OCXO Stable Reference and Control Voltages?

[FlyingHacker]

Were there other people, besides the two people on time-nuts (where only some units were defective) who had defective units?

I mean is there evidence to show that this is widespread?

[Vgkid]

Over 27 hours
Min: 7.03573,max: 7.03588, 150uVP-P, mean:7.03581
Variance :   .449580E^-9.
Most likely more stable then some 5.5 digit multimeters.

[FlyingHacker]

Over 27 hours
Min: 7.03573,max: 7.03588, 150uVP-P, mean:7.03581
Variance :   .449580E^-9.
Most likely more stable then some 5.5 digit multimeters.

Thanks. I saw similar performance on the CTS, and I was lagging the voltage after my cheap trim pot divider.

So it seems like these reference voltages are designed for stability, which would make a lot of sense given their intended use. 

[uncle_bob]

Over 27 hours
Min: 7.03573,max: 7.03588, 150uVP-P, mean:7.03581
Variance :   .449580E^-9.
Most likely more stable then some 5.5 digit multimeters.

Thanks. I saw similar performance on the CTS, and I was lagging the voltage after my cheap trim pot divider.

So it seems like these reference voltages are designed for stability, which would make a lot of sense given their intended use. 

Hi

The reference is just a 4.1V zener hooked to the +5 rail through a resistor.

Bob

[uncle_bob]

Were there other people, besides the two people on time-nuts (where only some units were defective) who had defective units?

I mean is there evidence to show that this is widespread?

Hi

When you have multiple "tray full" buyers each find that all the ones in the tray have the same issue .... that's cause for concern. These all appear to have come from the same "sale" (same part number / same pictures / same time on eBay) somewhere in China.

Bob

[FlyingHacker]

Over 27 hours
Min: 7.03573,max: 7.03588, 150uVP-P, mean:7.03581
Variance :   .449580E^-9.
Most likely more stable then some 5.5 digit multimeters.

Thanks. I saw similar performance on the CTS, and I was lagging the voltage after my cheap trim pot divider.

So it seems like these reference voltages are designed for stability, which would make a lot of sense given their intended use. 

Hi

The reference is just a 4.1V zener hooked to the +5 rail through a resistor.

It seems to be a pretty decent zener and resistors ;-) Also since they remain at the same temp as the rest of the OCXO I assume that helps as well. I suspect the stable temperature is the most useful aspect.

[FlyingHacker]

Were there other people, besides the two people on time-nuts (where only some units were defective) who had defective units?

I mean is there evidence to show that this is widespread?

When you have multiple "tray full" buyers each find that all the ones in the tray have the same issue .... that's cause for concern. These all appear to have come from the same "sale" (same part number / same pictures / same time on eBay) somewhere in China.

Ah. Thanks. I did not see that there were trayfulls defective. All I saw in a web search was a couple people with defective units on time-nuts. I either didn't find the right place in that thread, or this is someplace else.

The two I got are from Hong Kong, BTW. Of course they could have walked across the border. Need to find a friend in St. Louis with a spectrum analyzer.

[uncle_bob]

Were there other people, besides the two people on time-nuts (where only some units were defective) who had defective units?

I mean is there evidence to show that this is widespread?

When you have multiple "tray full" buyers each find that all the ones in the tray have the same issue .... that's cause for concern. These all appear to have come from the same "sale" (same part number / same pictures / same time on eBay) somewhere in China.

Ah. Thanks. I did not see that there were trayfulls defective. All I saw in a web search was a couple people with defective units on time-nuts. I either didn't find the right place in that thread, or this is someplace else.

The two I got are from Hong Kong, BTW. Of course they could have walked across the border. Need to find a friend in St. Louis with a spectrum analyzer.

Hi

If you go the spectrum analyzer route, you need one that will go down -120 dbc. That's a really good analyzer. A phase noise test set is a much easier way to look for this stuff.

More than you ever wanted to know about what is inside:

http://syncchannel.blogspot.com/2016/03/schematic-of-cts-1960017-10mhz-ocxo.html

Bob

[John Heath]

Hi,

What do you folks think is best practice for the reference and control voltages for fine tuning an OCXO? Is a passive divider network off VCC sufficient, or should I be looking at a reference IC, and then a passive divider off the reference? Or something else entirely?

The oscillator itself is quite stable, but if the voltages drift (relative to one another) then any tuning is not so useful.

I'd like to maximize long term accuracy, and have to recal only a couple times a year.

The OCXO is this guy:

http://www.ctscorp.com/wp-content/uploads/2015/12/OCXO196.pdf

Thanks in advance.

Hi FlyinHacker

From one hacker to another stay away from carbon resisters. I knew a guy who used a 1K carbon at 5 volts to generate a 400 Hz tone with nothing but band pass filters for a lunch time alarm in factory.  This had my antenna up so I asked him why he would use such a esoteric approach to generate a 400 hertz tone. He said just to make a point as there is so little to do. With this in mind I would go with a constant current source and zener diode with a op-amp voltage follower if precision is a concern. Failing this a 12 volt source then 7808 followed by a  7805 is about a stable as it is going to get assuming temperature is constant. Did not want to probe but what are you on to with long term precise time measurements?

[uncle_bob]

Hi,

What do you folks think is best practice for the reference and control voltages for fine tuning an OCXO? Is a passive divider network off VCC sufficient, or should I be looking at a reference IC, and then a passive divider off the reference? Or something else entirely?

The oscillator itself is quite stable, but if the voltages drift (relative to one another) then any tuning is not so useful.

I'd like to maximize long term accuracy, and have to recal only a couple times a year.

The OCXO is this guy:

http://www.ctscorp.com/wp-content/uploads/2015/12/OCXO196.pdf

Thanks in advance.

Hi FlyinHacker

From one hacker to another stay away from carbon resisters. I knew a guy who used a 1K carbon at 5 volts to generate a 400 Hz tone with nothing but band pass filters for a lunch time alarm in factory.  This had my antenna up so I asked him why he would use such a esoteric approach to generate a 400 hertz tone. He said just to make a point as there is so little to do. With this in mind I would go with a constant current source and zener diode with a op-amp voltage follower if precision is a concern. Failing this a 12 volt source then 7808 followed by a  7805 is about a stable as it is going to get assuming temperature is constant. Did not want to probe but what are you on to with long term precise time measurements?

Hi

Given that you can now see the schematic of what's in the OCXO, just put a very normal multi turn pot between the Vref and ground. Attach the wiper to the EFC pin and you have the best you possibly can have. It's referenced to and tracking to  the internal bias as best you can. That (and the "zero TC" of a normal pot) will do as good as anything you can do.

Bob

[FlyingHacker]

Given that you can now see the schematic of what's in the OCXO, just put a very normal multi turn pot between the Vref and ground. Attach the wiper to the EFC pin and you have the best you possibly can have. It's referenced to and tracking to  the internal bias as best you can. That (and the "zero TC" of a normal pot) will do as good as anything you can do.

This is exactly what I did as soon as I figured out that the vRef was an output not an input! Doh!

I ran some tests, and it was quite stable for a few days.

John Heath: I may have misspoke in terms of "long term." I am really looking to make a lab frequency standard. However, one of the cheap GPSDO from ePay may be a better option (into some distribution amp).

[suicidaleggroll]

Use that vref to reference a dac and run it off an MCU or CPLD that's also measuring the output frequency vs a GPS PPS and you have an easy GPSDO.

[FlyingHacker]

Use that vref to reference a dac and run it off an MCU or CPLD that's also measuring the output frequency vs a GPS PPS and you have an easy GPSDO.

I was wondering about something like this. Do you think a nav grade GPS would be sufficient with a long gate time? I think one of the other threads had something about this.

[uncle_bob]

Use that vref to reference a dac and run it off an MCU or CPLD that's also measuring the output frequency vs a GPS PPS and you have an easy GPSDO.

I was wondering about something like this. Do you think a nav grade GPS would be sufficient with a long gate time? I think one of the other threads had something about this.

Hi

Simple answer -- no.

More complicated answer: On a nav grade receiver, time is an afterthought. They don't put much work into it. The result is that the error in the timing solution gets pretty high. As the constellation shifts around that error moves plus and minus at a 1 to 12 hour-ish rate. If it's a few microseconds of error (as it is on some parts), that's a big deal.

With LEA-6T's dirt cheap, why not use the "right part"?

Bob

[Cerebus]

That (and the "zero TC" of a normal pot) will do as good as anything you can do.

I did a bit of experimenting recently with inexpensive multiturn pots (i.e. not the £20 a piece Vishay ones but ordinary Bourns et. al.). With the pots adjusted to mid scale I was seeing a worst case tempco on the ratio of voltage applied to the voltage at the wiper of about 25% of the tempco of the end to end resistance. e.g. For a 200 ppm track tempco I was seeing a ratiometric tempco of 50ppm. This is only a small sample (I had about 5 different pots to try) but it does suggest that you can't rely on the assumption that using a variable resistor as a potentiometer will cancel the temperature coefficient of the track.

[DimitriP]

With LEA-6T's dirt cheap, why not use the "right part"?

Due to "out of sight out of mind disease".
Probably because most can't measure the difference and if they dont' see it, it doesn't exist.

[DimitriP]

Use that vref to reference a dac and run it off an MCU or CPLD that's also measuring the output frequency vs a GPS PPS and you have an easy GPSDO.

I was wondering about something like this. Do you think a nav grade GPS would be sufficient with a long gate time? I think one of the other threads had something about this.

Yes or no will depend on your requirements and definition of "Sufficient".

"Sufficient", depends on how many decimal digits your counter can display past 10000000 Hz

[uncle_bob]

That (and the "zero TC" of a normal pot) will do as good as anything you can do.

I did a bit of experimenting recently with inexpensive multiturn pots (i.e. not the £20 a piece Vishay ones but ordinary Bourns et. al.). With the pots adjusted to mid scale I was seeing a worst case tempco on the ratio of voltage applied to the voltage at the wiper of about 25% of the tempco of the end to end resistance. e.g. For a 200 ppm track tempco I was seeing a ratiometric tempco of 50ppm. This is only a small sample (I had about 5 different pots to try) but it does suggest that you can't rely on the assumption that using a variable resistor as a potentiometer will cancel the temperature coefficient of the track.

Hi

Based on tests by the guys who make them (and that repeat when done in the field) the radiometric numbers on a properly made wire wound 20 turn is nearly zero (< 5 ppm). A properly done 20 turn cermet part with a TC of 100 ppm / C will come in below 10 ppm. In both cases it has to be a linear (not a log taper) part. It also assumes you have zero current through the wiper and that the pot is in good shape.

Bob

[FlyingHacker]

Use that vref to reference a dac and run it off an MCU or CPLD that's also measuring the output frequency vs a GPS PPS and you have an easy GPSDO.

I was wondering about something like this. Do you think a nav grade GPS would be sufficient with a long gate time? I think one of the other threads had something about this.

Hi

Simple answer -- no.

More complicated answer: On a nav grade receiver, time is an afterthought. They don't put much work into it. The result is that the error in the timing solution gets pretty high. As the constellation shifts around that error moves plus and minus at a 1 to 12 hour-ish rate. If it's a few microseconds of error (as it is on some parts), that's a big deal.

With LEA-6T's dirt cheap, why not use the "right part"?

I was (apparently erroneously) thinking the LEA-6T was a nav grade receiver. That was the one I was thinking of using, and yes they are cheap. I have not done much research into the topic yet, as I am sure is obvious.

I was going to order a LEA-6T and give it a shot. How much advantage is there to disciplining an OCXO with that GPS source vs. simply using the GPS source itself? Some of the models appear to use TCXOs. I get that OCXOs are more accurate, but the GPS model would be running off TCXO anyway.

I suppose you have to look at the GPS lock status and stop any disciplining of the OCXO when the GPS is not locked? But what feedback voltage would you apply to the OCXO vC at those times when the GPS was not locked? The last voltage before you lost the lock, or some predetermined middle ground setting, or what? I am guessing the last voltage before you lost the lock, or a (decaying?) moving average of recent values.

Sounds like fun.

[DimitriP]

How much advantage is there to disciplining an OCXO with that GPS source vs. simply using the GPS source itself?

Here is the short version pasted from  http://www.ko4bb.com/Timing/FAQ-2.php
 

 
  [url]http://www.ko4bb.com/Timing/FAQ-2.php#]http://www.ko4bb.com/Timing/FAQ-2.php#] [url]http://www.ko4bb.com/Timing/FAQ-2.php# [/url]

1. GPSDO: GPS Disciplined Oscillator, a precision oscillator that is compared to GPS timing signals and periodically adjusted to compensate for long term drift.
    GPS signals have good long term stability (> several hours) but short term stability is affected by propagation conditions, atmospheric conditions, receiving antenna effects, multipath and other short term impairements.
    A quality, oven controlled crystal oscillator such as the HP 10811, has excellent short term stability but is affected by drift due to aging and other effects.
    A GPSDO uses a stable crystal oscillator and a GPS receiver optimized for timing, and combines the good short term stability of the crystal with the good long term stability of the GPS signal to provide the best of both worlds.

There are plenty of designs out there, attempting to meet various degrees of "sufficiency".

[Vgkid]

GPSDO circuits.
1pps:
Brooks Shera(PLL); http://home.teleport.com/~oldaker/10mhz_construction.htm
http://www.rt66.com/~shera/index_fs.htm
Ve2zaz(FLL): http://ve2zaz.net/GPS_Std/GPS_Std.htm
ZL1BPU(PLL) ; http://www.qsl.net/zl1bpu/MICRO/SIMPLE/SimpleGPS.htm
KE0FF (PLL); http://www.rollanet.org/~joeh/projects/GPSlave_writeup.pdf
10khz input
J.R. Miller(PLL) ; http://www.jrmiller.demon.co.uk/projects/ministd/frqstd0.htm
____________________Heres one I just stumbled upon, will read it later.
https://hackaday.io/post/21322

[DimitriP]

Since we are posting lists
 http://www.jackson-labs.com/index.php/products/fury
 http://www.jackson-labs.com/index.php/products/lte_lite

For anyone building one primarily for educational purposes, I believe nothing beats the Miller GPSDO for simplicity ,ease of construction , performance and price (if you are into that sort of thing) .
Here is a comparison to put things in perspective: http://www.leapsecond.com/pages/gpsdo/

[edpalmer42]

I was going to order a LEA-6T and give it a shot. How much advantage is there to disciplining an OCXO with that GPS source vs. simply using the GPS source itself? Some of the models appear to use TCXOs. I get that OCXOs are more accurate, but the GPS model would be running off TCXO anyway.

I suppose you have to look at the GPS lock status and stop any disciplining of the OCXO when the GPS is not locked? But what feedback voltage would you apply to the OCXO vC at those times when the GPS was not locked? The last voltage before you lost the lock, or some predetermined middle ground setting, or what? I am guessing the last voltage before you lost the lock, or a (decaying?) moving average of recent values.

Sounds like fun.

A bare GPS receiver (even a timing-grade one) will wander around in the e-8 area when averaged over 1 sec.  So a 10 MHz output might change by 0.1 Hz.  If you average over 10 sec., you're now in the e-9 area and so on.  A reasonably good OCXO will get you to the e-11 range at 1 sec.  The GPSDO merges these two as described in the quote from KO4BB.

GPSDOs behaviour during a loss of lock vary depending on the sophistication of the design.  The smartest ones learn the drift rates of the OCXO and shift the EFC voltage accordingly.  Less sophisticated ones freeze the voltage.  Some DIY versions (often analog-based circuits without a processor) don't take any special action.  But remember that a timing-grade GPS receiver typically only needs one satellite to maintain lock.

[FlyingHacker]

Great info, everyone! Thank you.

[uncle_bob]

Hi

Just in case anybody else has missed the uBlox "magic code":

A LEA-6 is a nav receiver *unless* it is a LEA-6T. The LEA-6T is a timing receiver. The same convention applies to other parts in their lineup in the same fashion.

The "T" enables a number of features. It is *not* clear that it changes the basic function of the unit. The features it enables make it much easier to use in a GPSDO or other precise timing (sawtooth corrected) application. The net result is that you are below 1x10^-9 at one second rather than up around 1x10^-8 at 1 second.

Bob