Arhi is designing a 10 MHz Rubidium frequency standard output driver:
Frequency standards have stable frequency output but they usually have 50 R impedance sine output with some low Vpp values (100 mV or 400 mV or in best cases 1 V). Most devices that require frequency standard (frequency counters, frequency generators etc etc) have input that can cope with this low power low level analog signal but if you want to drive few devices from same frequency standard things tend to get weird as often the output from a standard can’t handle it…
Via the forum.
Just to be perfectly clear and put attribution where one is due, this schematic I took from http://www.ka7oei.com/10meg_rubidium1.html it’s not my design.
Using that switching regulator in this application is a bad idea. Why have such a nice reference then pollute it with power supply noise? Read the TI application note, SLWA066, Nov. 2011, “Supply Noise Effect on Oscillator Phase Noise”.
I am not using switching supply. I’m using only the LM7171 part of the schematic from KA7OEI, the PSU is
One rail: 20VAC -> GRETZ -> 10000uF 36V LOW ESR -> 7818 -> 10000uF 25V LOW ESR -> 7815 -> 10000uF 25V LOW ESR.
Second rail:
12VAC -> GRETZ -> 10000uF 25V LOW ESR -> 7809 -> 10000uF 16V -> 7805 -> 4700uF -> 7803 -> 1000uF
The rubidium is powered from 18V until the lock is achieved and then is switched to 15V. The noise is .. there is no noise :D
I would like to see a schematic of what your are planning. Is your 10MHz Rb osc output sine or square? The LM7171 lives with dual-rail supply. Consider one LM7171 per output. The LM7171 may become unstable with capacitive loads e.g., coaxial interconnections with some mismatch. My experience is to put baluns between the reference source and instrument ref input connections (solves ground loop problems too).
If memory serves, if your reference is digital (or even sine wave) there are some nice single-ended and/or differential clock distribution level/convert parts (LVDS?) out there that may out-perform what the LM7171 can do. One thing is certain, with the high slew rates necessary to reduce jitter, your board layout and especially power supply storage caps (ideally non-polar with very low ESR) on the rails will require much consideration. Use batteries (but watch out for ground loops).
10MHz reference distribution designs have been discussed widely here:
The Time-Nuts mail-list
http://www.leapsecond.com/time-nuts.htm
And here:
http://tapr.org/
One more thing… I’ve run into 10MHz reference distribution problems more than once. Mostly cabling issues, ground loops, and/or mismatches between the reference output and instrument ref inputs. You can try a quick fix by taking some 10MHz crystals and bridging the connections with the crystals in either T (unbalanced) or Pi (balanced) topologies. YMMV.
I would like to see a schematic of what your are planning.
Only bits and pieces are done. PSU is done, it’s home etched board that I drawn with a marker on pcb and then etched. Huge ground and 2 rails.
12VADC -> GRETZ -> 10000uF -> 7809 -> 10000uF -> 7805 -> 10000uF -> 7803 -> 2000uF
20VADC -> GRETZ -> 10000uF -> 7818 -> 10000uF -> 7815 -> 10000uF
Parallel to 10mF there is bunch of 100nF (on both rails) and 1uF ceramic (on 9/5/3V rail)
18V and 15V go to pin1 trough relay that switches power to rubidium, initially it’s 18V and after lock it’s moved to 15V .. after relay there’s a bank of 4.7mF and few 100nF ceramic
9V should power LM7171, 5V is going to pin 4 of the rubidium to enable output, 3V is powering input 74hc14 and cpld, and the output 74hc14 is powered via switch switching it between 3V and 5V hence I get 5V or 3V TTL output from cpld.
The output from rubidium is sine, it goes into lm7171 just like on the schematic on the post (50R to gnd, 100n trough, 1k, inverting input) , the output from lm is just as on the same schematic, only that one of the outputs goes into 74hc14 and then into cpld … so I have few sine and few square outputs. Sine outputs directly from lm and square from 74hc14 after cpld.
I will draw schematic when I finish and test it.
I’ve run into 10MHz reference distribution problems more than once
I plaid with 10MHz signal for a looooong time and 10MHz works even with a simple phone cable, with huge mismatch between cables, t-connectors and terminators … all mixtures of 50, 75 and 1M with and without termination work. I assume it’s because all devices I was feeding in 10MHz signal had some good input signal conditioning that really didn’t care about all that. Also 10MHz is just too slow for any serious reflections to influence the signal quality .. Few days I was running 5 devices with 10MHz signal without any “driver”, with 2 1mm twisted pair wires, without termination, and it all worked perfectly :D so I’m not scared about pcb layout, 7171 issues with output etc etc .. Initially I was thinking about double sided board with huge ground plane, transistor driving the output fed trough series xtal .. something like http://www.ka7oei.com/FE-5680_diag_1b.gif but I gave up as I found that to be overkill for what I need :)