Design Suggestion Requested

Bridging these only get you 1140Vpp

Maybe a tube amplifier will work.

D from BC myrealaddress(at)comic(dot)com BC, Canada Posted to usenet sci.electronics.design

sheet/PA89U_I.pdf

Look again: a single unit with +/- 600 V supplies will give 1140 Vp-p. Bridged, the load sees twice as much.

But before you reinvent the wheel, consider this:

-- Joe

1140Vpp

Oops..that's right.. Must be getting tired.. :P Bridging the PA89s would yield 2280Vpp.

D from BC myrealaddress(at)comic(dot)com BC, Canada Posted to usenet sci.electronics.design

Any idea of the actual load impedance?

How accurate does the drive have to be?

John

You'll need an opamp followed by two common base/gate stages. These in turn control a cascade of P-channels from V+ and N-channels from V-. To get the full swing you'll probably need +/- 1050V stabilized, or higher if non-stabilized so that they never dip below +/-1050V.

For the N-channels side of the power stage you might get by with a single FET because those can be had with 2500V rating (but I'd still go with a cascade to get some margin here). P-channels don't come that high so cascading will be necessary. 500kHz is not that easy. I haven't seen your transducer in the document at quick glance. Tell that company to make their PDFs text searchable. Anyhow, depending on the transducer's capacitance you may need a lot of power since single sinewave bursts will not allow you to use resonance.

Setting the bias current is not trivial. Can either be adjusted or needs some sort of servo.

And be very, very careful. One wrong move and bzzzt ... cardiac arrest or uncontrollable fibrillation ... poof.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

1140Vpp

Only 800Vpp.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

Right, but look how much more there is to deal with than just the output voltage. The transformer option looks interesting.

-- Joe

This is a possibility:

V+----------+-----------------------+ | | | | L L | | | | +--------load-----------+ | | | | | | | | | | ---| | nfets or npn's | |----- | | | | | | | | gnd gnd

where V+ is a hundred volts or so.

John

I'd have suggested a transformer but the OP wants 10Hz lower cutoff. I don't really know why, that makes it more elaborate and expensive.

Usually pulsed ultrasound at least in the medical and NDT situations I dealt with is done with hard pulsing. Either unipolar or bipolar. The natural bandwidth of the transducer takes care of wave shaping and if it has the usual PZT bandwidth of 30-40% the slight riniging doesn't make much of a difference because receive will be bandwidth-limited about the same way. We usually send 2-3 pulses per burst.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

Maybe, maybe not. The feedback can come from a low-voltage winding if you use a transformer; depending on the power requirement, the larger pot cores could do it all with ease. It will take quite a lot of fine-gage wire, though.

The scientist I am doing it for says it is important it be driven with a sine wave. I do not know how much filtering by transducer bandwidth there would be if I pulsed square wave. The scientist wants the same pulse output by the Ritec Ram 1000 in the lab and do it with a much smaller package.

The 10 Hz was a rough guess. The bandwidth needs to be wide enough in the direction of 0 Hz so there not be significant ringing.

The pulse is to be one sinewave cycle in the vicinity of 450kHz. The sinewave cycles are to be pulsed at a rate of 100Hz.

The Ritec uses a transformer. It appears the answer for me is to use one also. I hope there is one available off the shelf that can do this. Otherwise I must design it and wind it myself.

Joerg wrote:

--
To reply directly remove the sj. from my email address. This is a spam 
jammer.

The accuracy might 5% or more. The scientist says that accuracy is not as important as repeatability.

--
To reply directly remove the sj. from my email address. This is a spam 
jammer.

--
To reply directly remove the sj. from my email address. This is a spam 
jammer.

I do not have a good idea of load impedance. I do not have a way to measure how much current is being supplied. The best I can do is duplicate the Ritec RAM 1000's output specs.

--
To reply directly remove the sj. from my email address. This is a spam 
jammer.

Can you respond below or within the text of posts? Makes it easier for others to follow.

If the transducer is from PZT material it'll ring out the way it wants to no matter whether you drive it with a sine wave or a pulse. Except that sine is a whole lot more expensive and space consuming to do. How much ringing there will be depends, for the most part, on the backing material.

If you want to try that out here is a simple test: Couple the transducer and a wideband hydrophone via water or some ultrasound gel. Excite it with a function generator. Use a single 450kHz sine cycle, record the hydrophone response with a digital scope. Now excite with a unipolar pulse, then with a bipolar pulse and record all that. Now compare the recordings. The pulse width for the square excitation should be around

600-700nsec, in the bipolar case separated by 1.11usec.

The amplitude doesn't matter since you'll only be looking at wave shapes.

That contains one heck of a power amp. If you need only one 450kHz sine wave cycle every 10msec you can make it smaller. But what about all the other stuff in there, like the receiver?

Well, the RAM 1000 cannot go that low, not below 50kHz according to the spec in your link. Quote, page 1 "Three Standard Frequency Ranges covering frequencies from 50 kHz to 40 MHz."

Then you'll lose the ability to go below 10kHz or so, just like the RAM

1000 cannot do that. I think you do not need to provide a bandwidth down to almost DC. Not that it's impossible but it'll be a whole project unto itself, building something that probably doesn't exist yet.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

Not much info in those pages either. No impedance, no bandwidth, no backing material info and the RTS cert is nearly unreadable because this whole document seems to be a scan of a sales brochure, not a real datasheet. Also, I have no idea what they mean with "gain". Passive piezo transducers are, well, passive. They have no gain.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

I have contacted the manufacturer of the piezoelectric driver. It appears I got some information wrong.

It is a narrow band device. So it may work to square wave pulse it and depend on its narrow band to filter. But the Ritec RAM 10000 does synthesize a sine wave.

Mechanical loading shifts the resonant frequency, but it will still be narrow band.

Tech support could not tell me what kind of electrical load it will present. They will get back with me on this.

I am not necessarily going to duplicate the specs for the Ritec RAM

10000 because it is capable of an enormous amount of pulsing power which might not be needed.

--
To reply directly remove the sj. from my email address. This is a spam 
jammer.

The 10MHz version in the plot on page 7 looks like regular wideband PZT to me. A narrowband transducer would ring out for a long time.

Interesting, that data is usually supplied with transducers. You can measure it. Since you seem to work in a university/research environment you could ask around if someone has a HP-4191A or another impedance analyzer. Best to bring the transducer there because those analysers are very heavy.

Power amps are usually of very low output impedance. While the spec may say 50ohms that usually means you should not load it with anything less at high power levels to avoid blowing the final stage transistors.

Just do the pulser plus hydrophone test versus the Ritec. If you get the same results (or close enough for the scientist) life will become much easier :-)

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

I wonder if the low frequency requirement is because the person specifying the system is going to generate short sine wave pulse trains, and thinks DC is needed to kill the motion of the transducer. Trying to think out of the box here, could you kill the amp then slam some fet switch across the transducer to kill the sensor movement? If so, then maybe a transformer scheme could be used for the sine wave if this crowbar scheme works.

This is kind of fun. I designed a part of a laser measuring system eons ago and wished I could have seen the whole project working. [I had a contract for a very small part of the circuitry.] The guy designing the measurement system had an entire floor of a multistory building to play in. Everything in the building seemed normal, until you get out of elevator on the top floor and they had their own little Skunkworks lab..