crystal debugging continued

[Mykle James Hansen]

Hi,

I just gotta share this, I find it entertainingly strange.

Tom Hudson kindly lent me a 21-st century oscilloscope, which

has no trouble picking up the oscillations from the crystals

on my working v6 board and my not-working v8 board.  I just

place the probe tip on one leg of the crystal and leave the

ground clip disconnected.

The working v6 board signal seems to show a bit of noise, but overall 

it’s sinusoidal & regular:

The not-working v8 board signal has less noise, a lower level, but also it has this strange

tachycardia in it: 

… I’ve never seen that!  Would anybody have a clue whether this is a symptom

of not enough drive, too much drive, wrong impedance, too much cholesterol,

not enough exercise … or operator error?

Cheers,

-mykle-

[Jon Boro]

I haven’t been following along with everything/not sure what you’re doing, so forgive any assumptions…

You really can’t directly probe a crystal and get reliable, if any results. You definitely can’t measure the frequency (accurately, if at all) that way.  The probe capacitance loads down the crystal oscillator too much and de-tunes it, potentially to the point of not working. You can see if anything is happening via non contact probing with coax or maybe the scope probe tip, just picking up the electric field from a small distance. 

OR, your crystal shunt capacitance (combo of added shunt/load capacitor + PCB capacitance) could be so far off that it won’t oscillate reliably without the scope probe capacitance. In which case your shunt/load capacitance is too small (but I’d guess it’s too large if you don’t have a tight layout)

You definitely need to use COG/NPO capacitors.

Your problem is not the crystal tolerance. I suppose it could be the crystal if you found some real counterfeit or no name trash. 

 I don’t know what you’re working on, but you need to probe a buffered version if this clock.   assuming this is a crystal for an MCU clock, The way you usually tune the shunt caps is you need to set up the MCU to output a frequency on a pin via a timer/counter, or however it can. You probe THIS, as it’s buffered, not loaded down by your probe, and you can get an accurate result and representation of what the MCU is seeing when your not sticking things on the OSC pins (the result of course would be divided by however the counter/timer, or whatever, is set up). 

An oscilloscope is generally not going to be accurate enough to do this unless it has a dedicated frequency counter built in… You can usually only “pull” a crystal frequency Maybe a few hundred ppm by adjusting (or by accidental design) the load capacitance  before it doesn’t oscillate reliably or at all. A scope cursor measurement or on screen automatic measurement probably won’t be accurate to a few hundred ppm. 300ppm at 12Mhz is 3600Hz. I’m not looking at it, but pretty sure my siglent scope (a step above what your showing in your pictures) doesn’t even display 100s of Hz at 12Mhz, and the 1000s certainly isn’t accurate. Maybe with a bunch of averaging you could get there, but I personally would not rely on this. Not without thoroughly reading the scope manual and understanding how it gets it’s frequency measurement results anyway. 

This might sound more complicated than you were bargaining for, I hope it helps. If you want more help and want to email me direct feel free. 

-Jon

Sent from my iPhone

On Feb 4, 2023, at 10:51 PM, Mykle James Hansen <my...@mykle.com> wrote:

Hi,

I just gotta share this, I find it entertainingly strange.

Tom Hudson kindly lent me a 21-st century oscilloscope, which

has no trouble picking up the oscillations from the crystals

on my working v6 board and my not-working v8 board.  I just

place the probe tip on one leg of the crystal and leave the

ground clip disconnected.

The working v6 board signal seems to show a bit of noise, but overall 

it’s sinusoidal & regular:

<IMG_6110.jpeg>

The not-working v8 board signal has less noise, a lower level, but also it has this strange

tachycardia in it: 

<IMG_6111.jpeg>

… I’ve never seen that!  Would anybody have a clue whether this is a symptom

of not enough drive, too much drive, wrong impedance, too much cholesterol,

not enough exercise … or operator error?

Cheers,

-mykle-

--
You received this message because you are subscribed to the Google Groups "dorkbotpdx-blabber" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dorkbotpdx-blab...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/dorkbotpdx-blabber/942FA751-4106-4C4B-910E-AE2FFB490AA4%40mykle.com.

[Erik Lane]

Maybe check for a bad solder joint? Just a guess though. I don't usually try to scope crystal lines, though I have a few times. Problem is I don't have a really high speed probe, so the probe itself has plenty of capacitance to overwhelm the regular operation of a crystal circuit. I mean, seeing if it's running or not, or whether adding the probe capacitance causes it to start running can be helpful, but it doesn't really show the normal operation of the circuit. (I think, at least. Like I said, not something I have much experience with.)

If not a solder joint, maybe the crystal itself has a problem/loose connection internally? But that's probably a pretty low likelihood. I'm just completely guessing here.

Good luck!

Erik

--

[Phil Hochstetler]

This thread might be useful:

https://www.eevblog.com/forum/beginners/how-to-debug-crystal-oscillator-startup-failures/

Phil

On Sat, Feb 4, 2023 at 10:51 PM Mykle James Hansen <my...@mykle.com> wrote:

--

[p.fer...@gmail.com]

Did your scope have an isolation plug.

If you use one of those it gets rid of house power noise and the scope displays the voltage difference(potential) between the "ground" clip and the scope probe tip. 

[Andrew M. C. Dawes]

How often does the blip happen? If you zoom out, do you see another one come into view? 

--

You received this message because you are subscribed to the Google Groups "dorkbotpdx-blabber" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dorkbotpdx-blab...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/dorkbotpdx-blabber/31c6ca52-3dfc-4046-aece-87ba923c434cn%40googlegroups.com.

[Tom Moxon]

Sound like it might be defective, did you try another on Crystals can be temperamental beasts. Some designs require a series resistor on one leg, most require the correct capacitor on each leg. Maybe the wrong value caps got installed?

--

You received this message because you are subscribed to the Google Groups "dorkbotpdx-blabber" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dorkbotpdx-blab...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/dorkbotpdx-blabber/942FA751-4106-4C4B-910E-AE2FFB490AA4%40mykle.com.

[Paul Stoffregen]

While the crystal blip is on-screen, turn on the scope's other channel and touch the other probe to every wire on the PCB starting with those closest to the crystal (or running underneath).  You're looking for a signal that does something at the same time or right before the blip.  Of course this is much easier said than done if you only have 2 hands to awkwardly fumble the PCB and both probes.

To view this discussion on the web visit https://groups.google.com/d/msgid/dorkbotpdx-blabber/CAJBFORBJvCne5yUdXtTB0Pm-tGSoKn%2BPQkwb-0y8mJSOQtuAFw%40mail.gmail.com.

[p.fer...@gmail.com]

In Japan, if they have two probes, they hold them in one hand like chopsticks. :-)

It's frustrating looking at scopes with disconnected ground clips. Bad things can happen if you hook them up with no isolation plug, though. I've seen exploding capacitors and ground clips welded in place. I'm retired and I looked at scopes for 50+ years.

The bottom trace has a larger DC component and I have no clue why. The closest it gets to ground is below the the trigger point(yellow triangle on the right) so you see it as a glitch. How often it's triggered is 104.1 something(units not visible).

The top trace has a low frequency wave on the bottom peaks (60hz because it uses chassis ground??)that shows as a peak detect maybe. It triggers at 2.2 mhz(you can see the units).

Another Paul