Author Topic: Solder metal can of a crystal to GND or not?  (Read 474 times)

0 Members and 1 Guest are viewing this topic.

Offline RoGeorgeTopic starter

  • Super Contributor
  • ***
  • Posts: 7089
  • Country: ro
Solder metal can of a crystal to GND or not?
« on: January 25, 2025, 01:31:01 pm »
The quartz is same size as in the picture, but a 20MHz THT, scrapped from a defective HDD (manufactured 1991).  Want to use it instead of the stock 8MHz crystal, to test some software ideas.  Nothing metrology grade, weekend tinkering but with a need for low phase noise.  The lower the clock jitter, the better .



Usually the can is never soldered to GND in production PCBs, but I remember some DIY ham radio gear (hand made) where the metal can of a 2 pins quartz was soldered to GND.  Is that a mistake?

For minimum phase noise/clock jitter, would it help to solder the can to GND?

Online CaptDon

  • Super Contributor
  • ***
  • Posts: 2244
  • Country: is
Re: Solder metal can of a crystal to GND or not?
« Reply #1 on: January 25, 2025, 01:51:43 pm »
Many of my salvaged can crystals had a grounding lead soldered to the can. Most often this was done when there were several oscillators of different frequencies near each other. I have also seen this done if the unit was designed to operate in a noisy environment. In your operational situation I doubt if the grounded or ungrounded difference could be measured but it surely wouldn't hurt to ground it as long as the crystal isn't damaged by heat. During the common TTL crystal oscillators of days gone by if two or more oscillators were next to each other you could see a certain amount of pulling or FM'ing between each other which was about 90% caused by poor power rail decoupling and about 10% do to the traces of each oscillator acting like an antenna picking up signal coming from the adjacent oscillator.
Collector and repairer of vintage and not so vintage electronic gadgets and test equipment. What's the difference between a pizza and a musician? A pizza can feed a family of four!! Classically trained guitarist. Sound engineer.
 
The following users thanked this post: RoGeorge

Online Wallace Gasiewicz

  • Super Contributor
  • ***
  • Posts: 1396
  • Country: us
Re: Solder metal can of a crystal to GND or not?
« Reply #2 on: January 25, 2025, 02:11:18 pm »
Lots of Xrals have case soldered to GND. Even in production assembly in old radios.  Lots do not have case soldered to Gnd. Many Xtals  in old Channel radios have one Xtal per channel that is not soldered to Gnd.    Soldering to Gnd helps shield the Xtal from other signals but also can change the capacitance that the metal Xtal case adds to the circuit.

Check the chip first and see if it will support 20 MHZ.. I cannot read the chip number.
The biggest problem I see here is that the circuit the Xtal is it may not support 20 MHZ oscillation.     
I would stick your 20 MHz Xtal in the circuit and see if it works. Otherwise you may have to change the adjoining capacitors to a lower value.   
Apparently you wish to use the entire board but at 20 MHz, otherwise a 20 MHz 4 pin oscillator is very cheap.
 
The following users thanked this post: RoGeorge

Offline RoGeorgeTopic starter

  • Super Contributor
  • ***
  • Posts: 7089
  • Country: ro
Re: Solder metal can of a crystal to GND or not?
« Reply #3 on: January 25, 2025, 03:37:47 pm »
The MCU is an ATmega328P-AU at 5V, it supports max. 20MHz.

Did the quartz replacement already.  So far the MCU seems to be working fine at 20MHz.  The already existing capacitors are 2 x 20pF (according to the schematic for that 8MHz PCB), but I didn't measure them.

Don't have any instrument to directly measure the phase noise.  Maybe I'll improvise later some counters, and compare the output pulses against another clock generated from a Rigol DG4102 AWG.  The AWG is supposed to have a decent phase noise.  My guess is the AWG should produce a smaller time jitter than the MCU, but I didn't try yet.

Back to the crystal can, in case grounding it doesn't hurt, I'll probably add a 3'rd wire to GND for better mechanical sturdiness.

Offline temperance

  • Frequent Contributor
  • **
  • Posts: 726
  • Country: 00
Re: Solder metal can of a crystal to GND or not?
« Reply #4 on: January 25, 2025, 04:59:45 pm »
Those old style xtals are very often not soldered to GND. But indeed I have seen it being done and always wondered if it would make a difference. The can of an SMD versions HC49 XTALS can't be soldered to GND because of the plastic base underneath.

But smaller SMD xtals with 4 connections do have two connections which connect the metal top lid and I connect those to GND. You also have variants of those where the housing is ceramic and those two extra connections don't serve any purpose.
 

Online PCB.Wiz

  • Super Contributor
  • ***
  • Posts: 2144
  • Country: au
Re: Solder metal can of a crystal to GND or not?
« Reply #5 on: January 25, 2025, 08:53:00 pm »
For minimum phase noise/clock jitter, would it help to solder the can to GND?
It cannot do much harm - use a fine wire, and lead based solder, so you can use least heat.

Quote
Nothing metrology grade, weekend tinkering but with a need for low phase noise.
If you chase phase noise, other things may dominate.
MCU oscillators are forced to share the GND and power pins, so they have significant MCU activity crosstalk.

Low ESR certainly helps.
If your crystal was from a HDD, it was likely cheap, rather than high spec, but it should have 'aged' by now

You may be better using an external oscillator, either one spec'd for low phase noise, or you could try a crystal with this part :
https://www.nisshinbo-microdevices.co.jp/en/products/quartz-crystal-oscillator-ic/spec/?product=nju6385
That is a Xtal Oscillator that is specified to have low phase noise.
Nisshinbo have other (cheaper) siblings, but that one is spec'd to have low phase noise, and they give the Xtal spec's ( ESR = 17.7 ohms).


Abracon do spec a special low ESR crystal, in the ABLS-LR series (same large HC-49/US case as your xtal), but they are somewhat rare and no stock shows for 20Mhz
Mouser have 16,24,25MHz models, with the 24MHz one done in lower ppm.

It would be interestieng to compare a 'best discrete' solution like (say) NJU6385 + ABLS-LR-24.000MHZ-B-1-G-T
with a 'good' oscillator (this one 24.576MHz )
https://www.lcsc.com/product-detail/Crystal-Oscillators_Seiko-Epson-X1G0041710029_C255994.html  ( 40c / 100+)


Addit: If you really want low phase noise discrete designs, I found this link that shows how important crystal variations are
https://www.diyaudio.com/community/threads/the-well-tempered-master-clock-building-a-low-phase-noise-jitter-crystal-oscillator.261651/page-150#post-6385649

If you are making these in higher volumes, I saw recently lcsc have a new category.
https://www.lcsc.com/products/Temperature-Sensing-Resonators_13490.html
These parts are designed for high performance oscillators, so they have low initial ppm and low aging, and high min Q ( 75,000 )  & come in 2520 packages.


 
« Last Edit: January 25, 2025, 11:53:51 pm by PCB.Wiz »
 
The following users thanked this post: RoGeorge


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf