An AT-cut crystal will have a fundamental frequency of 1664 divided by the thickness in micrometres. I guess that one will be a 48 MHz 3rd overtone, so the fundamental frequency will be around 16 MHz and the crystal will be around 100 micrometers thick. The circuit needed for that is very simple, just two inverters and a resistor. The IC may include dividers for lower frequencies but they would not be used for 48 MHz. The crystal's silver electrodes will have been plated on in vacuum by evaporating. It's not sputtering, as no voltage is used to propel the silver atoms, only heat. The heat comes from a low voltage filament, and there is no high voltage. After that, the crystal is attached to the oscillator with the silver-loaded epoxy. A flexible epoxy is used to stop thermal stresses on the crystal. In the video, Tim was able to lift the crystal a long way before it came off, because of the flexibility of the epoxy. The adjustment shape is the faint silver rectangle that isn't very well aligned in this example. To adjust the oscillator, it will have been operated in vacuum, complete except for the lid. More silver would have added while the frequency was measured. As the correct frequency was approached, the silver would have been stopped, probably by moving the oscillator out of the way as that would be quicker than letting the filament with molten silver cool down, and also because there were probably many oscillators in the one vacuum chamber, so when this one was adjusted, the next one would be moved into place. The ceramic package is to prevent out-gassing. A tiny package like that, or even many of them, won't stop a good vacuum being achieved in the production equipment. The reason that outgassing would be a problem is that the gasses would react with the silver and change the mass, which would affect the frequency the same as the adjustment silver does. I think that the lid is resistance welded onto the package. Rollers are rolled along the edges while passing a current between them, causing the bottom of the lid to weld to the top of the package.
Thanks for those details, I've now got a better die photo and found the datasheet for the die, more details here: ruclips.net/user/postUgkx-P1Bvt0LztyGCv-cCmT4MA4Oa2P8HjAq It does indeed use the third overtone and has other mask selectable dividers on board for different frequency versions from 22 to 70 MHz along with CMOS or TTL output and output loads versions.
I have hardly any understanding of how this stuff works. I can identify basic components and explain how they work. That’s about it. But I am fascinated with how much detail goes into these components and how complex it is.
Well, thank you for your efforts in tearing one of these apart and filming the interior. I suspected pretty much what we saw but it is alway interesting to see the exact implementation and placement of the components. 73...
Of course there is a video out there that shows the inside of these. I was just struck by the thought.. what is in there and this video was awesome! Well done mate!
My old and probably already obsolete understanding has been that any crystals more than some 50 MHz would actually be operated at third or higher harmonic overtone. Then it would require a filter of some sort to pick the desired harmonic frequency and attenuate the base frequency. If you in any case do laser trimming on the crystal, why not trim it to the 48 MHz directly, instead of picking some odd harmonic of a lower frequency and then dividing it with another odd number for the final output? I guess, if getting motivated, I could get one of those oscillators, remove the top and then operate it in the open. Then I could use a small field probe and my spectrum analyzer to probe what (other) frequencies there are. Could be fun, but do I need the info?
I’ve looked into buying a metallurgical microscope, but the good ones are well over the budget for this channel, I’ll buy one eventually but probably not anytime soon.
Hi! The tech behind Crystal Oscillators is pretty new to me, so I was hoping to get a couple of basic questions answered, for which I was I was struggling to find answers online. 1) How exactly are the fixed output frequencies of an SPXO set? We have many 'standard frequencies' such as 10,12,16,20 MHZ, etc. Is it only due to the way the crystal is cut, or does it also depend on the other supporting circuitry within the Oscillator? 2) How does the circuitry for a programmable crystal oscillator (Such as Epson's) differ from a fixed frequency oscillator? I know these involve 'blank' oscillators that can be programmed to a desired frequency using a handler. Does it contain any circuitry such as programmable non-volatile memory? Thanks in advance!
I don't work in the crystal oscillator industry so I can't give you a very good answer but I can tell you the crystal oscillators field is very big and gets pretty complex. In the case of the most simple SPXOs they just use a feedback inverter to get the crystal oscillating, the cut of the crystal (eg. AT cut and 0.1 mm thick), the region in which the crystal is oscillator (eg. third overtone) and load capacitance are the main parameters that set the frequency. The numbers used for frequency are typically used because they are either a "preferred number" often E12, have some mathematically important property like 32.768 kHz for time keeping or from a standard like 5.6448 MHz often in CD players. Epon's SG-8101 and SG-9101 series programmable crystal oscillators use a fixed frequency crystal that is then divided by dividing circuity or multiplied by a PLL (phase locked loop) depending on what frequency is requested, those are much more complex circuit as thus much more expensive but also have their own tradeoffs in order to accomplish this (eg. PLL phase noise/jitter) which make SPXOs more preferred in most applications. They appear to have OTP (one time programmable) memory to keep settings.
@@WizardTim Great! Thank you so much for your answer! It definitely helped! I was having a look at the Epson SG8101-CG series and that made me wonder how programming was possible. I also really liked this teardown. It's actually the first time I'm looking inside of a crystal Oscillator, so pretty cool!
I guess they might use a crystal of a complete different frequency - and the target frequency is derived digitally through synthesis (fractional divider). I´m not sure if that´s correct, but this would allow to use one standard size crystal and avoid mechanical calibration.
Yea, I suspect they use an 80 MHz crystal with a divide by 2 (current similar models from KDS list f0 as 80 MHz). However, if this were true, I would expect the quartz strip to be less than 0.05 mm thick but it’s 0.10 mm thick alluding to it being a lower frequency (around 16 to 25 MHz) however I’m not sure if the quartz frequency constant works the same way as conventional round disks as with strip resonators. To get 40 MHz from a lower frequency they would need a PLLs which are much more complex and thus a lot more expensive to implement than a simple divider and they tend to have limitations especially in phase jitter and start up performance so I would assume KDS would avoid using one as much as possible, I’m wishing I powered and probed it without the lid, would have made it a lot easier.
are you whispering so you dont wake the crystal demon up? i can hear you but it sounds like you arent putting in much effort to voice your words as if your trying not to alert the police who are searching for dead crystals while you film the dissection video under a bed.
I was being quiet so my temperature compensated and ovenized crystals couldn’t hear me giving more attention to some cheap oscillator, didn’t want them getting jealous and drifting out of spite. (Reality: I just sound boring IRL, the cheap microphone and DIY XLR preamp didn’t do me any favours either, the newer videos are somewhat better)
@@WizardTim i like your quiet naration style. Its so comforting and soothing. However, (sorry if grossed out), the saliva noise is quite distracting. Great thing its sorted out in your latest video
Yea I noticed it as well and was grossed out by it so I tried to reduce it in the newer videos. I have a new found respect for voice actors after realising how difficult it is just to record a simple consistent and clean voiceover.
Good demonstration. What's the matter with your voice? Can you speak more loud and clearly please? For moments I can barely understand what are you saying without having to rewind and raise the volume of my speakers. Thanks!
An AT-cut crystal will have a fundamental frequency of 1664 divided by the thickness in micrometres. I guess that one will be a 48 MHz 3rd overtone, so the fundamental frequency will be around 16 MHz and the crystal will be around 100 micrometers thick.
The circuit needed for that is very simple, just two inverters and a resistor. The IC may include dividers for lower frequencies but they would not be used for 48 MHz.
The crystal's silver electrodes will have been plated on in vacuum by evaporating. It's not sputtering, as no voltage is used to propel the silver atoms, only heat. The heat comes from a low voltage filament, and there is no high voltage. After that, the crystal is attached to the oscillator with the silver-loaded epoxy. A flexible epoxy is used to stop thermal stresses on the crystal. In the video, Tim was able to lift the crystal a long way before it came off, because of the flexibility of the epoxy.
The adjustment shape is the faint silver rectangle that isn't very well aligned in this example. To adjust the oscillator, it will have been operated in vacuum, complete except for the lid. More silver would have added while the frequency was measured. As the correct frequency was approached, the silver would have been stopped, probably by moving the oscillator out of the way as that would be quicker than letting the filament with molten silver cool down, and also because there were probably many oscillators in the one vacuum chamber, so when this one was adjusted, the next one would be moved into place.
The ceramic package is to prevent out-gassing. A tiny package like that, or even many of them, won't stop a good vacuum being achieved in the production equipment. The reason that outgassing would be a problem is that the gasses would react with the silver and change the mass, which would affect the frequency the same as the adjustment silver does.
I think that the lid is resistance welded onto the package. Rollers are rolled along the edges while passing a current between them, causing the bottom of the lid to weld to the top of the package.
Thanks for those details, I've now got a better die photo and found the datasheet for the die, more details here: ruclips.net/user/postUgkx-P1Bvt0LztyGCv-cCmT4MA4Oa2P8HjAq
It does indeed use the third overtone and has other mask selectable dividers on board for different frequency versions from 22 to 70 MHz along with CMOS or TTL output and output loads versions.
This video deserves more appreciation.
Very true
I have hardly any understanding of how this stuff works. I can identify basic components and explain how they work. That’s about it. But I am fascinated with how much detail goes into these components and how complex it is.
Such a nice looking component.. great video 📹
Well, thank you for your efforts in tearing one of these apart and filming the interior.
I suspected pretty much what we saw but it is alway interesting to see the exact implementation and placement of the components.
73...
Of course there is a video out there that shows the inside of these. I was just struck by the thought.. what is in there and this video was awesome! Well done mate!
Thank you. I do not understand what is going on at the PCB but it is very interesting.
@Jokreher
3 years ago
This video deserves more appreciation. ++++++++++++++++++
My old and probably already obsolete understanding has been that any crystals more than some 50 MHz would actually be operated at third or higher harmonic overtone. Then it would require a filter of some sort to pick the desired harmonic frequency and attenuate the base frequency. If you in any case do laser trimming on the crystal, why not trim it to the 48 MHz directly, instead of picking some odd harmonic of a lower frequency and then dividing it with another odd number for the final output? I guess, if getting motivated, I could get one of those oscillators, remove the top and then operate it in the open. Then I could use a small field probe and my spectrum analyzer to probe what (other) frequencies there are. Could be fun, but do I need the info?
Nice, now I know why I pay almost $1 for these in small quantities
Thank you Tim, very cool!
Glad you liked it!
Very fun video to watch, thank you
Hi Tim, ur video provides excellent info, can u plz use advanced lens to b able to zoom in more!!!!!
I’ve looked into buying a metallurgical microscope, but the good ones are well over the budget for this channel, I’ll buy one eventually but probably not anytime soon.
Hi! The tech behind Crystal Oscillators is pretty new to me, so I was hoping to get a couple of basic questions answered, for which I was I was struggling to find answers online.
1) How exactly are the fixed output frequencies of an SPXO set? We have many 'standard frequencies' such as 10,12,16,20 MHZ, etc. Is it only due to the way the crystal is cut, or does it also depend on the other supporting circuitry within the Oscillator?
2) How does the circuitry for a programmable crystal oscillator (Such as Epson's) differ from a fixed frequency oscillator? I know these involve 'blank' oscillators that can be programmed to a desired frequency using a handler. Does it contain any circuitry such as programmable non-volatile memory?
Thanks in advance!
I don't work in the crystal oscillator industry so I can't give you a very good answer but I can tell you the crystal oscillators field is very big and gets pretty complex.
In the case of the most simple SPXOs they just use a feedback inverter to get the crystal oscillating, the cut of the crystal (eg. AT cut and 0.1 mm thick), the region in which the crystal is oscillator (eg. third overtone) and load capacitance are the main parameters that set the frequency. The numbers used for frequency are typically used because they are either a "preferred number" often E12, have some mathematically important property like 32.768 kHz for time keeping or from a standard like 5.6448 MHz often in CD players.
Epon's SG-8101 and SG-9101 series programmable crystal oscillators use a fixed frequency crystal that is then divided by dividing circuity or multiplied by a PLL (phase locked loop) depending on what frequency is requested, those are much more complex circuit as thus much more expensive but also have their own tradeoffs in order to accomplish this (eg. PLL phase noise/jitter) which make SPXOs more preferred in most applications. They appear to have OTP (one time programmable) memory to keep settings.
@@WizardTim Great! Thank you so much for your answer! It definitely helped! I was having a look at the Epson SG8101-CG series and that made me wonder how programming was possible.
I also really liked this teardown. It's actually the first time I'm looking inside of a crystal Oscillator, so pretty cool!
Thank you
I guess they might use a crystal of a complete different frequency - and the target frequency is derived digitally through synthesis (fractional divider). I´m not sure if that´s correct, but this would allow to use one standard size crystal and avoid mechanical calibration.
Yea, I suspect they use an 80 MHz crystal with a divide by 2 (current similar models from KDS list f0 as 80 MHz). However, if this were true, I would expect the quartz strip to be less than 0.05 mm thick but it’s 0.10 mm thick alluding to it being a lower frequency (around 16 to 25 MHz) however I’m not sure if the quartz frequency constant works the same way as conventional round disks as with strip resonators.
To get 40 MHz from a lower frequency they would need a PLLs which are much more complex and thus a lot more expensive to implement than a simple divider and they tend to have limitations especially in phase jitter and start up performance so I would assume KDS would avoid using one as much as possible, I’m wishing I powered and probed it without the lid, would have made it a lot easier.
Thanks
I love it good
4:44 that small crystal looks kinda... suspicious.
(im sorry)
You should be sorry... But I assure you this crystal is hermetically sealed, it cannot possibly vent.
are you whispering so you dont wake the crystal demon up?
i can hear you but it sounds like you arent putting in much effort to voice your words as if your trying not to alert the police who are searching for dead crystals while you film the dissection video under a bed.
I was being quiet so my temperature compensated and ovenized crystals couldn’t hear me giving more attention to some cheap oscillator, didn’t want them getting jealous and drifting out of spite.
(Reality: I just sound boring IRL, the cheap microphone and DIY XLR preamp didn’t do me any favours either, the newer videos are somewhat better)
@@WizardTim i like your quiet naration style. Its so comforting and soothing. However, (sorry if grossed out), the saliva noise is quite distracting. Great thing its sorted out in your latest video
Yea I noticed it as well and was grossed out by it so I tried to reduce it in the newer videos. I have a new found respect for voice actors after realising how difficult it is just to record a simple consistent and clean voiceover.
Good demonstration. What's the matter with your voice? Can you speak more loud and clearly please? For moments I can barely understand what are you saying without having to rewind and raise the volume of my speakers. Thanks!