BJT or MOSFET for low noise voltage follower

[cloudhopper]

I'm using a JFET for it's very high input impedance and low noise and DC coupling it to a voltage follower. I'm considering either a MOSFET like a BS170 or a BJT like a 2SC1815. Which would be lower noise when running around 30V supply? The way I have the JFET configured, it's output impedance would be around 350K, not a problem for the MOSFET but not sure about a BJT. Suggestions on either these two choices or other transistor possibilities?

 

[Bigun]

It depends on many things. What you need to drive with the output of your voltage follower will influence the load impedance placed on the JFET.

A power MOSFET would be needed if you have to drive a lot of current into the load. This type of MOSFET has a fair bit of input capacitance so although it has a high input impedance at dc this falls as the frequency goes up. It's also a non-linear impedance. A small signal MOSFET will have a much smaller input capacitance but can't drive a low impedance load. A 350k output impedance from your JFET may not be able to drive a power MOSFET at high frequency.

A BJT has a lower input impedance than a MOSFET at dc but it doesn't suffer from as much input capacitance as a power MOSFET. The input impedance of a BJT is also non-linear and has a strong dependence on the load being driven by the BJT.

You need to provide more context around your application need to get good answers to your question - and there are many skilled people that can help you (I'm not one of them though !).

good luck.

 

[cloudhopper]

It depends on many things. What you need to drive with the output of your voltage follower will influence the load impedance placed on the JFET.

The follower is to convert to a lower impedance output, between 200 and 600 ohm. I've used a BS170 to do this before and it works well, I'm just wondering if a BJT would have lower noise. The JFET output is still very low amplitude, maybe 250mV and when I use the MOSFET, I have the MOSFET only drawing 1 or 2 mA.

A BJT has a lower input impedance than a MOSFET at dc but it doesn't suffer from as much input capacitance as a power MOSFET.

Where would I find the input impedance of say a 2SC1815? I may be just overlooking it when I look at the datasheet but I don't see it. I'm using a signal MOSFET and the input capacitance isn't an issue when using in a unity gain configuration.

 

[Mooly]

The input impedance of an emitter follower depends on many factors... it's not a single transistor specific value.

For example,
Emitter Follower

1 or 2 ma in a source follower won't drive a low impedance load to more than a few 10's of millivolts... as I think you have found.

For optimum noise figure don't discount an op amp correctly used... it will be quieter than a discrete EF configuration with the correct choice of device.

 

[CBS240]

IMHO, if you're intent is for unity voltage gain, i e source follower from the J-fet, BJT EF is the way to follow it up in terms of noise. In the realm of small signal followers with a high input Z, I prefer a J-fet driving a small signal BJT. The drain and collector are connected together, and the source of the J-fet connects to the base of the BJT with a resistor from base to emitter of the BJT determaining the current in the J-fet. Id is determained by 0.6V/R. Very simple but quite effective. Of course for lowest noise in the J-fet Id should be close to Idss for the input J-fet.

Also you could use a J-fet with much lower goss as a cascode device like this.... Cascode does has its advantages.

 

[CBS240]

EDIT.......

Also you could use a J-fet with much lower goss as a cascode device like this.... Cascode does has its advantages.

My bad, I mean higher goss (output conductance), i e larger Vgs at the bias current. A 4091 type works pretty well for the cascode with a few mA. I also have used this arrangement with the collector of the BJT connected to +V supply for larger BJT current.

 

[cloudhopper]

1 or 2 ma in a source follower won't drive a low impedance load to more than a few 10's of millivolts... as I think you have found.

Actually, I've found that a 32V supply to a MOSFET pulling 1mA can drive a lot more voltage than that. I found that a LND150 seems to perform about the same as a BS170 and the datasheet for the LND150 does go down to that low of current for the LND150. (I wish someone would test and publish data for the BS170 at low current). The gm of a LND150 at 1mA is about 2mS so the output impedance of a follower would be 1/gm or 1/.002 which is only 500 ohms which is plenty low and with an idle voltage of 16V at the source, the signal can swing several volts while still maintaining a low output impedance.

I do realize one is a depletion mode and the other an enhancement mode but in a follower it doesn't matter.

I've convinced myself that the MOSFET is the better choice for follower in this circuit.

CBS240, I've considered cascoded JFET's for the first stage and I've also considered bootstrapping from the source of the MOSFET follower to the JFET input to get more gain from the first stage but I'm not sure I really need that much gain from the first stage. I do understand that cascode can also provide a lower noise stage and I may have to play with that.

Thanks for all the comments as I work thru this.

 

[jcx]

mosfets often have very high 1/f noise corner - they are seldom used in low noise cirucits at low frequencies - discrete mosfet, power devices in particular will likely be built on lines with no steps taken to control 1/f noise

some op amp and ACD/DAC process do pretty good on noise with cmos but I wouldn't rely on being able to do as well with available discretes

if you have substantial V gain in the jfet front end then low noise in the follower is not really important

 

[cloudhopper]

mosfets often have very high 1/f noise corner - they are seldom used in low noise cirucits at low frequencies - discrete mosfet, power devices in particular will likely be built on lines with no steps taken to control 1/f noise

some op amp and ACD/DAC process do pretty good on noise with cmos but I wouldn't rely on being able to do as well with available discretes

if you have substantial V gain in the jfet front end then low noise in the follower is not really important

Yes, I agree that the signal MOSFET can have a lot of noise, the JFET offers substantial gain, especially when bootstrapped so the MOSFET doesn't have to do much except provide impedance conversion.

 

[seanvn]

The BS170 has low input capacitance of 40pf or something like that. Transistors have a high base diffusion capacitance that depends on the base current and which can easily exceed 40pf. It is not necessarily the case that there is a big difference between transistors and mosfets in terms of input capacitance. Also simple follower circuits have surprisingly high distortion figures. It is better to use a complementary feedback pair CFP with a small stopper resistor in the base of the non-input transistor (which stops rf oscillations). Are CFP circuits as bit noisy? I think they are somewhat. Designers choice.

 

[Bigun]

CFP tends not to behave so well in ClassAB as it does in ClassA - the switching behaviour is not very smooth. D Self found distortion of CFP to be lower - at least this is his general comment and oft repeated - but he also looked at low signal levels and found CFP is worse than EF. I use CFP as a Class A driver stage for ClassAB BJT output.