What frequency standard is sufficient?

[tkamiya]

With price of GPSDO with OCXO being what it is, there is no need to even think - I *think*. 

Mine was more of a want than a need.  It's a good thing to have.  Better have it than not to have it.  It was a fun thing to play with.  I can't think of myself needing any more than what I have now.  Couldn't go much cheaper either by going to TCXO. 

Rubidium, I understand, is worse for short term stability than OCXO?  With GPS taking care of long term, I wonder why would I actually want Rubidium?  I've seen a AT&T product that has GPS/Rubidium/TCXO.  Can someone chime in and elaborate on relative merits? 

[nctnico]

Like others said: with a Rubidium reference you don't need to wait for a GPS lock (and have a place of an antenna). If you put a cleanup oscillator behind a Rubidium reference then you'll get a very stable frequency reference. But you still don't know if the frequency from the Rubidium clock is accurate while a GPSDO output is -sort of- synchronised with the clocks at (IIRC) the US Naval observatory.

[coppercone2]

I want to know if what I have is useful for anything in particular or just be certain that I have a 'incomplete device'. I know you can do it, but I am curious if anyone has academic answers.

[nctnico]

I want to know if what I have is useful for anything in particular or just be certain that I have a 'incomplete device'. I know you can do it, but I am curious if anyone has academic answers.

You'd have to check the short and long term stability of the time reference in your instrument compared to those of a Rubidium clock and the GPDSO. I don't know the numbers from the top of my head but you can find the numbers for the BG7TBL GPDSO in the (long) thread about it. For the Rubidium clock they are in the datasheet.

[coppercone2]

I glanced at specs like that before but there is not that much information given . I will try to determine some error related time constants with that information

[ArthurDent]

A good GPS, even an older one like a Trimble Thunderbolt, steers the OCXO to exactly 10Mhz by using a large time constant of 500 seconds or more so any minor corrections are pretty smooth. For comparing 2 frequency standards, I trigger my scope off one GPS standard and connect the 10Mhz GPS, rubidium, or OCXO I’m testing to a vertical channel and check for drift at a sweep rate of 5ns/DIV. Almost all the GPS standards I’ve tested stay within +/-5ns over a long period of time. I’ve tested about 200 Thunderbolts plus several other telcom makes from HP, Lucent, and other companies and they are all pretty good. Any GPS standard that can be monitored on a PC with the ‘Lady Heather’ program is a plus because you can easily see how the GPS standard is doing over long periods of time. I also monitor the output of the unit I’m testing on a good counter. (see attached photo)

Both rubidium and crystal oscillators have warm-up characteristics and after they have been on for several hours or days are pretty good but both have to be adjusted against a higher accuracy standard like a GPS. I have one OCXO frequency standard that has been running (other than battery replacements) for over 35 years. All crystal oscillators exhibit aging and generally speed up by about 2 parts in 10E10 every day and are quite predictable. If an OCXO is powered down then turned back on there is an error caused by ‘retrace’ where the oscillator won’t come back to exactly the same frequency. They also may take several weeks or longer to reach maximum stability. Rubidium warms up fast and comes back to almost the same frequency each time but some take a day or two to reach maximum stability. The Rb standard I use the most is about 25 years old and very stable. It uses a FRK-HLN standard and is built into a unit that has buffered sine wave outputs of 0.1Mhz, 10Mhz, plus 1Ghz, and is great for checking counters.

Just buying a rubidium or crystal oscillator from Ebay may not be the best way to go because either type could have drifted so far out from a long ago calibration that they won’t be adequate for what you want them to do. In the past I sold LPRO-101 and other Rb units that I had carefully tested and adjusted over time so I knew they were reliable and adjusted to be on frequency. Buying one from China is kind of a crap shoot because a lot of these sold are pulled out of scrap and may be untested or near end of life. Attached is a graph showing an LPRO-101 being tested/adjusted. 

Even if you don’t have requirements for really high accuracy I’d still go with a Rb unit calibrated by the seller or for better long term stability, if you have a good spot for an antenna, I’d definitely go with a GPS. If you have no way to check calibration I’d avoid an OCXO even though it will probably be the cheapest way out. The GPS units probably range from about $100-$300 and the Rb oscillators like the LPRO-101 are probably up to $275 for a good one. If you don’t want to experience the ‘joy’ of trying to build a unit from collected parts, probably the small UCCM GPSDO boards from any of the various makers built into a small case would be the quickest (but maybe not the absolute best) way to get a GPS receiver up and running.

[cdev]

I leave my GPSDO on all the time when I am using it. 24/7

Thats what I've been told is the best way to have it operate both for accuracy and longevity.

Like others said: with a Rubidium reference you don't need to wait for a GPS lock (and have a place of an antenna). If you put a cleanup oscillator behind a Rubidium reference then you'll get a very stable frequency reference. But you still don't know if the frequency from the Rubidium clock is accurate while a GPSDO output is -sort of- synchronised with the clocks at (IIRC) the US Naval observatory.

[Wolfgang]

Tcxo, Ocxo, rubidium, GPS disciplined and all combinations thereof.

I can really see the use in having a frequency standard on the bench.  The problem for me is what level and cost is right for me?    I know it is whatever you are working on will dictate it, but I'd rather just buy what I need for future uses now. 

I'm studying now for ham, and am looking to start repairing various things for resale.  Older radios, test equipment, the odd repairable tv,   God knows what.  Calibrating test equipment is a good reason, too.

What is a good balance type standard that is good for the medium beginner and will still be useful a ways down the road and not break the bank?

Hi,

I do it this way:
- a homebrew unit with a really good 10MHz OCXO from MORION. Aligned to GPS and Rubidium every year. Drift was never more tha ca. 1Hz. For my everyday lab use.
- a GPS receiver from Leo Bodnar for other frequencies than 10MHz
- a source built around an EFRATOM Rubidium standard (this is a Rubidium governed OXCO) in case super high accuracy is needed.

[james_s]

You can get a distribution amplifier for next to nothing. I've seen guys using old analog video distribution amps. If you'd rather build something, I have several reels of Maxim video amp ICs, I'd send you a strip of them for the cost of a stamp and envelope.

[cdev]

I use an Extron video amp, worked great out of the box. Cost me almost nothing.

[texaspyro]

Some GPSDOs (like the Thunderbolt) let you disable disciplining and let the OCXO free run.  In this mode the phase noise will be lower than when disciplined.   For phase noise critical applications (like up-converting to microwave frequencies) this can be quite useful.   A typical application would be to let the GPSDO discipline the oscillator to get the freq dead-nut on and then disable disciplining when you need low phase noise.

Note that disabling disciplining is different than putting the GPSDO into "holdover" mode.  In holdover mode most GPSDOs still control the OCXO freq using predicted aging / temperature corrections.   Generally (but not always) holdover mode causes less phase noise than disciplining to GPS.

[ArthurDent]

Here is a photo of an Inline model 1N3021 video amp I modified to use as a 10Mhz distribution amp. Each output has its own wide bandwidth opamp to give good isolation and produce a good clean 10Mhz sine wave, except the bottom six outputs which are square wave. The 75 ohm resistors on input/output were paralleled with 150 ohm resistors to change the in/out impedance to 50 ohms.

Under the distribution amp is a Thunderbolt GPS I built into a case, then a modified HP designed GPSDO that has a built-in power supply and display, then a Datum GPSDO with multiple frequency outputs. Each GPSDO drives one of the 3 groups of 6 outputs. These three GPSDOs are powered 27/7 and receive their signal from one of two outside active GPS antennas feeding 1575.42Mhz L1 distribution amps that allow me to run or test about 16 GPS units at once. To the left of the GPSDOs is a nice Ball/Efratom FRK-HLN based rubidium standard with 0.1Mhz, 10Mhz, and an added 1.0Ghz buffered sine wave output. It produces good stable and accurate outputs even though it is about 25 years old.

I use one sine 10Mhz output to trigger my scopes, another for my HP 53132A counter, and another for the timebase in my modified FY6600 function generator. The scope and counter make it easy to check the frequency drift of any device under test or comparing one GPSDO against another. The GPSDO drift stays within 5ns over a long period of time.

[Housedad]

This has been a very informative thread.   I really appreciate the replies.

It looks like I'll find a GPSDO  and just use that.   Probably way more than I'll ever need, but I'll just consider it TEA. 

I see from what you guys are saying is that if I ever need to distribute the signal, it is best to use  Sine wave.   Interestingly, the RF generator I'm working on now has a sine wave on the INT Ref OUT line, but uses a square wave internally.   I'm thinking that a reference that puts out both would be useful.   Thoughts?

[tautech]

This has been a very informative thread.   I really appreciate the replies.

It looks like I'll find a GPSDO  and just use that.   Probably way more than I'll ever need, but I'll just consider it TEA. 

I see from what you guys are saying is that if I ever need to distribute the signal, it is best to use  Sine wave.   Interestingly, the RF generator I'm working on now has a sine wave on the INT Ref OUT line, but uses a square wave internally.   I'm thinking that a reference that puts out both would be useful.   Thoughts?

The link I posted earlier to Leo's GPSDO offers a US seller:
https://v3.airspy.us/product/lb-gpsdo-1/

$219, two configurable outputs, antenna and all ready to go !

[Housedad]

When you use a video distribution amp, how do you reconcile that the video amplifier ouputs are 75 ohm vs the test equipment being 50 ohm impedance?

[jpb]

The two options are to replace the 75 ohm resistors with 50 ohm ones, or do as ArthurDent posted above and parallel them with 150 ohms.

[Gyro]

This has been a very informative thread.   I really appreciate the replies.

It looks like I'll find a GPSDO  and just use that.   Probably way more than I'll ever need, but I'll just consider it TEA. 

I see from what you guys are saying is that if I ever need to distribute the signal, it is best to use  Sine wave.   Interestingly, the RF generator I'm working on now has a sine wave on the INT Ref OUT line, but uses a square wave internally.   I'm thinking that a reference that puts out both would be useful.   Thoughts?

The link I posted earlier to Leo's GPSDO offers a US seller:
https://v3.airspy.us/product/lb-gpsdo-1/

$219, two configurable outputs, antenna and all ready to go !

Leo's single output one is even cheaper, from the same site... https://v3.airspy.us/product/lb-gpsdo-mini/

[Housedad]

The two options are to replace the 75 ohm resistors with 50 ohm ones, or do as ArthurDent posted above and parallel them with 150 ohms.

Gotcha.  So a very, very  mild mod of the amp.

[ArthurDent]

If you try to remove the 75 ohm resistors to replace them with 50 ohm resistors you run the risk of lifting pads or doing other damage to the board. It is much easier and safer to just solder a 150 ohm resistor across the existing 75 ohm resistors.

Not all video amps are created equal. The one I bought came with one amp driving all six outputs through resistors so it wouldn't have the best isolation or drive. It did have the 8-pin IC layout for individual amps for each output and I installed amps with a wide bandwidth in those locations to get good isolation and better protection against shorted outputs, etc.. The problem is you don't know what the video distribution amp has inside until you buy it and take it apart.

[cdev]

The Leo Bodnar unit doesn't have an oven, right?

What makes it different than, say a ublox timing GPS, some of which can be found on ebay for $20.

They have two internal user programmable square wave outs which can be adjusted typically one is set to 1 pps and the other is set to 10 MHz.

I think with adjustable duty cycle.

One could slap one of them in a box with a Raspberry Pi and a screen or a serial port and have a standalone adjustable DDS.

It has the expected amount of jitter.
 
Am I wrong about something here?

[hamster_nz]

The Leo Bodnar unit doesn't have an oven, right?

What makes it different than, say a ublox timing GPS, some of which can be found on ebay for $20.

They have two internal user programmable square wave outs which can be adjusted typically one is set to 1 pps and the other is set to 10 MHz.

I think with adjustable duty cycle.

One could slap one of them in a box with a Raspberry Pi and a screen or a serial port and have a standalone adjustable DDS.

It has the expected amount of jitter.
 
Am I wrong about something here?

A GPS's PPS signal jumps up and down 'like crazy' (maybe a bit of an overstatement), I collected 10 hours of readings and got +/- , due to the noise in the receiver's location. It is also usually at the nearest cycle of a 48MHz internal clock - But over time it is exactly one pulse per second.

A crystal is relatively stable in the short term - from second to second, but over the long term will not be X pulses per second, and can wander with changes in temperature, but it's frequency isn't that accurate.

I did an experiment, and with a 100MHz nominal crystal as a reference, over 10 hours I got  between 100,000,489 and 100,001,193 counts between the rising edges of the PPS pulses. With an average of 100,000,877 counts I am sure that my Xtal is about 8.77 parts per million fast at standard lounge operating temperatures.

If I was able to feed back this measured error to adjust the speed of the crystal, it would be on frequency (thanks to the GPS's PPS reference), stable over the short term in both frequency and phase (due to the nature of the crystal), and also stable over the long term (once again due to the GPS input).

[metrologist]

I've been happy with a couple UCCM trimble units by this guy. This one is a symmetricon for under $120.

https://www.ebay.com/itm/GPSDO-Symmetricom-Inside-GPS-10MHz-1PPS-GPS-Send-GPS-antenna/263458624856?hash=item3d575b9558:g:bk8AAOSw9vlaael4:rk:42:pf:0

I am also building the Lars GPSDO project posted here. It uses an Arduino Uno type controller, OCXO and 1pps GPS of your choice, and a few cheap components.

Lars was very nice to have provided all of the project details: https://www.eevblog.com/forum/projects/lars-diy-gpsdo-with-arduino-and-1ns-resolution-tic/?all

[MrW0lf]

The Leo Bodnar unit doesn't have an oven, right?
What makes it different than, say a ublox timing GPS, some of which can be found on ebay for $20.

It's from Leo, therefore it rocks. That's for starters. Second look here:
http://www.leobodnar.com/shop/index.php?main_page=product_info&products_id=234
450 Hz to 800 MHz independent output with tunable (common) level and phase shifts (decent square). It has proven massively useful on occasions.
Then look overviews which include Allan deviation etc:
http://leobodnar.com/files/Microsoft%20Word%20-%20Investigation%20of%20Leobodnar%20GPSDO_rev2.pdf
http://leobodnar.com/files/Informal%20Evaluation%20of%20a%20Leo%20Bodnar%20GPS%20Frequency%20Reference.pdf
and this:



Looks like $20 device or perhaps bit better? My only trouble with it that I have nothing to check it's accuracy myself Besides possibly Aim TTi TF960.

Edit: If read between lines it seems he can offer OCXO version also, sine TCXO one is referred as "standard version".

[macboy]

The two options are to replace the 75 ohm resistors with 50 ohm ones, or do as ArthurDent posted above and parallel them with 150 ohms.

Those are two options but not the only two. Other options include:
- Use 75 Ohm cable, and let the 50 Ohm termination at the instrument cause some reflection back toward the amp. It should be absorbed by the amplifier's 75 Ohm impedance and end there.
- Use a 50 Ohm cable, and let the instrument terminate the signal. The level will be a little lower than nominal due to the 75 Ohm source, but there should be minimal reflection back into the cable.
In either case, I'd do a sanity check for signal quality, but you will likely not cause any significant impairment of the signal at just 10 MHz. The fact that it is a single fixed frequency sine wave is even better, as reflections will only cause some amplitude error rather than distorting the shape of the waveform.

Personally, I have two 1-to-4 video amps that I will use for distribution. I also have many very high quality 75 ohm cables which are not suitable for general 50 Ohm use. I plan to use them in this application, and I'm not worried about the mismatch in the least. I am open to learn why I'm wrong, if someone knows better...

[jpb]

On the subject of 50/75 ohm termination - a lot of instruments aren't either for the ref in anyway. I think my counters are more like a few kohms so you can locally make them either 50 ohm or 75 ohm with suitable additional resistors.