My Time Projects

[metrologist]

I thought instead of spreading my stuff all over the forum in little bits, maybe I should try and collect it all in one thread?

This is my first "Two-Transistor Atomic Frequency Standard" a design I found from Charles Wenzel:
http://www.techlib.com/electronics/atomic.html

I used a 2N3906 and 2N5152. I added the trimmer after fiddling with a bunch of resistors because I was interested in seeing the locking steps, and sure enough the circuit seems to step through various divider ratios. I have a small pile of very old transistors from my grandfather that I'd like to try out. I also have a some old binary 4-bit counter chips - how to make a /60?

I was trying to understand how it works. It seems C1 will charge, Q1 will turn on Q2 to dump the charge. It almost makes sense at 60Hz, but not so much at higher division ratios.

I had it set to 1 Hz with series resistors to the pot and could determine when the divider went to 59, 60, and 61. At 1 Hz, I could monitor the frequency drift somewhere between 0.999 and 1.000, or take a 20mHz jump up or down to the next divide ratio. As Charles said, the circuit is less stable at higher divide ratios, so naturally 60Hz seems to work very well and my HP5314A was meandering between 59.9 and 60.1Hz. I also had it counting seconds with the circuit in 1pps mode and timed it for about 10 minutes and each minute came up on the dot. It would be interesting to see how stable the power grid really is.

[DaJMasta]

Takes 6 bits of counter and some and gates (4 inputs worth), but if you count to 111100 on those 6 bits, read them out into and gates, and pipe that into the counter reset pin, you should have your solution.  Will count 0 to 59 and will reset to 0 on 60.

That being said, there's a limit to how much accuracy you're going to see using normal parts.  The 60Hz from the mains is a great time reference, not as good as GPS, but quite good... but when you're adjusting your resistor to change the output frequency, you're relying on your own calibration point and the stability of the parts.  For better stability and repeatability, you want low tempco resistors and a high stability, low leakage cap like a polypropylene (as mentioned).  Then it's just up to how closely you can do the initial calibration of the RC time constant.

[metrologist]

Another installment, but this one will take a little longer since it did not lock up right out of the gate 

This is a Wenzel design GPSDO using a 1pps input and output a tune voltage to an external 10MHz OCXO.

For a second, I thought it was locked with a bit less than 100ns jitter, but looking at the 1MHz output from the HC190 IC it was running fast. One thing I do not understand is the Q1 output is a nice square, but input to the CD4016 is a lump. Is that due to the 100pF cap?

Here is Wenzel's design: http://www.techlib.com/electronics/GPSstandard.htm

[CharlesWenzel]

You have the lump because the 10k is working against your scope probe capacitance to low-pass the square wave. Both sides of the switch are hard to observe, especially the output side. I don't remember the specific reason I split the resistor to have one on each side and it could probably be a 22k on the output side of the switch. Perhaps the switch capacitance will round the edges the switch sees a bit the way I have it - it probably just seemed like a good idea at the time.

Is the duty-cycle of the sampling pulse 1 uS? The flip-flop side of that 10k should have a sliver less than 1 uS wide. If the oscillator is too far off, that sliver changes width too quickly for the PLL. You can also observe the "beat note" at the output of the first op-amp. I have the slew switch to help the unit lock the first time (before the mechanical tuning is properly adjusted). You probably need to be within a fraction of 1 PPM to start but if you tune the oscillator around you should find a point where the meter starts moving slowly enough to make it from one end to the other.

[CharlesWenzel]

The transistor circuit is a little fussy but the CD4069 version is so solid I use it all the time without thinking about it (albeit a bit silly). My "Deluxe Keyboard Wedgie" is in use right this second. I don't even mess with the antenna. I should take it outside to see how far away I can get from the house wiring.

[metrologist]

The sampling pulse is just slightly less than 1us, if I recall correctly. I think fundamentally I missed adding a trimmer to my oscillator; I just connected direct to its tuning pin. The tuning range is 0-10V, but it tunes very close to 2V. I think that the circuit is designed to be stable around 2.5V so it pulls the osc high. At least I do not see much change in voltage on the osc tuning output, but I do see the op-amp pin 1 swing ~1-4V. I could see the 1MHz sweep past the tuning point after manipulating the slew, so I was thinking of a trimmer on the tune output or on the non-inverting input of the op-amp.

I'll build the CD4069 version soon.

Thank you!

[CharlesWenzel]

The tuning sensitivity needs to be right for the PLL components (basically the 66 megohms). If it's all electrical tuning, I'd use a good 10-turn trimmer to do coarse tuning with a large resistor feeding the tuning line for the electrical tuning. The PLL is set up for 0.66 Hz/volt (at 10 MHz). Tell me your total tuning range and I'll come up with a pot/resistor combo. If you leave the tuning range high the loop might not be stable.

[metrologist]

Thanks, Charles.

I measure +1Hz at 4.2V and -20Hz at 0V on the osc output. That is referenced to the pps. I can use the slew to move the tuning voltage through that range, but otherwise it does not seem to move.

I've attached a few photos of other circuit pins as named. I may have a wiring issue that I just cannot see, so might break it down, or it is because I am at the top end of the tune range. This osc is offset from my other one that I've mounted on a PCB.

Yellow trace is pps
Turquoise is 1MHz
Purple is pin

Last image is animated

[CharlesWenzel]

Your numbers suggest you have 5 Hz per volt sensitivity and that's just about 10 times too high. It also looks like your oscillator is on frequency at a very low voltage, around 200 mV. Try this: Connect a pot in series with a resistor to give you about 0 to 1 volt (using a stable voltage). If you have 10 volts you might use a 1k pot and a 10k resistor to plus. Connect the wiper to the tuning input through a 10k resistor. Now also connect the PLL output to the tuning input through a 100k resistor. The 100k working against the 10k should give you a reasonable tuning sensitivity for the PLL. Slew the voltage near half-scale and tune the pot to put the oscillator on frequency. That should move the needed voltage to within swing range of the PLL. It's possible that the PLL just can't swing low enough (and it won't be stable with so much tuning sensitivity).

[CharlesWenzel]

Your images make me think it's working but the frequency is off. Is the oscillator a stable oven type? I bet the PLL can't reach the low voltage required to lock. I just looked up the LMC6582 and it can only swing down to about 400 mV and your numbers suggest you need 200 mV. Oh, here you go : as a simple test you could just add a voltage divider to the output of the op-amp, say a 10k in series with a 1k to ground. That will still give you plenty of voltage to tune the oscillator and also allow the op-amp to swing down to around 40 mV. Plus it will give you the right tuning sensitivity.

[CharlesWenzel]

Oops, I read your voltages incorrectly, thinking it was on frequency at the lower end. You would be on frequency around 4 volts (not 200 mV) and that's just right. But the last idea won't work. I'd try the pot idea but set up to be centered around 4 volts. Connect a 10k pot across the 10 volts (or whatever you have) and connect the wiper to the tuning through 100k . Then connect the PLL through a 1 megohm. (I upped the resistors arbitrarily to save current.) Then use the slew switch to slow the beat note. You should see a big swing on the output of the first op-amp, slowly swinging the full voltage range.

[metrologist]

Hi Charles,

This oscillator series: http://www.mti-milliren.com/hirel_220.html

I think I should expect +/- 12 Hz through the 0-10V range.

I'll be able to experiment tonight. Thanks for the help.

[Kalvin]

How sensitive this is to mobile phone's RF field?

[CharlesWenzel]

I've not noticed any sensitivity to my phone. When you have the mechanical tuning added see if you can slow that variable duty-cycle change rate to under once per second. The 100 pF capacitor and 10k resistor act like a 1 Hz low-pass so you need a pretty slow beat note. You should see a nice sawtooth waveform swinging from zero to 5 volts on the output of the top op-amp.

I looked at the spec. sheet and I see "mechanical" tuning but it's apparent that means electrical tuning since there's no hole. : ) That oscillator should do a great job.

[metrologist]

3 oscillators failed, I have only one left so might get an update tonight.

[metrologist]

I've not been able to get the circuit to lock so I tried something different until I can more clearly think about it. I may just rebuild the circuit on a perf board with a fresh set of components.

Meanwhile, I built the 4016 clock next to the previous two-transistor clock (http://www.techlib.com/electronics/atomic2.html). This new clock is much more stable, so stable that small changes if component value do not seem to have any affect. Larger changes, such as swapping values cause the circuit to be unstable. For some reason, I am getting 15 and 7.5 MHz on the second and third stage. 

[CharlesWenzel]

Did you try switching the flip-flop output in case your tuning slope is opposite mine? It will not lock if the slope is upside-down.

[metrologist]

The analog GPS tuning circuit appears to be working now  I was watching the OCXO roll back and forth on the scope compared to the reference. I was reading further down the article about adding the coarse tuning pot and the tuning sensitivity being high, and that was likely the case. I added a 100K resistor to the tuning input and the OCXO slowly rolls back and forth, if at all. This OCXO tunes at 2.644V.

I think the main problem I had was starting out with a failed oscillator, and then not adding the mechanical tuning and reducing sensitivity. Since 3 of the 4 oscillators I have are failed (they fail outside of tuning range), I have an opportunity to procure a proper oscillator for this build and put it in a box as demonstrated. A nice large oscillator like shown would look good on the mantle next to my crystal radio set 

This is after about half an hour with infinite persistence.

[CharlesWenzel]

If it's working properly, pin 1 of the PLL op-amp should settle in at 2.5 volts DC - no wandering.

[metrologist]

Pin 1/2 are mostly 2V on my scope. I notice when the OCXO starts to move, Pin 1/2 take discrete ~0.5V max steps (usually less) above/below 2V until the OCXO stops.

The OCXO has stayed in-phase for more than an hour now. That is different than if I just tune it with a trim-pot, where it would usually start drifting one way...

I'm sure something could be drastically improved, however.

Edit: I increased the 100k series tuning resistor to 220K, like mentioned in your article, and that has drastically improved the stability. I'm totally amazed! I tried a 1K and the stability went to pot...