jensenmann
Well-known member
Can anybody recomment me a good reading about OSI (optimum source impedance). I only heard about this subject but never really learned or understood the basics around that subject.
:thumb:
Jens
:thumb:
Jens
OSI
- Thread starter jensenmann
- Start date
Help Support GroupDIY Audio Forum:
Samuel Groner
Well-known member
All the Jung books (including the one that has recently got posted as PDF) have a section about it IIRC.
Samuel
Samuel
[quote author="jensenmann"]Can anybody recomment me a good reading about OSI (optimum source impedance). [/quote]
YES - a really good book is "Low Noise Electronic Design" by Motchenbacher & Fitchen. I bought my copy back in 1979, there is later, slightly changed edition that I have seen but the orginal one is better in my opinion.
There is a copy on ebay, currently at $9.99 - even at twice that it would be a bargain if you're serious about the subject. Buy it!
YES - a really good book is "Low Noise Electronic Design" by Motchenbacher & Fitchen. I bought my copy back in 1979, there is later, slightly changed edition that I have seen but the orginal one is better in my opinion.
There is a copy on ebay, currently at $9.99 - even at twice that it would be a bargain if you're serious about the subject. Buy it!
bcarso
Well-known member
[quote author="cuelist"][quote author="jensenmann"]Can anybody recomment me a good reading about OSI (optimum source impedance). [/quote]
YES - a really good book is "Low Noise Electronic Design" by Motchenbacher & Fitchen. I bought my copy back in 1979, there is later, slightly changed edition that I have seen but the orginal one is better in my opinion.
There is a copy on ebay, currently at $9.99 - even at twice that it would be a bargain if you're serious about the subject. Buy it![/quote]
The later book is Motchenbacher and Connelly, and I agree adds little to the original.
YES - a really good book is "Low Noise Electronic Design" by Motchenbacher & Fitchen. I bought my copy back in 1979, there is later, slightly changed edition that I have seen but the orginal one is better in my opinion.
There is a copy on ebay, currently at $9.99 - even at twice that it would be a bargain if you're serious about the subject. Buy it![/quote]
The later book is Motchenbacher and Connelly, and I agree adds little to the original.
[quote author="bcarso"]The later book is Motchenbacher and Connelly, and I agree adds little to the original.[/quote]
Actually, the later book is called "Low Noise Electronic System design and has omitted certain parts of the original book.
I was directed to this book from an article in Wireless World at the time. At the time, the book was very expensive but the value is still there long after the price has been forgotten.
Actually, the later book is called "Low Noise Electronic System design and has omitted certain parts of the original book.
I was directed to this book from an article in Wireless World at the time. At the time, the book was very expensive but the value is still there long after the price has been forgotten.
bcarso
Well-known member
Agree---either one is well worth the investment. The curves on e sub n and i sub n are very valuable even though they are a bit dated. The general discussions and examples of specific circuit designs with their derivations are very good indeed.
Another book that has some very good information is Arbel's Analog Signal Processing and Instrumentation. It was breathtakingly pricey when I bought it but has proven its worth for many years now.
Another book that has some very good information is Arbel's Analog Signal Processing and Instrumentation. It was breathtakingly pricey when I bought it but has proven its worth for many years now.
jensenmann
Well-known member
:thumb:
Thanks for your replies. I´ll start with the Jung book.
:guinness:
Jens
Thanks for your replies. I´ll start with the Jung book.
:guinness:
Jens
NewYorkDave
Well-known member
I don't know if any of those references mention it, but DON'T fall into the trap of adding resistive build-outs to try to "match" a low source Z to an input device having a higher OSI. You'll end up with worse noise performance than if you just accepted the mismatch. The only right way to do it, passively, is with a transformer.
jensenmann
Well-known member
Actively it might be done by paralleling the input transistor, I guess, or am I wrong? What I´m trying to find out is why the A*mek 2500 Mikepre has these four paralleled transistors in the inputstage:
http://www.gyraf.dk/schematics/Amek_2500_micamp.gif
:roll:
http://www.gyraf.dk/schematics/Amek_2500_micamp.gif
:roll:
the reason to parallel transistors in a TL input stage is to lower the base spreading resistance (r sub b) by shunting it with another transistor's r sub b.
if the transistor's rb is 25R, then 2 means 12R5, 4 means 6R25, etc.
The base spreading resistance is one of the noise generators at the input.
BUT... you can't just go indiscriminately paralleling devices, b to b, c to c, e to e unless you match them because there's no guarantee that they'll split the currents equally. The usual dodge is separate emitter resistors, but these have to be minimized too because they are in series with rb.
Yes, you can just parallel even more transistors.
That's the beauty of the LM394; the monolithic construction guarantees the matching of the individual transistors. BUT... there are discretes that cost less and have lower rb.
if the transistor's rb is 25R, then 2 means 12R5, 4 means 6R25, etc.
The base spreading resistance is one of the noise generators at the input.
BUT... you can't just go indiscriminately paralleling devices, b to b, c to c, e to e unless you match them because there's no guarantee that they'll split the currents equally. The usual dodge is separate emitter resistors, but these have to be minimized too because they are in series with rb.
Yes, you can just parallel even more transistors.
That's the beauty of the LM394; the monolithic construction guarantees the matching of the individual transistors. BUT... there are discretes that cost less and have lower rb.
jensenmann
Well-known member
Does OSI and base spreading resistance correlate in any form or are the completely different mechanisms?
And which parameter do I have to match for paralleling transistors?
:roll:
btw, I´ve never before heard of base spreading resistance. Maybe I´m only confused by the english word. :roll: :roll: :roll:
Is it the dynamic B-E resistance rbe?
And which parameter do I have to match for paralleling transistors?
:roll:
btw, I´ve never before heard of base spreading resistance. Maybe I´m only confused by the english word. :roll: :roll: :roll:
Is it the dynamic B-E resistance rbe?
ok. some errors to correct since I was too lazy last night to look it up.
the base spreading resistance is R sub bb, and the intrinsic emitter resistance is r sub e. This is sometimes called the Shockley emitter resistance, and at room temps, it (re) approximates to 0.026/I sub c.
from Motchenbacher/Fitchen: The base spreading resistance is the resistance of the lightly doped base region between the external base contact and the active base region. This is a true resistance, and therefore exhibits thermal noise. Fluctuations in base current I sub B and collector current I sub C are responsible for shot noise at the respective junctions. The flow of base current I sub B throught eh base-emitter depletion region gives rise to 1/f noise.
the OSI for a given amplifier is found by using ohms law on the voltage and current noise specs for the part and then correcting the result for bandwidth.
the base spreading resistance does correlate with OSI in that it is one of the voltage noise sources at the input.
The parameter to match is most likely beta, but I am not 100% sure of this. Again, the emitter resistor helps obviate this, and provides a useful source of degeneration in the interest of linearity as well as a convenient node for feedback. But remember that the emitter resistor is in series with Rbb, and it's thermal noise becomes a part of the voltage noise spec.
the base spreading resistance is R sub bb, and the intrinsic emitter resistance is r sub e. This is sometimes called the Shockley emitter resistance, and at room temps, it (re) approximates to 0.026/I sub c.
from Motchenbacher/Fitchen: The base spreading resistance is the resistance of the lightly doped base region between the external base contact and the active base region. This is a true resistance, and therefore exhibits thermal noise. Fluctuations in base current I sub B and collector current I sub C are responsible for shot noise at the respective junctions. The flow of base current I sub B throught eh base-emitter depletion region gives rise to 1/f noise.
the OSI for a given amplifier is found by using ohms law on the voltage and current noise specs for the part and then correcting the result for bandwidth.
the base spreading resistance does correlate with OSI in that it is one of the voltage noise sources at the input.
The parameter to match is most likely beta, but I am not 100% sure of this. Again, the emitter resistor helps obviate this, and provides a useful source of degeneration in the interest of linearity as well as a convenient node for feedback. But remember that the emitter resistor is in series with Rbb, and it's thermal noise becomes a part of the voltage noise spec.
books:
in addition to Motchenbach & Fitchen (later Motchenbach & Connely),
Gray & Meyer, "Analysis and Design of Analog Integrated Circuits" has a good treatment, as does Horowitz & Hill, "The Art of Electronics."
There was also a good applications note from Analog Devices (at the time, it was PMI). The Horowitz & Hill treatment is particularly approachable.
Look at AN253 and AN358 from the Analog Devices website. They don't directly deal with discretes, but the information there might be useful.
further thoughts about emitter resistors and rbb.
Although these resistances are in series, for noise purposes their noise voltages do not directly sum because they are uncorrelated. For this reason, their noise voltages are RMS summed.
Just think about how quiet it could be if we could bring the lead singer and his microphone down to within a couple of degrees K of absolute zero... food for thought!
in addition to Motchenbach & Fitchen (later Motchenbach & Connely),
Gray & Meyer, "Analysis and Design of Analog Integrated Circuits" has a good treatment, as does Horowitz & Hill, "The Art of Electronics."
There was also a good applications note from Analog Devices (at the time, it was PMI). The Horowitz & Hill treatment is particularly approachable.
Look at AN253 and AN358 from the Analog Devices website. They don't directly deal with discretes, but the information there might be useful.
further thoughts about emitter resistors and rbb.
Although these resistances are in series, for noise purposes their noise voltages do not directly sum because they are uncorrelated. For this reason, their noise voltages are RMS summed.
Just think about how quiet it could be if we could bring the lead singer and his microphone down to within a couple of degrees K of absolute zero... food for thought!
bcarso
Well-known member
[quote author="rickc"]
Just think about how quiet it could be if we could bring the lead singer and his microphone down to within a couple of degrees K of absolute zero... food for thought![/quote]
Unfortunately carrier freezeout occurs long before absolute zero. Bipolars go south ahead of silicon JFETs, which still operate to 77K (liquid nitrogen) but work best about 100K.
Germanium, and intrinsic semiconductors like GaAs keep on ticking at liquid helium temps, but have the drawbacks of excessive low-frequency noise, at least at this point in their evolution.
Just think about how quiet it could be if we could bring the lead singer and his microphone down to within a couple of degrees K of absolute zero... food for thought![/quote]
Unfortunately carrier freezeout occurs long before absolute zero. Bipolars go south ahead of silicon JFETs, which still operate to 77K (liquid nitrogen) but work best about 100K.
Germanium, and intrinsic semiconductors like GaAs keep on ticking at liquid helium temps, but have the drawbacks of excessive low-frequency noise, at least at this point in their evolution.
Samuel Groner
Well-known member
I would have voted for Vbe, but nor sure either...
Samuel
bcarso
Well-known member
[quote author="Samuel Groner"]
Samuel[/quote]
Match Vbe. The desire is mostly to equalize collector currents. Roughly 60mV delta Vbe entails a current ratio of 10:1, all other things equal. It's nice to have (high!) betas matched too, but unless one of them is extremely low it's the base-emitter voltage relative to that transistor's particular characteristic that's going to determine its collector current.
But nature is forgiving in this case in terms of the noise due to the component of e sub n associated with r sub e. If it were only that we wouldn't care too much, as long as the device wasn't getting too hot etc.---in fact we'd just put more current through a single device until we got down to where r sub bb' was significant. But the base current noise starts to get an excess low freq component at higher currents, and it flows as mentioned above in r sub bb', and as well r sub bb' starts to get larger itself, so things get suboptimal.
If we wanted we could have separate emiiter current sources (not necessarily current sources, say just good-sized resistors) and separate feedback networks (including separate coupling caps), still tying the collectors together, if there was just no way to match the transistors satisfactorily. But parts these days out of a good tight process and with the same date code are pretty decently matched for this kind of work.
I would have voted for Vbe, but nor sure either...
Samuel[/quote]
Match Vbe. The desire is mostly to equalize collector currents. Roughly 60mV delta Vbe entails a current ratio of 10:1, all other things equal. It's nice to have (high!) betas matched too, but unless one of them is extremely low it's the base-emitter voltage relative to that transistor's particular characteristic that's going to determine its collector current.
But nature is forgiving in this case in terms of the noise due to the component of e sub n associated with r sub e. If it were only that we wouldn't care too much, as long as the device wasn't getting too hot etc.---in fact we'd just put more current through a single device until we got down to where r sub bb' was significant. But the base current noise starts to get an excess low freq component at higher currents, and it flows as mentioned above in r sub bb', and as well r sub bb' starts to get larger itself, so things get suboptimal.
If we wanted we could have separate emiiter current sources (not necessarily current sources, say just good-sized resistors) and separate feedback networks (including separate coupling caps), still tying the collectors together, if there was just no way to match the transistors satisfactorily. But parts these days out of a good tight process and with the same date code are pretty decently matched for this kind of work.
jensenmann
Well-known member
:thumb: :guinness: :thumb: :guinness: :thumb: :guinness:
Thanks, you guys ROCK!
Stil it´s hard to understand all this stuff but I started with serious reading. I definitely need to know much more about this stuff.
Sorry to say, but DIY is a heavy addiction, it even makes me read tons of formulae...
:green:
Jens
Thanks, you guys ROCK!
Stil it´s hard to understand all this stuff but I started with serious reading. I definitely need to know much more about this stuff.
Sorry to say, but DIY is a heavy addiction, it even makes me read tons of formulae...
:green:
Jens
