Watch Crystal dissipation

I don't know how the megabucks commercial DSP and PLL guys do it (they aren't saying?), but I'd use a "marginal oscillator," e.g.,

This is a design technique created for NMR probes, and not a "marginal" oscillator circuit, e.g. as referred to in Microchip's excellent appnote

:>)

I have a nice marginal oscillator design we used for semiconductor lifetime measurements, and thought I'd once discussed it and posted an ASCII schematic here on s.e.d., but couldn't find it using Google.

--
 Thanks,
    - Win

So the issue is not so much that you want to build any old oscillator and then measure the crystal dissipation. It sounds like what you want to do is build an instrument, possibly an oscillator, that measures crystal dissipation?

Usually when you set about to build an RF oscillator circuit you want to aim for the fewest number of active components you can, to reduce the amount that they'll screw you up. You also want to reduce phase noise. Both of these goals can be met if you design your bias networks so that your active element squirts about the same amount of charge into your resonant network (your crystal in this case) on each cycle. This is good for getting nice stable oscillation, but not for getting information about the crystal's load impedance.

Building a bridge wouldn't be a bad idea except that you would need to keep the excitation frequency just so. Another approach would be to build an op-amp oscillator with very explicit AGC action:

___ .------------------|___|-------------------. | _ | | | | | o----|| ||----. ___ | | |_| | .--|___|--o | | Multiply | | | |\\ | _ | | '----|-\\ | / \\ ___ | |\\ | | >--o--o-->| X |--|___|---o--|-\\ | .--|+/ | \\_/ | >---' | |/ | ^ .-----|+/ === | | | |/ GND | | === | | GND V '--------. - | Level | .------. | Detect | | | | Output .--o-o-->| H(s) |----o------------------ | | | | .-. | '------' | | --- | | --- '-' | | | | | === === GND GND created by Andy´s ASCII-Circuit v1.24.140803 Beta

This circuit depends on the fact that with a 32kHz crystal you can find opamps with relatively low phase shift -- so make sure you do, and consider putting a bit of lead into the oscillator's loop somewhere to make it nominally zero.

The idea behind the circuit is that it keeps everything linear, so you should get a very clean sinusoid out of the crystal. The level out of the oscillator should be steady when you find a damping proportional to the crystal load resistance. Since the line marked "output" will be proportional to the damping as long as you have an AGC loop filter that achieves stability and servos the level to the same value every time you'll always have a good reading out of the thing.

If this were me I'd implement the loop shaping in a microcontroller and the multiplier in a multiplying DAC; this would not only allow me to play games with the inherently nonlinear effect of regulating the oscillator output by changing it's damping but it would also let me easily communicate the output level directly, since it would be available internally to the processor. It has the added advantage that

1 quadrant multiplying DAC's are easier to get these days than good analog multipliers.

You may feel more comfortable implementing the compensation in analog with a good analog multiplier. Since the oscillator itself will have a more or less integrating effect as you change it's damping you should be able to do this with just a proportional gain, or perhaps a PD. You may get faster settling overall with a PID controller; you'll have to experiment to find out.

--

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

Hey! That's a cool idea! It's potentially cooler than my circuit, but only if you need to detect _really_ small changes in the resonator Q. As described such a circuit would depend on the effective oscillator regeneration changing with signal level, and the slope of Q change vs. amplitude change would depend on this dependency. It seems like it'd be very sensitive to changes in temperature and just about everything else, though. It sounds like it'd be fussy and finicky to adjust, and that it'd require it as part of the operation of the thing.

Since that reminds me of a regenerative receiver for shortwave, I wonder if you could do something like that using a superregenerative oscillator, where you quench the oscillations periodically then measure the amount of time it takes for them to come back as an estimate of the circuit Q. You'd have to make sure to start it with a bang or otherwise feed it some pilot tone, but it would make a simple circuit and because you could guarantee enough regeneration for oscillation it shouldn't be as fussy.

--

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

This sort of thing is done in optical spectroscopy using a "ring-down" approach. There's an electronic version which is a very useful method of calibrating log amps, and is also good here, I think:

1. Build a garden-variety Colpitts crystal oscillator, tuned to the series resonance of the crystal (which will require a series inductance). This maximizes the actual mechanical oscillation amplitude for a given electrical swing. Put a small resistance in series with the grounded end, and attach a buffer amp there.
2. Add a diode switch that ac-grounds the hot end of the crystal when a current is applied. This will stop the oscillator and allow the crystal to ring down at its natural frequency.
3. Measure the oscillation amplitude as a function of time. You can do this by sampling with an ADC, or with a DLVA. You get about Q cycles per decay time, so you can do this quite conveniently and accurately.
4. The excess loss goes as 1/Q-1/Q_0, where Q_0 is measured before whatever guck you're using gets put down.

This trick makes a great on-line calibrator for DLVAs, since the decay is very accurately exponential with time after the switch transient dies away, so a good DLVA will produce a straight line graph.

Cheers,

Phil Hobbs

there is tons of info re crystals published by the US Army Research lab in Fort Monmouth NJ.

Mark

Intrigued by the application. Various thoughts arise from the fact that the fork flexure crystal is not the lowest loss resonator though the cheapest and quite convenient. Remember that the oscillator exactly equals the crystal loss in negative resistance. The uncontrolled oscillator does this by varying negative resistance with time/phase through the cycle. The agc oscillator varies negative resistance with a time constant of many cycles to give a linear region so the multiplier should be the best approach but complex. Most oscillators start off with too much activity and then back off a long way. You will encounter a problem with contamination where the crystal activity reduces due to particles , surface films. A convenient means of achieving small gain variation for agc is to vary the current through a bipolar transistor oscillator transistor after a precise level is reached. Works well if you don't start with too high gain. Finally the flexure fork may not be the best resonator config. The lower the resonator loss (the higher the Q) the smaller the tip force that can be detected. Fork flexure Q 30000? (In air) Fundamental Shear mode Q 200000 Overtone Shear mode Q 2000000 An extensional resonator not normally encountered could be a practical compromise.

Change "gr" for "h"

--
dd

Hmm.

32Khz.

I notice that 96Khz soundcards are getting cheap. I wonder if at that sampling rate, they will work up to nyquist. If so, 32Khz wouldn't really be a stretch.

Simply output a 32Khz tone on one channel through a largish resistor, with another channel monitoring the amplitude (through maybe a FET probe stuck next to the crystal). Latency will suck, I'd be surprised if you could close the loop at better than 10Hz.

Thanks to all those who replied, I did get one private email that pretty much hit the nail on the head, Jim Williams has published several oscillators with closed loop feedback using bridges. The first thing I built today uses a classical lamp stabilized bridge, and I'll look at the signal across the lamp with a diff amp. If that doesnt work, I'll try Phil Hobbs suggestion on ringdown, and a multiplier based unit isn't far off, as I have a lot of AD633 setting around.

What I'm doing is working through a range of problems on a instrument that has been under construction for two years, works some of the time and has random failures while taking data. We use a atomic force microscope as the heart of the system, and while ran in in AFM mode, its perfect. We have two ion lasers, one kryton, set up for visible red and two hard to get IR lines, and a argon, set up for 8 lines in the blue green, both of these are set up to be tunable with prisms. A tracking stage moves a focused probe beam from the laser to follow where the AFM tip is, and also hauls around collection optics and a adaptive optics rail that keeps the scattering from the illuminated sample properly fed into a monochromator with dual Fabry- Perot spectrum analysers fed into a cryocooled CCD detector.

The Raman scattering from the sample is enhanced by surface plasmon effects using gold and silver plated AFM tips. So the system is called "Imaging Nanoraman" I'm responsible for the lasers and the motion stages, having built a 6 axis microstepping drive that drives huge surplus Aerotech motion stages.

For various reasons, we've used the watch crystals with micromachined silicon tips on them as sensors, for samples that dont "tunnel" very well. Often times we get good results, then for weeks, NOTHING. We're in a lab that really is a adapted office building design, with huge airflows for chemistry, that shakes at a 7 hertz resonace and has glass sides. So far I've knocked out problems such as a lifted ground wire in the 3 Phase, a flourescent lighting system that runs at 32.6 Khz and no PFC or filtering in the ballasts, magnetic fields from the motor drive inducing currents in the preamps, magnetic fields from the PC monitors, (bought new LCDs, almost no emissions, yea!). We have a building full of RF and Xray based machines, so the line power is crud. My fellow technician has brought in a dehumidifier which has helped, and now it looks like barometric pressure is now a factor, with a active building pressure control system ran by a computer across campus using 8 ~70 HP fans. The lasers average 7 kilowatts each on 208, but can draw up to

1. They are 6 inches from the experiment, and I can't run the light through fiber as that would add a Raman signal. To find crud, I have a labview FFT running on the electrical and magnetic fields in the room at all times sampling at 5 Mhz.

All this floats on a optical table. 9/10ths of it is commecial gear using labview or proprietary software and hardware that is out of my control.

To backcheck the hardware, I'm building a simple AFM clone, without the data collection. I have a piezo autofocus stage out of a old urine analyser, a one axis motion stage and a crystal holder. A basic stamp (we had one setting around) is running as a fast frequency counter with

1 hz resolution, looking at a 4046 PLL runing as a 4x multiplier and the circuit (to be determined, whatever works) will be used to look at the dissipation and AM. I'm gonna manually fly the watch crystal with tip (control signal from 10 turn pot) into a piece of brass on the piezo and see what happens on my trusty TEK 7834.

So yes, to answer Tim's question, I'm building a cheap instrument to test a instrument. I'm only in this lab two days a week, and no downtime is allowable on the main rig.I have very little control on how the grad students set up the main rig, yet it falls on me to find the mistakes. So this is my little seat of the pants- desktop test rig. So can I get a grasp of whats going on with the tips. I suspect poor quality control as we glue them in house. This is a polymer lab, so the usual EE type test equipment is sparse and borrowing from other departments is a pain. I do have a good HP programmable oscillator and a sound card running GRAM. I'm a Ham, and have a decent milliwatt meter, hence the question about bridges.

Another option is the standard crystal test rig, which consists of a 8 Db RF attenuator on each side of the crystal, a RF source and a detector.

If I get something working well, I'll publish it on line.

So if you need a good sensor to measure very tiny distances, gnat's wing lift, baro pressure,humidity, coating thickness in vacuum chambers, detect various chemicals, or magnetic fields, and/or a vacuum guage, take a look at the humble 69 cent watch crystal, it may suprise you.

NEWS FLASH, I just got loaned a "Q" meter!

Yes, I do love my job!

Again thanks to all that replied,

Steve Roberts