Simple 60dB discrete low noise amplifier (lna)

[iko]

This subject started in another thread

Super Regulator, collecting the facts

but it seems a discussion on its own, so I moved it here. The idea is to build a low noise amplifier with 60dB (1000x) to use it as for noise measurements. Initially I built the circuit from the Linear Technology application note 83

http://cds.linear.com/docs/Application%20Note/an83f.pdf

but I wanted something with even lower noise so I put together a simple discrete circuit using paralleled jfets (2sk170v). However, I used no feedback so setting the gain to 60dB with low tolerance is something that relies on a very good AC voltmeter. Jan recommended using feedback to be able to calculate and implement the gain accurately. What comes next has been moved from the other thread.



This is what I built

The subtler issues escape me, as you can see.

[jan.didden]

Originally Posted by ikoflexer

This is what I built

The subtler issues escape me, as you can see.

Yes that's open loop. Is this the measurement amplifier you mentioned before?

jd

[iko]

Originally Posted by janneman

Yes that's open loop. Is this the measurement amplifier you mentioned before?

jd

Yes. I might have to stick with the opamp one, to avoid all those issues.

[jan.didden]

Originally Posted by ikoflexer

Yes. I might have to stick with the opamp one, to avoid all those issues.

Try this: multiply the two collector resistor values by 10. Replace Q2 with an NPN. Feedback the output signal via 6.8k ohms to the input stage sources.
Rough maths says you've got about 59.92dB gain.
Then put 400 or 500 ohms in parallel with the 6.8 ohms source resistor, should get you at exactly 60dB.

I haven't looked at DC stuff so try first with cap coupling at all the sensitive points

jd

[AndrewT]

Originally Posted by janneman

multiply the two collector resistor values by 10.

is that r5 & r12

q2 is a common collector (= emitter follower).
Are you suggesting that it becomes a common emitter to generate some voltage gain?

[jackinnj]

Omit C5

[jan.didden]

Originally Posted by AndrewT

is that r5 & r12

q2 is a common collector (= emitter follower).
Are you suggesting that it becomes a common emitter to generate some voltage gain?

yes R5 & R12. I think you need to invert the output signal to make the feedback negative, and in my stupidity I said change Q2 to an NPN which, as a common collector, doesn't invert the signal of coure. Leave Q2 as-is, use a collector resistor equal to the emitter resistor and take the feedback from the collector. If the output sticks to the supply I've got it wrong and then take the feedback from Q2 emitter

jd

[jan.didden]

Originally Posted by jackinnj

Omit C5

Sure? That'll kill the gain again, I think.

jd

[iko]

Just got back from getting some parts. I'll try your suggestions and let you know.

OK, so I simulated the changes. Adding 6k8 feedback from between the collector of Q2 and a 100R resistor to the source of jfets in stage one doesn't change the gain. Making R5 = 6k and R12 = 9k1 kills everything. Are you sure about this?

[jackinnj]

Originally Posted by janneman

Sure? That'll kill the gain again, I think.

jd

Yes, but look at the response curve IKO has provided -- and where does 6.8R x 470uF work out?