Episode 1020 I get some oscillators off eBay and do a shoot out. I look at drift after 24 hrs 1. OSC5A2B02 oven 2. ISOTEMP 0910 temperature compensated 3. ILSI temp compensated Be a Patron: / imsaiguy
All those ovens need to be inside thermal housings to protect against temperature changes and drafts. That will improve the performance by an order of magnitude. The third (smallest) oscillator appears to be running an internal phase locked loop for frequency stability and correction. Suitably insulated against outside temperature influences, it should stay locked up rock solid, only jumping a single cycle when the error is too great for the loop to handle. Older crystal standards where the crystal has aged can actually prove to be much more stable even if the frequency is off by a couple of cycles per second. It all comes down to whether you are looking for stability or absolute measurement accuracy.
13 seconds to drift one cycle at 10 MHz, means that the Rubidium emitted 130M cycles (13s x 10MHz) while the largest TCXO (on the left) was off by Qty 1 cycle. 1/130M = 7.7 ppb. So it's drifted by about 0.077 Hz in 24 hours, in suboptimal conditions (garage, not in a chassis). I'd give it an A+ for effort.
Hi, in my Racal 1992, I change the inside 10Mhz by the small square oscillator you show in the video, and the result is better. Just I have to wait about 2 minutes to stabilised. Also changed all the switches ans one of the relay. HPIB is also fine working, just a small jumper to configure correctly. Thanks a lot for the video, very interesting. Juan from France
I have a few of those OCXOs (the square can type) which I picked from ebay. I plan on using them - and since I don't have my own rubidium standard, I plan on using the government's; in other words, build a GPSDO. :) Though, if you are OK with another frequency reference instead of 10MHz (like 6, 8 or 12 MHz) you can get a very stable and almost jitter-free signal directly out of some GPS modules (like Neo8 or newer) by configuring a divider. You can get 10MHz too but it will have noticeable jitter. That's because the internal reference frequency is 48MHz and that does not divide exactly by 10MHz. Integer divisors will get you very low jitters.
I replaced the oven in my hp 53132a counter & 3 years latter it's drifted about 5 mhz, yes milli! As checked against a gpsdo. Quite impressive I think considering its been power cycled a few times due to power cuts
I wonder how much is due to the crystal drifting and how much due to the pots. Also, maybe a tupperware on top would help to stabilize them against small air currents (such as waving your hand arround).
Interesting comparison. Thinking of getting the OSC5A2BO2 to replace my standard drfty timebase in my Racal Dana 1991...btw, you have got the part number mixed up in the text description.
Considering grabbing SiTime MEMS OCXO, specifically SiT5721, for the reference MCXO (microcomputer controlled crystal oscillator) source for the accurate 10 MHz clock that I would obviously need to be able to test oscilloscopes, something like that. Of course that digitally programmable OCXO is very expensive. Still kinda cheaper than brand new Rubidium atomic clock source though.
ike schrodingers cat, for you make sure your measurements are exact you have to measure the device 24hrs, connecting and disconnecting the oscilloscope changes the circuit even if only slightly.
How long were they powered on for before testing? I would have thought a more precise test would be power on for 24hrs to become thermally stable then make adjustments and re test 24hrs later. Many high end test equipment require at least an hour powered on before use to stabilise and some calibration equipment is recommended to *not turn off* to keep the entire instrument at its thermally stable operating condition. Also, the test bed should be in a somewhat insulated container to minimise temperature variations from day to night...unless of course your garage is nicely insulated, heated above ambient and draught free :)
On the 1st Osc you used a voltage ref, but on the 2nd you did not. Should that not have also had a voltage ref? Maybe use the same ref? I would think they would both be subject to the drift of the voltage feeding the adjustment pot. I have a standard I built years ago that has a voltage ref, but I installed it in a insulated box with the ovened Osc. I used WWV to adjust it as it was all I had. It was very stable after 1 hour, and I left if run 24x7 in the lab. I have compared it to the two GPS units I now have, and its still a great standard. Even my two GPS units do to agree all the time and drift a bit. What's the old saying? A person with more than 1 watch never knows what time it really is!
What happens to the output of the OCXO if there is no 50 Ohm load? Also, how did you determine that it needed a 50 Ohm load? The only datasheet available on the internet does not provide any sample circuits involving this OCXO. I've used this OCXO as a reference without adding a 50 Ohm resistor at the output, and it seemed to work just fine. Am I missing something? Thanks.
Where can through "analog proto board" that you use be purchased? I have acquired much knowledge from your posts and experiment with your circuits and reviews! Rich
That would require access to gps signals and use of antennas etc,the rubidium standard is a self contained reference, as are other atomic clocks. The gps time signals are derived from such static time sources such as rubidium/atomic standards.
Yesterday I made an eBay order to queen*s land for the osc5A2B02 (largest of your three test oscillators). I asked the seller for the pin out but the seller simply referred me to his spec sheet (English annotated Chinese specs). Upon reviewing your video a second time, I see the pin out that I need. Thanks. Do you have a good English spec sheet? Thanks
All those ovens need to be inside thermal housings to protect against temperature changes and drafts. That will improve the performance by an order of magnitude. The third (smallest) oscillator appears to be running an internal phase locked loop for frequency stability and correction. Suitably insulated against outside temperature influences, it should stay locked up rock solid, only jumping a single cycle when the error is too great for the loop to handle. Older crystal standards where the crystal has aged can actually prove to be much more stable even if the frequency is off by a couple of cycles per second. It all comes down to whether you are looking for stability or absolute measurement accuracy.
13 seconds to drift one cycle at 10 MHz, means that the Rubidium emitted 130M cycles (13s x 10MHz) while the largest TCXO (on the left) was off by Qty 1 cycle.
1/130M = 7.7 ppb. So it's drifted by about 0.077 Hz in 24 hours, in suboptimal conditions (garage, not in a chassis).
I'd give it an A+ for effort.
Hi, in my Racal 1992, I change the inside 10Mhz by the small square oscillator you show in the video, and the result is better.
Just I have to wait about 2 minutes to stabilised.
Also changed all the switches ans one of the relay.
HPIB is also fine working, just a small jumper to configure correctly.
Thanks a lot for the video, very interesting.
Juan from France
I have a few of those OCXOs (the square can type) which I picked from ebay. I plan on using them - and since I don't have my own rubidium standard, I plan on using the government's; in other words, build a GPSDO. :)
Though, if you are OK with another frequency reference instead of 10MHz (like 6, 8 or 12 MHz) you can get a very stable and almost jitter-free signal directly out of some GPS modules (like Neo8 or newer) by configuring a divider. You can get 10MHz too but it will have noticeable jitter. That's because the internal reference frequency is 48MHz and that does not divide exactly by 10MHz. Integer divisors will get you very low jitters.
I replaced the oven in my hp 53132a counter & 3 years latter it's drifted about 5 mhz, yes milli! As checked against a gpsdo. Quite impressive I think considering its been power cycled a few times due to power cuts
I wonder how much is due to the crystal drifting and how much due to the pots. Also, maybe a tupperware on top would help to stabilize them against small air currents (such as waving your hand arround).
Interesting comparison. Thinking of getting the OSC5A2BO2 to replace my standard drfty timebase in my Racal Dana 1991...btw, you have got the part number mixed up in the text description.
Well, if someone wants to be more precise than a Swiss watchmaker, that's impossible.
Bulova Precisionist watches are pretty neat. The 'Time & Frequency' crowd might like them.
Considering grabbing SiTime MEMS OCXO, specifically SiT5721, for the reference MCXO (microcomputer controlled crystal oscillator) source for the accurate 10 MHz clock that I would obviously need to be able to test oscilloscopes, something like that. Of course that digitally programmable OCXO is very expensive. Still kinda cheaper than brand new Rubidium atomic clock source though.
Hello IMSAI can you do a video on ferrite loopstick antenna (used in AM reception) with nanovna and tinysa...would be interesting
ike schrodingers cat, for you make sure your measurements are exact you have to measure the device 24hrs, connecting and disconnecting the oscilloscope changes the circuit even if only slightly.
How long were they powered on for before testing? I would have thought a more precise test would be power on for 24hrs to become thermally stable then make adjustments and re test 24hrs later. Many high end test equipment require at least an hour powered on before use to stabilise and some calibration equipment is recommended to *not turn off* to keep the entire instrument at its thermally stable operating condition. Also, the test bed should be in a somewhat insulated container to minimise temperature variations from day to night...unless of course your garage is nicely insulated, heated above ambient and draught free :)
they were temp stable before, then I did measure 24hrs later (did you watch the video?)
U'r the clock man!! they do make mini/micro crystals. Mom loves white castles! that's a cheese burger!
On the 1st Osc you used a voltage ref, but on the 2nd you did not. Should that not have also had a voltage ref? Maybe use the same ref? I would think they would both be subject to the drift of the voltage feeding the adjustment pot. I have a standard I built years ago that has a voltage ref, but I installed it in a insulated box with the ovened Osc. I used WWV to adjust it as it was all I had. It was very stable after 1 hour, and I left if run 24x7 in the lab. I have compared it to the two GPS units I now have, and its still a great standard. Even my two GPS units do to agree all the time and drift a bit. What's the old saying? A person with more than 1 watch never knows what time it really is!
yes that would make it better. I was trying to show a range of solutions.
How do you know the TL431 isn't responsible for (some of) the first one's drift?
What about using a PLL Synthesizer chip to produce the control voltage and allow you to adjust the output of the VXO to exactly 10,000,000hz
One tenth of a hertz in 10Mhz - one part in 10^8 - still pretty good?
very good
What happens to the output of the OCXO if there is no 50 Ohm load? Also, how did you determine that it needed a 50 Ohm load? The only datasheet available on the internet does not provide any sample circuits involving this OCXO. I've used this OCXO as a reference without adding a 50 Ohm resistor at the output, and it seemed to work just fine. Am I missing something? Thanks.
can be used either way
Thanks for the great video! Love the 10 MHz stuff.
Where can through "analog proto board" that you use be purchased?
I have acquired much knowledge from your posts and experiment with your circuits and reviews!
Rich
www.pcbway.com/project/shareproject/DIY_Protoboard_Analog_Circuits.html
Depending on your space (and budget) it may be better to get a GPS based master clock, and use GPS Disciplined Oscillators.
That would require access to gps signals and use of antennas etc,the rubidium standard is a self contained reference, as are other atomic clocks. The gps time signals are derived from such static time sources such as rubidium/atomic standards.
Can you give your source for the green prototype board that you used on the oscillators? I’ve search eBay without any luck. Thanks
www.pcbway.com/project/shareproject/DIY_Protoboard_Analog_Circuits.html
Yesterday I made an eBay order to queen*s land for the osc5A2B02 (largest of your three test oscillators). I asked the seller for the pin out but the seller simply referred me to his spec sheet (English annotated Chinese specs). Upon reviewing your video a second time, I see the pin out that I need. Thanks. Do you have a good English spec sheet? Thanks
no, only the Chinese one you can find by searching.
dl6gl.de/media/files/ocxo-cti-osc5a2b02-datasheet.pdf
I picked up a GPS disciplined 10MHz oscillator which seems really accurate, but I'd need something more accurate to verify!
you might be interested: th-cam.com/video/p-NUnCnOHHI/w-d-xo.html
Shouldn't the output be a nice sine wave? This looks more like a square wave.
there are different types, some have sinewave output, most have square.
@@IMSAIGuy I see, thank you. Can I ask how does one calculate the output power? I can't seem to find it in datasheets
@@thebloodychoppers voltage and current, if you intend on using these in a 50 ohm system. they might not be able to drive that
@@IMSAIGuy what do you mean by not being able to drive a 50 ohm system
@@thebloodychoppers these clocks are for digital systems not RF
Merci Guy, what is your rubidium oscillator please?
FE-5680A
@@IMSAIGuy Okay! Thank you for the answer. Interesting. Have you carried out a filter to have such a clean sinusoidal signal?
@@bobdoritique7347 no
good morning
1/12 Hz in 10 MHz is pretty small drift. Of course, it was only one day... but that is eight significant digits.
It's always fun to chase zeros but you are connect, not many applications care about 1/12 of a hertz