Cheap chinese 10 MHz OCXO Frequency Reference Modules

[Transradio]

On various e-commerce platforms chinese 10 MHz Frequency Reference Modules with OCXO have recently been offered at a favorable price. These are mostly equipped with used oscillators that have been removed from decommissioned GSM/UMTS base stations. These surplus OCXO are in in most cases OSC5A2BO from chinese manufacturer CTI (a data sheet of a comparable OXCO from CTI is attached). These Voltage Controlled OCXO have an input for a tuning voltage with a span of 0 to 4V. With the help of this tuning voltage the Quartz Crystal can be adjusted by a few Hertz to the exact target frequency to compensate manufacturing scatter or the effects of aging.

I ordered one of those black modules, branded as "sz-plaza.com Rev:0.0.2". Frequency adjustment is done by a 25 turn Cermet potentiometer. Unfortunetaly the OXCO could not be adjusted to 10MHz. The adjustment range of the tuning voltage was not enough to get the thing on frequency. To find a possible fix i reverse engineered the module.
It turned out to be simple to extend the limited range of the tuning voltage. It needs just a shortcut of one resistor and removing of another resistor. Please check: only the inner connector of the SMA Connectors of my module were properly soldered to the PCB, the ground connector needed resoldering. Despite the typical chines quality issues, IMHO this modules are a good bargain, a new OCXO with the same performance would cost a lot more.



The modules schematics with the proposed modifications is attached. The reverse engineering was done by best efforts, there is no guarantee for correctness.

[CaptDon]

So it seems like it was 'returned from the field' as defective and resold as tested/working and yet could not be adjusted to the stated frequency. I guess if you paid over $10 with no assurance the unit could meet original specs you paid to much?

[Transradio]

The OCXO is of high quality and works perfect, the module around the OCXO has deficiencies with regards to the voltage range of the tuning voltage. You need to know that in China and elsewhere many thousands of GSM/UMTS Basestations are going to be replaced by new 5G Stuff. The OCXO are taken out of those. Good example for recycling of electronic goods.

[807]

I'm waiting for one of these to arrive.

What did you use to measure the 10MHz? How far off frequency was it before the mod?

[EEEnthusiast]

For a 10MHz reference, you could buy a cheap GPS stabilized source. You don't need to adjust it for the drift, unless they shutdown the civilian GPS signals.

[strawberry]

I'm waiting for one of these to arrive.

What did you use to measure the 10MHz? How far off frequency was it before the mod?

two channel oscilloscope and 10MHz reference
both waveform wont drift if frequency is equal

[807]

I'm waiting for one of these to arrive.

What did you use to measure the 10MHz? How far off frequency was it before the mod?

two channel oscilloscope and 10MHz reference
both waveform wont drift if frequency is equal

Yep. That's how I used to do it.
Then I got a Racal Dana 1991 frequency counter with an external 10MHz reference from a Leo Bodnar GPSDO. Can read down to 1mHz using that combo.

[Transradio]

What did you use to measure the 10MHz?

I used a tinyPFA against a good 10 MHz GPSDOCXO as Reference. A great tool:
https://www.tinydevices.org/wiki/pmwiki.php?n=TinyPFA.Homepage

How far off frequency was it before the mod?

To to adjust to the target frequency of exactly 10 MHz a tuning voltage of 1.4V is required for my specific OCXO. The adjustment range of the Module's Potentiometer without Modification reaches only from 0V to 0,8V. So i was approximately 2Hz off. Because of aging, the tuning voltage to reach 10 MHz is a bit different for each OCXO specimen.

For information:
the control voltage range of a new CTI OCXO ist 2V ± 2V ( 0...4V)
that voltage range corresponds to a total frequency variation of approximately ±10Hz max.

[GigaJoe]

had some play with ocxo itself,  not the board, as it doesn't make sense to buy it.
this is AC type of quartz cut, technically more deviation from manufacturer due to larger residue stress.
i use lt1021 in plastic as voltage source for compensation, thermally glued to ocxo case, and isolate everything by heavy styrofoam insulation box
able to adjust to -11 precision,   drift per year approx:   -5-10 degree, quite sufficient as reference for  most stuff.

one ocxo die in a few month - the heater went off.

[807]

I bought the same module as the OP, as it was the quickest way to obtain one in UK. There's 1 UK seller on ebay for just under £19. Could have got the OCXO itself cheaper from China, but then I don't want to wait.

I want to use it to replace the internal 10MHz on my frequency counter, using it's 5v supply. I'll see what voltage I need to trim it to frequency before deciding on which precision voltage reference to use. Need to replace the LT1009 with something tighter, 5ppm or less.

[Detlev]

I had also ordered two such OCXO. Yesterday I had one matched the OCXO to the frequency counter (synced to a Leo Bodnar GPSDO).

Today I ran a measurement over 5 hours (18000 measurements with a gate time of 1s). Before that I let the OCXO warm up for 3 hours. The oscillator was on the workbench and the average temperature was 20 degrees Celsius.

In those 5 hours it has a P-P deviation of 16.09 mHz. The absolute deviation was 21.68mHz. For a 20€ OCXO I think that's OK. Now it comes in a nice case and gets an output driver.

Best regards
Detlev

[807]

Been playing with my module past couple of days. Like you Transradio, I couldn't get it to trim to exactly 10MHz. As it came, it settled down very quickly to 10.00000313MHz (+3.13Hz high). The range of the preset set the frequency to +1.80Hz to +3.49Hz. Looking at the Vref pin on the OCXO, this corresponded to a range of 1.940v to 2.1085v. I was puzzled as to why the voltage didn't drop to 0v when the preset was fully CW as, according to your schematic, it should have grounded the Vref voltage. It turns out that the tuning arrangement on my module is different to yours. Rather than needing an increase in Vref voltage, I need to decrease mine.

I've attached the schematic of my module. I was unsure about R2 & R6, as I thought they were marked 688, which is 6.8GOhms! Then I realised that the marking was actually 68B, which is the EIA-96 code for 4.99kOhms.

Within 5 minutes of switch on, it was within 180mHz of it's final settled frequency (within around 30 minutes). I havent done extensive measurements, but 5 hours later it hadn't changed.

[RCL]

Gentle Readers:

I felt guilty just sitting on the side-lines and letting you solve my problems.  So here is my contribution:

     1.945 volts at the V/C pin = 10,000,005.8 Hz
     2.114 Volts at the V/C pin = 10,000,007.4 Hz

     The temperature of the little TO-220 voltage regulator (12 Vdc applied).  = 105° F

  My calculations indicate that at 1.332 Volts we can achieve 10 MHz.

[fourfathom]

Been playing with my module past couple of days. Like you Transradio, I couldn't get it to trim to exactly 10MHz. As it came, it settled down very quickly to 10.00000313MHz (+3.13Hz high). The range of the preset set the frequency to +1.80Hz to +3.49Hz. Looking at the Vref pin on the OCXO, this corresponded to a range of 1.940v to 2.1085v. I was puzzled as to why the voltage didn't drop to 0v when the preset was fully CW as, according to your schematic, it should have grounded the Vref voltage. It turns out that the tuning arrangement on my module is different to yours. Rather than needing an increase in Vref voltage, I need to decrease mine.

I've attached the schematic of my module. I was unsure about R2 & R6, as I thought they were marked 688, which is 6.8GOhms! Then I realised that the marking was actually 68B, which is the EIA-96 code for 4.99kOhms.

Within 5 minutes of switch on, it was within 180mHz of it's final settled frequency (within around 30 minutes). I havent done extensive measurements, but 5 hours later it hadn't changed.

I just got two of these boards. They match your posted schematic (thanks for that!), and they also needed a lower tuning voltage (about 1.89V for one of them).  I ended up replacing R3 (it was 36K) with a 3.3K resistor,  I didn't calculate an optimum value, I just had a 3.3K handy, but the tuning range calculates out to 0.947 to 2.125 (assuming no loading from the OCXO Vref input.)  I put the OCXO output, and a Bodnar GPSDO output on the 'scope, and it looks quite stable.

[rhb]

I'm currently playing with a pair of these with 3 more in transit and 10 bare modules.  The lattr will be tested in a fixture with an LM399 voltage reference and low temperature coefficient trimmer pots in as stable a temperature oven as I can make.  I've not needed to make any changes to the first 2.

I'm not able to get better than +/- 0.01 Hz with it in an insulated craft beer six pack filled with fiber batting. 

Next step will be a simple thermistor and pot biasing the base of a 2N3904 with a 240 ohm resistor as a heating element to form a second oven inside an insulated box.  The hard part will be finding the thermistors and a suitable trim pot in the chaos.

I also need to load tinyPFA on a dedicated nanoVNA 4H.  I'm alos using one of Leo's GPSDOs.

Have Fun!
Reg

[radiolistener]

these OCXO were in use and desoldered from some equipment. There is some risk that they were stored in bad condition or improperly desoldered (with blowtorch or by frying PCB on the fire)

[rhb]

The plane might get hit by a meteorite  on the way.

At this point I think I'm fighting the Vref on the board.  If I trim it, it's very tight for 20-30 minutes and then it's off by the same 0.012 - 0.021 Hz.  I've been through this exercise many times now.  I think the Vref shifts when it reaches thermal equilibrium.

I have some LM399 voltage references so I'm going to box one to use as the Vref for testing all the OXCO modules with TL082 followers to drive the trimmer circuits which will be thermally isolated from the OXCO.  I'll also heat each OXCO to the point the internal heater turns off so that there are minimal heat gradients within the module itself.

Reg

[wasedadoc]

these OCXO were in use and desoldered from some equipment. There is some risk that they were stored in bad condition or improperly desoldered (with blowtorch or by frying PCB on the fire)

The 5 that I have were still soldered to their PCBs.  The PCBs had been cut a few millimetres from the edges of the OCXO can.  Cuts were clean. No sign that they were crudely done with a large scissor-like tool.

[807]

these OCXO were in use and desoldered from some equipment. There is some risk that they were stored in bad condition or improperly desoldered (with blowtorch or by frying PCB on the fire)

The 5 that I have were still soldered to their PCBs.  The PCBs had been cut a few millimetres from the edges of the OCXO can.  Cuts were clean. No sign that they were crudely done with a large scissor-like tool.

I have 2 like that, the other 4 I bought were desoldered. Need to get an idea of the Vref for each of them before deciding which reference voltage chips to buy.

[rhb]

Buy a good quality reference and use a TL082 and temperature stabilized foil trimmer and similar grade  fixed resistor to trim the value for Vctl.  Then put that in a thermally controlled enclosure.

[artag]

I bought one mounted on a board (black solder resist, schematic approximately as 807's) and 5 unmounted (desoldered). I found I couldn't adjust the frequency at all (it settles to about 4Hz high in less than 2 minutes, vs a GPSDO). This seems to be because the adjustment pot - which is 200K in this example - doesn't seem to have the centre pin connected to the end, so it functions as a fixed 200k resistor,

Linking 2 pins of the pot (the two nearest the adjustment screw : the pin near the edge of the board is ground, though  I guess that would work too) got me down to a +2Hz error with the pot at 0 ohms so a resistor change was necessary. 18k in parallel with R3 got it in range though it's at the low resistance (somewhat sensitive) end of the travel.

After adjustment to 0.08 Hz high it's drifted up to .15 Hz high in 2 days but is pretty solid there. It's in a small cardboard box full of polystyrene peanuts to let the voltage reference stabilise, heated by the oven and the linear regulator, which is supplied with 10V.  However a couple of days to stabilise isn't much use for a portable instrument. This may just be settling out in the environment and after some time powered off (and probably abused mechanically). I'll have to see what it now achieves from cold.

After some thermal cycles and resets but no adjustment the frequency had increased to 10M + 0.19. However it's stuck there (about 5x10^-10 variation) ever since, despite another 12 hour cool/heat cycle, a brief interruption and 2-3 days on test. After cooling the oscillator is stable in a couple of minutes but the entire board and enclosure has to warm up before the voltage source is at temperature.

Very grateful for the  schematic. Opaque solder resist is a crime .
Interested to see what people come up with for an improved pcb and reference.

[807]

Did they base this circuit on just 1 sample of the crystal & thought the rest would be the same? Seems like they didn't allow enough adjustment to cover the spread of Vref.

I calculate that I need somewhere around 1.76v for 10MHz. Halving the values of R1 & R3 should give a range of 1.63v to 1.92v. Haven't got round to doing it yet. Those resistors are very small. Might need a digital microscope for my ageing eyes. 

[rhb]

The temp co on Vishay's *best* foil trimmer is 5 ppm/C. That will require temperature compensation of its own to get them to <0.01 ppb and a good LM399 reference.

The last 3 finished boards are all several Hz high. Haven't checked on the 10 bare module shipping lately, but hopefully they will arrive soon and I can start working out how to thermally stabilize these more tightly.

Reg

[MathWizard]

How good/bad are the reference oscillators/clocks, that are found on common computer hardware PCB's, or say like a digital cable TV box ?

[Gyro]

Once upon a time, digital TV boxes had reasonably decent 27MHz VCXOs, then the market cheapened, and skip and repeat became the standard way of maintaining broadcast synchronisation.