Thinking about a 80MHz crystal - as a detector

[Winfield Hill]

I'm thinking about the possibility of using high-Q 80MHz crystals
in a sensitive electric-field detector.* You know, consider the
usual amplifier in a high-performance crystal oscillator, but
without ANY feedback path, and further modified so the amplifier
doesn't excite the crystal, not even a little.

Looking at crystal models, the loss-resistance element may be on
the order of 10 ohms, which implies a Johnson noise density on the
order of 0.4nV per root-Hz. So an optimum kT-sensitive amplifier
would need a similar noise level. Hmm, that could imply a rather
large JFET with excessively-high capacitances. Certainly the
JFET will be part of the crystal's tuning capacitance, but I'm
probably going to be limited to say 10pF or less. A 2sk146 JFET
has 1nV noise, but has 40pF of capacitance. A 2sk152 has 1.8nV
with 8pF, that's getting closer. But 1.8nV is 4.5 times kT for
a 10-ohm crystal...

Maybe a low-noise BJT amplifier would be better... I have some
2sd786 transistors, which state 0.55nV on the datasheet, for
Ic = 10mA. Oops, then r_e = 2.5-ohms and with a beta of 500 Zin
would be only 1.2k, not good enough to avoid loading down a sensor
with Q = 20,000. BJT base-current noise would be another issue.
Hah, JFETs suffer from a bit of current noise at RF frequencies,
according to AoE, but no doubt much less than a BJT amplifier.

Another issue, fundamental vs overtone mode crystals. I've read
that fundamental-mode crystals have much lower Qs than overtone
mode, e.g., Corning lists a 4:1 improvement for 3rd versus 1st.

* Don't ask about the application just now. It's a bit exotic,
the same experiment that's getting the 10kV 1us precision ramp.

 

[Stephan Goldstein]

I'm thinking about the possibility of using high-Q 80MHz crystals
in a sensitive electric-field detector.* You know, consider the
usual amplifier in a high-performance crystal oscillator, but
without ANY feedback path, and further modified so the amplifier
doesn't excite the crystal, not even a little.

Looking at crystal models, the loss-resistance element may be on
the order of 10 ohms, which implies a Johnson noise density on the
order of 0.4nV per root-Hz. So an optimum kT-sensitive amplifier
would need a similar noise level. Hmm, that could imply a rather
large JFET with excessively-high capacitances. Certainly the
JFET will be part of the crystal's tuning capacitance, but I'm
probably going to be limited to say 10pF or less. A 2sk146 JFET
has 1nV noise, but has 40pF of capacitance. A 2sk152 has 1.8nV
with 8pF, that's getting closer. But 1.8nV is 4.5 times kT for
a 10-ohm crystal...

Maybe a low-noise BJT amplifier would be better... I have some
2sd786 transistors, which state 0.55nV on the datasheet, for
Ic = 10mA. Oops, then r_e = 2.5-ohms and with a beta of 500 Zin
would be only 1.2k, not good enough to avoid loading down a sensor
with Q = 20,000. BJT base-current noise would be another issue.
Hah, JFETs suffer from a bit of current noise at RF frequencies,
according to AoE, but no doubt much less than a BJT amplifier.

Another issue, fundamental vs overtone mode crystals. I've read
that fundamental-mode crystals have much lower Qs than overtone
mode, e.g., Corning lists a 4:1 improvement for 3rd versus 1st.

* Don't ask about the application just now. It's a bit exotic,
the same experiment that's getting the 10kV 1us precision ramp.

Any possibility of inserting a transformer at the front?

steve

 

[Fred Bloggs]

Any possibility of inserting a transformer at the front?

steve

Or low impedance series resonance...

 

[Fred Bloggs]

Looking at crystal models, the loss-resistance element may be on
the order of 10 ohms, which implies a Johnson noise density on the
order of 0.4nV per root-Hz.

Prove it- what makes you think it is Johnson noise, because it is
represented as a resistor symbol?

 

[Winfield Hill]

Fred Bloggs wrote...

Prove it- what makes you think it is Johnson noise, because
it is represented as a resistor symbol?

Acckk, how am I going to do that? I'm making an assumption.
Maybe someone else can give us an answer, references, etc.

 

[Fred Bloggs]

Fred Bloggs wrote...



Acckk, how am I going to do that? I'm making an assumption.
Maybe someone else can give us an answer, references, etc.

It's back to basics for you. The book by Nye on the physics of crystal
should have the answer. Physical Properties of Crystals, Oxford, still
going through late editions, will be in nearly any technical library.

 

[Tim Shoppa]

* Don't ask about the application just now. It's a bit exotic,
the same experiment that's getting the 10kV 1us precision ramp.

Just speculating, not sure what you're actually detecting, but do some
Google searches for "microcalorimeter" and see if this is what you're
trying to do. To oversimplify they are looking for energy deposited
into a crystal giving some tiny phonons of quantum vibration, and
looking for this at extremely low noise levels.

I know the nuclear physics side (hint: these experiments are usually
buried deep in mines or inside mountains), but always got lost in the
solid-state phonon stuff.

Tim.

 

[John Larkin]

I'm thinking about the possibility of using high-Q 80MHz crystals
in a sensitive electric-field detector.* You know, consider the
usual amplifier in a high-performance crystal oscillator, but
without ANY feedback path, and further modified so the amplifier
doesn't excite the crystal, not even a little.

Looking at crystal models, the loss-resistance element may be on
the order of 10 ohms, which implies a Johnson noise density on the
order of 0.4nV per root-Hz. So an optimum kT-sensitive amplifier
would need a similar noise level. Hmm, that could imply a rather
large JFET with excessively-high capacitances. Certainly the
JFET will be part of the crystal's tuning capacitance, but I'm
probably going to be limited to say 10pF or less. A 2sk146 JFET
has 1nV noise, but has 40pF of capacitance. A 2sk152 has 1.8nV
with 8pF, that's getting closer. But 1.8nV is 4.5 times kT for
a 10-ohm crystal...

Maybe a low-noise BJT amplifier would be better... I have some
2sd786 transistors, which state 0.55nV on the datasheet, for
Ic = 10mA. Oops, then r_e = 2.5-ohms and with a beta of 500 Zin
would be only 1.2k, not good enough to avoid loading down a sensor
with Q = 20,000. BJT base-current noise would be another issue.
Hah, JFETs suffer from a bit of current noise at RF frequencies,
according to AoE, but no doubt much less than a BJT amplifier.

Another issue, fundamental vs overtone mode crystals. I've read
that fundamental-mode crystals have much lower Qs than overtone
mode, e.g., Corning lists a 4:1 improvement for 3rd versus 1st.

* Don't ask about the application just now. It's a bit exotic,
the same experiment that's getting the 10kV 1us precision ramp.

I don't understand this. Is the crystal itself going to be exposed to
the e-field, or is there an antenna of some sort, connected to the
crystal?

If the crystal is used directly, is its metallization the antenna?

Is there any advantage over using the crystal in the front-end, as
opposed to matching+amplification followed by a narrowband, maybe
crystal, filter? A well-matched fet amp can have a noise figure well
below 1 dB.

And of course, the crystal series resistance behaves, from a Johnson
noise perspective, just like any other resistor. Conservation of
energy, again. Ignoring microphonics.

John

 

[David DiGiacomo]

Another issue, fundamental vs overtone mode crystals. I've read
that fundamental-mode crystals have much lower Qs than overtone
mode, e.g., Corning lists a 4:1 improvement for 3rd versus 1st.

Wouldn't 80MHz fundamental crystals be mesa crystals?
How does their Q compare to bulk crystals?

P.S. I think you'll get more grant money if you use MEMS resonators
instead of boring old crystals. Nanotech would be even better.

 

[Phil Hobbs]

On 31 Mar 2006 03:08:55 -0800, Winfield Hill
<[email protected]> wrote:

I don't understand this. Is the crystal itself going to be exposed to
the e-field, or is there an antenna of some sort, connected to the
crystal?

The crystal's mechanical properties are actually somewhat anharmonic,
which is why the overtones aren't in exact harmonic relationship. This
could be used as a detection mechanism. Otherwise it sounds more or
less like it's being used as a high-Q tank circuit for impedance
transformation, which isn't altogether unreasonable.

And of course, the crystal series resistance behaves, from a Johnson
noise perspective, just like any other resistor. Conservation of
energy, again. Ignoring microphonics.

Providing that there's a dissipative process going on, this is right,
and that's what's going on in the crystal. The fluctuation-dissipation
theorem again.

<dim-memory>
There was that guy ten years or so ago who made a big stir with a
"lossless resistor" for switching circuits. I never really learned the
details, but he got a big IEEE medal or something for it. Does anybody
remember?
</dim-memory>

Cheers,

Phil Hobbs

 

[Tim Wescott]

Prove it- what makes you think it is Johnson noise, because it is
represented as a resistor symbol?

AFAIK Johnson noise happens to any dissipative 'thing' that shows up as
an electrical effect -- all that matters is the apparent resistance and
the temperature. AoE actually mentions this.

--

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

Posting from Google? See http://cfaj.freeshell.org/google/

 

[Tim Wescott]

I'm thinking about the possibility of using high-Q 80MHz crystals
in a sensitive electric-field detector.* You know, consider the
usual amplifier in a high-performance crystal oscillator, but
without ANY feedback path, and further modified so the amplifier
doesn't excite the crystal, not even a little.

Looking at crystal models, the loss-resistance element may be on
the order of 10 ohms, which implies a Johnson noise density on the
order of 0.4nV per root-Hz. So an optimum kT-sensitive amplifier
would need a similar noise level. Hmm, that could imply a rather
large JFET with excessively-high capacitances. Certainly the
JFET will be part of the crystal's tuning capacitance, but I'm
probably going to be limited to say 10pF or less. A 2sk146 JFET
has 1nV noise, but has 40pF of capacitance. A 2sk152 has 1.8nV
with 8pF, that's getting closer. But 1.8nV is 4.5 times kT for
a 10-ohm crystal...

Or an impedance matching network. One adjustable coil and a pair of
caps will do almost anything you want them to, and you don't have to
worry about the assembly being too narrowband compared to the crystal.
If that doesn't float your boat then you can get 3:1 trifilar broadband
transformers from Mini-Circuits that'll bring that optimal match up to
90 ohms -- still not good, but better at least.

Check out "Radio Frequency Design" by Wes Hayward -- he goes into detail
about noise matching in an RF environment (which you should already
know) but also about using transformer feedback to get a power match at
the same impedance as a noise match.

--

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

Posting from Google? See http://cfaj.freeshell.org/google/

 

[Ken Smith]

Looking at crystal models, the loss-resistance element may be on
the order of 10 ohms, which implies a Johnson noise density on the
order of 0.4nV per root-Hz.

Is there gas around the crystal? There will be noise from the atoms
smacking into it, if there is.

Crystals aren't very linear at very small or very large signals will this
be a problem?

Someone at Berkeley was using a tiny ring core to detect the magnetic
field that the electrostatic one implies. You may want to google on this
to see if it worked.

 

[Winfield Hill]

Terry Given wrote...

Wasn't it a transmission line followed by a dc-dc converter or
somesuch?

I vaguely recall reading a paper at the time, but forget the
details. ISTR his examples were all puny wee things, << 1kW

Puny, under 1000 watts? Some of us resemble remarks like that.

 

[Terry Given]

Terry Given wrote...



Puny, under 1000 watts? Some of us resemble remarks like that.

I wouldnt want a belt from one of your 10kV gadgets though
(electroporesis perhaps?)

I just scanned thru the last 6 years of IEEE professional comics, and
couldnt find a mention of it. but upon further recollection I think my
explanation is roughly correct, and the examples he gave were about 50W
or so. the lack of mention this century suggests it wasnt so great after
all though.

I'm doing some design work at the moment for a 200KVA converter. Only
problem is, there is nowhere to plug them in at home Even my
neighbour (a joinery place) only has a 60A 400V outlet. So I'm going to
make a 2kVA prototype first....

fault currents get quite exciting with the big stuff. IGBTs limit fault
current to around 10x their rated current (depending, of course, on Vg)
so my 300Arms converter has to cope with a peak fault current of about
4,000A. turning that off without going bang is a good trick.

I tried to estimate last night how much money has been spent training me
over the last 15 years, its a LOT. Tens of millions of dollars. No
wonder smart companies work hard to retain staff. I once destroyed
$1,000,000 worth of fancy flywheels, one at a time (OK, its not that
bad, I tried to operate them at their rated load etc, and they all
failed in various ways)

on an analogous note:

Air NZ is looking at laying off a whole lot of engineering workers in NZ
(they do all their own aircraft maintenance, and contract to others
too). I wondered why, until a friend who is in the military set me
straight. A few years back, we neutered our air force - we no longer
have planes with guns; any offensive air capability will only come about
through hiring saudi pilots so the air force no longer trains all the
associated personnel. So Air NZ cant hire pre-trained technical staff,
but has to train them itself. which it cant afford....

Cheers
Terry