Teardown of a quartz crystal oscillator and the tiny IC inside | Hacker News Reader

Teardown of a quartz crystal oscillator and the tiny IC inside

righto.com

123 pointsivanvas25 comments

> I was expecting a gem-like quartz crystal inside, but found that oscillators use a very thin disk of quartz.

Interesting - watches usually use a tuning fork shape. I wonder why they use a disc here?

> Wristwatches were revolutionized in the 1970s by the use of highly-accurate quartz oscillators.

If you're into highly accurate timepieces you can pick up some amazing quartz pieces from brands like Rolex (the OysterQuartz) and Audemars Piguet (the AP Royal Oak 6005) from this time. Since brands wanted to distinguish their quartz movements from their mechanical ones the cases the quartz pieces come in are very distinct.

> I wonder why they use a disc here?

It has been a long time since I've thought about it, so I might be misremembering, but the crystal geometry can influence skew related to orientation and external vibration. A crystal + IC design is using the crystal as more of a bandpass filter than an actual signal source, so it would make no sense to introduce the downsides of a tuning fork for a design that makes no use of the actual tuning fork quality.

It's a function of size. Tuning fork crystals generally have frequencies in the range of a few tens of kilohertz (generally 32768 Hz).

Quartz discs vibrate at many megahertz. A tuning-fork like structure that oscillated at 100 MHz would be far to small to easily manufacture.

------

One fun note: crystals like this are used for measuring deposited material in vacuum processes like PVD [1] or similar [2]. The crystal is exposed inside the vacuum chamber where it will be deposited on at the same rate as the target, and the change in frequency is a direct function of the mass (and therefore thickness) of the deposited material.

Eventually, the exposed crystal has to be replaced as it no longer oscillates properly.

1: https://en.wikipedia.org/wiki/Physical_vapor_deposition 2: https://www.semicore.com/news/72-thin-film-deposition-contro...

Note 32768 Hz is chosen because it is a power of 2 (the 15th power).

And therefore you can get a 1 Hz clock from a series of binary frequency dividers (flip-flops). That is, having the crystal oscillate at a power-of-2 frequency simplifies the circuitry a great deal.

The first power of two that is outside the range of (most) human hearing.

A PVD sensor is a very clever use case! Thanks for highlighting.

Also the Seiko Spring Drive.

. . . the Tri-synchro regulator uses three types of energy to regulate the moving parts and establish synchronicity:

  * Mechanical power, from the mainspring
  * Electrical power, creating a reference signaling via an IC/quartz oscillator
  * Electromagnetic power, to apply a brake via a rotor/stator.

These three forces work in harmony to regulate the way the spring unwinds and to make possible the precise movement of the second hand.

https://www.grand-seiko.com/sg-en/collections/movement/sprin...

Two common frequencies used at the introduction of the PC were 3.579545 MHz. & 4.43361875 MHz

Chosen because of the existing supply of cheap xtals manufactured for NTSC & PAL color television sets.

NTSC : https://en.m.wikipedia.org/wiki/NTSC

PAL : https://en.m.wikipedia.org/wiki/PAL

Edits to fix typos :( And add;

Crystal Oscillator Frequencies : https://en.m.wikipedia.org/wiki/Crystal_oscillator_frequenci...

For that reason: cheap available quartz crystals, since the 1970s, the shortwave radio frequency 3.579Mhz has been a spot on the dial where radio hams use low power home made morse code transmitters. At night, a range of a couple of hundred files is usually possible.

That is why the IBM PC has used a 14.31818 MHz oscillator, i.e. 4 times the NTSC frequency of 3.579545 MHz, and which divided by 3 provided a clock frequency for the CPU that was close enough to 5 MHz (the maximum clock frequency of the 1st version of Intel 8088).

While the current computers with Intel/AMD CPUs do no longer care about NTSC, all still have a timer equivalent with the Intel 8253, which uses the clock frequency of 14.31818 MHz, to ensure compatibility with the original IBM PC.

For early microprocessors, the application manual would often devote 20 pages to selection and support of a crystal oscillator. For wide-temperature range use, as in industrial circuits, it could be a good rule-of-thumb to budget for a self-contained powered oscillator in the BOM.

And lots of them had built-in oscillators only requiring a crystal, or even a resonator for low cost applications where timing wasn't all that critical.

I think the first paragraph explaining the oscillator operation is not quite right. Corrected, below, maybe. I swapped "increases" and "decreases". But the fix might be something different.

"In more detail, as the voltage across the crystal decreases, the transistor turns on, feeding current into the capacitors and boosting the voltage across the capacitors (and thus the crystal). But as the voltage across the crystal increases, the transistor turns off and the current sink (circle with arrow) pulls current out of the capacitors, reducing the voltage across the crystal. Thus, the feedback from the drive transistor strengthens the crystal's oscillations to keep them going."

It's like pushing a swing. You want to give it a push when it's moving forward to keep the oscillation going.

I am corrected.

Almost 40 years old technology and we are still unable to "print" these dies at home.

Can you imagine how much fun that would bring if you could design your own IC and just hit print, then load the die into the carrier and solder the pins?

There isn't even a service like JLCPCB where you could get your custom ICs made and shipped to you.

To be fair, it's not like the semiconductor industry has been standing still.

Lithography requires a lot of space and clean rooms.

I think there might be in a few years but it's super hard.

The technology is insane, the design skills are non-trivial, and the software is terrible

There are a few channels on YouTube about DIY lithography, it's not as far away as you might think.

Yes there are services where you can get custom ICs made - they’re called ‘multi project wafer’ services. They start around $10K (e.g. efabless does one about that much with the SkyWater 130 nm proces) which is pretty good given how much it costs to do the lithography, not to mention actually making the ICs!

But in general it’s a very difficult process. The cleanliness required, the scale at which all this happens, etc. are just orders of magnitude harder than a regular PCB. That’s not even considering the nasty chemicals needed to etch silicon which are way worse than PCB etching chemicals!

There are cheaper options than $10K (see Mosis or Europractice price tables) but they have smaller areas (to be cheaper, you can have larger chips if you pay more) and might require commercial tools and NDAs. The process options range from much older than the Skywater one to reasonably recent technologies (with proportionally higher prices).

Btw, am I the only one that would love to see FW-91W?

I can see one right now, on my wrist :-)

I mean in the context of Ken’s blog :-)

Reply on news.ycombinator.com