Lower Than Expected Input Impedance

[ChristianFletcher]

I have been working on a circuit to increase the input impedance of the popular Velleman K7000 audio signal tracer.

The original signal tracer suffers from relatively low input impedance and lacks a detector stage for demodulation. So I have built a simple FET probe but I am measuring the input impedance at just over 2 Meg Ohm and I was expecting for something nearer to 10 meg.

I think the input impedance should be governed by the FETs input capacitance and R1. I have built the full circuit on strip board but I have cut back the strips around the FET so it is in free space. The probe actually work great and is super sensitive and compares well to the Heath kit tube based signal tracer. But why is my input impedance lower than I expected where have I gone wrong ??

[Vintage_RC]

How are you measuring the input impedance?

[ChristianFletcher]

Hello Alan

I install a resistance in series with the probe input in the form of a decade box. Starting at 0 ohms I record the output at the drain of the FET. I then increase the resistance until the output is at half of its original value. The set resistance at the half voltage point is equal to the input impedance of the amplifier.

Thanks Regards Chris

[Silicon]

Could it be due to the 'Miller effect'?

What would the input impedance be if you disconnect the drain?

You would then have the problem of measuring the input impedance because you can't use the previous method.

[GrimJosef]

Miller effect raising the effective input capacitance ? (Just a thought, I haven't worked the numbers.)

Cheers,

GJ

EDIT Post crossed with Silicon's.

[ChristianFletcher]

Is miller capacitance something different to the CIS value that I calculated ? Either way I am surprised given the very low frequency I am testing at. I did the test at 1Khz but the probe still works at over a 100Mhz but with much reduced output. I thought the input impedance really would have been dominated by the value of R1. Could I have enough stray capacitance in my build to cause the problem.

[GrimJosef]

Yes. There will be capacitances between the gate and drain, and also between the gate and source. In a 'common source' amplifier the source is tied essentially to zero volts, so the input signal will simply be charging and discharging the gate-source capacitance via the gate terminal, without anything complicated going on at the other (source) terminal.

The same is not true of the drain, however. Imagine the input signal tries to raise the gate voltage. It will need to supply some current to charge the gate-drain capacitance to do that. But as soon as the gate voltage rises, the drain voltage will fall and (because the device has gain) it will fall a lot more than the gate rises. The falling drain voltage will tend to pull the gate voltage down, via the gate-drain capacitance, and the only way the input signal can counteract that is by driving a lot more current into the gate to charge the capacitance. The effect - needing a lot more current to achieve a given voltage rise - is the same as if the input capacitance were a lot bigger. This is the Miller effect.

If your circuit has a voltage gain of, say, 10 then the input capacitance will look like 10 times its actual value. I don't know what Cgd for your FET is, but I'd believe it's similar to Cgs i.e. ~7pF. In that case the effective input capacitance would be 70pF. You've calculated that 7pF has a reactance of 22.7Mohm at 1kHz. So 70pF would have a reactance of 2.3Mohm. Put that in parallel with the 10Mohm resistor and you won't be far away from the 2Mohm you're measuring. It all adds up.

Once you get above very low frequencies you can find, in common source (or common cathode, where I'm much more familiar with this), that it is the Miller capacitance which is dominant, not the (grid leak) resistance.

Cheers,

GJ

[dave cox]

GJ has it right.

If you need gain in the input you could add a cascode stage (NPN in common base).

dc

[ChristianFletcher]

Fantastic not encountered the miller effect before, we live and learn. OK thanks I am going to have to read GJ post a few times for me to understand that point about miller effect. I don't have my values of R2 R3 to hand as I drew the sketch from memory but I measured my gain at 1khz as 2.3. Its interesting that many similar online circuits and descriptions claim impedance in excess of 10M.. Based on the above would a standard follower configuration have higher impedance.

Just Reading online: https://www.circuitstoday.com/cascode-amplifier

Miller effect is actually the multiplication of the drain to source stray capacitance by the voltage gain. The drain to source stray capacitance always reduces the bandwidth and when it gets multiplied by the voltage gain the situation is made further worse. Mulitiplication of stray capacitance increases the effective input capacitance and as we know, for an amplifier, the increase in input capacitance increases the lower cut of frequency and that means reduced bandwidth. Miller effect can be reduced by adding a current buffer stage at the output of the amplifier or by adding a voltage buffer stage before the input.

[jjl]

The Miller effect is an issue with all inverting amplifiers whether the active component is a valve, FET, bipolar transistor or whatever. The original paper on the effect was published in 1919!

John

[ChristianFletcher]

The first circuit I built was just a simple follower. I found it advantageous to have a little gain in the circuit but I am wondering if I did the wrong thing now. I may have another go at building the follower and measuring its impedance. errm kind of wondering now in a signal tracer probe do I want gain or high impedance. The impedance of the vellemen kit is only 50K

[dave cox]

Yup, a follower (x1 gain) will do better, but i think the cascode will be better again - probably!
The cascode example circuit you show should have a gain ~ 10.

The way to think of 'miller capacitance' is that you have to charge up that capacitor BEFORE the input voltage reaches the target voltage. Unfortunately, something is moving the voltage at the other end of that capacitor in the OPPOSITE direction so you need to push even more charge into it!! The cascode solves that problem by FIXING the voltage at the other end of that capacitor.

The effect of 'stray' capacitance will has quite an impact as well, over and above that of the input FET. Getting a 10M ohm input at dc is quite easy, I have a voltmeter here with a 50G ohm input at low voltage dc! Not so easy with ac, even at 1KHz

dc

[ChristianFletcher]

Yeh I probably will build up that cascode circuit but with a reduced gain. I was trying to build a signal tracer that didn't load up the RF section in a transistor radio.

[dave cox]

For a x1 buffer, you could also try 2 FETs in a 'current sourced' source follower.

dc

[G0HZU_JMR]

But why is my input impedance lower than I expected where have I gone wrong ??

It would be interesting to know what the values of R3, C3, R2 and C2 are in your original circuit. Also the load impedance of the detector would be useful to know although I assume this will be a very high impedance. With this info I think the input impedance at 1kHz can be predicted using a basic model for a process 50 JFET. This can then be compared to the results you are seeing.

[ChristianFletcher]

Full Circuit

The input impedance of the AF amplifier is 50K. Diode charge pump detector

Info here: http://www.zen22142.zen.co.uk/Theory...diode_pump.htm

I have done another check tonight and the impedance is actually closer to 4 meg ohm so I think thats probably closer including for the gain miller effects

[G0HZU_JMR]

Thanks. Apart from using them with crystal microphones in my youth I haven't done much with JFETs at AF.

However, your circuit values are close to what I was expecting so I'm surprised you only see a 2 Meg input impedance at 1kHz.
I tried modelling it with a process 50 JFET model in genesys and I also tried a basic VCCS model based on the datasheet of the MPF102. The worst case capacitance values in the datasheet would show a 6 Meg input impedance at 1kHz but if I use the capacitances from a typical process 50 JFET then I'd expect to see close to 10 Meg ohm at 1kHz.

I'm not sure why the MPF102 has such a high 'max' feedback capacitance of 3pF. A typical process 50 JFET would have a capacitance of about 0.7pF here.

The input impedance of your circuit will be a lot lower by 10kHz because this is a decade away. If you want high impedance all the way through the AF band then maybe consider a bootstrapped JFET circuit. However, the risk with a typical bootstrap circuit is that it could produce a lot of negative resistance at the input up at RF frequencies.

[ChristianFletcher]

Thanks Jeremy thats really great work!

I had one of my friends build the probe as I didnt want to publish something that hadnt had a bit of peer review. So the input impedence was reported to me a 2M. However checking my circuit tonight I measure it at closer to 4.5M. So I think we are in the ball park for your calculated value. I did suspect it could be sensitive to build variation. I am using a 2N5457 I swapped out the MPF102 for something cheaper and more available but the capcitance is around the same

[ChristianFletcher]

I have just measured the -3db point as 5Mhz but it is still useable at 120Mhz using the K7000 amplifier. The sensitivity is -50dbm at 80% modulation to produce a good audio level on the speaker. The K7000 now has the same performance as the heathkit tube based signal tracer.

[G0HZU_JMR]

See below for a classic bootstrap circuit using a JFET. I'm not sure this is what you want but you may find it interesting. It will produce a lot of negative resistance at the input up at RF so this is not an unconditionally stable circuit as drawn. It does offer a very high input impedance well past the AF band as you can see in the text in the circuit below. It claims >100Meg in parallel with just 0.25pF.