Epson crystal oscillators

[Nigel Goodwin]

Hi, anyone have any history with Epson Crystal oscillators, like this one:

Epson, 20MHz XO Oscillator, ±50ppm CMOS, 4-Pin PDIP Q3204DC21000500 | RS

uk.rs-online.com

The guy in the unit next next door came round this afternoon, and he's got number of new/old PCB's that don't work.

They use the PIC16C73, which as I recall is an OTP chip from last century

Anyway, the company who originally made them went bust, a new company has taken them over (I've no idea how long ago), and they were having problems with some of the last boards. I think he had about ten duff ones, and a good working one - and by copying the code from the working one (not protected luckily) he managed to get most of them working - however, he's since had three returned as having stopped working again.

So I went round to help, and the PIC is fed from the Epson crystal oscillator mentioned above, and there's no output from the faulty boards, where the working one has a nice sine wave. One of the boards even reads 100 ohms to chassis from the output pin of the oscillator, and I've taken that one out (and it's 100% the Epson chip duff), as well as one of the others where it reads high. As well as the PIC the oscillator feeds a 74LS393, and an output from that (presumably divided down?, feeds to a PAL (which he's also been able to read and copy).

So my question is, are these devices renowned for being unreliable?, it seems a bit coincidental that he's had multiple failures in a short time - or could something be killing them? - the 5V rail reads 4.98V (on his meter).

 

[JimB]

Looking at the datasheet for the oscillator there appears to be two distinct types, one for a 5v supply and one for a 3.3v supply.
Could the problem be 3.3v parts inserted in a 5v circuit?

JimB

 

[ZipZapOuch]

If the oscillators are used in "rugged" environments without shock/vibration damping, the indium/gallium solder used to bond the quartz can fail, the crystal can crack and so on. In this case, rugged means any vibration that can damage the unit because resonances in x, y or z plane can all screw it up. Not common, but it happens.

Also, TI did a study about how different crystals or oscillators survive better in a given application because the guts are oriented differently or the crystal is cut in a different plane. There is no good way to predict the failure as of their paper (about 20-years ago).

 

[Nigel Goodwin]

Looking at the datasheet for the oscillator there appears to be two distinct types, one for a 5v supply and one for a 3.3v supply.
Could the problem be 3.3v parts inserted in a 5v circuit?

JimB

That's an interesting idea, I'll pop round and see the guy tomorrow.

From what I can gather the boards were assembled by people who didn't really know what they were doing, they were old boards left over from the previous company - so I can easily see them ordering the wrong voltage version.

 

[Diver300]

I used to work for a company that made the metal-cased ones, like this:-
https://uk.rs-online.com/web/p/crystal-oscillators/7960520
We bought the bases and added the crystals, adjusted the frequency and welded the lid on. It allowed custom frequencies is a couple of days.
Anyhow, we very, very rarely saw problem with the oscillator part, which was really simple anyhow. It was a couple of transistors, or a couple of inverters. There were also some that had dividers in them.
The ones we made had decoupling capacitors across the supply, but I am not sure if the plastic ones have those. I know that the small ceramic ones tend not to contain decoupling capacitors.
The failures were generally the crystals. They could lose activity, which effectively meant that they went open circuit, or they could break and go short circuit, but it didn't make much difference to the behaviour in that most designs without dividers would end up with the output at mid voltage and the oscillator taking quite a lot of current, more than it did when working properly.

The crystals that we used were circular, so the orientation was random. Crystals in the 4 - 80 MHz range are called AT-cut, with those above about 30 MHz being run in third-overtone mode. The cut of the crystal is the angle which the disk of quartz is cut from the single crystal of quartz that has been grown in an autoclave. The angle affects how the frequency changes with temperature, and there is only a narrow range of angles that is used to keep the frequency stable.

AT-cut crystals oscillate in thickness-shear mode, like a jelly wobbling side to side. The direction of oscillation isn't important in a circular crystal, but in the plastic oscillators, the crystal is inside a cylindrical metal can, like a watch crystal, and the rectangular crystals have to be cut so that they oscillate along their length, or they would not oscillate properly.

I don't think that the orientation of the crystals can vary inside the plastic oscillators. However there are lots of other ways that crystals can be made badly so that they fail prematurely. If the crystals aren't properly cleaned before the electrodes are added, or the silver electrode material is added when there is not enough vacuum, the electrodes can come detached. If the crystals are chipped or cracked, the cracks can grow until the crystal fails. The metal cans are welded, but if that isn't done correctly, oxygen can get in and corrode the electrode.

 

[Nigel Goodwin]

I used to work for a company that made the metal-cased ones, like this:-
https://uk.rs-online.com/web/p/crystal-oscillators/7960520
We bought the bases and added the crystals, adjusted the frequency and welded the lid on. It allowed custom frequencies is a couple of days.
Anyhow, we very, very rarely saw problem with the oscillator part, which was really simple anyhow. It was a couple of transistors, or a couple of inverters. There were also some that had dividers in them.
The ones we made had decoupling capacitors across the supply, but I am not sure if the plastic ones have those. I know that the small ceramic ones tend not to contain decoupling capacitors.
The failures were generally the crystals. They could lose activity, which effectively meant that they went open circuit, or they could break and go short circuit, but it didn't make much difference to the behaviour in that most designs without dividers would end up with the output at mid voltage and the oscillator taking quite a lot of current, more than it did when working properly.

The crystals that we used were circular, so the orientation was random. Crystals in the 4 - 80 MHz range are called AT-cut, with those above about 30 MHz being run in third-overtone mode. The cut of the crystal is the angle which the disk of quartz is cut from the single crystal of quartz that has been grown in an autoclave. The angle affects how the frequency changes with temperature, and there is only a narrow range of angles that is used to keep the frequency stable.

AT-cut crystals oscillate in thickness-shear mode, like a jelly wobbling side to side. The direction of oscillation isn't important in a circular crystal, but in the plastic oscillators, the crystal is inside a cylindrical metal can, like a watch crystal, and the rectangular crystals have to be cut so that they oscillate along their length, or they would not oscillate properly.

I don't think that the orientation of the crystals can vary inside the plastic oscillators. However there are lots of other ways that crystals can be made badly so that they fail prematurely. If the crystals aren't properly cleaned before the electrodes are added, or the silver electrode material is added when there is not enough vacuum, the electrodes can come detached. If the crystals are chipped or cracked, the cracks can grow until the crystal fails. The metal cans are welded, but if that isn't done correctly, oxygen can get in and corrode the electrode.

Interesting to hear, obviously the one that reads 100 ohms from output to ground has some kind of driver failure, so probably not the actual crystal in that one, but perhaps in the others?.

 

[Diver300]

Interesting to hear, obviously the one that reads 100 ohms from output to ground has some kind of driver failure, so probably not the actual crystal in that one, but perhaps in the others?.

I would guess that is the driver failure. They might all be electronic failures, as there is an oscillator part and a driver part. An electronic failure in the oscillator would be difficult or impossible to distinguish from a crystal failure without opening one up.

Have you cut open a dead one to find the form factor of the crystal itself?

 

[Nigel Goodwin]

I would guess that is the driver failure. They might all be electronic failures, as there is an oscillator part and a driver part. An electronic failure in the oscillator would be difficult or impossible to distinguish from a crystal failure without opening one up.

Have you cut open a dead one to find the form factor of the crystal itself?

I don't have them, it's the guy next door - but they are DIL plastic construction, so not as easy to open as a metal one. His new ones have arrived, they came first thing this morning from RS - as did a number of parcels for us. He's got a Zoom meeting to attend, but he'll let me know later how it goes.

 

[fourtytwo]

where the working one has a nice sine wave.

Should not be a sine-wave output unless it is severely overloaded! It is a CMOS logic output so should swing from near ground to it's supply with according to it's spec 4nS rise or fall time. 20Mhz period is 50nS so half period 25nS so at least 20nS of flat waveform should be visible between edges. I have used this style of oscillator for years, never any problems. As others have said maybe the part# is wrong, a photo wouldn't go amiss, also there are several types of that PIC? Could be just heavily loaded due to some other faulty or mis-inserted part connected to it (you mentioned several).

 

[Nigel Goodwin]

Should not be a sine-wave output unless it is severely overloaded! It is a CMOS logic output so should swing from near ground to it's supply with according to it's spec 4nS rise or fall time. 20Mhz period is 50nS so half period 25nS so at least 20nS of flat waveform should be visible between edges. I have used this style of oscillator for years, never any problems. As others have said maybe the part# is wrong, a photo wouldn't go amiss, also there are several types of that PIC? Could be just heavily loaded due to some other faulty or mis-inserted part connected to it (you mentioned several).

It was a nice sine on his scope, I did wonder if it should be square - I've not heard back from him yet, and I don't have any of the parts.