I am looking for help on wideband amplifier design for a battery operated device for ultrasonic testing. The ideal wanted performaces are:
- high (> 100kOhm) input impedance suitable for a piezo pick-up
- signal bandwidth from 50kHz to 250kHz
- gain 40 dB minimum (better if 60dB or 80dB)
- quiescent current < 50uA
- 3.3V single supply

Here is the whole story of my troubles, I think there could be enough to start a PhD thesiys... everybody's help would be appreciated!

--- OP-AMPs gain blocks ---
A quick overview of the available micropower monolithic op-amp finds out that the lower the quiescent current is, the slower the amplifer becomes.
For example a MCP6144 (0.6uA) has a GBW of 100kHz, too small to get any gain in the desired band.

1-A) If anyone can suggest op-amps that could fit both the GBW (>25MHz) and quiescent current budget (50uA total) it would be great!

--- discrete design ---
I found interesting paper in "Horowitz & Hill - The Art of the Electronics" pp. 950-952. It explains that when working with such small supply current all the amplification is obtained enlarging the voltage swings and in this case the parasitic capacitance of the transistors in the stages that follow the first one become the bandwidth limiting factor due to "Miller" effect.
The same book suggests a "Miller" effect workaround by using RF transistors with lower collector-base stray capacitance and even a ready-for-lab schematic (see picture attached) capable of selectable (!) 80dB amplification with a 20kHz bandwidth and only 10uA of supply current. Although the bandwidth is 1/10th of what desired and the supply voltage is 5V istead 3V I considered this a good starting point, hoping that, using better modern transistor and increasing the bias current I might be able to match my requirements.

I was not able to test the exact circuit of the picture since the 2N4957 pnp is obsolete and cannot be found anywhere. Anybody knows where 2N4957 can be found???

Then I tried to follow the circuit of the picture substituting the obsolete parts with modern low-capacitance microwave BJT, such as BFR505 and BFT93 and:

2-A) SPICE simulations seem not working, I suppose because the circuit is working far below the minimum collector current that is tabled in the models of the transitors. Is there any way to work around this problem?

2-B) even experimental test seem not working on the circuit of the schematic I derived from the original one (see pdf) on bench test it reported a maximum voltage gain x2 at 100 kHz, and tampering with the gain resistors/capacitrs had no success. What am I wrong with moving electrons??

2-C) from the same text the suggested schematic is mentioned as based on series-feedback pair gain stages but no detailed explanation of how to design this topology is given. The only extensive work I was able to found regarding series feedback pairs (M.S.Ghausi - "Optimum Design of the Shunt-Series Feedback Pair with a Maximally Flat magnitude Response" - IRE transactions on circuit theory) is really of little practical help... anybody would be able to help with design procedures for series-feedback pair topology???

If you have any alternative idea for such a terrific amplification task please advise!!

 

1) try a few video amps, particularly CFB opamps. they are blindingly fast.
2) your design will work, but I would use a jfet upfront (a differential pair) driving a cascoded VAS, for speed and for gain.

take a look at ths4012, only with a jfet front-end rather than bjt.

 

1) try a few video amps, particularly CFB opamps. they are blindingly fast.
2) your design will work, but I would use a jfet upfront (a differential pair) driving a cascoded VAS, for speed and for gain.

take a look at ths4012, only with a jfet front-end rather than bjt.

Thanks for the fast reply!

1) I am not familiar with current feedback amplifiers, however I had quick look yet and as far as I have seen there is nothing with a supply current of few microamps. OPA683 for example has a GBW of 1400MHz but draws more than 2mA. If you know any CFA running at 50uA please advise, this would be the best solution for sure.

2) regarding the design... I was not able to have it working in the real world... how can I find complemetary jfet? I am not able to find p-channel jfet. What do you mean for VAS (I am not too familiar with achronyms)? Do you have any schematic example I can study?

3) THS4012 draws 11mA and won't go below 9V... the smaller brother OPA2889 would go at 3.3V with 75MHz GBW but drawing 700uA, 20 times the power budget I can afford.

 

1) most cfb opamps have a diamond buffer output stage so they do draw more in idle current.

the 50ua iq requirement is likely going to kill any realistic opamp idle.

2) p-channel jfets do exist. 2sj/j family for example. you will unlikely find a true complementary but you will be able to find n and p channel jfets.

3) the topology used by ths4012 is a very fast one. the most interesting thing is its VAS. you can implement that with discrete components and adjust the idle current to suit your needs.

 

2) p-channel jfets do exist. 2sj/j family for example. you will unlikely find a true complementary but you will be able to find n and p channel jfets.

I agree. They do exists. But they seem to be unavailable from common suppliers such as digikey and mouser. Only p-jfet I can find are optimized for choppers and not for amplification such as MMBFJ175 (ON-semi). Can you suggest me other accessible sources? Thanks in advance.

I will try to study more the topology of THS, thanks for the advise.

 

I don't use those guys much but even the switching jfets have reasonably high gm. you just need to watch out for Ciss (less than 5pf should be sufficient for you).

alternatively, you can try those rf jfets (1-3pf Ciss typical).

you can also try mosfets - they have higher Ciss so that may not work out.

audio jfets are also very fast, fast enough for this operation. in the worst case, you can just use a jfet follower and then amplify the rest with bjts.

finding the jfets isn't going to be an issue. I think the likely issue is to control the idle current down to 50ua.

 

how about opa347/348 (and their dual/quad) versions? they are micropower cmos opamps, idling at less than 20ua/45ua per amp. But they are on the slow side so it may not be sufficient to get to your gain unless you serial them - then you are effectively cheating.

worth a try at least. and you can explore it a little bit further from there.