter's input).
Just out of idle curiosity, I calculated a 60 hour autonomy from a new, freshly charged 7AH SLA back up battery. Using my selected 12 and 9 volt wallwarts, the mains consumption is just a fraction over 2 watts making its energy demand on a UPS protecting a desktop computer and a file server almost invisible by comparison. Of course, in a parody of Asda's advertising slogan, "Every little helps", in this case, it's more a case of "Every little hinders.".
As I mentioned, I've spent this past week checking out my newly acquired pre-owned Efratom LPRO-101 (warranty expired way back in December 2001 - normally a 2 year warranty, implying that it's now some 21 years old). The lamp voltage now reads 4.963v. It had been 4.975v when I initially tested it on a 24v supply a week ago. I've been running it from a 19 volt laptop charging brick (19.65v actual) which may account for the small drop in voltage (elevating the temperature with towelling to test its susceptibility to temperature changes had increased the lamp voltage to 4.980v and using a Poundland USB powered fan to cool it had reduced the voltage).
Whilst these rubidium frequency standards (RFS) have some susceptibility to temperature variation, the real problem is in their susceptibility to barometric changes which are harder to compensate for without the use of a low retrace barometric pressure sensor and a compensation circuit calibrated to the EFC tuning slope.
DIYing an effective thermostatic control of the base plate heat spreader temperature seems a somewhat easier modification than a barometric frequency compensation controller for the hobbyist to concoct. Of course, that impression may be due to the seeming absence of any on-line DIY barometric compensation projects simply because I've never sought out any such barometric sensing projects - it might be a good idea to do some research on this.
Apparently, I'm not alone in my ignorance of barometric compensation measures to stabilise RFS modules. I came across a very recent research paper published only two years ago on this subject of quantifying and analysing the factors involved in the sensitivity of rubidium vapour gas cell RFSs to barometric changes for seemingly the very first time.
My main problem at the moment lies with the apparent complete lack of suitably sized extruded aluminium enclosures on Ebay or elsewhere to house my new RFS module into. I spent a couple days searching till my eyes bled before giving up this Holy Grail like search for rocking horse droppings (or decent 3 rail smpsus with a 1A rated 5v rail and half amp rated +/- 12 or 15v rails - another surprising rarity also on a par with a Unicorn droppings hunt).
There seems to be a huge gap between the maximum size (which still aren't quite big enough) of reasonably priced extruded aluminium enclosures and the much rarer more than large enough but overpriced enclosures being offered for sale. So much so, I think I'd rather build a customised case of my own either from my tiny stock of redundant aluminium enclosures or even from a local dealer if a tour of the local scrap merchants fails to produce what I need rather than pay through the nose for a not quite the size and shape I'd like ready made enclosure.
I've tried looking for used non working kit just for their cases as an alternative way of acquiring a more suitable enclosure at a reasonable price but even here, what I've looked at has still been grossly overpriced. However, I suspect that given the right sort of search term, I might find a cheap source of suitable ready made cases in the form of "For repair or spares" sales. Suggestions anyone?
JBG