HF FET Q-switch for pulsed NMR: broadband design?

I think we just use back to back diodes to protect the front end during the pulse. (I didn't design it... I looked at the schematic.. 1n4148's.) The NMR signals are small and the input tuned to 50 ohms, so after the pulse the impedance of the diodes can be ignored. I can sell you some 1n4148's for ~$1 each.. plus shipping :^)

George H.

reddit.com/r/ece:

e that reduces the Q of a (normally) high-Q resonant tank during a big tran sient. Following the pulse, the Q-switch effectively disengages, making the Q of the tank once again large.

cessively following the pulse, drowning out the NMR.

0 KHz and 5 MHz, or better.

lots of optical ones, but none for HF.

uld I use?

he pulse.

lse

for

that's a different issue, the common setup is anti-parallel diodes in serie s with the transmitter so it doesn't load the receive signal, and anti-para llel diodes in parallel with the receiver, with a quarter wave stub from th e probe to the diodes to transform the short into and open seen from the transmitte r.

Q switching is for stopping the probe ringing from the transmit pulse, so i t'll be ready to recieve

-Lasse

What's the bottom of Q1 connected to? (Ground?)

of reddit.com/r/ece:

ice that reduces the Q of a (normally) high-Q resonant tank during a big tr ansient. Following the pulse, the Q-switch effectively disengages, making t he Q of the tank once again large.

excessively following the pulse, drowning out the NMR.

500 KHz and 5 MHz, or better. 7

nd lots of optical ones, but none for HF.

could I use?

the pulse.

pulse

's for

ies with the transmitter so it doesn't load the receive signal, and anti-pa rallel diodes in parallel with the receiver, with a quarter wave stub from the probe

ter.

it'll be ready to recieve

OK Thanks Lasse, So you short the line for a short time after the end of th e spin flipping pulse? 5MHz is not that fast.... I wonder high much voltage? Maybe an analog switch could do it?

George H.

Q switched NMR probes have been commercially available for many years but the market is very small - seldom needed. For the high power levels involved (especially with low gamma nuclei where kilowatt level pulses are routine) FETs are a bit out of their element and PIN diodes are used (switch in a resistor). For broadband applications the diode should be close to the sample coil and must be in a non-magnetic package. For fixed frequency applications, the diode can be part of the T/R duplexer placed some multiple of 1/4 wave away. Typically, the damping is turned on during or after the transmit pulse and turned off just before the receiver is turned on. MRI system coils have a very high Q - simply because they're large and active damping is routinely needed. Analytical systems generally don't need active damping save the few cases where a spin/quadrupole echo technique can't be used.

Are you dealing with low gamma dilute quadrupoles, some abundant spin with potential radiation damping problems or NQR?

Suitable PIN diodes can be gotten from Microsemi.

UFL NHMFL?

--
Grizzly H.

I guess a big ole R in line with a bypass set of contacts will do it!

Really don't take me serious there, but the concept is there, I think you need some a lot more robust components of course.

Switching time I am sure is very important, do you have that data?

Jamie

Microseconds. The transmit pulses have to be short enough to provide uniform excitation over the spectral range. There are resonances which may be very narrow (order 0.1 Hz) spread over 10s of kilohertz for spin

1/2 liquids (1H, 13C, 15N...) - an acceptable pulse width might be 3-5 us. In the case of solids, interactions between spins might give a linewidth of the order 100 kHz. An acceptable pulsewidth in this case might be 1-2 us. The power levels required depend on the sample size - (flux density of the RF field). To achieve optimal excitation for a 5 mm dia. sample at the 1H frequency, a power level order 100W would be required. To achieve optimal excitation for a non-spin 1/2 nucleus (broad line, strong interaction with nearby spins, quadrupolar splitting...) at a much lower frequency a power level > 1kW might be required.

The issue presented by the OP is sometimes encountered with very low frequency spin systems - the ringdown time of a high Q resonator is comparable to the ringdown time of the nuclear response. Messes up the spectrum something awful. Can't do good science when that happens.

There are spectroscopic techniques to get around this issue but for certain classes of interesting systems you can't always use them - the only alternative is whack the resonator ringdown in something less than a us so you can see the NMR (or NQR - nuclear quadrupole resonance) response in the following 10-20 us.

--
Grizzly H.

Hah, I've had that exact problem in the distant past, a coupling stage in my amp was ringing... somewhere near the resonant freq.. which made for weirdness in the 180 deg pulse sequences.

I had no spectrum analyzer back then... looking at the spectrum even with a DDS FFT scope might help... you might use the time window to look at different parts of the pulse. (Well I'm assuming the OP is having a problem similar to mine.) George H.

There're good reasons to broadband the entire spectrometer and bandlimit only where it's absolutely necessary - the occasional matching network, keeping PA noise on one channel from leaking into the receiver of another, respecting nyquist etc.

You might find this interesting:

No room on the chip for analog filters but the spins don't care. I believe he's using a 60 GHz process so phase shift inaccuracy and timing jitter are insignificant.

$10 NMR on a USB stick - probe and magnet not included ;)

I think the OP is actually looking at NQR signals - depending on the isotope he could be working anywhere from 500 kHz up to ~30 MHz. NQR resonances tend to be moderately broad - kHz - a Hahn echo sequence may not find enough coherence left to produce anything useful after the LC response settles.

--
Grizzly H.

Nice thanks,

I don't know much about NQR... way back when a (retired) professor I knew was growing crystals to try and do NQR. I think detecting the resonance CW... maybe one of those marginal oscillator type of things.

George H.

We have tried to create a super-regenerative detector with FETs, then one with vacuum tubes to see if it performed better, but our S/N is very low with this implementation, after years of trying.

The pulsed rig has much better S/N but is not good for sweeping/searching for unknown NQR.

Yes, UF-NHFML here. I am @ UF , not tallahassee however.

This is for pure QR spectra of 14N samples (as well as solids which we susp ect happen to have 35Cl resonance at low frequencies).

In any case, many of the comments here are have been dead on - it seems eve ryone understands my application.

Indeed, it is the ringdown of the coil after the pulse that is the problem. The coil rings down for "about 20 time constants" of 2Q/omega. This isn' t a problem for VHF, even for F = 30 MHz and Q = 50, this is pretty sho rt compared to the T2* of the samples (~ 500 us).

But say F = 3 MHz. If you still want Q = 50, I calculate around 100-20

0 us of coil ringing, and indeed I have found that is the case.

Here is a plot of the data taken from that run.

.png

Here is a signal obtained in hexamethylenetetramine @ 3 MHz. http://i.imgu r.com/PX2MLEH.png

The window is 2 ms long, so you can see there is at least 200 us of interfe rence from the coil ringing following the pulse, which is why the FID has a weird shape for the part following the pulse:

Fun photgraphs: nice FID I got the other day in NaClO3 at ~29 MHz - no coil artifacts very visible. Nice FFT (red trace) too.

J

One user asked: How many volts? probably some kV - its a 50 W amplifier an d a hard problem to determine exactly how many volts, but kV is a good gues s for a parallel resonant probe with a Q of around 50.

:

e:

r of reddit.com/r/ece:

evice that reduces the Q of a (normally) high-Q resonant tank during a big transient. Following the pulse, the Q-switch effectively disengages, making the Q of the tank once again large.

g excessively following the pulse, drowning out the NMR.

n 500 KHz and 5 MHz, or better.

027

ng

ound lots of optical ones, but none for HF.

s could I use?

ng the pulse.

e pulse

48's for

eries with the transmitter so it doesn't load the receive signal, and anti- parallel diodes in parallel with the receiver, with a quarter wave stub fro m the probe

itter.

so it'll be ready to recieve

the

e?

Finally ...

"So you short the line for a short time after the end of the spin flipping pulse?"

That is exactly what I want to do. I believe the paper in OP is doing some thing else - q-damping during the entire pulse. It is better to short it q uickly following the pulse.

I have an idea to use the logic pulses I have access to from my rig, differ entiate them (resulting in peaks at the corners of the pulse) and using thi s to drive a switch.

The details are fuzzy of course.

eddit.com/r/ece:

that reduces the Q of a (normally) high-Q resonant tank during a big transi ent. Following the pulse, the Q-switch effectively disengages, making the Q of the tank once again large.

ssively following the pulse, drowning out the NMR.

KHz and 5 MHz, or better.

ots of optical ones, but none for HF.

d I use?

Re PIN diodes and an MRI application, see MicroSemi's nice PIN Diode handbo ok, pp.81-83:

They make PIN diodes rated at 2kV (e.g. GC4604, GC4605).

-Rich S.

--


*** 
This e-mail message is intended only for the designated recipient(s) named  
above. The information contained in this e-mail and any attachments may be  
confidential or legally privileged. If you are not the intended recipient,  
you may not review, retain, copy, redistribute or use this e-mail or any at 
tachment for any purpose, or disclose all or any part of its contents. If y 
ou have received this e-mail in error, please immediately notify the sender 
 by reply e-mail and permanently delete this e-mail and any attachments fro 
m your computer system. 
***

I need a Q-switch for HF (not optical frequencies).

Have you looked at using an echo sequence to get away from the breakthrough? A little extra effort to set up and there's some loss of sensitivity but you can avoid a lot of hardware work for now - e.g.:

Go Ota, Hideo Itozaki, Nuclear quadrupole resonance echoes from hexamethylenetetramine, Solid State Nuclear Magnetic Resonance, Volume

30, Issues 3?4, October 2006, Pages 135-140, ISSN 0926-2040,

--
Grizzly H.

I have personally never observed an echo with my rig, but I could since it is already set up to deliver sequences of up to 3 pulses.

It never occurred to me to try since echo trains are normally used to measu re T1 and I never has a need to do so, but you are correct that it would im prove my SNR to have echo signal present. That is a good idea, and attacks the problem of poor sensitivity from another direction.