A nice toplogy for low level amplifiers

Linear and non linear

A nice toplogy for low level amplifiers

Postby fusordoug » Fri Jul 16, 2010 11:31 am

I have been using variations on a theme for quite some time with this basic idea. I think I saw it first used in an old Marantz tone control amplifier, well before this came out as AN-222 in the National Apps book, but have since seen it and variations all over the business. It works real well, is cheap, and popular.
The topology uses two opposite sex devices (for bipolars that would be PNP and NPN), DC coupled, with feedback. One of the cleverer implementations is shown in the national apps handbook, used in this case for low noise preamplification of moving coil pickups that have very low output impedance, but the circuit adapts easily to lots of other cases, including the version I'm now using on my 3He neutron detector, which is not linear, but has an adjustable threshold, instead. This topology tends to slew much quicker in turn-on than turn-off, which makes it nice for signals that have that anyway -- geiger tubes, proportional tubes, phototubes and so on. With a threshold set (bascially via bias) it tends to improve the rise times and fall times both for those devices. But first, lets look at the National circuit which is pure linear and very high grade for audio kinds of things. This version is loaded with mainly AC/Audio features, but with slight variations (which turn out to be cheaper) is also a great one for physics metrology applications.

National's version
Natpreamp.gif (4.9 KiB) Viewed 2233 times

Notice the transistors are connected so that when the first one turns on, so does the second. Also note that there is a negative feedback loop to the emitter of the first NPN. For audio, and to make biasing easier the emitter of the PNP is bypassed, and for low noise at low input impedances, the first transistor is actually two special NPN's in parallel. Most other uses of this circuit don't have those. For normal line level audio kinds of things, we skip the bypass, use cheaper transistors, and increase all the R values to suit that application. And instead of deriving our bias from the PNP emitter, we might just use the supply rail for gains on the order of 10 or so -- with essentially perfect linearity. This circuit is a just a gem. The second attachment is their words gushing about how nice it is, and having built a few of these -- they really are that good, and fast even with the slow transistor they used.

National Semiconductors words on this

For you audio guys, just use cheaper NPN and PNP resistors, use the NFB loop R values to set total gain, making sure the the NFB each transistor already sees due to emitter R to collector R gives you a fat gain margin when the stage gains are multiplied. I like to use lower noise transistors than 2n390X types here when I care about the quality.

Now, for the physics crowd who is usually using negative going input pulses from various sources, what you do is make the first transistor PNP, and the second NPN. If you use no bias (just an R to ground on the input) then the circuit draws zero power when there is no pulse -- it can run on batteries for a long time that way with no on-off switch needed. This gives a threshold of one Vbe, of course, or slightly more as the first transistor has to turn on enough to turn on the second one as well. For this case, I run the collector of the second transistor to ground through it's load R and skip the negative feedback altogether if I just want an output that will drive a counter. For a higher threshold, you can add a diode in the emitter of the first transistor as well. Biasing this diode with about 100k to the supply rail makes the threshold "tight", and for current inputs, using a variable R to ground on the input lets you set the actual current threshold for signals like from phototubes. That bias resistor does draw some power though, and for uses here I omit it and allow the threshold to be a little more squishy. In real use as a preamp for a 3He tube, played then through an audio amplifier, you can then tell a little more about the input signal -- for one thing, you get a wider and taller pulse if the tube was hit by multiplie neutrons withing its time window (a few uS) vs just one, so if you are looking for bursty kinds of things, you adjust the input threshold pot to just hear or count them, not single hits. Since there is no NFB in this variation, it's probably not good for feeding into a PHA, and in fact for that signal the main thing that changes with one vs many hits is the pulse width, not so much the height anyway. Believe it or not, the circuit I show here drives coax just fine -- it has low output impedance during the risetime, but then is essentially open circuit during the fall, so you just see the decay time lengthened a bit by the coax parallel C, but really nothing much in ringing at all. Nice, as this one can go in a tiny box with a pair of Li cells and run for years with no on-off switch. I am aware that the threshold does vary a little with temperature in this design variation, but in practical use the difference hasn't been noticeable at all, as it's not a big fraction of the 2 Vbe at human being in the room kinds of temperature swings. It might bite you if the thing saw a lot of heat from some local source, but so far, that's not the case in my applications of this :)

Here is the circuit I am using now. You can use the 2N390X transistors here, or some higher gain audio class PNP as you like. All are plenty fast enough for a 3He tube with its microsecond pulse widths, I'd look for something faster for short phototube pulses. You will note I'm using a fairly higher volts than the norm on the 3He tube, this is to get into the region of a little bit of gas-gain so I have a stronger signal in the first place -- always fix up things as near the source as possible! The emitter resistor in the NPN is not there for NFB -- it's merely to limit peak currents in both transistors in cases of huge pulse noises getting into this.

My preamp design for 3He tubes.
mypre.gif (7.69 KiB) Viewed 2229 times
Why guess when you can know? Measure!
User avatar
Site Admin
Posts: 45
Joined: Wed Jul 14, 2010 1:59 pm

Return to Analog

Who is online

Users browsing this forum: No registered users and 1 guest