10MHz GPSDO by "BH3SAP"

[mawyatt]

Hi,

reading the discussions about U5 and its somewhat negative effect on the output signal and the mitigation mods, why us U5 anyways?

Wouldn´t it be better to ommit U5 and feed the output of the OCXO directly to the output?

I don´t have the datasheet, but I expect the OCXO to be able to drive a 50ohms load with a reasonable level.

Regards
Chris

Good question!!

The OCXO we have is specified with a CMOS logic level output (squarewave) and not 50Ω likely with sine wave as the output. Think the ISOTEMP OCXOs are also CMOS Logic squarewave.

If your GPSDO has a OCXO that has a 50Ω output sinewave, then bypassing the Driver/Buffer U5 Amp (as originally implemented with OPA692) should improve the output waveform unless you make some of the mods we've suggested which will work with either type OCXO and produce a cleaner output from U5.


The "intent" of U5 as an OCXO driver/buffer is good, the original implementation is not!!

Best

[KungFuJosh]

Looks good!! When you get U5 powered from ~9V rather than 5V you can use a smaller resistor, or even replace with a short, since you'll have enough voltage headroom to keep U5 input and output away from saturation with larger waveforms.

Just a thought/question: is there something to replace U5 with that would work better with the the circuit as it is?

Also, could you please draw your LM7810 mod? It's hard to see in the photo everything that's going on.

Thanks,
Josh

[mawyatt]

Looks good!! When you get U5 powered from ~9V rather than 5V you can use a smaller resistor, or even replace with a short, since you'll have enough voltage headroom to keep U5 input and output away from saturation with larger waveforms.

Just a thought/question: is there something to replace U5 with that would work better with the the circuit as it is?

Also, could you please draw your LM7810 mod? It's hard to see in the photo everything that's going on.

Thanks,
Josh

Here you go!!

10uF are 1206 and mounted directly to Reg Pins, Center Pin is Ground, so one 10uF on front and one on back. Reg Output pin is directly soldered to L2 pad, and Reg Center Pin is soldered directly to ground side of decoupling cap next to L2 inductor. Wire runs from Reg Input to 12V Switch.

Best

[KungFuJosh]

Here you go!!

10uF are 1206 and mounted directly to Reg Pins, Center Pin is Ground, so one 10uF on front and one on back. Reg Output pin is directly soldered to L2 pad, and Reg Center Pin is soldered directly to ground side of decoupling cap next to L2 inductor. Wire runs from Reg Input to 12V Switch.

Thank you!

Since I'll need to order anyway, I think I'll get some THT 10uF MLCCs, as they'll give me a more reliable/mechanical connection to the reg pins. 10% X7R should be fine, right?

Thanks,
Josh

[mawyatt]

Thru Hole should work, but the 1206 or 0805 ceramic 10uF can be soldered directly to the regulator leads as they come out of the TO-220 plastic body bottom. That's a pretty reliable connection!!

Just solder the caps directly to the regulator, cut the input lead short and solder insulated wire, cut the ground and output leads to fit. Spread the Center and Output leads to fit and solder directly to PCB. Route 12VDC input wire under SMA connectors to Power Switch as shown. Position the regulator TO-220 vertically between the BNC case and the Blue Pill connector (TO-220 tab is ground). Shouldn't take more than a few minutes to add this modification. Don't for get to remove L2 and change C9 to ~1.2pF. Also change the series resistor if you want a larger output PP amplitude.

Best

[fredo_]

It was mentioned earlier that the baud rate setting is only saved to battery backed RAM in the GPS module.  So after some days the setting will be reverted to default.

Make it a few minutes...

[Ringmodulator]

It was mentioned earlier that the baud rate setting is only saved to battery backed RAM in the GPS module.  So after some days the setting will be reverted to default.

Make it a few minutes...

What is the default baudrate, that the ATGM336H gps module reverts to after the battery is empty?
I think, it would be advised to set the baud rate in the bluepill to this value to avoid problems after the unit has not been powered for longer time periods.

Chris

[fredo_]

What is the default baudrate, that the ATGM336H gps module reverts to after the battery is empty?
I think, it would be advised to set the baud rate in the bluepill to this value to avoid problems after the unit has not been powered for longer time periods.
Chris

Should be the case...
Default value is 9600.
Maybe I should update the documentation to advise against changing this setting unless you really know what you are doing !... 

[Kean]

Maybe I should update the documentation to advise against changing this setting unless you really know what you are doing !... 

You make a big assumption that people read documentation...

[mawyatt]

More bad news!! The Bowie OX256B OCXO has now decided to go bonkers and spit out garbage! Time to order a ISOTEMP OCXO and/or maybe a CTI.

Best

[KungFuJosh]

Maybe I should update the documentation to advise against changing this setting unless you really know what you are doing !... 

You make a big assumption that people read documentation...

I mean, there was only 254 results for "RTFM" in the Test Equipment section. 

[KungFuJosh]

I got the 2nd GPSDO from Mike's link. It was a little wonky with the stock firmware (more than normal), so I switched it to 1.0.14, and reset it a few times, and it's working now. The only mod I've made so far on unit 2 (besides fw) was to remove C9.

I have both GPSDOs plugged in to the Magnova to see how they compare, and how tight they are in phase with each other using the scope's internal ref. It was entertaining seeing how crazy out of phase they were when one was already warmed up (and triggered on) and the changes as the other one counts down and begins to lock in. I'll assume that makes sense, I'm tired.


Here's an idea/request for the next firmware version:

If you can have the firmware synchronize the saved baud rate at startup with the module, I would assume that should get rid of the issues we're having after being powered off for a while.

Thanks,
Josh

[mawyatt]

I got the 2nd GPSDO from Mike's link. It was a little wonky with the stock firmware (more than normal), so I switched it to 1.0.14, and reset it a few times, and it's working now. The only mod I've made so far on unit 2 (besides fw) was to remove C9.

Did you change the Blue Pill crystal connection wire?
Best

[KungFuJosh]

I got the 2nd GPSDO from Mike's link. It was a little wonky with the stock firmware (more than normal), so I switched it to 1.0.14, and reset it a few times, and it's working now. The only mod I've made so far on unit 2 (besides fw) was to remove C9.

Did you change the Blue Pill crystal connection wire?
Best

I did this morning.

The amplitude is better on the more modified version with the 220Ω resistor in series, but I dunno if I think one square wave is significantly better than the other. I'll post a screenshot after they've been running for a while- let me know what you think when I do.



Thanks,
Josh

[mawyatt]

The amplitude is better on the more modified version with the 220Ω resistor in series, but I dunno if I think one square wave is significantly better than the other. I'll post a screenshot after they've been running for a while- let me know what you think when I do.



Thanks,
Josh

The 220Ω won't help much if the 10k bias pair aren't replaced with 1k. Leaving them at 10k causes the center bias voltage to drop below VCC/2 due to U5 bias current. This will be evident on the lower input signal swing dropping below U5 input lower limit. Then you still have the U5 VCC decoupling and the output limiting due to saturation.

Providing a 9V rather the 5V U5 VCC allows more headroom on the input and output, which should yield better output waveforms.
Best

[KungFuJosh]

Fair enough. I haven't placed that Mouser order yet, but I probably will today and try the rest of the mods on unit 1 later this week.

Unit 2 is pretty stable, but obviously not locked perfectly with unit 1. Do you think aging will help that?

Thanks,
Josh

[Ringmodulator]

Fair enough. I haven't placed that Mouser order yet, but I probably will today and try the rest of the mods on unit 1 later this week.

Unit 2 is pretty stable, but obviously not locked perfectly with unit 1. Do you think aging will help that?

Thanks,
Josh

Josh,

there is an effect called "retracing" for XOs. After beeing powered on, it takes some time until all parameters of the quartz are setteled in.
This can show as small frequency jumps up and down.

Aging is a different effect. It describes the almost constant changing or drift of the resonant frequency in one direction over long time periods.
Disciplining with a gps unit takes care of this.

Be warned, you can easily slip into "time nut" territory 

There is an eccelent foundation article from Ulich Bangert  DF6JB: https://www.bartelsos.de/_media/messtechnik/amsat-journal.pdf
I have not found an english version. You might run it through your favorite translator.

Chris

[mawyatt]

Fair enough. I haven't placed that Mouser order yet, but I probably will today and try the rest of the mods on unit 1 later this week.

Unit 2 is pretty stable, but obviously not locked perfectly with unit 1. Do you think aging will help that?

Thanks,
Josh

Josh,

there is an effect called "retracing" for XOs. After beeing powered on, it takes some time until all parameters of the quartz are setteled in.
This can show as small frequency jumps up and down.

Aging is a different effect. It describes the almost constant changing or drift of the resonant frequency in one direction over long time periods.
Disciplining with a gps unit takes care of this.

Be warned, you can easily slip into "time nut" territory 

There is an eccelent foundation article from Ulich Bangert  DF6JB: https://www.bartelsos.de/_media/messtechnik/amsat-journal.pdf
I have not found an english version. You might run it through your favorite translator.

Chris

Good points!!

We should also remember these are Frequency Locked and not Phase Locked, which means the long term frequency will approach ideal, however short term frequency and/or phase is not locked and can jump around some. Continuous Phase Locking should provide the best overall short term effects for both phase and frequency.

Best

[KungFuJosh]

Yes, I assumed they probably wouldn't line up perfectly, though that would be cool.

What I'm looking to see is stability between the two once everything is settled.

If there's an offset of ~57ns, am I being unrealistic for that to remain fairly consistent?

That delta shifted up to 64ns in the time it took me to type that sentence.

Thanks,
Josh

[Ringmodulator]

Yes, I assumed they probably wouldn't line up perfectly, though that would be cool.

What I'm looking to see is stability between the two once everything is settled.

If there's an offset of ~57ns, am I being unrealistic for that to remain fairly consistent?

That delta shifted up to 64ns in the time it took me to type that sentence.

Thanks,
Josh

Provided everything else is equal, you would need identical cable lengths for the 10 MHz cable.
This would take care for the constant offset, not the shifting over time.

Chris

[KungFuJosh]

Provided everything else is equal, you would need identical cable lengths for the 10 MHz cable.
This would take care for the constant offset, not the shifting over time.

The cables in use are a matched pair.

Thanks,
Josh

[Ringmodulator]

Provided everything else is equal, you would need identical cable lengths for the 10 MHz cable.
This would take care for the constant offset, not the shifting over time.

The cables in use are a matched pair.

Thanks,
Josh

Well, I overlooked the obvious  :

The BH3SAP units are frequency locked, not phase locked.

Chris

[mawyatt]

The Frequency Lock update is discrete not continuous with an update rate once per second.

Best

[KungFuJosh]

The Frequency Lock update is discrete not continuous with an update rate once per second.

Does that mean I should expect a potential shift every second? They've been running continuously, and not too far off.

[mawyatt]

The Frequency Lock update is discrete not continuous with an update rate once per second.

Does that mean I should expect a potential shift every second? They've been running continuously, and not too far off.

It's a sampled data system and the sample rate is the GPS 1PPS that is used to determine the OCXO frequency. So the frequency error in counts can only be assessed every second and the PWM used to correct the OCXO and the PWM filter has a long time constant as well. This is not like a typical PLL where the VCO is constantly (well almost) being updated by the loop filter based upon the phase error.

Best