Here is a tested design for a simple, relatively low-noise, JFET-input, 3-transistor audio preamp, designed to satisfy various goals:
The input transistor Q2 is an N-channel JFET. Each particular JFET sets a particular output bias voltage, so the JFET must be selected from among multiple devices for the desired bias value. Fortunately, due to the low current involved, desirable devices can come from a wide range of device types.
JFETs also vary with respect to gain and noise level, so best operation requires some selection there as well. However, other than exceptionally bad noise in particular devices, a wide range of JFETs function surprisingly similarly.
The input resistor R1 just drags the JFET gate to ground potential and so reverse-biases the FET. The value should be at least 10x the signal source impedance, but probably could be as high as 100 megohms, if that was useful for some reason. Larger resistors do produce somewhat more Johnson noise, but that is normally "shunted" or "shorted out" by a low-impedance signal source.
The second transistor Q3 is a bipolar PNP. One might think to make it a low-noise low-current-linear device, but in practice that does not seem to matter much. Many different PNP devices can be used with no circuit change and remarkably little difference.
The resistor across the base-emitter junction of the PNP R2 sets the "standing" or "quiescent" current for the input transistor. The PNP operates on the "knee" of turn-on, at about 0.6V. The recommended value of 5K allows about 0.6/5k = 120uA to flow through the JFET before the PNP starts to turn on, which then increases the output voltage and starts to turn off the input transistor. Since feedback keeps the current through the input transistor almost constant, resistor R2 can be seen as a particularly simple form of constant-current source.
The low operating current allows the JFET gate to operate at a bias voltage much lower than the specified "Gate-Source Cutoff Voltage," thus making a wide range of devices usable. The best JFET current for lowest noise can easily be checked by varying the PNP base-emitter resistor value.
The current in the second transistor typically is much lower than in the input transistor. That typically reduces the gain we can expect from the second transistor.
The PNP load resistor is missing. The original design, and every other design I have seen, had a load resistor to ground. However, removing that resistor seemed to produce more gain and less noise. The current in the second transistor apparently works against a larger load resistance to produce a larger voltage result. The results seem fairly independent of device characteristics, and the circuit without the resistor simulates just fine.
The output transistor Q1 is a bipolar NPN as an emitter-follower, and is also noncritical. For example, a high-gain Darlington device increased the open-loop gain by all of 3dB, and had even less effect on the closed-loop gain.
The emitter-follower output does reduce the maximum positive output by 0.6V. So if we expect to work well with a 4V supply (a 9V battery at end-of-life), we have a maximum output of 3.4V. That means, ideally, we would want the output bias to be around 1.7V, although we can vary that with minimal impact.
The output pull-down resistor R3 provides the energy for negative-going parts of the audio output. Since the emitter-follower can only source current, this resistor will define the maximum load which can be driven in a negative direction. Thus, it should be small, typically under 1/10 the input resistance of the next stage. However, this resistor also is one of the main contributors to power consumption, so it also should be large. If the output bias is set (by selecting the FET) at 1.5V, a value of 1K will sink about 1.5mA which the battery must then supply whenever the preamp is operating.
Battery life for a new 9V alkaline battery with a 2mA load should be about 330 hours (13.75 days or about 2 weeks) of continuous operation. Less current means better battery life.
APPROXIMATE 9V ALKALINE BATTERY LIFE
Load Hrs Dys Wks 1mA 750 31 4 2mA 330 14 2 3mA 220 9 1+ 4mA 160 7 1 5mA 120 5 0 6mA 100 4 0 7mA 90 4- 0 8mA 80 3+ 0 9mA 70 3 0 10mA 64 3- 0
Power comes in at the upper left. The JFET is at the lower left, the PNP at the top, and the emitter-follower NPN at the right. Here each transistor is in a little terminal strip so different devices could be tried fairly easily. The diodes at the bottom are the remains of other experiments. The blue component at the right is a cubical 1uF film capacitor on the output. Note that few interconnect wires are required.
The most direct way to set the output bias is to drop in different FETs until some produce an output bias in the desired range. These could then be soldered directly into other preamp boards, if several were being constructed.
The feedback network feeds back returns all of the DC bias, but
only a fraction of the AC voltage, which sets the gain, here 30dB.
What amplification is that? Well, first we know that 30dB is just
20dB plus 10dB.
We also know that 20dB is an amplitude ratio of 10.
So 30dB would be 10 x SQRT(10), for an amplitude gain of about 31.6
times.
Alternately, since 60dB is 1000x, an amplitude ratio of 30dB is
SQRT(1000) or 31.6x.
Mid-band feedback is about 14dB.
By reversing the power source, reversing polarized capacitors, and
exchanging PNP and NPN, a very similar circuit will support a P-channel
JFET as the input device.
Unfortunately, it appears that the P-channel devices are considerably
more noisy than N-channel devices.
One goal of the feedback system is to present the output DC bias level to the input JFET source lead, with minimal attenuation. Another goal is to provide a known attenuation for AC signals within the desired bandwidth; the inverse of this becomes the closed-loop gain. There are various ways to structure the feedback:
The first design had resistor R4 from the emitter-follower output to the JFET. Then a much smaller resistor R5 from the JFET went to ground through a large electrolytic C1. This is a very common arrangement in feedback audio amplifiers. The amount of mid-band feedback is the ratio between the two resistors. The low-frequency roll-off is set by the filter formed by the second resistor and the capacitor. The second resistor should be small, because the Johnson noise is essentially in series with the input signal, which means the capacitor must be large. We can measure the open-loop gain by shorting the second resistor.
The alternate design essentially taps the emitter-follower load for the desired feedback ratio. (Actually, the small resistor R4 is added at the bottom.) That signal is coupled to the JFET by capacitor C1. If the feedback resistor R5 is increased, capacitor C1 can be decreased, with the same low-frequency response. But increasing resistor R5 also lowers the output bias voltage due to JFET current. So if resistor R5 is high (thus allowing a smaller feedback cap), a JFET with a lower pinch-off voltage is required. This could be a way to use the low-pinch-off devices, but shows up as a lower-than-desired output bias in the simulation.
The original goal for all this was a fixed and known 60dB of
audio amplification in two stages.
At first it would seem that 30dB should be fairly easy for three
transistors, but here one is an emitter-follower and another a JFET,
so most of the gain is probably from the second transistor.
Unfortunately, that transistor runs at very low current levels,
which reduces the gain there as well.
Experiments seem to indicate that 30dB probably is pretty close
to the edge so perhaps 20dB would better guarantee substantial
amounts of feedback.
Simulated Frequency Response
While the simulated frequency response is fine, it is not quite
as broad as the original circuit.
That is probably because of the higher load resistor R3 and lower
current operation.
The Prototype Board
Although I tried various other preamp designs, the JFET approach
was one of only two successful development branches.
(The other successful branch used modern low-noise low-voltage op
amps.)
The JFET approach originated from a year-2000 Usenet
From: Win HillSubject: Re: The definitive guitar -> soundcard preamp. Date: 2000/07/24 Message-ID: <397C3DC1.9C420342@mediaone.net>#1/1 Newsgroups: sci.electronics.design [...] OK folks, here's my improved preamp circuit. It will have very low distortion due to an unusual design feature: each device in the signal chain operates at a constant current, independent of output voltage. Consider, that's better than class A operation! In addition it has negative feedback. [...] ,------+--------+----+------- +9v R2 | | | 2.7k | 330 7.5k | E | | +--- B +-||-+ 10uF | Q2 C Q3 | | (R8+R9) R10 | pnp | C | G = 1 + --------- = 20 2N5457 | +----- B | R9 * R8 etc. | R3 E | Q1 |-' 2.7k Vx | | 1 uF In ---+----->| ,-+--------+-------||---+----- Out | |-, R10 Q4 | | | | | 1.8k C | | | +----+ B -+ 47k 10M | | E | | | R8 | R9 | diode | | ** 750 100 ,---+ | | | select | | | 680 1.5k | | for Vx | 100uF 22uF | | | GND ---+--------+----+------+---+----+-------+----- GND
In this design the FET gate is a ground referenced signal input, which may avoid an input capacitor in some cases. Normally, however, the preceding stage will have some output voltage bias, so at least one DC blocking capacitor generally is needed between stages. But we do not need both an input and an output capacitor on each stage.
Unless we know the bias voltage levels for the preceding stage output and this input, we may not know which direction a polarized capacitor should face (and I would prefer to avoid using non-polarized electrolytics). That argues for a high input impedance, which allows a smaller value non-polarized film capacitor. The JFET design has that high input impedance.
Although the current design would seem to be in some sense
optimal, perusing the literature will reveal a variety of
designs with similar or related topologies.
This was my original implementation of the Win Hill article. It was also my first acceptable preamp. It still has a lot of parts I eventually discarded.
On this board, power comes in at the upper right, through a
battery-protect resistor (47 ohms), and a reversed-voltage
protection transistor (a P-channel power MOSFET used in reverse).
The clips short out the 2.7k resistor, thus removing Q3.
That provided much better RFI isolation, presumably by reducing
loop gain.
In later designs, I kept Q3 but removed Q4 and replaced it with
a resistor.
I also removed R3 and R8.
In many cases I was able to actually hear a noise reduction
when resistors were removed.