10Mhz GPSDO OSCILLOQUARTZ OSA OEM GPS GPSDO 10MHz 1PPS (STAR GPS Clock) tear dow

[Yrrah]

Hello,

I just received this GPSDO and it seems to work fine. I am curious and a bit suspicious so I wanted to open the unit. Just removing the 8 screws is not enough. The front and back plates (pcb material) remain secured. Do I need to loose the connectors? Any one opened this particular unit? Any hint what to do? I am running the GPSDO in test in comparison with a free running MV89A and so far on my HP5345A (A vs B) there is a 4 nsec difference, quite stable. Ok for now I would say. The spectrum however is only reasonable. I don't know what I can expect. Just some first impressions.
Please advise.
kind regards,
Y

[testpoint1]

if you need a stable reference, you can use the GPS discpined Rubidium Oscillator, if you use a GPSDO, please make sure that is from a brand factory (some small factory from China no testing condition, and do not know the tracking algorithm). and if your GPSDO is a old one (1995-2005), they just use the Motorola GPS, I am doing the research to replace them by Ublox M8T.

[Yrrah]

Yeah I know many of the alternatives. I have a rubidium. I had two HP Z3801A but both failed on the Motorola GPS receiving module. So I needed something..... Still curious: any one opened the Chinese one? Please let me know. It does work well by the way....
Y

[FransW]

Hi Harry,
I had no problems at all with 2 of them.
Maybe the print is too tight in the enclosure.
Happened to me once before with a BG7TBL distribution box.
Frans

[notfaded1]

The inside of mine is in this thread but it's not oscilloquartz:
https://www.eevblog.com/forum/testgear/bg7tbl-gpsdo-master-reference/msg2893616/#msg2893616

Bill

[FransW]

The inside of mine is in this thread but it's not oscilloquartz:
https://www.eevblog.com/forum/testgear/bg7tbl-gpsdo-master-reference/msg2893616/#msg2893616

Bill

Thanks Bill,

The problem has been solved by PM.
The exact version did matter.
Just loosen the RS232 connector.

Regards,  Frans

[mhsprang]

I jut received such a GPSDO with an Oscillquartz board inside. And boy, has this amoj1010 guy done a bad job.
The sine wave output of the unit shows an LF modulation, but only on the downgoing slope of the sine.
Also, the 10MHz output has a DC offset, so the sine goes from 0V to appr. 4V instead of being AC coupled.

The 1PPS output shows a ringing twice as high as the pulse amplitude on both edges, of appr. 14MHz. Totally unusable to drive a receiving circuit. When this 1PSS output is terminated with 50 Ohm, the ringing on the rising edge is gone, but on the falling egde, a negative ringing can be seen going from high level, spiking down to less than half the amplitude but never being higher than the high level. Once a 0V, no ringing can be seen.

So I decided to take the unit apart to see what's going on. To my surprise, amoj1010 hasn't used the sine wave output of the Oscilloquartz at all. Not even the sinewave available on the MCX connector! Instead, he took the square wave output of the Oscilloquartz board and fed it into two opamps and some filtering to create a sinewave again.

The 1PPS signal from the Oscilloquartz board is buffered by a 74LS244 and then fed to the 1PSS output through a small SMD choke. There's the source of ringing! The guy did the same with the 1PPS signal to the RS232 connector: straight from the LS244, through a choke, onto the connector. He did not even use one of the unused RS232 buffers to give the signal proper levels.

Funny enough, this guy has used 3A (!) LDO regulators to feed the 1PSS and 10MHz buffers.

I will do some more measurements so find out where the modulation and noise on the 10MHz output come from and see if I can improve this. But as-as, this is one bad GPSDO when it comes to signal quality.

Oh and the specified supply voltage on the front panel is plain wrong (12-24V). The first regulator from the power input to the Oscilloquartz board is a switcher and this outputs only 8,5V to the Oscilloquartz board if you supply the box with 12V. It definitely needs 15V from the supplied power brick. Which is unusable next to my ham radio gear as it sort of floods the 28 MHz band with noise. Very bad power brick!

To be continued...

Meindert

[FransW]

Pictures for comparison please.
We might be talking about totally different pieces of equipment.
Frans

[mhsprang]

This is the unit I have.

I have now removed the inductors on the 1PPS output and the 1PPS signal to the RS232 connector and replaced them with 100 Ohm resistors. This makes the output short-circuit proof and delivers 1V into a 50 Ohm load with just a slight ringing and overshoot and 2.5 ns risa/fall times.

In the mean time I did some more "research" on the board designed by Amoj1010. The 3A regulators I mentioned earlier, supply a total of 30 mA to the board!

Then the 10 MHz signal path: the  The Amoj1010 carrier board takes the square wave output from the Oscilloquartz board, puts this through an inverter of a 74ACT244, then feeds it into a very fast opamp via a 50 Ohm resistor. The output of the opamp goes through a 50 Ohm resistor into another 74ACT244 inverter and from there through a simple LC low pass filter to the BNC connector. It is beyond me why Amoj1010 did it this way. The Opamp is specifically meant to be a 50 Ohm line driver. Why on earth add an inverter which cannot even drive 50 Ohm.

The Oscilloquartz board has an MCX connector that delivers a very clean and stable 10 MHz sine wave. This could have been directly taken to the BNC connector or, if you must, use this nice opamp in between and connect its output to the BNC. This whole mess of using inverters only adds to the noise and it even picks up the 1PPS somewhere, changing the offset of the 10 MHz output.

[jpb]

Interesting work on this.
I have a couple of these which I did open up but only to check that they were Star 4+ boards inside.
I had noticed that the sine wave output had a dc offset but I've been doing ADEV measurements and for those they are very stable - I haven't done anything to measure the spectrum.

I think perhaps I should do some mods to mine - I'd be interested to have more details on the mods you've done.

I feed mine with a linear power supply (it didn't come with a brick anyway) and have attached it to an aluminium block to even out temperature changes.
As far as ADEV measurements are concerned they are very stable if the time constant is set to 3600 secs (the default is very short).

I guess that the high current capacity is for when the OCXO is warming up when the current draw should be higher.

[jpb]

I've done a mini-teardown on one of mine - here are the pictures:

[FransW]

Oscilloquartz specification:

[jpb]

Oscilloquartz specification:

Thanks for that.
I see in the drawing the smb connectors are horizontal  while on my board they are vertical - that may have been changed to allow the other board to piggy back more easily.

Here is the output of mine (no mods). The sine wave has a dc offset and is perhaps a bit distorted but it doesn't seem extreme.
The 1pps doesn't have a lot of ringing into 50 ohms.
 I also attach a picture of the underside of the board (for completeness, it is not very interesting).

[jpb]

The voltage regulators on the main board are 500 mA ones (LO5 78M05s). The ones on the daughter board 5V supply are up to 5A. (LM2678)
My current draw from 15V was around 0.6A at start up dropping to 0.33A on warming up so two lots of 500mA are needed. I guess the daughter board ones are just convenient.

I might decide to change to an external linear supply at some point that bypasses the daughter board to reduce noise.

[jpb]

I stuck a filter on the sine output and it does look rather better as well as getting rid of the dc offset.

[mhsprang]

Apparently there are different designs of the carrier board. Yours takes the 10Mhz from the MCX connector of the Oscilloquartz board, which delivers a nice clean sinewave. I don't know what happens further with the 10Mhz on your board but the presence of the 74244 worries me 

Regarding the regulators: apparently Amoj1010 uses what he can get his hands on. The small add on board with the two switchers receives 15V from the external supply. It generates 12V for the Oscilloquartz board and 8V for the carrier board. On the carrier, you have 5V/0,5A regulators. One feeding the 1PPS/RS232 circuits and the other feeding the 10Mhz circuit. On my board, these are 3A(!) regulators. Both of them supply no more than 15mA each. Shere overkill.

The power brick that came with my unit (15V/5A!), is too noisy and it wouldn't pass any EMC test. So I took out the small regulator board from the carrier board, added two wire  bridges from the center pin to each pin on the end (12V to the Oscilloquartz board and 8V to the circuits on the carrier board). So everything runs off 12V now and I use a small12V/1.2A wall wart too feed the unit.

I have removed the LC low pass filter at the 10MHz BNC connector to disconnect it from the 74244 and added a piece of coaxial cable to connect the BNC directly to the MCX connector on the Oscilloquartz board. I now have a clean jitter free output of 330mV rms. A bit less than I hoped for but enough to feed my HP 53151A counter.

Best,
Meindert

[jpb]

Thanks for the details. Until you reported your findings it had not occurred to me to try modifying my Star 4s.

It almost makes sense to get rid of the extra boards altogether and just supply power externally. I'm not bothered about trying to fit everything into the small case.

Though given all my other projects and the fact that I don't find time to do any of them I think it may be a while before I get around to doing anything.

I'm not sure what the extra boards really do apart from supply power and mess with the sine output from the Star board!

Presumably the RS232 RX and TX could be wired from the Star board and some sort of buffering could be supplied externally.

I have a second Star 4+ (or rather a first one, which was the first one I got and performs a bit better I think). I might do a tear down on that and see if I can spot any differences. At present I've been running it for a few months and don't want to stop it.