Don't confuse things yet with the actual second stage. Let's look first at the first stage. Once again, the first stage device has several noise sources inside it. What I call 'second stage' shot noise is from the collector to the emitter, rather than from the input current (first stage). You could also call it output stage shot noise, does this make it any clearer? This has only to do with ONE device, not other devices.
Ultimately, the tradeoff is between the input current related shot noise and the output stage related shot noise. Reduce one, increase the other. These is where the concept of NOISE FIGURE which is defined by dividing one noise source by the other to find the optimum source resistance is derived.
Another factor is the tradeoff between very high beta, which would REDUCE input stage shot noise by reducing input current, and r(bb') which is the intrinsic base resistance, as it tends to increase with very high beta devices.
What a hassle! There are some fairly ideal devices out there, but most of you could not purchase them easily. I don't know why they are not more available.
With an inductive source, IT IS IMPOSSIBLE to have an optimum noise figure with a transistor over a range of frequencies, but with a transformer and a resistive source, or just a higher value resistor source alone, it is possible to get an optimum noise figure with a bipolar transistor. Noise figure with a fet does not really apply at audio frequencies, in general, because the noise current is so small, that the optimum matching is in the megohm region, but under these conditions a good fet will beat ANY transistor.
Ultimately, the tradeoff is between the input current related shot noise and the output stage related shot noise. Reduce one, increase the other. These is where the concept of NOISE FIGURE which is defined by dividing one noise source by the other to find the optimum source resistance is derived.
Another factor is the tradeoff between very high beta, which would REDUCE input stage shot noise by reducing input current, and r(bb') which is the intrinsic base resistance, as it tends to increase with very high beta devices.
What a hassle! There are some fairly ideal devices out there, but most of you could not purchase them easily. I don't know why they are not more available.
With an inductive source, IT IS IMPOSSIBLE to have an optimum noise figure with a transistor over a range of frequencies, but with a transformer and a resistive source, or just a higher value resistor source alone, it is possible to get an optimum noise figure with a bipolar transistor. Noise figure with a fet does not really apply at audio frequencies, in general, because the noise current is so small, that the optimum matching is in the megohm region, but under these conditions a good fet will beat ANY transistor.
Edit: I see that John has posted more info while I was writing
this. Having a brief look at it he seems not have given any
answer to his original question, but only given some more hints,
so I don't delete this post.
Not sure if I understand correctly, but I'll give it a try.
The second stage shot noise will increase by a factor sqrt(10) =
10 dB, as we have already agreed. I am not quite clear if
we are talking noise current or noise voltage here, or if it matters
which. I will assume below that it can be treated as a voltage
source whose output resistance will not change (not sure if
that is a valid assumption).
If we treat the input shot noise as a current source and first
assume that this noise current is independent of Ic, then the
voltage contribution of this noise would decrease by a factor
10, since Re = 1/gm decreases by a factor 10. John didn't give
any clue about how to calculate the input shot noise, but said
it will decrease, so we will thus end up with more than a factor
10 decrease, ie. more than 20 dB decrease. Since we don't have
any actual figures for the decrease in input shot noise, let's say
its contribution to the overall noise decreases by a factor x,
where x > 10.
Let a be the noise voltage contribution by the second-stage
shot noise in the first case, and let b be the contribution by
the input shot noise. Then, the overall noise will be reduced
if
sqrt((sqrt(10)a)^2 + (b/x)^2) < sqrt(a^2 + b^2)
which is true iff
b > 3xa/(x^2 - 1)
which is approximately
b > 3a.
That is, if the input shot noise contributes a noise voltage at
least 3 times higher than the second-stage shot noise, then
we will benefit from decreasing Ic by a factor 10.
this. Having a brief look at it he seems not have given any
answer to his original question, but only given some more hints,
so I don't delete this post.
Not sure if I understand correctly, but I'll give it a try.
The second stage shot noise will increase by a factor sqrt(10) =
10 dB, as we have already agreed. I am not quite clear if
we are talking noise current or noise voltage here, or if it matters
which. I will assume below that it can be treated as a voltage
source whose output resistance will not change (not sure if
that is a valid assumption).
If we treat the input shot noise as a current source and first
assume that this noise current is independent of Ic, then the
voltage contribution of this noise would decrease by a factor
10, since Re = 1/gm decreases by a factor 10. John didn't give
any clue about how to calculate the input shot noise, but said
it will decrease, so we will thus end up with more than a factor
10 decrease, ie. more than 20 dB decrease. Since we don't have
any actual figures for the decrease in input shot noise, let's say
its contribution to the overall noise decreases by a factor x,
where x > 10.
Let a be the noise voltage contribution by the second-stage
shot noise in the first case, and let b be the contribution by
the input shot noise. Then, the overall noise will be reduced
if
sqrt((sqrt(10)a)^2 + (b/x)^2) < sqrt(a^2 + b^2)
which is true iff
b > 3xa/(x^2 - 1)
which is approximately
b > 3a.
That is, if the input shot noise contributes a noise voltage at
least 3 times higher than the second-stage shot noise, then
we will benefit from decreasing Ic by a factor 10.
"He's Greek, Mr. Fawlty...."
"Great.....they're the ones who started that stuff."
Pardon the interuption in the lecture session, but this is for SY.
From The Real Frank Zappa Book, page 15:
"My ancestory is Sicilian, Greek, Arab, and French. My mother's mother was French and Sicilian, and her Dad was Italian (from Naples). She is first generation. The Greek-Arab side is from my Dad. He was born in a Sicilian village called Partnico, and came over on one of the immigrant boats when he was a kid."
So.....if you count Sicilian as Italian, he is at least 3/8 Italian, and at most 1/4 Greek. And since his Dad's family is from Sicily, there may be some more Italian blood mixed in that is not mentioned.
So I win. So there.
Back to John's lesson. I wonder if the good Prof. Krell is attending today????
Jocko
"Great.....they're the ones who started that stuff."
Pardon the interuption in the lecture session, but this is for SY.
From The Real Frank Zappa Book, page 15:
"My ancestory is Sicilian, Greek, Arab, and French. My mother's mother was French and Sicilian, and her Dad was Italian (from Naples). She is first generation. The Greek-Arab side is from my Dad. He was born in a Sicilian village called Partnico, and came over on one of the immigrant boats when he was a kid."
So.....if you count Sicilian as Italian, he is at least 3/8 Italian, and at most 1/4 Greek. And since his Dad's family is from Sicily, there may be some more Italian blood mixed in that is not mentioned.
So I win. So there.
Back to John's lesson. I wonder if the good Prof. Krell is attending today????
Jocko
Attachments
When I am debating someone who doesn't seem to know how things interact, let me say that: As idle current is dropped, what we define as 'voltage noise' will increase by the square root of the current change. The input current generated 'shot noise' contribution will reduce by the square root, IF the BETA remains constant. However, the beta usually drops, so the current noise will NOT drop as fast, as the voltage noise increases. Enough for now.
Close enough.........
Ever hear any of the stories of how they would only go to Italian doctors, the old man calling the WASPs "hillbillies", or them treating Frank's asthma by shoving radium up his nose.
You should have mentioned that they moved away from Baltimore because Frank was always sick. Wonder why......
(I think with the deck stacked against him, he would have been sick anywhere.)
Jocko
Ever hear any of the stories of how they would only go to Italian doctors, the old man calling the WASPs "hillbillies", or them treating Frank's asthma by shoving radium up his nose.
You should have mentioned that they moved away from Baltimore because Frank was always sick. Wonder why......
(I think with the deck stacked against him, he would have been sick anywhere.)
Jocko
Yes
The worm.......on someone else's computer.........was deactivated on the 10th of the month. And for those of you who care:
The virus was not on the computer of one of my DIY buddies. It was more than likely on some lurker's machine. You know the kind of guys who would e-mail me wanting a complete schemamtic of X circuit, along with a detailed parts list. Oh yeah, they also wanted me to send them pictures of both sides of the PCB. I guess if your only impression of America is seeing our presidents on TV, you may think that we are all brain dead hillbillies who think about sex 24 hours a day. I am not that stupid, and thinking about sex is about all I can do about it, so I don't............
So.....since it was only my DIY e-mail that I had to fight, I quarantined the whole gang. The irony is that the jerks I had a bone to pick with probably never saw my rant. I only brought it up so that the guys who had a legitimate need to contact me would not freak out.
Back to our regularly scheduled entertainment.
Jocko
The worm.......on someone else's computer.........was deactivated on the 10th of the month. And for those of you who care:
The virus was not on the computer of one of my DIY buddies. It was more than likely on some lurker's machine. You know the kind of guys who would e-mail me wanting a complete schemamtic of X circuit, along with a detailed parts list. Oh yeah, they also wanted me to send them pictures of both sides of the PCB. I guess if your only impression of America is seeing our presidents on TV, you may think that we are all brain dead hillbillies who think about sex 24 hours a day. I am not that stupid, and thinking about sex is about all I can do about it, so I don't............
So.....since it was only my DIY e-mail that I had to fight, I quarantined the whole gang. The irony is that the jerks I had a bone to pick with probably never saw my rant. I only brought it up so that the guys who had a legitimate need to contact me would not freak out.
Back to our regularly scheduled entertainment.
Jocko
fets look nice in theory, but you guys are forgetting one thing, the source impedance... now we have a pre amplifier whit a typical 100 to 200 ohm output impedance... this impedance is reflected upon the input of the amplifier an that is the impedance the input sees
so an bipolar input amplifier will performer much better.
One bigger problem you guys are forgetting is the input capacitance much bigger in a fet, and non linear!
so mike, you are right, bipolar is the best device for audio applications...
Cheers
so an bipolar input amplifier will performer much better.One bigger problem you guys are forgetting is the input capacitance
much bigger in a fet, and non linear!so mike, you are right, bipolar is the best device for audio applications...
Cheers
Re: going by the book
Hi Christer...and Joe Berry...
I fear you've taken the remarks in Solomons paper out of their proper context......FET's in monolithic op. amps. were adopted because of the inherent limitations in the lateral pnp process, which severely curtailed design flexibility.......
ALL the comparisons you've quoted above are made in relation to these monolithic pnp's, whose current gains and Ft's barely approached 50, and 4MHz respectively..........Note that no mention is made of noise performance, DC precision,......etc
At any rate, these 'advantages' do not apply to discrete or dual-monolithic non-lateral process BJT's, normaly used in power amp. construction.
Oh..and No....JFET's simply cannot produce less noise than bJT's at audio frequencies, and low source impedances.....sorry.
...they are also intrinscally less linear...by a vast margin.....
Christer said:May I join the fun?
While I don't have the knowledge or experience to have any
strong personal opinion on the JFET vs. BJT matter myself, I
get the impression that National had some favourable things
to say about JFET inputs already 30 years ago, but what do
they know?
http://www.national.com/an/AN/AN-A.pdf p. 10
Joe Berry said:
Hi Christer...and Joe Berry...
I fear you've taken the remarks in Solomons paper out of their proper context......FET's in monolithic op. amps. were adopted because of the inherent limitations in the lateral pnp process, which severely curtailed design flexibility.......
ALL the comparisons you've quoted above are made in relation to these monolithic pnp's, whose current gains and Ft's barely approached 50, and 4MHz respectively..........Note that no mention is made of noise performance, DC precision,......etc
At any rate, these 'advantages' do not apply to discrete or dual-monolithic non-lateral process BJT's, normaly used in power amp. construction.
Oh..and No....JFET's simply cannot produce less noise than bJT's at audio frequencies, and low source impedances.....sorry.
...they are also intrinscally less linear...by a vast margin.....
Dear John.
---quote
With an inductive source, IT IS IMPOSSIBLE to have an optimum noise figure with a transistor over a range of frequencies
---quote
You are right, but typical pickup data is Rs=650 Ohm, Ls=0.7H, Rl=47kOm, Cl=250 pF. It is not purely inductive source. With this source we got optimum noise figure for BJT with dc bias around 100 uA. As for JFET there is no optimum current (because the input noise current is absent), and noise figure will be less with higher dc current. The FET should have Io/Vp>3ma/V for our source. When we take in account that we have unavoidable cartridge dc resistance Rs=650 Ohm we don’t need to design the amp with “a 10 ohm overall equivalent noise level” and feed the cat with a cream. BJTs here feel right at home.
---quote
With an inductive source, IT IS IMPOSSIBLE to have an optimum noise figure with a transistor over a range of frequencies
---quote
You are right, but typical pickup data is Rs=650 Ohm, Ls=0.7H, Rl=47kOm, Cl=250 pF. It is not purely inductive source. With this source we got optimum noise figure for BJT with dc bias around 100 uA. As for JFET there is no optimum current (because the input noise current is absent), and noise figure will be less with higher dc current. The FET should have Io/Vp>3ma/V for our source. When we take in account that we have unavoidable cartridge dc resistance Rs=650 Ohm we don’t need to design the amp with “a 10 ohm overall equivalent noise level” and feed the cat with a cream. BJTs here feel right at home.
I design around 2 ohm sources, worst case, so I need everything that I can get. But then, I don't need to buy step-up transformers, or use cheap moving magnet cartridges. I went from bipolar to fets, in 1968, 35 years ago, when I made a quieter reproduce stage with fets than bipolar devices operating between 50-100ua, even then, when I worked for the audio division at AMPEX.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.