A Regenerative divider board for HP 10811- oscillator modules for 10 -> 5 MHz

[Uky]

Having read some comments on the time-nuts mail lists regarding the 10 -> 5 MHz divider board featuring 7474 as divider, I wanted to see if a regenerative divider could be implemented in order to improve the phase noise of the 5 MHz signal. A suitable design by Luciano Paramaithiotti was found on the internet and I decided to try it. Using slightly more modern types as the original MSA07 and MSA1105 was unobtainable I layed out a new circuit board designed as a drop-in replacement in HP105B, 5065A etc. The 10811- adapter in the picture was something I had to fabricate myself as I had to have the "Coarse" adjustment accessible from the frequency standard front panel. I did not have the original mechanism available. I also added a 10 MHz sniffer port on a PCB that I also designed for this adapter.

The IC in one of the corners was later replaced by an LM431 precision reference (for the +15V "Fine" adjustment)

The design workes like a charm and fulfills the results measured by Luciano.

[S57UUU]

Did you achieve better phase noise than with 7474?  (I guess 74HC74?)  Can you share the results here? Maybe a link to Luciano's design?

Sorry for the barrage of questions, but I'm also playing with similar stuff.

[buta]

Luciano's design from Google:
http://www.timeok.it/wp-content/uploads/2015/08/low-noise-regenerative-divider1.pdf

[Uky]

I have a ordered a few professionally made boards.
(Solder mask, thru-plated holes, silk screen)

Since I am retired, I will have to ask my former employer
if I could use the time & frequency measuring equipment.

If the design improves the phase noise, I will share the results
with the eevblog community.

 

[KE5FX]

Nice-looking solution!  However, consider whether you really want a cleaner 5 MHz signal, as opposed to direct access to the internal 10 MHz signal from the 10811.  I've always found it more useful to repurpose an unused jack for 10 MHz output. 



If I were going to spin a boad like this, I'd take the opportunity to add a well-isolated 10 MHz output jack that could be routed to the front panel if desired.  Something to think about for the next revision, maybe.

[jonpaul]

just get leo Bodnar GPS DSO set to 5 MHz, perfect clock, no divider....
Jon

[Gerhard_dk4xp]

I work in the opposite direction. I have lots of MTI-260 5 MHz oscillators
from Lucent GPS receivers; the plugins without the GPS receiver boards
did not sell good, so I got them for a friendly price.
These plugins were intended as hot standby units for redundancy.
Built by HP/Agilent for Lucent and still in the unopened box.

But I need 10 MHz everywhere. So I decided to lock them all to
another Lucent Z3811A == KS24361, with GPS receiver.

The locking is sloooow, then i'll combine 4 or 16 of them with
Wilkinson combiners which averages the phase noise away
with n=4 or 16.  That should make the difference between
quite good  and excellent. Every oscillator has a push-pull
frequency doubler whose noise will also average out.

I have a semi-finished Eurocard 220 * 100 mm for 4 oscillators;
I could then use 4 cards for 16 oscillators with healthy output power.
Or 2 * 2 cards for the 3-cornered hat.

The picture is the circuitry for 1 oscillator. 2FF phase discriminator
to the left, frequency doubler to the right.

Cheers,
Gerhard  DK4XP

[Uky]

I have implemented a 10 MHz "sniffer" SMB-connector on a redesigned 10811- adapter board that connects the XO and the divider board to the chassis.

Will use the 1PPS output connector on the right side of the 5065 after having built a suitable sniffer-amplifier/buffer module.

As have been pointed out: Just get a GPS-locked frequency standard.
Could not agree more. I have that. The 5065A is like a vintage car: A continously ongoing "project".

 

[mawyatt]

Did you achieve better phase noise than with 7474?  (I guess 74HC74?)  Can you share the results here? Maybe a link to Luciano's design?

Sorry for the barrage of questions, but I'm also playing with similar stuff.

Yes, same question here. Always thought the Regenerative and Passive Dividers were used at frequencies beyond where Flip-Flop type dividers operate, but not because of lower Phase Noise. Long ago we used what was called "Tuned Dividers" for extending the divider frequency operation with custom IC SiGe ECL/CML dividers. These used an inductive load to extend the divider frequency range, recall something on the order of 35%.

Would be interesting to see how the various CMOS flavor Flip-Flop Phase Noise compares with an ECL/CML FF based divider, a TTL divider, and how these compare to the Regenerative/Passive types. ECL/CML should produce the better close in PN than CMOS because the bipolar device has a much lower 1/f corner than MOS devices, and since ECL/CML doesn't saturate the active device this should be better than TTL types.

Best,

[Gerhard_dk4xp]

... but I'm not really convinced. At what center frequency, for example???

(Crawford, Artech House)


Gerhard

[mawyatt]

Interesting, thanks for posting!!

Recall our SiGe based ECL/CML dividers achieved baseline better than -166dBc/Hz. The ECL/CML was custom logic just for the dividers, had different size devices, bias and such, and not the usual Motorola MECL type.

Where we used ECL/CML was towards 100GHz, the available CMOS (BiCMOS SiGe Process) wasn't fast enough for the prescaler. After the prescaler when the frequency was low enough we utilized CMOS dividers as well for the rest of the custom synthesizer portion of the SOC, which utilized a Multi-MASH based Noise Shaped based synthesizer. This work was back in ~2004 with some of the first experimental RF/MW/MMW SOCs.

Best,

[ch_scr]

... but I'm not really convinced. At what center frequency, for example???

(Crawford, Artech House)


Gerhard

Please find the article mentioned attached.

[KE5FX]

Good info, thanks -- here's a single .PDF for ease of browsing/archiving.

Although it's a little debatable whether he should have relied on a 90-degree hybrid ahead of the dividers to establish quadrature at the PN analyzer inputs, as opposed to adding 90 degrees' worth of coax in one of the measurement arms.  90 degrees at 24 MHz corresponds to about 10 ns of time difference at the divider inputs.  This time difference will be preserved at the divider outputs since their output edges are generated only in response to input edges, but now it's only 22.5 degrees since the frequency has been divided by 4. 

This will cause the measured PN to be 20*cos(22.5) dB low, not a huge deal since it's only about 1 dB.  But it will also have the effect of impairing the analyzer's AM noise rejection, so he may have been measuring the DUT's PSRR to some extent.  It would be nice to reproduce the measurements with a modern digital PN test set, and include newer logic families as well.

I had a customer a couple of years ago who had some unusual measurement requirements at 6 GHz, so I built a 'limited edition' low-noise divider module for them.  Was really impressed at how well the HMC983LP5E and HMC394LP4 did in this application:



So natürlich the HMC983LP5E is now EOL.  Never change, Hittite.  Never change. 

[Gerhard_dk4xp]

Thanks for removing a lot of doubts! I have published an article in DUBUS last year:
frequency synthesizer with LMX2594 / 95. But I could not measure the phase noise
noise with the Timepod  in X-band, so I decided to build a stereo down converter.

1st LOs are lmx2594, LO driver is HMC451, 1st mixer is HMC220 for the 1st test;
maybe I'll find something better. 1st IF is 900 MHz. I found SAW filters that have
a passband from 900 to 928 MHz. (EPCOS, RF360). 2nd LO is 100MHz *3*3,
filtered with the same SAWs. second Mixer is 17 dBm MiniCircuits, 2nd IF=0 to 28 MHz.

The plane with the downconverter boards from China, HK, Chinchinatty has just landed
in Germany today. If you want some, just tell me. also, the synth. The 20 GHz LMX2595
chip will probably work, too.

The TI software and USB dongle can't tell my synthesizer from the eval board.
Unluckily, TI has horrible lead times.
I have 2 LMX2594 chips + the eval board, just enough for a test.

Cheers, Gerhard

[KE5FX]

Thanks!  Yes, it'd be great to have a couple of the downconverter PCBs if you have some to spare -- my address is in the .pdf link for the LNDIV manual above.  Can I pay you for the shipping via PayPal, at least? 

I can try them out with the LMX2820 boards that I used previously (I assume you saw my 53100A app note, which of course is FB for the TimePod as well.)

Availability on the LMX parts is very frustrating.  Clearly they are being snapped up by brokers the minute they reach the distribution channel, if not sooner.  I see that WinSource alone is sitting on 23000 pieces of the LMX2594, for which they are asking US $176/1. 

Good luck to those guys.  May the winds of capitalism always blow at their back.  Hard.

[Uky]

Received boards from Aisler and populated one of them.
Worked OK.

[bnz]

Had you some chance to measure the phase noise of the replacement board? I measured the original open loop in the 5065A and the OCXO outside on a carrier board. On the close in side the 5065A board is quite good but on the noise floor side there is room for improvement, especially if you take in acount that the external measurement was for 10MHz not 5.