OCXO locking circuit component swaps

[metrologist]

Hi All,

I wanted to build this GPSDO circuit and wondered if I could swap a few components. If they won't work, what other suitable common part might suffice that could see if I have and if I'd need to change some other component values?

http://www.techlib.com/electronics/GPSstandard.htm

74HC190 to SN7490A (the author says it will)

74HC74 to DM74S74N (I have an HC version in SMD package if needed)

CD4016 to MAX4066A or MM74HC4066M

LMC6582 to AD8042

I think I have the rest to give it a shot.

[edpalmer42]

The only substitution that might give you trouble is the LMC6582.  It's input current is spec'ed at 20 pa while the AD8042 is 1.2 ua.  That bias current will probably affect the integrator.  Notice the 66 Mohm resistor and the 10 uf nonpolar capacitor!

If the AD8042 is the best you've got in your junkbox, go ahead and try it.  See what the performance is and maybe then consider replacing it and see how the performance changes.

What measuring equipment do you have?  How are you going to measure the performance?  Do you have anything to compare it to?  Normally you'd use either another GPSDO, a good Rb standard that was known to be on frequency and low aging, or a Cesium.  The measurement would be to feed the two signals into a Time Interval Counter and capture the data in a computer using a program like Timelab.

Ed

[metrologist]

I found an LMC6772 and MAX412.

I ordered a trimble repackaged GPSDO and can test it at work too. Must have a couple to verify myself.

[metrologist]

How about TL072?

I don't think the LMC6772 will work, so was tending toward the MAX412.

[Cerebus]

How about TL072?

I don't think the LMC6772 will work, so was tending toward the MAX412.

I think any FET input OPA would be fine there. Slew rate requirement is about a volt a fortnight, input offset is non-critical, gain isn't critical - a few thousand would do it, about the only thing you want to ensure is low input bias current and lowish noise. A TL072 would be fine.

[edpalmer42]

The TL072 is spec'ed for a minimum power supply of +-5V.  You could try it.  The MAX412 is spec'ed for +-2.4V power, but neither is spec'ed for a single supply.  It's not hard to create a nice bipolar power supply, so maybe that would be a good idea.  Neither chip is rail-to-rail, but if you use a bipolar 15V power supply I don't think that will matter.

If you've got either chip, go ahead and try it.

I did a brief search for the original LMC6582.  Now I see why you're looking for alternatives.  Not the easiest chip to find, is it?  Unfortunately, this circuit pretty much requires an extremely low input bias current.  The 66 Mohm resistor only allows a maximum of 75 na @ 5V so the op amp bias current has to be much less than that.  Hence, the LMC6582 with its input bias current of 20 pa (max), 80 fa (typical).  The article states that the buffer amp needs to have an unusually high input impedance which is what drove the selection of the LMC6582.

Even if you find the LMC6582 or a good replacement, those current levels might require a bit of finesse to get the circuit working properly.  But I noticed that Wenzel didn't mention anything about this.  I don't understand why.  Maybe try normal circuit construction techniques first, but if the results don't seem good enough, you'll have to get drastic.  In that case, don't attach either op amp input directly to the circuit board.  The leakage through the board could be too high.  You'll either need guard traces, teflon standoffs, or just hang the components in mid-air.  If you do anything other than guard traces around the input pins, you'll have to bend the input pins so they're not touching the board or socket.

Although the circuit is 'simple', it's not exactly 'easy'.

Ed

[metrologist]

Thanks for the tips. This stuff is far beyond my understanding. I might have a very small PC supply with +/-12V and 5V.

[edpalmer42]

Speaking of weird components, do you have (or have a source for) the 10 uf nonpolar capacitor?  That's a very unusual component.

Ed

[metrologist]

I thought I measured a large red disc capacitor, about the size of two US quarters, at 5uF and thought of using two. Maybe the meter said nF 

[edpalmer42]

You should check them.  Look at the picture of his circuit.  The big red capacitor is the 10 uf nonpolar.  It's almost as big as a 9 volt battery.

If your GPS has a high frequency output, you should reconsider Miller's circuit.

http://www.jrmiller.demon.co.uk/projects/ministd/frqstd0.htm

It's simple and has been tested to give good performance.  Since it's so simple, you can tweak it to see the effect on its performance.  Depending on your oscillator, such tweaking may be able to improve it.

Ed

[metrologist]

That is one I was planning to build with my Jupiter GPS module.

K6JCA also started with an idea to use an Arduino clocked by the 10 MHz reference, but abandoned it. I think I would also like to try a uc approach.

http://k6jca.blogspot.com/2016/02/an-arduino-based-gps-disciplined.html

I have several good oscillators and I think one more good timing GPS (Trimble Condor).

I am also a little close to buying one of those NEC GPSDO modules from ebay.

[edpalmer42]

Okay, if you're interested in trying different ways to design a GPSDO, that's fine.  There are lots of ways to do it.

Wenzel went for a basically analog design with the expected complications of low currents, high impedances and somewhat hard to source components.

With a more digital design, you can do PLL or FLL  Both can be done well or poorly.  If you haven't checked out the Brooks Shera design, you should.  His is one of the best known designs.  I haven't studied his code, but I suspect that if there were any errors they've been found and fixed.

Just be careful not to fall into the trap that BG7TBL did and end up with a small frequency error.  While irrelevant to most users, it's annoying.  That's why I asked how you were going to measure the result.

Ed

[tautech]

Speaking of weird components, do you have (or have a source for) the 10 uf nonpolar capacitor?  That's a very unusual component.

Ed

10uF non polarized electrolytics are commonly used on speaker crossover circuits and shouldn't be too hard to find.

[edpalmer42]

Speaking of weird components, do you have (or have a source for) the 10 uf nonpolar capacitor?  That's a very unusual component.

Ed

10uF non polarized electrolytics are commonly used on speaker crossover circuits and shouldn't be too hard to find.

True, but for this application an electrolytic of any type won't work.  For use in an integrator, you need a capacitor with extremely low leakage and low dielectric absorption.  Electrolytics aren't good at either of those.  The best capacitors for this application are polypropylene film.

It looks like those capacitors aren't as hard to find as I thought.  Mouser has them in stock for $2.90 each.

http://www.mouser.com/search/ProductDetail.aspx?R=0virtualkey0virtualkeyMKP1848610704K2

They've also got some very nice op-amps.  No surprise there.

Warm up the credit card!

Ed

[CharlesWenzel]

Send me an email at my name (first @ last.com) and I'll help you find the parts. I recognized that those op-amps might be the among the last "good ones" (with actual legs) so I stocked up. Everything at work requires a microscope - I want to play with parts I can see. : )

[metrologist]

Wow, Charles, I'm very flattered you joined to forum to help. Thank you.

[CharlesWenzel]

Well, I'm the one who's flattered. My hobby page is just a lark and probably shouldn't be taken too seriously. The projects do work well but I tend to use parts from the 80's and 90's because I have them and I can actually see them. : ) And I tend to prefer an analog design, typically with a trick or two as opposed to using the "right" part or approach just for the fun of it (like using a flip-flop to divide by 5). Come to think of it I should have used that trick in this project! The 1 uS pulse generator is another example. It uses a "trick" with the timing capacitor that works great. That's the real fun for me.

The truth is that any new commercial design should be nearly 100% digital but that's too much like work, literally. You look inside our products now and it's a sea of digital ICs. I expect someone to ask, "so, how much memory does this 10MHz oscillator have." Sadly, the answer isn't zero. I wish I could post a picture of the insides of our latest oscillator. It looks like it probably runs Linux. : )

[CharlesWenzel]

The op-amp leakage isn't so critical that special precautions are necessary. I would think that 100's of pA would be fine. High-speed amps with uA currents will develop too much voltage across the 66 megohm. CMOS and JFET input op-amps are probably all fine as long as there aren't other issues (like insufficient supply voltage, etc).

[CharlesWenzel]

Oops, I forgot about the first op-amp acting as a buffer. 100 pA bias would cause a drift of 1 volt during the second. That's too much so I would want 10 pA or less. I still don't think special wiring is needed (like bending the leg up into the air) but a reasonably good CMOS op-amp is necessary.

[edpalmer42]

Charles:

Since you hadn't mentioned any special treatment for the input I suspected that it wasn't necessary.  But it never hurts to make a note of the possibility.

According to the board, you're a newbie.  If that's true, then the rest of us are at the level of single-cell organisms!   

Ed
Amateur Time-Nut