receiver / driver problem

Hello everyone.

I'm working on an ultrasonic detection system. We're driving a Piezo element on the end of a long coax cable, and we're driving it with a voltage spike anywhere from 100V to 500V, with a rise time around 15ns. The return signal is an ultrasonic pulse around 5MHz, and relatively weak (0.5mV).

I'm currently working on the receiver circuit, which listens to the same co-ax, next to the driver circuit. I'd like a way of passively surviving the spike.

I initially ran 5K in series with the input, and shunted that to gnd with 1N4148s (see below). The problem is the capacitance of the diodes draws too much current through the resistor and kills the signal. I could reduce the resistors, but then that starts to divert more power from the Piezo. I guess I could look for lower capacitance diodes, or run two in series, but it seems like there's probably a better way.

Anyone have any ideas?

..Piezo 5K ..Input>------////-----*------>LNA .. | .. | .. --- .. / 1N4148 .. --- .. | .. | .. | .. ----- .. --- .. -

Reply to
Pat
Loading thread data ...

I'd suggest a small series cap, a few pF maybe, a gated active clamp (a gaasfet would be ideal, or maybe a schottky diode pair) and then a fet amp. 5 pF is about 6K at this frequency. The clamp could run in the couple ohms range when on and maybe a couple pF when off.

Hmmm, maybe something could be done with a tuned circuit and a clamp. That would give better signal transmission into the fet, and the active clamp would kill the Q and shunt the signal to ground.

-----C--------+-------------+-------- amp | | L d | g------ control gnd s | gnd

something like that?

John

Reply to
John Larkin

Hi Pat,

Use a T/R switch in T-configuration (three diodes) with a few milliamps of quiescent current through them. This also avoids the rather large SNR loss in the 5K resistor which could go. Also, try impedance matching of the transducer to further increase your SNR. If you can't do that at the far end where the transducer is located you could also attempt to do that at the other end but that will come with a larger bandwidth penalty. If you opt for a toroid transformer here make sure that the core can handle the pulse power without much saturation. "Number 43" ferrite material is a good start and very available.

If you can get enough SNR with less pulse amplitude you may be able to build the T/R switch from diodes such as the BAS21. You could also build an active T/R switch (controlled FETs etc.) but that is usually too much hassle because you would have to generate the blanking pulse and this pulse would need to be very clean spectrally to avoid modulating into your signal.

SPICE is a great tool to model this whole scenario including the coax. The only thing that I found not to be handled well by SPICE is the transducer ringdown behavior, mostly because that is tough to model.

Regards, Joerg

formatting link

Reply to
Joerg

Ideas:

Replace the resistor with a small power MOSFET that you turn on after the spike is over.

The zero bias impedance of zener diodes is usually higher than the zero bias impedance of a recifier diode of the same size.

Design the LNA so that its input can swing several volts without damage and arrange the diode to be back biased instead of a zero bias. Normal diodes act a lot like variactors.

--
--
kensmith@rahul.net   forging knowledge
Reply to
Ken Smith

Hi, Joerg:

Thanks to you and everyone else who responded. There's a lot of good ideas here.

I must confess some ignorance here. I don't really know what a T/R switch in the T configuration is. I've seen T/R switches with four diodes used in a bridge configuration, but what is the three diode version?

I'm guessing it's something like this (plus some baising for D1 and D2):

.. D1 D2 ..Input>------|

Reply to
Pat

Hi Pat,

Almost, but D1 and D2 are anti-serial and must be supplied with current from the center into inductors at either end. You can leave off D3.

This article describes TR switches:

formatting link

It's for Bluetooth and you are dealing with much higher voltages but the architecture is what matters. Just don't use PIN diodes as these are for long carrier lifetime applications, typically in controlled RF switching. Here we are dealing with a passive TR switch and that needs fast diodes with enough breakdown voltage for the pulse. There are plenty of SMD available in SOT-23 package, some even as duals with common anodes which makes it ease to feed the current from a positive supply. Also, with a double diode you usually find nice matching. The diodes need to be fast so the receive path is open the nanosecond you have to read echoes. In my field that is the toughest task because we have to be able to receive immediately following the pulse.

Another important factor is the first receive amp. It needs to recover from saturation instantly. Since it does not have to be Hi-Z that won't be a big deal. I don't know what spectrum you are looking at but if it is below 25MHz or so check out the AD603 from Analog Devices which also features a nice and fast dB-linear gain control. In case you are higher than that in frequency you'd have to roll your own amp and if you need gain control that would be where PIN diodes come in.

Ringing is yet another issue. Your pulse should decay within a few cycles or your bandwidth could become too low. Also, ringing jeopardizes the ability to detect echoes right after a pulse.

Regards, Joerg

formatting link

Reply to
Joerg

Sorry about the belated responce, but I was pulled of this job for a while. I now have a circuit designed and recently bb'ed it. So far it seems to be working really well.

Thanks for the help Joerg.

Reply to
Pat

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.