EEVblog #1089 - XTAL Oscillator Teardown

Comments · 401

• AN

  @andrewbaker7839 · 8 years ago (edited)

  I used to design and repair test and control equipment at a major Quartz Crystal manufacturer. This was &quot;not recently&quot;. We did everything from growing the quartz through building TCXO, OCXO, DIL oscillators, SAW devices and Filters.<br><br>Our DIL oscillators used a real thick-film modules, with the chip bonded to it, and connections made with bond wire. Partly because (in big volume) it was cheaper, but also because it allowed you much more space to &quot;do things properly&quot; - with one of those things being proper decoupling.<br><br>Dividers were quite common, partly because it is hard to make good crystals at certain frequencies, and there was also a &quot;size&quot; thing - good 1MHz crystals are quite physically big.<br><br>Did you get a sample of the gas when you opened it? Mostly they&apos;re filled with Nitrogen with a tiny bit of helium. This is part of the controlling the ageing, the helium makes it easy to check them for leaks very thoroughly.<br><br>The mounts were crucial, get them wrong, and you get a lousy, rotten xtal. How you hold the crystal changes its characteristic. Get it wrong and you&apos;ll get a horrible product with issues like &quot;band breaking&quot;.<br><br>For crystals in conventional packages - those two clips means you hold it in just the right way, and with just enough &quot;spring&quot;. It&apos;s much hard with DILs and that&apos;s why you often see three posts.<br><br>If you were seeing substantial shifts and changes in ageing following a shock, then bluntly, your crystals were a bit crap. The phenomenon has several names in the industry, we called it &quot;sleepy crystals&quot;. It really was (and still it) one of the industry&apos;s dirty little secrets. &nbsp;It is a sign of a production problem. I&apos;ll explain that later.<br><br>In extreme cases you would have a product sat in a drawer for a few months where the Oscillator wouldn&apos;t start at power up. If you (very carefully) took it apart, you would find the ESR was through the roof, the Q was pathetic and if you did get it to start oscillating, it was likely to have been on a &quot;band break&quot; - so the frequency would be slightly (but significantly) offset.<br>Bang it on the table - and all would be back to normal again, until it was left for ages.<br><br>One of my first post graduation jobs was to design something which could detect this problem without the whole &quot;waiting a month or two to see what happened&quot; thing.<br><br>I designed a system, which used the then standard ISO methods of measure frequency and ESR, except at very low drive levels. You would then gradually increase the drive level, watching to see how F and ESR change, until you are at maximum power, then you wound the drive level back to the zero again. You plotted the results on a graph, and you would get a fairly classical hysteresis curve, and the bigger the area inside the graph, the dodgier the crystals. You only had to leave the crystals alone for a day or two, per-test to get a clear result.<br><br>I&apos;m being called out on an errand now, but I will return later and explain where the phenomenon comes from.

  409

• CO

  @CoolDudeClem · 8 years ago

  &quot;Hi, it&apos;s time for a quick 2 minute teardown&quot; * video goes on for 17 minutes *

  169

• NE

  @NetworkXIII · 8 years ago (edited)

  Fascinating video Dave. &nbsp;Crystals and resonators belong to the dark arts, just like fortune telling and antenna design.

  132

• XR

  @XRinger · 8 years ago (edited)

  Wow, that really took me back. About 20 years ago when I was working for NEC at a PC factory in MA. <br>We were getting our PC motherboards (MBs) from SCI Technology in Tampa FL.<br>We started seeing a lot of PCs fail on the assembly line. Some of the MBs were DOA, but most (90%) were just working badly. After looking at some clock signals on one of the poorly working MBs, I zeroed in on one of those little metal oscillator modules. The clock was off-frequency, so I replaced it and the MB worked! &nbsp;<br>We powered up about 20 of the failing PCs and I used an RF spectrum analyzer (with a probe antenna) to observe the signals from the boards. Since these boards had six oscillator modules, I mapped out the RF profile from a good MB and compared frequencies emitted by the failing MBs. &nbsp;Some of the failed MBs had more than one failing oscillator modules. The failure pattern of which module and the shift in frequency was random. <br>I decided it was a QC problem with the oscillator modules. However, since some of the off-freq modules came from different vendors, I deduced these modules were somehow damaged at the SCI factory in Tampa. SCI said Nope, no way.. <br>I had a collection of failed modules, a microscope with a Polaroid camera and a dermel tool.. <br>I did a presentation for the NEC VP of manufacturing and was quickly on a plane to Tampa. <br>The photos of the little cracks in the quartz disks told the story. <br>Because I&apos;m a Ham Radio guy, I knew that dropping a radio Crystal 4&quot; onto a desktop would break it. So I knew what to look for at the SCI factory. <br>I watched the guy that loaded the tubes for the PCB parts loading machines. Nothing unusual.<br>Then late in the afternoon, I saw him unloading the plastic tubes from the module vendor, he was dumping them into a hard plastic tray, from a height of about 6 inches!! &nbsp;Afterwards, he hand-loaded them into the special tubes for the parts-loaders. He didn&apos;t believe he was breaking them. I shown him the pictures. &nbsp;Another Case closed.. :)

  185

• JU

  @judeabrigo278 · 6 years ago

  I don&apos;t know jack shit about what this component is about or what to use it for, but was hooked with your narration style and the way you tear down baby electronics components, sorta reminds me of how I used to tear down my toys when I was a kid to see how it works 🙏🏻

  5

• 1A

  @1aicrag · 7 days ago

  Thanks for showing with facts, not only opinions, about the reasons to use bypass capacitors

• WO

  @wondledonkey · 7 years ago

  I just realized Dave is the Steve Irwin of electronics

  25

• YE

  @yendak · 8 years ago

  I would find it interesting to see temperature compensated (TCXO) , oven heated (OCXO) and voltage controlled (VCXO) oscillators aswell, in comparison to the regular oscillator.

  11

• HE

  @Herby-1620 · 8 years ago

  The reason for using an 8MHz crystal and divider is that if you used a 1MHz crystal, the crystal would be WAY to big. &nbsp;About the lowest frequency you can get in an HC-18 package is a little bit less than 4MHz, and they were pushing it when you got one at 3.579..MHz (NTSC colorburst frequency).<br><br>Now you know.

  7

• JA

  @jamilarifin5721 · 6 years ago

  Thanks for cutting up one and showing us all.

  1

• RT

  @rthomp_tb · 6 years ago

  As an engineer at Abracon, I loved this video

  1

• JF

  @JFirn86Q · 8 years ago

  Thanks for going the extra mile and figuring out what that noise actually was. I had no idea it was a 8Mhz crystal with a divide-by circuit. Good stuff!

  2