Crystal oscillator testing in a circuit using DMM

[crystal]

Hi,

what is a proper way to test a crystal in a circuit? Usually when I test a crystal with a multimeter the circuit gets disrupted probably because of a dmm internal capacitance. I read somewhere on the net that good way to test with a scope is with a X10 or X100 probe since it cancel the internal capacitance of a scope. I don't have a scope anyway, I wonder could I of course with a proper multimeter adapter use a X10 probe on a mutimeter and will it show frequency the crystal is on?

The other way is to dismantle a crystal from the board and to test it on this: http://www.ebay.com/itm/1Hz-50MHz-Digital-LED-Crystal-Oscillator-Frequency-Counter-Tester-Meter-Kits-/400832061253?pt=LH_DefaultDomain_0&hash=item5d5373df45

But I would prefer if I would not have to remove it from the board.

[TSL]

Well given a crystal is used in an oscillator circuit then you should be looking to measure the output of the oscillator after the first buffer so as to prevent loading on the oscillator. Most crystal oscillator circuits only require a few pf to change their frequency and is some circuits outright cause the oscillator to stop. So you don't really want to probe the crystal itself.

What is the frequency of the oscillator ?

What is the capability of your DMM ?

Most DMM+Freq counters I've seen have poor accuracy and poor resolution.

Do you have a schematic of the circuit in question ?

regards

Tim

[crystal]

The question is general. Sometime I stumble on a circuit that does not work and with a pure luck when I change the crystal it starts working again. It happened to me a few times, clearly the crystal was defective but I have no idea how to properly test it in a circuit.

I have a UT61E it goes up to 220MHz but so far I had absolutely no use of it.

[TSL]

Basically you don't test a crystal in circuit, you look for output of the preceding amp. If all the voltages around the oscillator appear correct, then the crystal itself is the obvious choice. An oscilloscope is a good friend here to ensure waveforms appear as they should.

regards

Tim

[crystal]

Usually the schematic is not available so you really don't know what voltages to expect around an oscillator. If the suspicion why the device is not working falls on a microcontroller then I believe in the absence of a scope, the best way would would be to  remove crystal from the board and test it. If it is ok, then the microcontroller is faulty. Or the passive components around  but that is less likely and can be tested.

Strangely I didn't find a single youtube video about this.

[Fraser]

There is much difference between diagnosing whether a crystal is oscillating and whether it is oscillating at the correct frequency. I tend to find crystals fail to oscillate rather than go wildly off frequency. The exception is a very unstable oscillator design that self oscillates even without the crystal in circuit ! If the crystal goes O/C then that sort of oscillator goes off on whatever frequency it best resonates at.

As has been stated, you should not really test for oscillation directly across a crystal and the loading will affect the frequency and in some cases stop the oscillation. It is, however possible to place a high impedance probe close to a crystal oscillator and pick up the oscillation without actually touching teh circuit  EMC probes will also work well for this. Used in conjunction with an oscilloscope, frequency counter or spectrum analyser much can be established about the oscillator.

A multimeter would NOT be my first choice as a diagnostic tool for the oscillation but I would check voltage around key components with it. Yours has a frequency counter though so is more useful. A VERY simple tool that any hobbyist should have in his/her toolkit is a RF probe for a multimeter. It is a simple Diode detector probe that may be seen on many DIY pages on the web. Te probed is AC couples and rectifies any signal within its capability, just like an AM radio detector stage. Such a toll was/is used to detect the presence of RF in a circuit. As such you would be able to see whether an oscillator is actually running or whether it is 'dead'. Best location for a first check is after the first buffer amplifier but you CAN look on the crystal itself but have to consider the effects of loading.

http://www.cliftonlaboratories.com/diodes_for_rf_probes.htm

At maritime college this was the first piece of test equipment that we built. We constructed it on a PCB that sat inside a syringe and used a blunted needle for the probe tip. Worked well to over 500MHz. It was used throughout our 3 year course and proved that a simple diode detector had many uses in RF diagnostics.

Aurora

[SeanB]

Simpleset check is a sensitive AM detector ( otherwise an AM radio) to detect oscillation. In most cases if a crystal is not oscillating simply turn it around, and it will work again.

[Fraser]

@SeanB

Very interesting comment regarding reversal of the crystal.

Once a crystal has become unreliable (i.e. it has stopped oscillating at some time) I have always fitted a new one. In some cases it was an issue with the drive from the surrounding oscillator components so a new crystal either covered up the issue or did not oscillate either.

With regard to desoldering and re-soldering an intermmitent crystal, my concern was that by de-soldering the crystal I could have initiated some change in the crystal mount that made it work again, but was not a true 'fix'. I had one crystal that worked fine on most occasions but now and then failed to oscillate unless you tapped its case. No dry joint issues just a dodgy crystal. A new part solved the problem.

I am no expert on crystals in terms of their failure modes but would love to know more from knowledgeable people like SeanB if they can enlighten me. I have seen and dismantled 1940's era crystals that could be frequency fine tuned using the application of graphite pencil to the crystal ! Those crystals were huge and thick in comparison to modern parts. I have dismantled modern crystals and found just a crystal wafer held in place by a pair of loop spring wire mounts. I have assumed that crystals that fail to oscillate have suffered from crystal contamination or physical defects in the crystal or its mounts. Having never studies crystal failure modes I (and likely others) would find this an interesting topic of conversation.

All comments welcomed as this thread is looking at diagnosing crystal oscillator faults with a multimeter and so is on topic  Understanding failure modes can lead to an intelligent choice of testing vectors.

Aurora

[SeanB]

The turn around trick was a procedure from the original Apple ][ manual, and I have used it a few times. It works because the crystal is stressed from being overdriven by the TTL gate, so turning it around evens out the stress and it will work for a long time thereafter. Eventually though it will fail from the overdrive and need to be replaced.

[Fraser]

@SeanB,

Thanks, that is really helpful information 

I did a Google on crystal failure modes and there appears little on the NET detailing any such topic. They are considered inherently reliable, but we all know that crystals do fail in service. Mind you I always suspect the drive circuit before the crystal as the active element gain reduction can often be the cause of no oscillation.

Aurora

[cncjerry]

Simpleset check is a sensitive AM detector ( otherwise an AM radio) to detect oscillation. In most cases if a crystal is not oscillating simply turn it around, and it will work again.

I wonder why this happens and have seen crystals not work one away in a poorly designed circuit.  But once running, I have had nearly zero crystals fail in bench equipment.  Portable, yes; bench, no.

[edavid]

Why would reversing the crystal do anything?  If that was a joke, I guess I didn't get it 

[SeanB]

Turning it around changes the driven end of the crystal, which in many cases will allow it to oscillate again. IIRC the issue is the TTL oscillator, using a single non CMOS gate, tends to drive the crystal with too much power, and this eventually weakens it. Using a CMOS inverter with a series resistor on the output to the crystal, and a series resistor to the gate input, reduces the drive level to prolong crystal life. You need short traces, low capacitance and attention to decoupling as well, otherwise it is very prone to picking up noise from the supply lines and ground plane, and from sharp signal;s nearby.

[cncjerry]

I never checked but I wonder if each leg to ground has different capacitance?

[edavid]

Turning it around changes the driven end of the crystal, which in many cases will allow it to oscillate again. IIRC the issue is the TTL oscillator, using a single non CMOS gate, tends to drive the crystal with too much power, and this eventually weakens it. Using a CMOS inverter with a series resistor on the output to the crystal, and a series resistor to the gate input, reduces the drive level to prolong crystal life. You need short traces, low capacitance and attention to decoupling as well, otherwise it is very prone to picking up noise from the supply lines and ground plane, and from sharp signal;s nearby.

I still don't get it.  The drive current has to pass through the crystal, since there is no other path for it.  Damage from excessive drive will take place at the weakest point of the crystal, not at whatever end happens to be closest to the driven electrode.  Turning the crystal around does nothing.

[Seekonk]

Where I used to work we had a policy...Now that it works, let's destroy it.  We had this arcing ancient cattle prod that we would arc to the boards inputs.  Most of our products used little Microchip controllers.  The ones with crystals would often reset.  The ones that used the internal RC oscillator seemed to always keep running.

Every crystal reacts to a DC bias.  Perhaps that is the factor in rotating it.

[cncjerry]

As to turn around, do this: as in a PIC master oscillator, reduce the leg capacitance maybe using a trimmer until the circuit fails to start.  Flip the crystal and do the same.  I would bet that you will find one that will/won't start one way vs the other.  I have a circuit setup as I was testing phase and harmonic noise of a crystal vs a modern DES chip and found the DDS to be superior. I'll try and post.

[dom0]

I have a circuit setup as I was testing phase and harmonic noise of a crystal vs a modern DES chip and found the DDS to be superior. I'll try and post.

And the reference for the DDS came from what kind of oscillator exactly...?

[cncjerry]

The crystal reference has little to do if anything with the much exaggerated DDS phase noise and spurs though your comment will extend my testing. The testing i was referring to was comparing the overall distortion of a DDS vs  crystal. I was surprised to see the DDS have lower harmonic distortion vs a crystal.

[dom0]

Didn't read your post correctly then – I thought you were saying that *both* phase noise and distortion were superior in the DDS. For the distortion part this should strongly depend on the topology + implementation of the crystal oscillator(*). The cut of the crystal also plays quite a role afaik.

Generally speaking it is possible and has been in fact done to use crystals for very distortion-sensitive applications. For example, the post-mixer fundamental filter in the HP 358x series is a four (or five?) stage design based on crystals.

(*) the typical CMOS/TTL oscillator generates an extremely clipped sine wave, if any, so yeah, that got a lot of distortion.

[cncjerry]

I became interested in the crystal v DDS issue as a member of the R2Pro receiver group.  The R2Pro is a high quality direct conversion receiver.  Though I and a number of others used DSP analog processors after the mixing and downconversion, many said they shied away from a DDS for the VFO due to phase noise, spurs and harmonic distortion compared to a good crystal oscillator. So I built a simple but good CXO and tested it vs my ad9854 and other DDS units having I think four different chips. The DDS won by quite a bit in harmonic distortion aand I was surprised. Overall noise was as good but the close-in spurs were higher with the DDS.  Given that the spurs are frequency dependent and can sometimes be eliminated or mitigated by an inaudible tuning change, I am going back to the DDS though I might try a DDS PLL hybrid. I don't know how I took us of subject here.

[TSL]

I became interested in the crystal v DDS issue as a member of the R2Pro receiver group.  The R2Pro is a high quality direct conversion receiver.  Though I and a number of others used DSP analog processors after the mixing and downconversion, many said they shied away from a DDS for the VFO due to phase noise, spurs and harmonic distortion compared to a good crystal oscillator. So I built a simple but good CXO and tested it vs my ad9854 and other DDS units having I think four different chips. The DDS won by quite a bit in harmonic distortion aand I was surprised. Overall noise was as good but the close-in spurs were higher with the DDS.  Given that the spurs are frequency dependent and can sometimes be eliminated or mitigated by an inaudible tuning change, I am going back to the DDS though I might try a DDS PLL hybrid. I don't know how I took us of subject here.

Start a new thread for this - I'm interested in what you did and how you tested it all!

cheers

Tim