10MHz GPSDO by "BH3SAP"

[Kurt_!]

One more question:
Pins PA0 and PA1 each have a LED connected to ground with a resistor.....

[mawyatt]

One thing to consider about the "Quality" of the output waveform is how the driver amplifier effects the result from the OCXO. If the driver is perfectly linear but band-limited then the output will just contain the band-limited OCXO signal and the added noise of the driver.

However if the driver is non-linear and band-limited then the output signal becomes the addition and multiplication of the OCXO signal and the driver. This multiplication implies the noise characteristics of the driver get "mapped into" the output signal rather than "summed". What this basically means is the driver noise gets superimposed on the OCXO signal such that the result is an output signal with more noise on the fundamental which can't be removed by filtering, whereas with a linear driver the driver noise can be removed/reduced by filtering.

Since all amplifiers linear or otherwise, have broadband noise and 1/f noise, the 1/f noise becomes a particular concern. This gets mapped onto the OCXO fundamental whether the OCXO signal is a sinusoidal, squarewave or something in-between, and can be "seen" as a heightened level of phase noise or jitter. The output signal close in phase noise due to the non-linear driver 1/f noise shows as random phase fluctuations in the output for a sinusoidal waveform and low frequency edge jitter in a square or squarish waveform. This is why the output signal phase noise/jitter should be better if the driver amplifier U5 doesn't saturate and become highly non-linear.

This is why the Local Oscillator phase noise is so important in well designed receivers, as the LO phase noise gets "mapped" directly onto the received signal reducing the SNR, and close in LO phase noise even more so in Direct Down Conversion (Zero IF) receivers.

We should mention that there's an area of advanced research (beyond the scope of this discussion) where the above (linear vs non-linear) doesn't completely hold true where an output power amplifier is under extreme non-linearity by design to improve the final output noise characteristics.

Anyway, folks that have access to the proper equipment should be able to "see" the results and effects of U5 being saturated and not with close in phase noise measurements of the 10MHz waveform fundamental. Not sure a standard DSO is going to show this jitter, or a standard SA as close in phase noise of the OCXO signal, which might be better than these instruments can resolve.

Best

[daddygo]

One more question:
Pins PA0 and PA1 each have a LED connected to ground with a resistor.....

PA0/PA1 as follows
@fredo_ (GitHUB)

"
GPS Lock and GPSDO Lock Outputs
Pins PA0 and PA1 can be used to drive leds showing GPS Lock and GPSDO Lock status.

PA0 will be pulled low as soon as the GPS module is locked and PA1 will be pulled low when PPB mean value goes above the set threshold.
"

I put a driver PCB here so as not to overload the MCU output. like: pict1

[daddygo]

Should the two software programs (u-center and VisualGPS...) work then?

Sure 

u-Center for uBlox base (NEO-6M)

GnssToolKit for ATGM336H base

[daddygo]

Same to me " Done "

Do you have a 3.3v USB-->TTL Uart converter ?

YES.

That was for Kurt 

Got it 

[bingo600]

I'm still getting the TTL-to-USB converter. I've already ordered it.
Should the two software programs (u-center and VisualGPS...) work then?

The Bluepill  "mirrors" the GPS Uart Rx/Tx on the Bluepill pins shown above.
So you could be able to use u-center etc.

[bingo600]

Anyway, folks that have access to the proper equipment should be able to "see" the results and effects of U5 being saturated and not with close in phase noise measurements of the 10MHz waveform fundamental. Not sure a standard DSO is going to show this jitter, or a standard SA as close in phase noise of the OCXO signal, which might be better than these instruments can resolve.

Best

Would this guy suffice ??   
http://wavecrestdts.com/dts2079.htm

Well it's packed down in the cellar right now, but jitter is it's keyword

[mawyatt]

Would this guy suffice ??   
http://wavecrestdts.com/dts2079.htm

Well it's packed down in the cellar right now, but jitter is it's keyword

Don't know, never used anything like this before or heard of the company. Since you have direct access best to just try it out and see 

Our interest is more in the close in phase noise performance.

Best

[Eric-H]

Hi,

I've made some tweaks to fredo's firmware that might help:
- Made the tuning faster, it will now lock under 15 minutes
- Fixed sporadic hangup at powerup
- Correctly display small negative ppb-values
- Set version to v0.1.14

I will make my changes public if they are useful to you. They are based on the latest version of fredo_ (v0.1.13).
Please let me know if these changes work on your units.

Eric

[fredo_]

Hi,

I've made some tweaks to fredo's firmware that might help:
- Made the tuning faster, it will now lock under 15 minutes
- Fixed sporadic hangup at powerup
- Correctly display small negative ppb-values
- Set version to v0.1.14

I will make my changes public if they are useful to you. They are based on the latest version of fredo_ (v0.1.13).
Please let me know if these changes work on your units.

Eric

Hello Eric,
Thank you for your contribution !
Can you please create a Pull Request on Github so that I can review and merge your changes ?
Best,
Frederic.

[daddygo]

One thing to consider about the "Quality" of the output waveform is how the driver amplifier effects the result from the OCXO.
Best

What if there was something like this at the output?

I'm just reading and got some good ideas for an AD8009 + GAIN block combination, even rail to rail (if it were designed, a new main PCB with a new PSU), since the AD8009 max. is 3.8-3.9V, and optionally xfrmr.

[mawyatt]

One thing to consider about the "Quality" of the output waveform is how the driver amplifier effects the result from the OCXO.
Best

What if there was something like this at the output?

I'm just reading and got some good ideas for an AD8009 + GAIN block combination, even rail to rail (if it were designed, a new main PCB with a new PSU), since the AD8009 max. is 3.8-3.9V, and optionally xfrmr.

That won't work as you have the + and - inputs shorted!!

Best

[mawyatt]

One thing to consider about the "Quality" of the output waveform is how the driver amplifier effects the result from the OCXO.
Best

What if there was something like this at the output?

I'm just reading and got some good ideas for an AD8009 + GAIN block combination, even rail to rail (if it were designed, a new main PCB with a new PSU), since the AD8009 max. is 3.8-3.9V, and optionally xfrmr.

Also consider that input to the AD8009 will be centered around VCC/2 or 2.5VDC, so loading the output to ground with a 49.9Ω resistor (R1) at the filter input will cause the AD8009 to constantly supply a ~50ma output current irrespective of what the 10MHz requires if output is not saturating.

Don't think you need to go to all this trouble with the 10MHz Output. This is intended as a Frequency/Phase reference and the amplitude isn't that critical, Frequency and Phase are. Also the BW of the output driver doesn't need to be GHz high and don't see a real need for the complex filtering.

In our view the 10MHz output (or as we intend to use such) is this will be used as a Frequency/Phase reference for other instruments like AWG, RF Source, SA or DSO and these don't require a pure sinusoidal waveform, just a clean, noise free waveform.

For example the Siglent SDG6000X AWG specifies a Reference Input of 10MHz with 1.4VPP minimum amplitude and has a Ref Input Z of 5kΩ and the Siglent SSA3000X Plus SA Reference Input Requirements are -5 to +10dBm (0.356 to 2VPP) with 50Ω reference input Z.

Not knowing the details of how these instruments 10MHz input reference circuitry works, but coming from experience one would expect a high frequency RF filter with a termination and overload/voltage protection followed by a limiting amplifier, so the internal instrument 10MHz "signal" eventually becomes a squarewave. This doesn't mean that the input waveform can be "sloppy" and have lots of perturbations and noise since these will get squared up inside the instrument and could cause excessive edge jitter. The reference input Phase Noise gets converted into edge jitter by the limiting amplifier in the same way.

With this in mind it seems that a simple buffer driver amp is all that's required and additional filtering isn't necessary. Since various OCXOs could have sinusoidal or squarewaves (or in-between), maybe a simple pot to the input of the buffer/driver amp from the OCXO output to allow one to "set" the proper level to keep the buffer/driver amp out of saturation.

Anyway, here's a simple concept that we believe should work as the OCXO Buffer/Driver Amp for most applications with no need for a wide-band IC amp, just an emitter follower with a 2N3904 and 2N3906. We aren't an expert in this area, so correct any of this if we aren't understanding things!!

Best

[daddygo]

One thing to consider about the "Quality" of the output waveform is how the driver amplifier effects the result from the OCXO.
Best

What if there was something like this at the output?

I'm just reading and got some good ideas for an AD8009 + GAIN block combination, even rail to rail (if it were designed, a new main PCB with a new PSU), since the AD8009 max. is 3.8-3.9V, and optionally xfrmr.

That won't work as you have the + and - inputs shorted!!

Best

Oh nope, -I haven't looked through the wiring diagrams deeply yet, and I haven't put the stages together; these are just ideas from the internet, from various designs that I took screenshots of.

This is just an idea(s), if anyone is familiar with these circuits and its theories.

like:
https://eu.mouser.com/datasheet/2/412/TQP369184_Data_Sheet-1500714.pdf

and this old one with BGA616:
https://huprf.com/huprf/wp-content/uploads/2016/03/DA1-8-Manual-V1.pdf

For example, the AD8900 can be cascaded and rail-to-rail powered, as it is a very fast amplifier and I have seen it used in astronomical instruments for 10MHz.

If were planning to design a new main PC, I would be looking for ideas to start implementing in September.

can have multiple 10 MHz outputs and sine and square waveforms....

[daddygo]

One thing to consider about the "Quality" of the output waveform is how the driver amplifier effects the result from the OCXO.
Best

What if there was something like this at the output?

I'm just reading and got some good ideas for an AD8009 + GAIN block combination, even rail to rail (if it were designed, a new main PCB with a new PSU), since the AD8009 max. is 3.8-3.9V, and optionally xfrmr.

Not knowing the details of how these instruments 10MHz input reference circuitry works, but coming from experience one would expect a high frequency RF filter with a termination and overload/voltage protection followed by a limiting amplifier, so the internal instrument 10MHz "signal" eventually becomes a squarewave. This doesn't mean that the input waveform can be "sloppy" and have lots of perturbations and noise since these will get squared up inside the instrument and could cause excessive edge jitter. The reference input Phase Noise gets converted into edge jitter by the limiting amplifier in the same way.


Best

For example, we use 10 MHz not only for instruments & lab env., so buffering is not the only factor that matters.
In most cases, we use it for audio clocks and for synchronizing FM radio transmitters.


https://en.wikipedia.org/wiki/Word_clock
https://download.thimeo.com/MicroMPX_manual.pdf

At the beginning of the thread, I mentioned that we purchased this GPSDO for SFN synchronization for FM radio transmitters. I also wanted to send it back because, like yours, ours couldn't find any satellites, but Ali refunded the price and we kept two of them. 

[bingo600]

I will make my changes public if they are useful to you. They are based on the latest version of fredo_ (v0.1.13).
Please let me know if these changes work on your units.

Eric

Please make the changes available.

If you don't know how to make a pull request .. I don't
You could run a : git diff
And attach the output here as a zip file

I'm sure @fredo_ could work wit that.

[Eric-H]

I,ve sent my changes to fredo_. He wil review them.

[fredo_]

Please make the changes available.

Yes, Eric already PM'd me his changes.
I made a quick review and we agreed that his algorithm should be something configurable as it may not suite ISOTemp OCXOs.
It would be interesting to have feedback from other users on his firmware.

[bingo600]

Please make the changes available.

Yes, Eric already PM'd me his changes.
I made a quick review and we agreed that his algorithm should be something configurable as it may not suite ISOTemp OCXOs.
It would be interesting to have feedback from other users on his firmware.

I have an IsoTemp , so it seems like i'll pass

[mawyatt]

Udate, the 3rd unit was just received. This one "seems" to behave better than the other two, altho we haven't loaded the latest firmware. The waveforms around U5 are better than the other two.

Took some DSO images of the U5 Output (Pin 6) and VCC (pin 7) with the Output connected to a long (6 feet) RG58 Cable to a AWG Counter (SDG2000X+) with a cheap 50Ω Thur connected on the AWG Rear Counter Input. Power from cheap 12V 1A SMPS.

Then we did some crude modifications, removed C9 to reduce U5 gain to AC and DC 1X. Then changed the Input Bias Resistors R5 and R6 from 10k to 1k to lower input impedance to U5.  Added a series R of 180Ω to C11 using same C11 pads but Totem Pole the series RC network to reduce OCXO input level to U5. Used the 0.1uF from C9 to help decouple U5 closer to the package.

BTW should mention the DSO (SPS3104X HD) traces were captured with DSO 10X Probe (SP2030A) with short spring wire ground, not using the long alligator clip ground wire.

Judge for yourself if these crude modifications are worthwhile!!

Best

[bingo600]

Udate, the 3rd unit was just received. This one "seems" to behave better than the other two, altho we haven't loaded the latest firmware. The waveforms around U5 are better than the other two.

Did you get an IsoTmp OCXO or "the square one"  ?

Was the u5 output "unloaded"  or 50R loaded ?

[mawyatt]

No, same OX256B-T-LU-V-10M as before.

Best

[Eric-H]

Please make the changes available.

Yes, Eric already PM'd me his changes.
I made a quick review and we agreed that his algorithm should be something configurable as it may not suite ISOTemp OCXOs.
It would be interesting to have feedback from other users on his firmware.

I have an IsoTemp , so it seems like i'll pass

I've tuned the algorithm for the  OX256B-T-LU-V-10M ocxo I have. It might work for the ISOTemp but I could not test that. Maybe someone with the ISOTemp can check my version of the FW. If needed the FW can be updated to add a setting for the timeconstant for the algorithm so you can choose the right setting depending on the ocxo.

[mawyatt]

Was the u5 output "unloaded"  or 50R loaded ?

Output was a 6 foot RG58 Cable to a Counter (High Z) with a 50Ω thru terminating at counter.

Edit: Here's what GPSDO Output looks like with short (500mm) RG178 cable directly into DSO (SDS3104X HD) at full bandwidth (>1GHz) and internal 50Ω DSO termination with the mentioned modifications.

For our meager needs this is a "keeper waveform", and we'll wait for the firmware folks to hone things before we attempt any firmware upgrades. We are cautious with firmware as this isn't an area we are totally comfortable with

Anyway, things are looking good from our perspective now

Best 

[Kurt_!]

Eric-H
I also have the OX256B-T-LU-V-10M in there...