Well, I have been pressed for a bit on mic amp design, so I have knocked
something up here.

http://www.anasoft.co.uk/EE/MicAmpDe...AmpDesign.html


Kevin Aylward

http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.

"Kevin Aylward" wrote in message
...
Well, I have been pressed for a bit on mic amp design, so I have knocked
something up here.

http://www.anasoft.co.uk/EE/MicAmpDe...AmpDesign.html

Good so far. You've listed some specs derived from our discussions on this
group:

Bandwidth 10Hhz - 100Khz +/-0.5db
Distortion 0.005% THD/IMD at all frequencies
and gains.
Minimum Signal 1 mv rms
Maximum Input Signal 20V rms
Maximum Output Signal 20V rms
Gain Range 0db - 60db
CMRR 100db at all frequencies and gains.
EIN (noise) -130dbv or 2.2nV/sqrthz
Rin 2Kohm
SR 20V/us

Bandwidth: Check.
CMRR: Check.
EIN, Rin, SR Check.

But --

The device was only sim-tested at a gain of ~31dB. No word how it performs
in terms of noise and distortion(s) at other gain settings. And it was only
tested at a maximum output of 3.5V (into an unstated load), or about +13dBu.
No word on what distortion products were present other than 2nd & 3rd. No
word on IMD performance. So its success in meeting the noise and distortion
specs is undetermined.

Good start, not even close to fulfilling the assignment.

Peace,
Paul

Paul Stamler wrote:
"Kevin Aylward" wrote in message
...
Well, I have been pressed for a bit on mic amp design, so I have
knocked something up here.

http://www.anasoft.co.uk/EE/MicAmpDe...AmpDesign.html

Good so far. You've listed some specs derived from our discussions on
this group:

Bandwidth 10Hhz - 100Khz +/-0.5db
Distortion 0.005% THD/IMD at all
frequencies and gains.
Minimum Signal 1 mv rms
Maximum Input Signal 20V rms
Maximum Output Signal 20V rms
Gain Range 0db - 60db
CMRR 100db at all frequencies and gains.
EIN (noise) -130dbv or 2.2nV/sqrthz
Rin 2Kohm
SR 20V/us

Bandwidth: Check.
CMRR: Check.
EIN, Rin, SR Check.

But --

The device was only sim-tested at a gain of ~31dB. No word how it
performs in terms of noise and distortion(s) at other gain settings.

Because those results are inherent. I chose some spot results to save
time. I posted the SuperSpice schematics so that the interested ones can
do those tests themselves.

And it was only tested at a maximum output of 3.5V (into an unstated
load), or about +13dBu.

But from a technical point of view, its all in the wash. Just change the
supply voltage and load resister. It wont effect the distortion, cmrr,
noise or anything else.

There is a standard routine set of trade off one does, like actual
collector current, choice of the RE range etc. Its all mundane stuff,
you know, I'm the consultant brain surgeon handing it off to the jr. dr.
Ones the topology is done, its a done deal. Seriously.

One issue, is if you overload it, it starts oscillating:-)

No word on what distortion products were
present other than 2nd & 3rd. No word on IMD performance. So its
success in meeting the noise and distortion specs is undetermined.

Noise is trivial. As in any decent mic amp, its essentially determined
by the noise of the source (200 ohm) and the resister between the
emitters. The noise will increase with the sqrt(RE), with the signal
handling going up with R, such that the S/N will increase from the
minimum by sqrt(RE). I can understand that some may not be fully
conversant with how to extrapolate these results, but I don't have the
time to go into stuff that a designer knows as a matter of course, and
isn't usually concerned with as its taken as read.

Imd, well, load up SS and switch the thd button to imd button to do the
sweep over frequency!

Ye have little faith my son.


Good start, not even close to fulfilling the assignment.


As I said, I don't have the time to go into a full design or full
simulation. I have shown a basic design idea that will do better in all
specs from the other typical topologies I have shown. Like *huge* cmrr,
and better then -90db basic distortion.

Have a play with files yourself.

Kevin Aylward

http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.

Kevin Aylward wrote:
Well, I have been pressed for a bit on mic amp design, so I have knocked
something up here.

http://www.anasoft.co.uk/EE/MicAmpDe...AmpDesign.html

Thanks, Kevin. I don't even play circuit designer on TV so
could I ask you if it is possible to run Full Monty off of
+/- 15V (or even +/- 9V) without compromising the specs
other than signal swing? If so could you please indicate
what resistor values need substituting to get those
operating points?

I'm obviously trying to allow for the most common (thus
least expensive) power supply and also see if it could be
operated on batteries.

For us circuit Dummies, it would be really helpful if you
filled in the components that would effect the current
sources in the circuit and could specify transistors that
can be easily obtained matched. Oh, and also indicate which
transistors must be matched to which others for best
performance.

Oh again, what is the expression for gain involving R5?


Thanks,

Bob
--

"Things should be described as simply as possible, but no
simpler."

A. Einstein

On Sun, 08 May 2005 16:19:28 -0700, Bob Cain
wrote:

Oh again, what is the expression for gain involving R5?

Differential gain is the ratio of differntial output
loading and differential input losses, which includes
semicon-stuff and that resistor, in series.

The semicon-stuff is horribly non-linear and to be
minimized in modern designs, and should (must!) be
dominated by the R5 value.

FWIW, this is not a "design"; it's a back of an envelope
sketch. Of about a tenth of an actual mic preamp.
Overall, several magnitudes away from anything interesting.
Even as a first rough cut.

Chris Hornbeck
"Clean, edgy, gutless, and lifeless." -Dan Kennedy

"Kevin Aylward" wrote in message

But from a technical point of view, its all in the wash. Just change the
supply voltage and load resister. It wont effect the distortion, cmrr,
noise or anything else.

Kev,

You must have 'auto-correct' on in your email editor, and it keeps
erroneously changing resistor to 'resister'. It's been doing it for a week
now.

geoff

Bob Cain wrote:
Kevin Aylward wrote:
Well, I have been pressed for a bit on mic amp design, so I have
knocked something up here.

http://www.anasoft.co.uk/EE/MicAmpDe...AmpDesign.html

Thanks, Kevin. I don't even play circuit designer on TV so
could I ask you if it is possible to run Full Monty off of
+/- 15V (or even +/- 9V) without compromising the specs
other than signal swing?

Supply voltages are not important at all, other then swings.

If so could you please indicate
what resistor values need substituting to get those
operating points?

The issue here is that, I would need to add some more of the routine
circuitery for this to be built, even as a test case.

For example, the cascode voltage sources would be replaced zeners

e.g. http://www.anasoft.co.uk/EE/MicAmpDesign/TheMicAmp.GIF

However, even this simply circuit wont achieve the full CMRR.


I'm obviously trying to allow for the most common (thus
least expensive) power supply and also see if it could be
operated on batteries.

For us circuit Dummies, it would be really helpful if you
filled in the components that would effect the current
sources in the circuit and could specify transistors that
can be easily obtained matched. Oh, and also indicate which
transistors must be matched to which others for best
performance.

Again, this is too much of a topology sketch to consider building yet.

I'll see what time I have to try and make the schematic above, to one
that is actually buildable.


Oh again, what is the expression for gain involving R5?


The single ended gain is the load resister/2.Re.

Note, that CMRR is determined by only *one* output. You don't have to
take the output differentially.


Note, that the intention is op-amp buffers be used at the collectors as
output drivers. There is no intent that any final outputs should be
taken directly from the collectors.

Kevin Aylward

http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.

Chris Hornbeck wrote:
On Sun, 08 May 2005 16:19:28 -0700, Bob Cain
wrote:

Oh again, what is the expression for gain involving R5?

Differential gain is the ratio of differntial output
loading and differential input losses, which includes
semicon-stuff and that resistor, in series.

The semicon-stuff is horribly non-linear and to be
minimized in modern designs, and should (must!) be
dominated by the R5 value.

FWIW, this is not a "design"; it's a back of an envelope
sketch.

I never do paper designs, my sketches are always virtual.

Of about a tenth of an actual mic preamp.

As I clearly explained. It is a topology. The hard part of any design is
deriving a decent topology. All the rest is doggy work that can be
parcelled out, well, hopefully.

Overall, several magnitudes away from anything interesting.

Oh? Show me a mic amp that has a basic cmrr on a par?

Even as a first rough cut.


Turning it into a real design, is routine, but time consuming. How much
do you want to pay me do that?

Kevin Aylward

http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.

Kevin Aylward wrote:

Noise is trivial. As in any decent mic amp, its essentially determined
by the noise of the source (200 ohm) and the resister between the
emitters. The noise will increase with the sqrt(RE), with the signal
handling going up with R, such that the S/N will increase from the
minimum by sqrt(RE). I can understand that some may not be fully
conversant with how to extrapolate these results, but I don't have the
time to go into stuff that a designer knows as a matter of course, and
isn't usually concerned with as its taken as read.

Noise isn't trivial. And what is interesting is first of all that it
differs with the source impedance (which is not always constant with
frequency), and that it the noise spectrum itself matters. Problem is
that folks differ on just how it matters.

The INA103 has a stated 1 nv/sqrt(hz) specification, but it can occasionally
be very noisy on microphones with fairly high output impedances. It's also
interesting to see that a lot of that noise isn't 1/f noise either.

Ye have little faith my son.

It's true that a lot of us have very little faith in simulation, because
somehow very small errors in device models turn into very large errors in
overall model behaviour. Admittedly device models are getting better and
better every day, but still it's only a first step toward actual testing
until we get perfect device models and perfect modelling of parasitics.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

Bob Cain wrote:
Kevin Aylward wrote:
Well, I have been pressed for a bit on mic amp design, so I have knocked
something up here.

http://www.anasoft.co.uk/EE/MicAmpDe...AmpDesign.html

Thanks, Kevin. I don't even play circuit designer on TV so
could I ask you if it is possible to run Full Monty off of
+/- 15V (or even +/- 9V) without compromising the specs
other than signal swing? If so could you please indicate
what resistor values need substituting to get those
operating points?

I'm obviously trying to allow for the most common (thus
least expensive) power supply and also see if it could be
operated on batteries.

For battery use, +/-18V is very easy because you can do it with
four 9V batteries. A very convenient level. You could run this
circuit at less than rated voltage because everything is biased using
constant current and voltage sources so the bias points won't change
with changes in rail voltages.

For us circuit Dummies, it would be really helpful if you
filled in the components that would effect the current
sources in the circuit and could specify transistors that
can be easily obtained matched. Oh, and also indicate which
transistors must be matched to which others for best
performance.

This isn's a constructable circuit yet. It still has things like
abstract current sources, which don't exist in the real world and have
to be replaced with real current sources built with a transistor and
a couple discretes. This then changes the noise calculation a lot,
by the way. I don't know how significant it will be since I haven't
done the math on the resistor noise of the front end for comparison.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

Scott Dorsey wrote:
Kevin Aylward wrote:

Noise is trivial. As in any decent mic amp, its essentially
determined by the noise of the source (200 ohm) and the resister
between the emitters. The noise will increase with the sqrt(RE),
with the signal handling going up with R, such that the S/N will
increase from the minimum by sqrt(RE). I can understand that some
may not be fully conversant with how to extrapolate these results,
but I don't have the time to go into stuff that a designer knows as
a matter of course, and isn't usually concerned with as its taken as
read.

Noise isn't trivial.

The calculation of noise is trivial in spice, not that I enjoy a large
dose of hiss in my sound.

And what is interesting is first of all that it
differs with the source impedance (which is not always constant with
frequency),

Of course. For example, shot noise across a guitar pickup is often an
issue. It is give by

Vn = 2.pi.F.L. Ibn

It is why a fet input is best for guitar, noise wise. It has very low
current shot noise.

and that it the noise spectrum itself matters. Problem is
that folks differ on just how it matters.

The INA103 has a stated 1 nv/sqrt(hz) specification, but it can
occasionally be very noisy on microphones with fairly high output
impedances. It's also interesting to see that a lot of that noise
isn't 1/f noise either.

I really don't see your point with all this. Its all in the wash. Spice
just crunches the numbers.

To design a low noise amp, one searches the data sheets and find the
lowest rbb' device with largest hfe that one can find. One might have to
trade off one from the other, depending on the souce impedance with
frequency. One runs simulations to optimises this.


Ye have little faith my son.

It's true that a lot of us have very little faith in simulation,

Indeed. However, this is by those that are not engaged in professional
high transistor count analogue circuit design.

because somehow very small errors in device models turn into very
large errors in overall model behaviour.

In principle, yes. In practise, hardly ever, if ever. I have never seen
this happen, and I know a bit about simulation. Models themselves are
not usually that sensitive to parameter values.

What is more important is given a model in *a circuit* has parameter
variations, how does the circuit *response* change. This is much more
likely an occurrence. However, such an obvious aspect does not go
unnoticed by any but the most inexperienced, and systems are well in
place to handle such issues as matter of routine. Typically, this is
dealt with by using a combination of Worst Case Models and performing
worst case simulations and Monty Carlo simulations.

A typically WC rerun setup can be seen here
http://www.anasoft.co.uk/WCSETUP.GIF

with an explanation here http://www.anasoft.co.uk/worstcase.html

No professional analogue engineer, ****es about on the bench getting a
one off to work, which is what you seem to be implying. We are not daft
you know.

Admittedly device models
are getting better and better every day,

I.c. models at companies are usually nuts on. The're brilliant. They
often have dedicated teams working on them, certainly at Texas
Instruments they do.

Its not uncommon to have a 10,000 transistor analogue circuit be
sellable on first pass through the fab, designed purely by simulation.

but still it's only a first
step toward actual testing until we get perfect device models and
perfect modelling of parasitics.

Now let me tell you how it is. Don't argue on it, cos it aint debatable,
and with all due respect Scott, it dont look like you know how it is,
mate. Its not a first step, for probably 95% of analogue ic design,
simulation is the only step. Sure, for some applications, parasitics and
so forth can be an issue, but much mainstream design works on the bench
first time, or it can do if the designer gets his act together.

1000's and 1000's, of 100-10,000 transistor count, analogue i.c.s are
designed by 1000's of analogue engineers every year, essentially, by
only using simulation. No one breadboads in the i.c. design universe
today. Its impossible to do so.

First, cadence analogue artist setups cost maybe $30,000 per years
licence, so like, companies do that for their fun?

It costs say, $100k, for a set of masks. 2 months to get the die back
from fab. Layout engineers, production engineers, all costing much cash,
resulting in maybe $300k in NRE for a design, it could be up to $1M for
some. One simply can't afford to **** about on the bench like a high
school boy, and we don't.

I can see where your coming from, but truly, your out of date, things
have moved on. Simulation to bench confirmation to data sheets is pretty
much the norm. There are incentives to make it that way. A first time
pass success can get you a $30k bonus, as it saves $300k.

Typically I have spent say, 2 months solid 40 hours a week simulation
for an i.c. design. Then, maybe a week in the lab checking it out, and
making data sheets. A lot of the lab work, is simply waiting for the
bloody thermal stream box to get up and down in temperate to confirm
specs.

Kevin Aylward

http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.

Kevin Aylward wrote:


Now let me tell you how it is. Don't argue on it, cos it aint debatable,
and with all due respect Scott, it dont look like you know how it is,
mate. Its not a first step, for probably 95% of analogue ic design,
simulation is the only step. Sure, for some applications, parasitics and
so forth can be an issue, but much mainstream design works on the bench
first time, or it can do if the designer gets his act together.

Absolutely, but what you're looking at here _isn't_ analogue IC design.
It's very different in philosophy and practice.

1000's and 1000's, of 100-10,000 transistor count, analogue i.c.s are
designed by 1000's of analogue engineers every year, essentially, by
only using simulation. No one breadboads in the i.c. design universe
today. Its impossible to do so.

Right, but what does that have to do with simple discrete designs? With
analogue ICs, device parameters are probably better characterized than
with discrete stuff, and parasitics are under a lot better control.

I can see where your coming from, but truly, your out of date, things
have moved on. Simulation to bench confirmation to data sheets is pretty
much the norm. There are incentives to make it that way. A first time
pass success can get you a $30k bonus, as it saves $300k.

Yup, that's a big disadvantage with IC design, that you basically lose
the whole successive refinement procedure.

Typically I have spent say, 2 months solid 40 hours a week simulation
for an i.c. design. Then, maybe a week in the lab checking it out, and
making data sheets. A lot of the lab work, is simply waiting for the
bloody thermal stream box to get up and down in temperate to confirm
specs.

I get 2N5088s in bags from On Semiconductor. I order a bag and
sometimes it comes marked "Made in Korea" and sometimes it comes marked
"Made in Thailand." Both of the bags contain parts that meet the published
specifications for the 2N5088, but the noise floor on the ones from Korea
are much lower than the noise floor for the ones in Thailand.

I get 2N5457s from On Semiconductor too. They all meet the specs for
the 2N5457, which are very, very lax. But they turn out also to have
much better noise performance than the specifications require, to the
point that if you sort them for Gm, they are actually not bad for audio
applications. BUT, the bias point still varies a lot from unit to unit
so any design has to take this into account.

Only by actually putting parts into circuits and trying them to do you
find this sort of thing out. The data sheet numbers are only a starting
point.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

Scott Dorsey wrote:
Bob Cain wrote:
Kevin Aylward wrote:
Well, I have been pressed for a bit on mic amp design, so I have
knocked something up here.

http://www.anasoft.co.uk/EE/MicAmpDe...AmpDesign.html

Thanks, Kevin. I don't even play circuit designer on TV so
could I ask you if it is possible to run Full Monty off of
+/- 15V (or even +/- 9V) without compromising the specs
other than signal swing? If so could you please indicate
what resistor values need substituting to get those
operating points?

I'm obviously trying to allow for the most common (thus
least expensive) power supply and also see if it could be
operated on batteries.

For battery use, +/-18V is very easy because you can do it with
four 9V batteries. A very convenient level. You could run this
circuit at less than rated voltage because everything is biased using
constant current and voltage sources so the bias points won't change
with changes in rail voltages.

For us circuit Dummies, it would be really helpful if you
filled in the components that would effect the current
sources in the circuit and could specify transistors that
can be easily obtained matched. Oh, and also indicate which
transistors must be matched to which others for best
performance.

This isn's a constructable circuit yet. It still has things like
abstract current sources, which don't exist in the real world and have
to be replaced with real current sources built with a transistor and
a couple discretes. This then changes the noise calculation a lot,
by the way.

It dosnt effect the noise at all.

What is of note is that the current soures have to be very, very good to
get the implied CMRR.

I don't know how significant it will be since I haven't
done the math on the resistor noise of the front end for comparison.

I have:-)


Kevin Aylward

http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.

here's a nice practical aspect to consider....

we need to add some RF EMI protection at the input and still maintain
the CMRR.

Mark

Scott Dorsey wrote:
Kevin Aylward wrote:


Now let me tell you how it is. Don't argue on it, cos it aint
debatable, and with all due respect Scott, it dont look like you
know how it is, mate. Its not a first step, for probably 95% of
analogue ic design, simulation is the only step. Sure, for some
applications, parasitics and so forth can be an issue, but much
mainstream design works on the bench first time, or it can do if the
designer gets his act together.

Absolutely, but what you're looking at here _isn't_ analogue IC
design.

Agreed.

It's very different in philosophy and practice.

Not really. Its volts and amps. A few differences, but mostly its all
the same.


1000's and 1000's, of 100-10,000 transistor count, analogue i.c.s are
designed by 1000's of analogue engineers every year, essentially, by
only using simulation. No one breadboads in the i.c. design universe
today. Its impossible to do so.

Right, but what does that have to do with simple discrete designs?

A lot. Its how I design discrete circuits. If I design a discrete
circuit, it *will* work on the bench, parasitics and all.

With analogue ICs, device parameters are probably better
characterized than with discrete stuff, and parasitics are under a
lot better control.

Parasitics may not be an issue. In this example, the main one is
probably board issued with getting good cmrr.

As I noted, I use worst case models that are invariable way worse then
what is manufactured. One uses factors of safety.


I can see where your coming from, but truly, your out of date, things
have moved on. Simulation to bench confirmation to data sheets is
pretty much the norm. There are incentives to make it that way. A
first time pass success can get you a $30k bonus, as it saves $300k.

Yup, that's a big disadvantage with IC design, that you basically lose
the whole successive refinement procedure.

Typically I have spent say, 2 months solid 40 hours a week simulation
for an i.c. design. Then, maybe a week in the lab checking it out,
and making data sheets. A lot of the lab work, is simply waiting for
the bloody thermal stream box to get up and down in temperate to
confirm specs.

I get 2N5088s in bags from On Semiconductor. I order a bag and
sometimes it comes marked "Made in Korea" and sometimes it comes
marked "Made in Thailand." Both of the bags contain parts that meet
the published specifications for the 2N5088, but the noise floor on
the ones from Korea are much lower than the noise floor for the ones
in Thailand.

Noise is a little tricky. Unless it is a low noise transistor
specifically guaranteed for low 1/f, it could be anything.

The next dominate noise source variation is due to hfe. Too small a hfe
will increase input shot noise, via Zs.sqrt(2.q.IC/hfe)

rbb' is all that is left for noise considerations. rbb' is chosen to be
small by design, say 20 ohms, which is low in relation to the mic
resistance, if it varies a bit, it wont really matter.


I get 2N5457s from On Semiconductor too. They all meet the specs for
the 2N5457, which are very, very lax. But they turn out also to have
much better noise performance than the specifications require, to the
point that if you sort them for Gm, they are actually not bad for
audio applications.

You dont sort for gm. gm is the same for all tramsisters, its:

gm = Ic/25mV

Do you mean sort vbe or hfe matching?

BUT, the bias point still varies a lot from unit
to unit so any design has to take this into account.

Vbe might vary, say up to 50mv. Any decent design will be such that Vbe
dont matter.


Only by actually putting parts into circuits and trying them to do you
find this sort of thing out.

I disagree. 99.9% of analogue design can be all done in the virtual
world. The bits that are hard to simulate with Spice, are things like
x-talk. However, there are other tools that can do this, expensive as
they may be.

The data sheet numbers are only a
starting point.

I don't agree at all. Of course models can be in error, but as I said,
one can, and does, make worst case models that include pretty much 99.9%
of all conditions.

I can guarantee that if I design discrete circuits by simulation, that
are not in the 2.5 Ghz league sort of stuff, it will work correctly.
Simulation has come a long way over the last 20 years.

Kevin Aylward

http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.

Mark wrote:
here's a nice practical aspect to consider....

we need to add some RF EMI protection at the input and still maintain
the CMRR.


You want fries on that as well?


Kevin Aylward

http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.

Kevin Aylward wrote:

I get 2N5457s from On Semiconductor too. They all meet the specs for
the 2N5457, which are very, very lax. But they turn out also to have
much better noise performance than the specifications require, to the
point that if you sort them for Gm, they are actually not bad for
audio applications.

You dont sort for gm. gm is the same for all tramsisters, its:

gm = Ic/25mV

Do you mean sort vbe or hfe matching?

Oh, sorry, it's a J-Fet. Sort for mu or transconductance instead of
for beta. I didn't make that clear.

Only by actually putting parts into circuits and trying them to do you
find this sort of thing out.

I disagree. 99.9% of analogue design can be all done in the virtual
world. The bits that are hard to simulate with Spice, are things like
x-talk. However, there are other tools that can do this, expensive as
they may be.

It's that 0.1% that is the real sticky point, though.

I can guarantee that if I design discrete circuits by simulation, that
are not in the 2.5 Ghz league sort of stuff, it will work correctly.
Simulation has come a long way over the last 20 years.

Actually, with the modern stripline simulation stuff, getting stuff in
the 2.5 GHz range _is_ probably easier to deal with in the virtual world.
--scott

--
"C'est un Nagra. C'est suisse, et tres, tres precis."

Kevin Aylward wrote:
Mark wrote:
here's a nice practical aspect to consider....

we need to add some RF EMI protection at the input and still maintain
the CMRR.

You want fries on that as well?

This weekend, I did a concert recording at a local church where I had never
worked before. Unfortunately in the small room where I set up, there were
28 cheap household dimmers for the lighting circuits. You could brush your
hand against the conduits and feel a tingle from the RF trash.

I set up a pair of Sennheiser MKH104s in a Jecklin disc. Now, the MKH104
basically drives only one leg of the output, and the other is tied to
ground. The source impedance is therefore imbalanced, but it is usually
not a problem because the source impedance of the mike is pretty low
anyway so the imbalance is not a huge one.

For some reason, the concertmaster really likes the sound of the recordings
done with the MKH104. When I have given him a choice between recordings
made at the same place with a couple mikes, he'll pick the MKH104 over
the others. So I use it, in spite of the fact that I think it sounds a
little smeary in comparison with the better omnis I have in the kit.

I ran about 200 feet of star-quad cable into the input of the Nagra
recorder, which is transformer-balanced and has really outrageously
good CMRR even by transformer standards. And what I got was buzz.

The first thing I found is that my moving the cables about three feet
from the wall, I could reduce the buzz substantially.

If I put a 15 dB pad at the recorder, the buzz was not reduce appreciably
but the signal was dropped 15 dB. BUT, if I put the same 15 dB pad at
the microphone, the buzz dropped completely. This is because the series
resistance of the pad was much higher than the output impedance of the
microphone, so effectively I was reducing the percentage of impedance
imbalance on the two legs with the pad.

With the buzz gone, I could hear the air conditioning system rumble
and the slap echo from the side walls perfectly. Anyway, I just thought
this was another interesting diversion into the wonderful world of
noise rejection.
--scott


--
"C'est un Nagra. C'est suisse, et tres, tres precis."

Scott Dorsey wrote:
I set up a pair of Sennheiser MKH104s in a Jecklin disc. Now, the MKH104
basically drives only one leg of the output, and the other is tied to
ground. The source impedance is therefore imbalanced
[snip horror story]

For a mic of that price, it seems surprisingly cheap not to have used a
resistor to ground of comparable value to the output impedance of the
hot leg. It's common practice in line level outputs, providing a
perfectly usable balanced output without the expense of inverting the
signal to drive the other leg.

--
Anahata
-+- http://www.treewind.co.uk
Home: 01638 720444 Mob: 07976 263827

anahata wrote:
Scott Dorsey wrote:
I set up a pair of Sennheiser MKH104s in a Jecklin disc. Now, the MKH104
basically drives only one leg of the output, and the other is tied to
ground. The source impedance is therefore imbalanced
[snip horror story]

For a mic of that price, it seems surprisingly cheap not to have used a
resistor to ground of comparable value to the output impedance of the
hot leg. It's common practice in line level outputs, providing a
perfectly usable balanced output without the expense of inverting the
signal to drive the other leg.

It is common practice now, but it wasn't back when that mike was new.
I think I'm going to be taking a soldering iron to the power supply
some time this week and changing that.
--scott

--
"C'est un Nagra. C'est suisse, et tres, tres precis."

"Scott Dorsey" wrote in message

It's true that a lot of us have very little faith in simulation, because
somehow very small errors in device models turn into very large errors in
overall model behaviour. Admittedly device models are getting better and
better every day, but still it's only a first step toward actual testing
until we get perfect device models and perfect modelling of parasitics.


Question, because I really don't know.

Do these SPICE simulations model complex moving waveforms with multiple hf
tones and harmonics varying in amplitude and time ? These are the factors
which IMO create the subtle (and not so subtle) differences in pre/amp
sound.

Steady state THD is only a small part of a sound quality.


geoff

Kevin Aylward wrote:
Bob Cain wrote:

Thanks, Kevin. I don't even play circuit designer on TV so
could I ask you if it is possible to run Full Monty off of
+/- 15V (or even +/- 9V) without compromising the specs
other than signal swing?

Supply voltages are not important at all, other then swings.

I'd hoped that was the case.

For example, the cascode voltage sources would be replaced zeners

e.g. http://www.anasoft.co.uk/EE/MicAmpDesign/TheMicAmp.GIF

However, even this simply circuit wont achieve the full CMRR.

But it's getting closer to something buildable. Thanks.

Again, this is too much of a topology sketch to consider building yet.

I'll see what time I have to try and make the schematic above, to one
that is actually buildable.

Any efforts will be appreciated. I'm involved in some DIY
groups that will likely jump all over this if you do find
the time and I guarantee you that your name will be
attached. Some of us (not me) have the gear to give
independant validation of the specs and probably would do so.

Oh again, what is the expression for gain involving R5?

The single ended gain is the load resister/2.Re.

Note, that CMRR is determined by only *one* output. You don't have to
take the output differentially.

Note, that the intention is op-amp buffers be used at the collectors as
output drivers. There is no intent that any final outputs should be
taken directly from the collectors.

Got it.


Thanks,

Bob
--

"Things should be described as simply as possible, but no
simpler."

A. Einstein

Geoff Wood wrote:
"Scott Dorsey" wrote in message

It's true that a lot of us have very little faith in simulation, because
somehow very small errors in device models turn into very large errors in
overall model behaviour. Admittedly device models are getting better and
better every day, but still it's only a first step toward actual testing
until we get perfect device models and perfect modelling of parasitics.


Question, because I really don't know.

Do these SPICE simulations model complex moving waveforms with multiple hf
tones and harmonics varying in amplitude and time ? These are the factors
which IMO create the subtle (and not so subtle) differences in pre/amp
sound.

Yes, and that's the neat thing about them. You can put an arbitrary waveform
into them, and get a waveform out.

Is it the same waveform the breadboarded circuit would have? Only your
hairdresser knows for sure.

Steady state THD is only a small part of a sound quality.

Steady state THD measures a whole bunch of different things together, some
of which are very important and some of which aren't important at all.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

Hello Kevin,

I'd also mind the roll-off at higher frequencies, to avoid
susceptibility to EMI. Big row of fluorescent being turned on,
transmitters etc.

Maybe caps across BE or ferrite beads in the base lines?

Regards, Joerg

http://www.analogconsultants.com

selecting a bipolar differential amplifier as a TOPOLOGY for a mic pre
is pretty obvious.

Mark

Geoff Wood wrote:
"Scott Dorsey" wrote in message

It's true that a lot of us have very little faith in simulation,
because somehow very small errors in device models turn into very
large errors in overall model behaviour. Admittedly device models
are getting better and better every day, but still it's only a first
step toward actual testing until we get perfect device models and
perfect modelling of parasitics.


Question, because I really don't know.

Do these SPICE simulations model complex moving waveforms with
multiple hf tones and harmonics varying in amplitude and time ?

Yes.

Spice will model pretty much anything electronic you want.

These are the factors which IMO create the subtle (and not so subtle)
differences in pre/amp sound.

Oh?

What do you base that conjecture on?


Steady state THD is only a small part of a sound quality.


Steady state distortion measurements of each harmonic, by and large
characterise an amplifier very well. Sure, I can artificially construct
an amplifier that has impressive thd on steady state, and fails on burst
tests, but I don't.

In the distant past, there was an awfull lot of cheating going on in
specs and design. For the most part, its still in the past. Steady state
tests on modern amps show up most, if not all audible artefacts.

Ok. Here is an example where a steady state test isn't quite the same as
a transient burst. Its the tube amp. Tube amps have coupling caps to the
main output valve grids. If you overload the valve so it clips, grid
current starts flowing, asymmetrically charging up the capacitor,
producing a different grid bias voltage then in non overload state. This
causes a steady shift in anode bias current. This might well be audible.
Its quite a large effect. The issue here though, is that this effect is
occurring in an overload condition. In normal operation, these types of
dynamic biasing effects are usually very minimal.

Can similar effects occur in conventional transistor amps, running
unclipped? In principle, yes, but in practice, I am sceptical that there
would be any audible effects. Asymetrical charging of the input
capacitance of output transistors, when a bust is applied, is a
possibility, in principle, as they are nonlinear. This could change the
distortion spectrum from steady state.

Certainly, in oscilloscope design, they have been fundamental issues
with thermal tails of devices, causing relatively long settling times,
so these potential accuracy issues have long ben invesigated,
indepandant of audio.

If an amplifier is designed such it has a slew rate margin, say 3-5
times what is actually, required from the signal, it is unlikely, in my
view, that potential dynamic biasing issues due to that slew will be
audible. As far as the above output bias effects, so long as the bias is
enough, it shouldn't be an issue. I suppose it all depends on how
competent the designer was:-)

Kevin Aylward

http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.

On Mon, 09 May 2005 10:13:40 GMT, "Kevin Aylward"
wrote:

FWIW, this is not a "design"; it's a back of an envelope
sketch.

I never do paper designs, my sketches are always virtual.

And I'm just the opposite. Maybe that explains a lot of our
differing perceptions.

Overall, several magnitudes away from anything interesting.

Oh? Show me a mic amp that has a basic cmrr on a par?

First you show me a mic amp.

Real-world CMRR is degraded by the real-world upstream
bandlimiting necessary in the real world. It's trivially
easy to *buy* an AD797 with 130dB CMRR. But a real amp
using it won't acheive that, or anything close (without
a transformer, or multi-staging instrumentation design, etc.)

Chris Hornbeck
"Clean, edgy, gutless, and lifeless." -Dan Kennedy

On Mon, 09 May 2005 20:04:32 +0100, anahata
wrote:

Scott Dorsey wrote:
I set up a pair of Sennheiser MKH104s in a Jecklin disc. Now, the MKH104
basically drives only one leg of the output, and the other is tied to
ground. The source impedance is therefore imbalanced
[snip horror story]

For a mic of that price, it seems surprisingly cheap not to have used a
resistor to ground of comparable value to the output impedance of the
hot leg. It's common practice in line level outputs, providing a
perfectly usable balanced output without the expense of inverting the
signal to drive the other leg.

It does contribute to the mic's noise output. Maybe *any* is
too much in a world of spec numbers competition.

Chris Hornbeck
"Clean, edgy, gutless, and lifeless." -Dan Kennedy

1000's and 1000's, of 100-10,000 transistor count, analogue i.c.s are
designed by 1000's of analogue engineers every year, essentially, by
only using simulation. No one breadboads in the i.c. design universe
today. Its impossible to do so.

Au contraire. One of the regulars on alt.binaries.schematics.electronic
(Jim Thompson) is a working IC circuit designer. Some of his designs
are likely running in your home as you read this. Just a week or two
ago he tells about taking an existing monolithic integrated circuit
design
and breadboarding it in discretes so that he could optimize the circuit
(which he did).

It has been years (decades?) since we breadboarded a major CPU
chip (back before the '386?), but analog circuits are more likely to
be breadboarded AND simulated.

Geoff Wood wrote:

"Scott Dorsey" wrote in message

It's true that a lot of us have very little faith in simulation, because
somehow very small errors in device models turn into very large errors in
overall model behaviour. Admittedly device models are getting better and
better every day, but still it's only a first step toward actual testing
until we get perfect device models and perfect modelling of parasitics.

Question, because I really don't know.

Do these SPICE simulations model complex moving waveforms with multiple hf
tones and harmonics varying in amplitude and time ? These are the factors
which IMO create the subtle (and not so subtle) differences in pre/amp
sound.

Steady state THD is only a small part of a sound quality.

Indead steady state THD is sadly blunt measurement tool. You can tell quite a
lot more by looking at the distortion components though.

OTOH designs with fabulously low THD usually sound very good. I think that's
simply a reflection of the care needed to acheive one goal being responsible
for good performance throughout.

Graham

On Sun, 08 May 2005 16:19:28 -0700, Bob Cain
wrote:



Kevin Aylward wrote:
Well, I have been pressed for a bit on mic amp design, so I have knocked
something up here.

http://www.anasoft.co.uk/EE/MicAmpDe...AmpDesign.html

Thanks, Kevin. I don't even play circuit designer on TV so
could I ask you if it is possible to run Full Monty off of
+/- 15V (or even +/- 9V) without compromising the specs
other than signal swing? If so could you please indicate
what resistor values need substituting to get those
operating points?

I'm obviously trying to allow for the most common (thus
least expensive) power supply and also see if it could be
operated on batteries.

For us circuit Dummies, it would be really helpful if you
filled in the components that would effect the current
sources in the circuit

Oh, gee, you've got to design your own current sources, or at least
google for them. ISTR Bob Pease writing about them in his Pease
Porrige. You can probably google for What's All This Current Source
Stuff, Anyhow? and find it.

Cheap low-end current source circuit description: Two
forward-biased diodes in series connected to V- with the other end
connected to a resistor (scaled for I dunno, maybe 40uA through the
diodes) that goes to ground. NPN base goes to the diode-resistor
junction, emitter goes to a resistor which goes to V-. VBE is the same
as one diode drop (0.6V), so the voltage across the emitter resistor
equals the drop across the other diode (also 0.6v). Emitter current
equals collector current (plus about 1 percent for base current, but
we're ignoring that), so scale the resistor for the current you want.
For 1mA, R = V/I = 0.6/0.001 = 600 ohms. The collector becomes a 1-mA
current sink from VCEmax down to about 0.8V (Vdiode + Vcesat) above
V-. Reverse the rail and polarities for a current source from V+.
Anyone familiar enough with the circuit and transistor parameters
knows this 'constant' current varies somewhat with supply voltage,
temperature differential between the transistor and diodes, and VCE.
Those can be fixed by throwing more transistors at it, but it works
well enough for a first approximation.

Doing those two five-volt floating supplies in the Full Monty
schematic are a little more complicated, but I at least have an idea
of how to do it (hoping it's not thought of as too crude).

and could specify transistors that
can be easily obtained matched.

I thought of asking Kevin this myself, but I actually have enough
knowledge to give an answer.

Offhand cheap, if not yet completely obsolete, is the old LM/CA3046
NPN array, originally in a 14-DIP. I'm sure there is or was a cheap
PNP array as well.
I bet the LM394 would work nicely, but at $5 per matched pair,
you'd want to pick the best place(s) for it. The data sheet has some
interesting schematics, specifically the "high performance
instrumentation amplifier" which uses three of the darn things (could
they have designed it to use more?). Use a few in your next
high-volume design, maybe with good volume the price will go down.
There are also high-quality (i. e. expensive) matched transistor pairs
from THAT, and (also ISTR quads) from Analog Devices. I haven't
actually used any of these, though I have a couple of LM394's.

A cheap (moneywise, not timewise) way of generating lots of
'matched pairs' is get a bunch (100 or more, the more you get the
closer the matching, you'll use a lot anyway) of cheap transistors of
sufficient Vbe and other specs (probably min beta of 50), and sort
them by beta. I found this test circuit for power transistors, and it
appears only Rc1-Rc7 need to be scaled for signal transistors to give
maybe a milliamp of collector current (and maybe replace the
darlington TIP142 with a single transistor for lower leakage into the
base of the DUT):

http://sound.westhost.com/project106.htm

The logistics: Number each transistor sequentially with a 'unit #'
(perhaps with a number on a piece of paper taped to it), then measure
each one's gain (handle with tweezers rather than fingers so you won't
warm up each transistor when installing in the tester, and don't let
the room temperature vary much while doing all the measurements, and
unless your room temp is 98.6F, don't breathe on anything), put unit #
and gain in adjacent spreadsheet columns, select both columns and sort
by gain. The two adjacent units with the closest gains are 'matched'.
I have 60-volt parts MPSA06 (NPN) and MPSA56 (PNP), they're cheap
enough from digikey. I use them for switching phantom power on the
circuit on my aging webpage.

I recall an Audio Amateur cover with an array of transistors laid
out on a sheet of paper, and a number (surely beta) written by each
one. Back then they had to sort the gain values manually, or even
worse, use a BASIC program to sort them.

Oh, and also indicate which
transistors must be matched to which others for best
performance.

What seems somewhat obvious to me is that, looking at the circuit
as a mirror image, each 'original' and 'reflected' transistor should
be matched. In the Full Monty schematic, that would be QP1 matched
with QP2, QN3 matched with QN4, and QN5 matched with QN6. Some of
these pairs may not need matching (it looks like QN5 and QN6 could
have wildly different gains with little effect on the circuit), but if
you've sorted all those cheap transistors, you might as well use up
your matched pairs.
While you're at it, arrange the PCB layout so each pair is
physically close, so when installed their backs can be glued together
so they are always very close to the same temperature.

Oh again, what is the expression for gain involving R5?


Thanks,

Bob

-----
http://mindspring.com/~benbradley

"Kevin Aylward" wrote in message
k...


These are the factors which IMO create the subtle (and not so subtle)
differences in pre/amp sound.

Oh?

What do you base that conjecture on?

Listening to a bunch of mic and power amps with ostensibly identical specs,
that sound very different, especially when paying attention delicate signals
and harmonics = 'detail'...

geoff

Ben Bradley wrote:

Oh, gee, you've got to design your own current sources, or at least
google for them. ISTR Bob Pease writing about them in his Pease
Porrige. You can probably google for What's All This Current Source
Stuff, Anyhow? and find it.

Er, I didn't _want_ something that I could (possibly) teach
myself in four easy Googles, I wanted to know what the
designer thinks is best in the context of _his_ circuit.

Ya know, I just don't get this new thing of telling people
who asks a question that they should have consulted Google
first to eliminate the need to ask it. Since when did it
become a requirement that people first search for printed
answers before asking people? That's nuts if you ask me and
it seems to have become the de rigor response to questions
just about everywhere in this medium. What's up with that?
In real life if you tell someone to "look it up" when they
ask you a question you're liable to get a boxing lesson.

But thanks for the useful information that followed the jab.
Really. :-)


Bob
--

"Things should be described as simply as possible, but no
simpler."

A. Einstein

Richard Crowley wrote:
1000's and 1000's, of 100-10,000 transistor count, analogue i.c.s are
designed by 1000's of analogue engineers every year, essentially, by
only using simulation. No one breadboads in the i.c. design universe
today. Its impossible to do so.

Au contraire.

Oh?

One of the regulars on
alt.binaries.schematics.electronic (Jim Thompson) is a working IC
circuit designer.

Yes. I know Jim very well as a poster.

Some of his designs are likely running in your home
as you read this. Just a week or two ago he tells about taking an
existing monolithic integrated circuit design
and breadboarding it in discretes so that he could optimize the
circuit (which he did).

If Jim did this, it is an extremely rare exception. Jim is a consultant
and designs to spec using PSpice. The idea of "optimising" an ic design
on a bench, is quaint, but generally unrealistic. Sure, there may be a
couple of issues to address, but on a % basis not much. An analogue ic
might have 1000 transistors to "optimise", doing this on a bench just
aint gonna happen.


It has been years (decades?) since we breadboarded a major CPU
chip (back before the '386?), but analog circuits are more likely to
be breadboarded AND simulated.

No. i.c. designs are never breadboarded today. Never means like
0.000000001% of the time.

As someone having the grandiose title of analogue ic design team leader
at Texas Instruments for 3 1/2 years, I never saw an analogue breadboard
once. People used to use kit parts, like 20 years ago, but not now. An
i.c. company has rows of engineers all sitting at desks simulating. When
the dies come back they are checked out in the lab. If there are issues,
that were not seen in simulation, new simulations are done to try and
identify the problem. Doing a change is a bit of work. Usually extra
transistors are placed on the die for prototypes so that they can be
fibed (electron beam wired up). Lasers are also use to cut metal routes.

Kevin Aylward

http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.

Geoff Wood wrote:
"Kevin Aylward" wrote in message
k...


These are the factors which IMO create the subtle (and not so
subtle) differences in pre/amp sound.

Oh?

What do you base that conjecture on?

Listening to a bunch of mic and power amps with ostensibly identical
specs, that sound very different, especially when paying attention
delicate signals and harmonics = 'detail'...


This proves nothing. You don't know what the real specs are, all you
have seen are spot ones, chosen to look good. You also have no way of
knowing whether difference, if any, are due to proposed transient
effects or steady state effects.


Kevin Aylward

http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.

Pooh Bear wrote:
Geoff Wood wrote:

"Scott Dorsey" wrote in message

It's true that a lot of us have very little faith in simulation,
because somehow very small errors in device models turn into very
large errors in overall model behaviour. Admittedly device models
are getting better and better every day, but still it's only a
first step toward actual testing until we get perfect device models
and perfect modelling of parasitics.

Question, because I really don't know.

Do these SPICE simulations model complex moving waveforms with
multiple hf tones and harmonics varying in amplitude and time ?
These are the factors which IMO create the subtle (and not so
subtle) differences in pre/amp sound.

Steady state THD is only a small part of a sound quality.

Indead steady state THD is sadly blunt measurement tool. You can tell
quite a lot more by looking at the distortion components though.

OTOH designs with fabulously low THD usually sound very good. I think
that's simply a reflection of the care needed to acheive one goal
being responsible for good performance throughout.

Yes.


Graham

But he wasn't really referring just to the thd. What he is actually
claiming is that steady state distortion specs don't tell the whole
story.

We know that thd, on its own is not enough to characterise the sound if
the distortion is not low enough. For example, 1% second harmonic 1% 3rd
harmonic, do not sound the same.

The implication was made that distortion changes as a function of
transient conditions. That is, consider a continuous on off burst of
sine waves. Are the sine wave cycles during the burst the same as in
steady state?

In principle this possible, because non-linear capacitors can charge up
giving a different average dc voltage but does a decent design actually
do this to any significance? Or are thermal tails significant in a
typical amp?

Kevin Aylward

http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.

Mark wrote:
selecting a bipolar differential amplifier as a TOPOLOGY for a mic pre
is pretty obvious.

Ahmmmm...

Kevin Aylward

http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.

Chris Hornbeck wrote:
On Mon, 09 May 2005 10:13:40 GMT, "Kevin Aylward"
wrote:

FWIW, this is not a "design"; it's a back of an envelope
sketch.

I never do paper designs, my sketches are always virtual.

And I'm just the opposite. Maybe that explains a lot of our
differing perceptions.

It was a bit tongue in cheek.


Overall, several magnitudes away from anything interesting.

Oh? Show me a mic amp that has a basic cmrr on a par?

First you show me a mic amp.

Real-world CMRR is degraded by the real-world upstream
bandlimiting necessary in the real world. It's trivially
easy to *buy* an AD797 with 130dB CMRR. But a real amp
using it won't acheive that, or anything close (without
a transformer, or multi-staging instrumentation design, etc.)


I did note in another post, that a mere 1pf on the emitter to ground
reduces the single ended cmrr from the ideal 145db to 100db at 20khz.
One can take the output differentially, and it should be infinite,
ideally:-)

Kevin Aylward

http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.

Joerg wrote:
Hello Kevin,

Hi Joerg. Haven't noticed you here before yet.


I'd also mind the roll-off at higher frequencies, to avoid
susceptibility to EMI. Big row of fluorescent being turned on,
transmitters etc.

Maybe caps across BE or ferrite beads in the base lines?

Regards, Joerg


Yes. At the moment, the raw response goes way out to 10Mhz!!!

from the outset, this is supposed to be more of design tutorial in the
sense of identifying configurations, and there issues etc.

Kevin Aylward

http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.

Bob Cain wrote:
Kevin Aylward wrote:
Bob Cain wrote:

Thanks, Kevin. I don't even play circuit designer on TV so
could I ask you if it is possible to run Full Monty off of
+/- 15V (or even +/- 9V) without compromising the specs
other than signal swing?

Supply voltages are not important at all, other then swings.

I'd hoped that was the case.

For example, the cascode voltage sources would be replaced zeners

e.g. http://www.anasoft.co.uk/EE/MicAmpDesign/TheMicAmp.GIF

However, even this simply circuit wont achieve the full CMRR.

But it's getting closer to something buildable. Thanks.

But don't try yet, as it still needs way more info. I'll try and do some
more, but the first go wont have the best of current sources. It will
take time to design a good one. I am more interested in how its
distortion performs then cmrr.


Again, this is too much of a topology sketch to consider building
yet. I'll see what time I have to try and make the schematic above,
to one
that is actually buildable.

Any efforts will be appreciated. I'm involved in some DIY
groups that will likely jump all over this if you do find
the time and I guarantee you that your name will be
attached. Some of us (not me) have the gear to give
independant validation of the specs and probably would do so.

Oh again, what is the expression for gain involving R5?

The single ended gain is the load resister/2.Re.

Note, that CMRR is determined by only *one* output. You don't have to
take the output differentially.

Note, that the intention is op-amp buffers be used at the collectors
as output drivers. There is no intent that any final outputs should
be taken directly from the collectors.

Got it.


However, if you do take a diff output, the cmrr will improve.

Kevin Aylward

http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.