Sorry for bringing this old thread back to life.
I bumped into something like this by chance.
In Spice it has a gain of 1000x, and has decent distortion figures.
(~0.001% at 2Vp-p output).
Supply is 30V.
I didn't bother to put a JFET follower after the 33k RL.
Any comments ?
Patrick
I bumped into something like this by chance.
In Spice it has a gain of 1000x, and has decent distortion figures.
(~0.001% at 2Vp-p output).
Supply is 30V.
I didn't bother to put a JFET follower after the 33k RL.
Any comments ?
Patrick
Attachments
At least it works in this :
http://www.diyaudio.com/forums/analogue-source/64028-jfet-srpp-riaa-preamp.html
Patrick
http://www.diyaudio.com/forums/analogue-source/64028-jfet-srpp-riaa-preamp.html
Patrick
I am really amazed that you can do something like that with just two active devices - other schemes use enough parts to build a medium size cyclotron
jan
A typical method for getting accurate AC gain in a discrete amp is to use series-shunt feedback.
An example is a Jfet input with 10 ohms Rsource from the source to ground.
The eventual output signal is ac-coupled and passed through a feedback resistor Rf back to the source of the Jfet.
The AC gain is then Rf/Rsource. This is a standard method used in one-transistor rf amps.
Another method is to use several ADA4528 op-amps in parallel to get the noise down to several nV/Rt-Hz. Keep in mind that 10 ohms represents about 0.4nV/Rt-Hz. 10 ohms will be needed as the ADA4528 input gain resistor.
An example is a Jfet input with 10 ohms Rsource from the source to ground.
The eventual output signal is ac-coupled and passed through a feedback resistor Rf back to the source of the Jfet.
The AC gain is then Rf/Rsource. This is a standard method used in one-transistor rf amps.
Another method is to use several ADA4528 op-amps in parallel to get the noise down to several nV/Rt-Hz. Keep in mind that 10 ohms represents about 0.4nV/Rt-Hz. 10 ohms will be needed as the ADA4528 input gain resistor.
The jfet loaded with the very high effective impedance of the upper CCS
will result in susceptibility of DC operating point to drifting significantly.
I think if you either thermally bond them, use a dual and also allow for a fine
adjustment trim on current then it may be OK.
I suppose if you want a lot of gain with not many bits this one way to do it.
T
I am really amazed that you can do something like that with just two active devices - other schemes use enough parts to build a medium size cyclotron
jan
Well it's just a simple matter of the effective AC current swing of the lower
device versus the load impedance. See my comments WRT stable DC
operating point.
At a guess the quiescent will be around 5mA. The total gm will be around
33+25=58 ohms so total load R must be 58,000 to get 60dB gain.
If we allow say 2% total current drift (pretty tight for discrete) then
0.005/50*58000 =5.8 V drift. That's pretty significant. However if the
devices are matched and thermally tied it should be OK. IME they track very
well if you super glue them face to face and wrap / solder some copper wire
around. Or use one of those dual wire cut aluminium thingies I've seen here.
It's perfectly feasible to get a -lot- of gain from 1 stage and I've gotten
up to 80dB from a single stage, with very low distortion to boot.
It's just gm versus load R.
Terry
I agree that the simulated results in Spice is ideal case and probably not realistic.
With not perfectly matched and thermally tracked devices, there is no guarantee of the DC operating point at mid-rail.
One can of course use a simple potential divider to hold the gate of the top FET at mid-rail.
Penalty is limited bandwidth at LF (around 60Hz).
One can also trim the resistor values to minimise even harmonics.
-3dB Bandwidth is about 60Hz / 100kHz.
THD is 0.0086% and mostly 2nd harmonics.
Patrick
With not perfectly matched and thermally tracked devices, there is no guarantee of the DC operating point at mid-rail.
One can of course use a simple potential divider to hold the gate of the top FET at mid-rail.
Penalty is limited bandwidth at LF (around 60Hz).
One can also trim the resistor values to minimise even harmonics.
-3dB Bandwidth is about 60Hz / 100kHz.
THD is 0.0086% and mostly 2nd harmonics.
Patrick
Attachments
One can of course use a self-biasing cascode to increase bandwidth.
Again the resistor values need adjusting for minimum distortion.
Also rail voltage needs to be raised to accommodate the extra headroom for the cascode.
Bandwidth is now almost 1.5MHz.
THD is about 0.0080%.
Patrick
Again the resistor values need adjusting for minimum distortion.
Also rail voltage needs to be raised to accommodate the extra headroom for the cascode.
Bandwidth is now almost 1.5MHz.
THD is about 0.0080%.
Patrick
Attachments
This is an almost textbook example of misleading simulation. Build it, at 1k gain you'll have some nasty surprises regarding gain fluctuations (output fluctuations near DC) which will make it unusable at low frequencies. PSRR will also be a big issue in the noise budget.
As you start addressing the stable gain and noise issues the circuit inevitably will no longer be so simple. I wanted something that I could use to measure the noise of my shunt voltage regulator, and the final circuit that was decent enough for me was that one in the attached image. If you look carefully you'll notice that it really a simple circuit. Stable gain precisely dialed in via two precision resistors (1k/1R). And it works in real life, as I built and measured it at about 0.4nV/rtHz. No idea what distortion it has, but that was no concern for me, as I intended to use it just for measuring noise.
Attachments
It works, yes, encountered no problems with it. Tried various opamps, parts, etc. However, I don't recommend anyone building this. There's another low noise amplifier on this site, even lower noise than this. Gerhard built it. I would recommend that one.
http://www.diyaudio.com/forums/equipment-tools/240910-220pv-sqrt-hz-preamp-measurements.html
Patrick made me curious what distortion my circuit might have, so I just ran a quick sim in ltspice. If only I could believe it.
http://www.diyaudio.com/forums/equipment-tools/240910-220pv-sqrt-hz-preamp-measurements.html
Patrick made me curious what distortion my circuit might have, so I just ran a quick sim in ltspice. If only I could believe it.
Attachments
Unfortunately, for that schematic, the input current noise will kill the noise performance for anything beyond a few ohms of source impedance.
Bottom line, at these noise levels, there is no one size fits all. Such a preamp should be designed having the scope of measurements in mind. For power supplies, that schematic would work perfectly. But for e.g. Measuring moving magnet cartridges it will not.
I agree. When I started this little project the intended use was very narrow.
Interesting application that I haven't come across yet. Why measure an MM cartridge? If one really cares, would one not just get a moving coil cartridge? Just thinking out loud...