A better mic preamp
 

On Wed, 14 Mar 2007 14:41:10 -0700, Jim Thompson
wrote:

On Wed, 14 Mar 2007 22:31:16 +0100, Fred Bartoli
fred._canxxxel_this_bartoli@RemoveThatAlso_free. fr_AndThisToo wrote:

Jim Thompson a écrit :
On Wed, 14 Mar 2007 22:19:32 +0100, Fred Bartoli
r_AndThisToo wrote:

Genome a écrit :
"Jim Thompson" wrote in
message ...
I got it backwards... that must be why Bartoli is laughing... swap
NPN-PNP and it's correct. The value is correct though, just the
naming was wrong.

I think NPN diff pairs too much :-(

...Jim Thompson
I still don't think you are spilling the full tin of beans but since all
those circuits like wot you learned them in the old days were upside down
before someone else discovered the NPN transistor it's probably an easy
mistake to make.

That Bartoli bloke sounds a bit Eyetalian to me so he probably has no
problem making the adjustment..... you know, tanks and gears and stuff.

One thing you could do to improve CMRR is to reference your U1 & U2
positive inputs to the Q1/Q2 emitters. (the output conductance of Q1/Q2
is multiplied by the Q3/Q4 stage current gain (eh Graham, it's not Q3/Q4
beta) and will thus be much lower than the common bases Q9/Q10 output
conductance.

Huh? A drinking Eyetalian ?:-)


Maybe you didn't enough? :-)

I guarantee that I drink more than you. Ask Spehro about my famous
"grow hair in your throat" martinis.

Spehro, now improved...

5 parts Sake
2 parts Frangelico
7 parts Vodka
2 parts R/O water ;-)

...Jim Thompson

Left out an important part... shake with crushed ice, strain into
chilled martini glasses, garnish with cucumber slices and/or olives

...Jim Thompson
--
| James E.Thompson, P.E. | mens |
| Analog Innovations, Inc. | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| Phoenix, Arizona Voice:(480)460-2350 | |
| E-mail Address at Website Fax:(480)460-2142 | Brass Rat |
| http://www.analog-innovations.com | 1962 |

I love to cook with wine. Sometimes I even put it in the food.

A better mic preamp - CompoundTransistor.pdf
 

On Thu, 15 Mar 2007 06:47:57 -0700, Jim Thompson
wrote:

On Thu, 15 Mar 2007 09:09:01 +0100, Fred Bartoli
fred._canxxxel_this_bartoli@RemoveThatAlso_free. fr_AndThisToo wrote:

Jim Thompson a écrit :
On Wed, 14 Mar 2007 16:20:56 -0700, Jim Thompson
wrote:

On Wed, 14 Mar 2007 22:47:56 +0100, Fred Bartoli
r_AndThisToo wrote:

[snip]
Aha! Finally speeding up :-)
Yep, max effective current gain is only 38, down to 30-35 with realistic
betas.
Aha! An anomaly!

In simulation, I get an "Alpha" of 0.997 (Iout/Iin), which is
equivalent to a composite Beta of 332. But simulation of Zout looks
like around 3 ohms.

That certainly is not what I expected.

I tried to do an exact solution, but it's nastily transcendental :-(

Time for a glass of wine ;-)

(It would appear to be the shunting input impedance on the NPN stage,
yet replacing the 680 ohm with an ideal current source still has an
error I can't account for.)

Time for another glass of wine... BURP ;-)

...Jim Thompson

Attached.

Another glass of wine will do it every time.

Optimization left to the student ;-)


Yes but:
*** the process you use is very nice to you since it kindly make
identical NPN and PNP betas; (going to the trouble of such a calculation
supposes you pretend to some accuracy(significance of beta/(beta+1)), at
any rate more than any possible beta matching) I tend to prefer simpler
equations which don't obscur the physical meaning:
Zin = reP /(RB.gmN+1)

*** then you can get a deeper by noticing that gm(PNP) is linked to RB,
i.e. to the NPN apparent beta, giving:
Zin = kT/q * RB/VBE /(RB.gmN+1)
or
Zin = kT/(q.VBE) 1/(gmN+1/RB)
which, with reasonable RB values give Zin almost not depending on RB
Zin = kT/(q.VBE.gmN)

With the usual values that is
Zin = 40/gmN (with gm in mA/V)

Since gm is linked to Ic (gm=Ic.q/kT), Zin expresses in a very simple
manner:
Zin = (kT/q)^2/(VBE.Ic)
but (kT/q)^2/VBE is remarkably close to 1mV (1.028mV for VBE=0.65V).

Then
Zin = 1/Ic(NPN) (with Ic in mA)

So you presumed Beta(PNP) is infinite?

Back in a few hours.

...Jim Thompson

If I let Beta be infinite, my equation reduces to the same as yours.

Suppose RB is a current source? (The case in my IC's)

...Jim Thompson
--
| James E.Thompson, P.E. | mens |
| Analog Innovations, Inc. | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| Phoenix, Arizona Voice:(480)460-2350 | |
| E-mail Address at Website Fax:(480)460-2142 | Brass Rat |
| http://www.analog-innovations.com | 1962 |

I love to cook with wine. Sometimes I even put it in the food.

A better mic preamp
 

On Thu, 15 Mar 2007 09:45:49 +0000, Eeyore
wrote:



John Larkin wrote:

Eeyore wrote:

In response to various posts please find attached a mic amp which has seen
volume production in commercial equipment.

Feel free to ask any questions / criticise whatever

Love those 1N4148 Reference Noise Generators.

How much do you think they leak ?They're shunted by the source Z of course.

The Fairchild datasheet has them typically 20 nA at 25 C. So on a warm
day, you could have a few hundred nA total. My trusty old SHOT.EXE
program claims about 0.3 pA/rt hz shot noise current, 44 pA total over
20 KHz, which is probably way under local Johnson noise sources.

Sorry, false alarm. But do keep them in the dark.


The reason they're there is to deal with 'hot plugging' microphones when phantom
power is on. Without them you take out the input devices instead.

What do you reckon the 4R7s are for ?

More noise?

John

A better mic preamp
 

John Larkin wrote:

Eeyore wrote:
John Larkin wrote:
Eeyore wrote:

In response to various posts please find attached a mic amp which has seen
volume production in commercial equipment.

Feel free to ask any questions / criticise whatever

Love those 1N4148 Reference Noise Generators.

How much do you think they leak ?They're shunted by the source Z of course.

The Fairchild datasheet has them typically 20 nA at 25 C. So on a warm
day, you could have a few hundred nA total. My trusty old SHOT.EXE
program claims about 0.3 pA/rt hz shot noise current, 44 pA total over
20 KHz, which is probably way under local Johnson noise sources.

Sorry, false alarm. But do keep them in the dark.

I think I explained their purpose in a reply to the above psot.


The reason they're there is to deal with 'hot plugging' microphones when phantom
power is on. Without them you take out the input devices instead.

What do you reckon the 4R7s are for ?

More noise?

LOL. No not that.

Clue: it turns out that in 0805 size they can sometimes fail open.

Graham

A better mic preamp
 

On Thu, 15 Mar 2007 20:52:16 +0000, Eeyore
Gave us:



John Larkin wrote:

Eeyore wrote:
John Larkin wrote:
Eeyore wrote:

In response to various posts please find attached a mic amp which has seen
volume production in commercial equipment.

Feel free to ask any questions / criticise whatever

Love those 1N4148 Reference Noise Generators.

How much do you think they leak ?They're shunted by the source Z of course.

The Fairchild datasheet has them typically 20 nA at 25 C. So on a warm
day, you could have a few hundred nA total. My trusty old SHOT.EXE
program claims about 0.3 pA/rt hz shot noise current, 44 pA total over
20 KHz, which is probably way under local Johnson noise sources.

Sorry, false alarm. But do keep them in the dark.

I think I explained their purpose in a reply to the above psot.


The reason they're there is to deal with 'hot plugging' microphones when phantom
power is on. Without them you take out the input devices instead.

What do you reckon the 4R7s are for ?

More noise?

LOL. No not that.

Clue: it turns out that in 0805 size they can sometimes fail open.



Nearly all metfilm small form factor resistors "fail open".

They nearly never "fuse shorted".

Small wirewound work? There are very low inductance versions.

A better mic preamp
 

On Thu, 15 Mar 2007 20:52:16 +0000, Eeyore
wrote:



John Larkin wrote:

Eeyore wrote:
John Larkin wrote:
Eeyore wrote:

In response to various posts please find attached a mic amp which has seen
volume production in commercial equipment.

Feel free to ask any questions / criticise whatever

Love those 1N4148 Reference Noise Generators.

How much do you think they leak ?They're shunted by the source Z of course.

The Fairchild datasheet has them typically 20 nA at 25 C. So on a warm
day, you could have a few hundred nA total. My trusty old SHOT.EXE
program claims about 0.3 pA/rt hz shot noise current, 44 pA total over
20 KHz, which is probably way under local Johnson noise sources.

Sorry, false alarm. But do keep them in the dark.

I think I explained their purpose in a reply to the above psot.


The reason they're there is to deal with 'hot plugging' microphones when phantom
power is on. Without them you take out the input devices instead.

What do you reckon the 4R7s are for ?

More noise?

LOL. No not that.

Clue: it turns out that in 0805 size they can sometimes fail open.

Graham

Well, 48 volts through 47 uF is a lot of zot. Do the caps need to be
that big?

And why not use opamps? Transistors are so last millenium.

John

A better mic preamp
 

Jim Thompson a écrit :
On Thu, 15 Mar 2007 09:17:27 +0100, Fred Bartoli
r_AndThisToo wrote:

Eeyore a écrit :
Fred Bartoli wrote:

The NPN stage current gain is at best gm(NPN)*680R = 38

gm*R is a CURRENT gain??


Sure it is.

Demonstration left to the student :-)

Ah, writing it the other way (i.e. R*gm) may help.

Or maybe have another bottle of Chardonay or two to clear your mind ;-)


--
Thanks,
Fred.

A better mic preamp - CompoundTransistor.pdf
 

Jim Thompson a écrit :
On Thu, 15 Mar 2007 06:47:57 -0700, Jim Thompson
wrote:

On Thu, 15 Mar 2007 09:09:01 +0100, Fred Bartoli
r_AndThisToo wrote:

Jim Thompson a écrit :
On Wed, 14 Mar 2007 16:20:56 -0700, Jim Thompson
wrote:

On Wed, 14 Mar 2007 22:47:56 +0100, Fred Bartoli
r_AndThisToo wrote:

[snip]
Aha! Finally speeding up :-)
Yep, max effective current gain is only 38, down to 30-35 with realistic
betas.
Aha! An anomaly!

In simulation, I get an "Alpha" of 0.997 (Iout/Iin), which is
equivalent to a composite Beta of 332. But simulation of Zout looks
like around 3 ohms.

That certainly is not what I expected.

I tried to do an exact solution, but it's nastily transcendental :-(

Time for a glass of wine ;-)

(It would appear to be the shunting input impedance on the NPN stage,
yet replacing the 680 ohm with an ideal current source still has an
error I can't account for.)

Time for another glass of wine... BURP ;-)

...Jim Thompson
Attached.

Another glass of wine will do it every time.

Optimization left to the student ;-)

Yes but:
*** the process you use is very nice to you since it kindly make
identical NPN and PNP betas; (going to the trouble of such a calculation
supposes you pretend to some accuracy(significance of beta/(beta+1)), at
any rate more than any possible beta matching) I tend to prefer simpler
equations which don't obscur the physical meaning:
Zin = reP /(RB.gmN+1)

*** then you can get a deeper by noticing that gm(PNP) is linked to RB,
i.e. to the NPN apparent beta, giving:
Zin = kT/q * RB/VBE /(RB.gmN+1)
or
Zin = kT/(q.VBE) 1/(gmN+1/RB)
which, with reasonable RB values give Zin almost not depending on RB
Zin = kT/(q.VBE.gmN)

With the usual values that is
Zin = 40/gmN (with gm in mA/V)

Since gm is linked to Ic (gm=Ic.q/kT), Zin expresses in a very simple
manner:
Zin = (kT/q)^2/(VBE.Ic)
but (kT/q)^2/VBE is remarkably close to 1mV (1.028mV for VBE=0.65V).

Then
Zin = 1/Ic(NPN) (with Ic in mA)
So you presumed Beta(PNP) is infinite?

Back in a few hours.

...Jim Thompson

If I let Beta be infinite, my equation reduces to the same as yours.


Sure. But even with beta=100 mine is in error of just a few percents.
Largely enough to get a good grasp of what's happening.

Suppose RB is a current source? (The case in my IC's)


Didn't I say 'with reasonable RB values' ?
I agree that's totally different with a CS (gmPNP isn't coupled anymore
to the NPN stage beta) but not everybody do IC design and there are
still people doing discrete designs.


--
Thanks,
Fred.

A better mic preamp
 

MassiveProng wrote:

On Thu, 15 Mar 2007 20:52:16 +0000, Eeyore
Gave us:



John Larkin wrote:

Eeyore wrote:
John Larkin wrote:
Eeyore wrote:

In response to various posts please find attached a mic amp which has seen
volume production in commercial equipment.

Feel free to ask any questions / criticise whatever

Love those 1N4148 Reference Noise Generators.

How much do you think they leak ?They're shunted by the source Z of course.

The Fairchild datasheet has them typically 20 nA at 25 C. So on a warm
day, you could have a few hundred nA total. My trusty old SHOT.EXE
program claims about 0.3 pA/rt hz shot noise current, 44 pA total over
20 KHz, which is probably way under local Johnson noise sources.

Sorry, false alarm. But do keep them in the dark.

I think I explained their purpose in a reply to the above psot.


The reason they're there is to deal with 'hot plugging' microphones when phantom
power is on. Without them you take out the input devices instead.

What do you reckon the 4R7s are for ?

More noise?

LOL. No not that.

Clue: it turns out that in 0805 size they can sometimes fail open.


Nearly all metfilm small form factor resistors "fail open".

But *WHY* would a 4R7 fail ?

Graham

A better mic preamp
 

John Larkin wrote:

Eeyore wrote:
John Larkin wrote:
Eeyore wrote:

What do you reckon the 4R7s are for ?

More noise?

LOL. No not that.

Clue: it turns out that in 0805 size they can sometimes fail open.


Well, 48 volts through 47 uF is a lot of zot.

It is.


Do the caps need to be that big?

Yes. Now why might that be ? Give that some consideration. Intruiged me when I first
saw it too. 2 reasons in fact.


And why not use opamps? Transistors are so last millenium.

Op-amps are (a) far noisier and (b) don't provide differential inputs.

Graham

A better mic preamp
 

On Fri, 16 Mar 2007 08:21:56 +0100, Fred Bartoli
r_AndThisToo wrote:

Jim Thompson a écrit :
On Thu, 15 Mar 2007 09:17:27 +0100, Fred Bartoli
r_AndThisToo wrote:

Eeyore a écrit :
Fred Bartoli wrote:

The NPN stage current gain is at best gm(NPN)*680R = 38

gm*R is a CURRENT gain??


Sure it is.

Demonstration left to the student :-)

Ah, writing it the other way (i.e. R*gm) may help.

Or maybe have another bottle of Chardonay or two to clear your mind ;-)

Or writing out your derivation in full, so that the intermediate steps
can be seen ?:-)

When I allow Beta - Infinite I have your equation, which is the ideal
case... doesn't happen in practice.

...Jim Thompson
--
| James E.Thompson, P.E. | mens |
| Analog Innovations, Inc. | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| Phoenix, Arizona Voice:(480)460-2350 | |
| E-mail Address at Website Fax:(480)460-2142 | Brass Rat |
| http://www.analog-innovations.com | 1962 |

I love to cook with wine. Sometimes I even put it in the food.

A better mic preamp - CompoundTransistor.pdf
 

On Fri, 16 Mar 2007 08:32:23 +0100, Fred Bartoli
r_AndThisToo wrote:

Jim Thompson a écrit :
On Thu, 15 Mar 2007 06:47:57 -0700, Jim Thompson
wrote:

On Thu, 15 Mar 2007 09:09:01 +0100, Fred Bartoli
r_AndThisToo wrote:

Jim Thompson a écrit :
On Wed, 14 Mar 2007 16:20:56 -0700, Jim Thompson
wrote:

On Wed, 14 Mar 2007 22:47:56 +0100, Fred Bartoli
r_AndThisToo wrote:

[snip]
Aha! Finally speeding up :-)
Yep, max effective current gain is only 38, down to 30-35 with realistic
betas.
Aha! An anomaly!

In simulation, I get an "Alpha" of 0.997 (Iout/Iin), which is
equivalent to a composite Beta of 332. But simulation of Zout looks
like around 3 ohms.

That certainly is not what I expected.

I tried to do an exact solution, but it's nastily transcendental :-(

Time for a glass of wine ;-)

(It would appear to be the shunting input impedance on the NPN stage,
yet replacing the 680 ohm with an ideal current source still has an
error I can't account for.)

Time for another glass of wine... BURP ;-)

...Jim Thompson
Attached.

Another glass of wine will do it every time.

Optimization left to the student ;-)

Yes but:
*** the process you use is very nice to you since it kindly make
identical NPN and PNP betas; (going to the trouble of such a calculation
supposes you pretend to some accuracy(significance of beta/(beta+1)), at
any rate more than any possible beta matching) I tend to prefer simpler
equations which don't obscur the physical meaning:
Zin = reP /(RB.gmN+1)

*** then you can get a deeper by noticing that gm(PNP) is linked to RB,
i.e. to the NPN apparent beta, giving:
Zin = kT/q * RB/VBE /(RB.gmN+1)
or
Zin = kT/(q.VBE) 1/(gmN+1/RB)
which, with reasonable RB values give Zin almost not depending on RB
Zin = kT/(q.VBE.gmN)

With the usual values that is
Zin = 40/gmN (with gm in mA/V)

Since gm is linked to Ic (gm=Ic.q/kT), Zin expresses in a very simple
manner:
Zin = (kT/q)^2/(VBE.Ic)
but (kT/q)^2/VBE is remarkably close to 1mV (1.028mV for VBE=0.65V).

Then
Zin = 1/Ic(NPN) (with Ic in mA)
So you presumed Beta(PNP) is infinite?

Back in a few hours.

...Jim Thompson

If I let Beta be infinite, my equation reduces to the same as yours.


Sure. But even with beta=100 mine is in error of just a few percents.
Largely enough to get a good grasp of what's happening.

Suppose RB is a current source? (The case in my IC's)


Didn't I say 'with reasonable RB values' ?
I agree that's totally different with a CS (gmPNP isn't coupled anymore
to the NPN stage beta) but not everybody do IC design and there are
still people doing discrete designs.

Really. Still people doing discrete designs ?:-) Wow!

Seriously, I know.

I do this mostly in CMOS with mirror bias top and bottom to get nice
stiff hi-Z analog buffers.

...Jim Thompson
--
| James E.Thompson, P.E. | mens |
| Analog Innovations, Inc. | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| Phoenix, Arizona Voice:(480)460-2350 | |
| E-mail Address at Website Fax:(480)460-2142 | Brass Rat |
| http://www.analog-innovations.com | 1962 |

I love to cook with wine. Sometimes I even put it in the food.

A better mic preamp
 

Jim Thompson a écrit :
On Fri, 16 Mar 2007 08:21:56 +0100, Fred Bartoli
r_AndThisToo wrote:

Jim Thompson a écrit :
On Thu, 15 Mar 2007 09:17:27 +0100, Fred Bartoli
r_AndThisToo wrote:

Eeyore a écrit :
Fred Bartoli wrote:

The NPN stage current gain is at best gm(NPN)*680R = 38
gm*R is a CURRENT gain??

Sure it is.

Demonstration left to the student :-)

Ah, writing it the other way (i.e. R*gm) may help.

Or maybe have another bottle of Chardonay or two to clear your mind ;-)

Or writing out your derivation in full, so that the intermediate steps
can be seen ?:-)


Which derivation?
That's so obvious.

Or maybe I didn't understand your
gm*R is a CURRENT gain??

When I allow Beta - Infinite I have your equation, which is the ideal
case... doesn't happen in practice.


But is often good enough for back of the envelop calculations.


--
Thanks,
Fred.

A better mic preamp
 

On Fri, 16 Mar 2007 09:47:02 -0800, John Larkin
wrote:

On Fri, 16 Mar 2007 07:55:05 +0000, Eeyore
wrote:



John Larkin wrote:

Eeyore wrote:
John Larkin wrote:
Eeyore wrote:

What do you reckon the 4R7s are for ?

More noise?

LOL. No not that.

Clue: it turns out that in 0805 size they can sometimes fail open.


Well, 48 volts through 47 uF is a lot of zot.

It is.


Do the caps need to be that big?

Yes. Now why might that be ? Give that some consideration. Intruiged me when I first
saw it too. 2 reasons in fact.


Can't imagine; tell me please.


And why not use opamps? Transistors are so last millenium.

Op-amps are (a) far noisier and (b) don't provide differential inputs.

Graham

Hmmm. I buy (a) 0.8 nv/rthz opamps that (b) have two inputs.

John

You would need instrumentation style to do differential to
single-ended.

Bang-for-the-buck, the composite PNP/NPN has a nice noise floor...
~1nV/rt-Hz!

...Jim Thompson
--
| James E.Thompson, P.E. | mens |
| Analog Innovations, Inc. | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| Phoenix, Arizona Voice:(480)460-2350 | |
| E-mail Address at Website Fax:(480)460-2142 | Brass Rat |
| http://www.analog-innovations.com | 1962 |

I love to cook with wine. Sometimes I even put it in the food.

A better mic preamp
 

On Fri, 16 Mar 2007 07:55:05 +0000, Eeyore
wrote:



John Larkin wrote:

Eeyore wrote:
John Larkin wrote:
Eeyore wrote:

What do you reckon the 4R7s are for ?

More noise?

LOL. No not that.

Clue: it turns out that in 0805 size they can sometimes fail open.


Well, 48 volts through 47 uF is a lot of zot.

It is.


Do the caps need to be that big?

Yes. Now why might that be ? Give that some consideration. Intruiged me when I first
saw it too. 2 reasons in fact.


Can't imagine; tell me please.


And why not use opamps? Transistors are so last millenium.

Op-amps are (a) far noisier and (b) don't provide differential inputs.

Graham

Hmmm. I buy (a) 0.8 nv/rthz opamps that (b) have two inputs.

John

A better mic preamp
 

On Fri, 16 Mar 2007 07:55:05 +0000, Eeyore
wrote:



John Larkin wrote:

Eeyore wrote:
John Larkin wrote:
Eeyore wrote:

What do you reckon the 4R7s are for ?

More noise?

LOL. No not that.

Clue: it turns out that in 0805 size they can sometimes fail open.


Well, 48 volts through 47 uF is a lot of zot.

It is.


Do the caps need to be that big?

Yes. Now why might that be ? Give that some consideration. Intruiged me when I first
saw it too. 2 reasons in fact.


And why not use opamps? Transistors are so last millenium.

Op-amps are (a) far noisier and (b) don't provide differential inputs.

---
Huh???

http://www.linear.com/pc/downloadDoc...,P1293,D161 2

Figure 1.


--
JF

A better mic preamp
 

John Larkin wrote:

Eeyore wrote:
John Larkin wrote:

And why not use opamps? Transistors are so last millenium.

Op-amps are (a) far noisier and (b) don't provide differential inputs.

Hmmm. I buy (a) 0.8 nv/rthz opamps

Insanely priced op-amps. Those 2SA1084s are under 10c ea IIRC.


that (b) have two inputs.

That doesn't make a true differential amplifier does it ?

Graham

A better mic preamp
 

John Larkin wrote:

Eeyore wrote:
John Larkin wrote:

Do the [input] caps need to be that big?

Yes. Now why might that be ? Give that some consideration. Intruiged me when I first
saw it too. 2 reasons in fact.

Can't imagine; tell me please.

If the value isn't that large, the amplifier won't 'see' the 200 ohms or so source
impedance at LF and flicker noise will rise.

As it is even with 47uF, Zc is 169 ohms @ 20Hz.

Secondly, if an electrolytic cap starts forming a filter with any signal level much over
100mV you'll get distortion.

Graham

A better mic preamp
 

"John Fields"
Eeysore


Op-amps are (a) far noisier and (b) don't provide differential inputs.

Huh???

http://www.linear.com/pc/downloadDoc...,P1293,D161 2

Figure 1.


** LOL

Very noisy due to those 1 kohms in series with the mic.

Gain is non adjustable.

Total waste of a nice op-amp.



........ Phil

A better mic preamp
 

On Fri, 16 Mar 2007 07:52:10 +0000, Eeyore
Gave us:



MassiveProng wrote:

On Thu, 15 Mar 2007 20:52:16 +0000, Eeyore
Gave us:



John Larkin wrote:

Eeyore wrote:
John Larkin wrote:
Eeyore wrote:

In response to various posts please find attached a mic amp which has seen
volume production in commercial equipment.

Feel free to ask any questions / criticise whatever

Love those 1N4148 Reference Noise Generators.

How much do you think they leak ?They're shunted by the source Z of course.

The Fairchild datasheet has them typically 20 nA at 25 C. So on a warm
day, you could have a few hundred nA total. My trusty old SHOT.EXE
program claims about 0.3 pA/rt hz shot noise current, 44 pA total over
20 KHz, which is probably way under local Johnson noise sources.

Sorry, false alarm. But do keep them in the dark.

I think I explained their purpose in a reply to the above psot.


The reason they're there is to deal with 'hot plugging' microphones when phantom
power is on. Without them you take out the input devices instead.

What do you reckon the 4R7s are for ?

More noise?

LOL. No not that.

Clue: it turns out that in 0805 size they can sometimes fail open.


Nearly all metfilm small form factor resistors "fail open".

But *WHY* would a 4R7 fail ?


Any number of reasons. Poor affixing can cause the termination(s)
to fail, an ESD event before assembly, A current surge that far
exceeds its power capacity.

The film is very thin, you know. And we all know that the
termination is the weakest point on an SMD resistor. Have you tried a
thick film version?

A better mic preamp
 

On Sat, 17 Mar 2007 10:08:32 +1100, "Phil Allison"
wrote:


"John Fields"
Eeysore


Op-amps are (a) far noisier and (b) don't provide differential inputs.

Huh???

http://www.linear.com/pc/downloadDoc...,P1293,D161 2

Figure 1.


** LOL

Very noisy due to those 1 kohms in series with the mic.

---
Doesn't have to be.

Their resistances can be lowered depending on what comes out of the
mic, and the ultimate gain can be adjusted by changing R3 and R4.
---

Gain is non adjustable.

---
Doesn't have to be. Make Rp a rheostat.
---

Total waste of a nice op-amp.

---
I disagree.

It seems to me that if that active discrete front end can be done
away with, so much the better.


--
JF

A better mic preamp
 

"John Fields"
"Phil Allison"
"John Fields"
Eeysore


Op-amps are (a) far noisier and (b) don't provide differential inputs.

Huh???

http://www.linear.com/pc/downloadDoc...,P1293,D161 2

Figure 1.


** LOL

Very noisy due to those 1 kohms in series with the mic.

---
Doesn't have to be.


** Fraid it does.

That topology is just not capable of good results.



Their resistances can be lowered depending on what comes out of the
mic,


** Goobbledegook.

Load the mic = loose output voltage = worsen the s/n.

That topology is always noisy.



Gain is non adjustable.


Doesn't have to be. Make Rp a rheostat.


** How utterly asinine.

That will only increase the noise gain.


Total waste of a nice op-amp.

I disagree.


** Makes you a colossal fool.


It seems to me that if that active discrete front end can be done
away with, so much the better.


** If pigs could fly ....



........ Phil

A better mic preamp
 

MassiveProng wrote:

Eeyore Gave us:
MassiveProng wrote:
Eeyore Gave us:
John Larkin wrote:
Eeyore wrote:

What do you reckon the 4R7s are for ?

More noise?

LOL. No not that.

Clue: it turns out that in 0805 size they can sometimes fail open.


Nearly all metfilm small form factor resistors "fail open".

But *WHY* would a 4R7 fail ?

Any number of reasons. Poor affixing can cause the termination(s)
to fail, an ESD event before assembly, A current surge that far
exceeds its power capacity.

I meant in operation. I know what causes it to fail. I was actually asking you.


The film is very thin, you know. And we all know that the
termination is the weakest point on an SMD resistor. Have you tried a
thick film version?

Actually, it was a mistake to use SMD for that part. I'm going to revert to a more
physically substantial leaded part.

Graham

A better mic preamp
 

Phil Allison wrote:

"John Fields"
"Phil Allison"
"John Fields"
Eeysore

Op-amps are (a) far noisier and (b) don't provide differential inputs.

Huh???
http://www.linear.com/pc/downloadDoc...,P1293,D161 2

Figure 1.

** LOL

Very noisy due to those 1 kohms in series with the mic.

---
Doesn't have to be.

** Fraid it does.

That topology is just not capable of good results.

Their resistances can be lowered depending on what comes out of the
mic,

** Goobbledegook.

Load the mic = loose output voltage = worsen the s/n.

Yup. Mics should ideally (in most cases) be loaded with around 10 times their source impedance.


That topology is always noisy.

Gain is non adjustable.

Doesn't have to be. Make Rp a rheostat.

** How utterly asinine.

That will only increase the noise gain.

Total waste of a nice op-amp.

I disagree.

** Makes you a colossal fool.

Fields specialises in playing the fool.


It seems to me that if that active discrete front end can be done
away with, so much the better.

** If pigs could fly ....

The IC mic amps that are around are internally quite like their discrete counterparts but just not quite as good and vastly more expensive.

They're a boon for 'designers' who can't actually really design though.

Graham

A better mic preamp
 

Eeyore wrote...
Winfield Hill wrote:

It's nice to see the diff-pair currents going directly to
the output opamp, with it's 6.8k resistors. But why'd ya
steal away loop gain so aggressively by choosing 680-ohms?

It doesn't. The + and - op-amp input nodes see almost no
differential ac signal voltage due to feedback action.

Hence no-need for active loading either. The 680Rs simply
set the DC condition.

Win wasn't speaking about this. Well, you've splendidly
succeeded in reducing the loop gain by a 1:11 factor.

Come on then. Let's see your explanation.

And you need more explanation than what Win and I said for
such an elementary matter? ROFL!

Ok, if we really have to spell it.
What's the opamp feedback loop made with?

I'm not playing any silly guessing game with you.
Explain yourself or shut up.

Graham

Of course the 680-ohm guys don't affect the gain equation.

But come on Graham, it's elementary. The output opamp has
to work harder with a fixed gain of 10 you've created with
the 680-ohm resistors. This creates distortion, because
only 1/10 of the already-poor TL072 loop gain is left for
the preamp to use, as the 3MHz TL072 bandwidth is reduced
to 300kHz. Sheesh! This also creates a high output noise
(10x 18nV = 180nV, which is referred back to the input as
180nV/gain, which is sizable at low gains). Need we go on?

If the 680R value has no effect, imagine making it 10 ohms?

Note, you can substantially increase this resistor without
affecting the preamp, using up a little common-mode range.
Nevertheless, the collector resistors are the bane of this
design approach.


--
Thanks,
- Win

A better mic preamp
 

Winfield Hill wrote:

Eeyore wrote...
Winfield Hill wrote:

It's nice to see the diff-pair currents going directly to
the output opamp, with it's 6.8k resistors. But why'd ya
steal away loop gain so aggressively by choosing 680-ohms?

It doesn't. The + and - op-amp input nodes see almost no
differential ac signal voltage due to feedback action.

Hence no-need for active loading either. The 680Rs simply
set the DC condition.

Win wasn't speaking about this. Well, you've splendidly
succeeded in reducing the loop gain by a 1:11 factor.

Come on then. Let's see your explanation.

And you need more explanation than what Win and I said for
such an elementary matter? ROFL!

Ok, if we really have to spell it.
What's the opamp feedback loop made with?

I'm not playing any silly guessing game with you.
Explain yourself or shut up.

Graham

Of course the 680-ohm guys don't affect the gain equation.

But come on Graham, it's elementary. The output opamp has
to work harder with a fixed gain of 10 you've created with
the 680-ohm resistors.

Yes OK, you've 'thrown away' ~ 20dB of loop gain and it detracts from the noise
figure at low gains since your second stage is now contributing ~ 180nV/sqrt Hz
of noise to the final output.

The fact is that with a TL07x op-amp, that 'thrown-away 20dB doesn't impact THD
levels one bit since it's a pretty linear op-amp in the first place.

Secondly, the impact at lower gains is not seriously at variance with the
inherent noise level of following stages.

So yes, in isolation it's possible to make some serious criticisms of that stage
but in context it actually performs quite well.


This creates distortion, because only 1/10 of the already-poor TL072 loop gain

Not that poor actually. It's actually quite linear to begin with you see.


is left for the preamp to use, as the 3MHz TL072 bandwidth is reduced
to 300kHz. Sheesh!

I know what you're saying but it works out in practice rather better than the
numbers alone might suggest.


This also creates a high output noise
(10x 18nV = 180nV, which is referred back to the input as
180nV/gain, which is sizable at low gains). Need we go on?

If the 680R value has no effect, imagine making it 10 ohms?

Addressed above. With hindsight I realised they could have their value increased
somewhat to decent effect.


Note, you can substantially increase this resistor without
affecting the preamp, using up a little common-mode range.
Nevertheless, the collector resistors are the bane of this
design approach.

See my later "improved mic amp".

I should perhaps point out that the "better mic amp" was one that went into my
'cheapest' products. The higher end ones get the "improved mic amp". Also that
the history of this topology and its variants goes back over 30 years in which
time various derivatives have been developed. There's others I haven't yet
posted here.

Graham

A better mic preamp
 

On Sat, 17 Mar 2007 15:15:20 +1100, "Phil Allison"
wrote:


** If pigs could fly ....

---
They can.

If you want to learn how, go to:

http://www.airaustralia.net/c-profile.htm


--
JF

A better mic preamp
 

On Sat, 17 Mar 2007 09:42:57 +0000, Eeyore
wrote:



MassiveProng wrote:

Eeyore Gave us:
MassiveProng wrote:
Eeyore Gave us:
John Larkin wrote:
Eeyore wrote:

What do you reckon the 4R7s are for ?

More noise?

LOL. No not that.

Clue: it turns out that in 0805 size they can sometimes fail open.


Nearly all metfilm small form factor resistors "fail open".

But *WHY* would a 4R7 fail ?

Any number of reasons. Poor affixing can cause the termination(s)
to fail, an ESD event before assembly, A current surge that far
exceeds its power capacity.

I meant in operation. I know what causes it to fail. I was actually asking you.


The film is very thin, you know. And we all know that the
termination is the weakest point on an SMD resistor. Have you tried a
thick film version?

Actually, it was a mistake to use SMD for that part. I'm going to revert to a more
physically substantial leaded part.

---
Good idea. Something like this:

http://www.ohmite.com/cgi-bin/showpa...duct=89_series

should keep you out of trouble.


--
JF

A better mic preamp
 

On Sat, 17 Mar 2007 09:48:03 +0000, Eeyore
wrote:

....

Fields specialises in playing the fool.

---
Yes, while you work hard being serious about it.


--
JF

A better mic preamp
 

On Sat, 17 Mar 2007 12:42:15 +0000, Eeyore
wrote:



Winfield Hill wrote:

Eeyore wrote...
Winfield Hill wrote:

It's nice to see the diff-pair currents going directly to
the output opamp, with it's 6.8k resistors. But why'd ya
steal away loop gain so aggressively by choosing 680-ohms?

It doesn't. The + and - op-amp input nodes see almost no
differential ac signal voltage due to feedback action.

Hence no-need for active loading either. The 680Rs simply
set the DC condition.

Win wasn't speaking about this. Well, you've splendidly
succeeded in reducing the loop gain by a 1:11 factor.

Come on then. Let's see your explanation.

And you need more explanation than what Win and I said for
such an elementary matter? ROFL!

Ok, if we really have to spell it.
What's the opamp feedback loop made with?

I'm not playing any silly guessing game with you.
Explain yourself or shut up.

Graham

Of course the 680-ohm guys don't affect the gain equation.

But come on Graham, it's elementary. The output opamp has
to work harder with a fixed gain of 10 you've created with
the 680-ohm resistors.

Yes OK, you've 'thrown away' ~ 20dB of loop gain and it detracts from the noise
figure at low gains since your second stage is now contributing ~ 180nV/sqrt Hz
of noise to the final output.

The fact is that with a TL07x op-amp, that 'thrown-away 20dB doesn't impact THD
levels one bit since it's a pretty linear op-amp in the first place.

Secondly, the impact at lower gains is not seriously at variance with the
inherent noise level of following stages.

So yes, in isolation it's possible to make some serious criticisms of that stage
but in context it actually performs quite well.

---
Do you often have waffles for breakfast?
---

This creates distortion, because only 1/10 of the already-poor TL072 loop gain

Not that poor actually. It's actually quite linear to begin with you see.


is left for the preamp to use, as the 3MHz TL072 bandwidth is reduced
to 300kHz. Sheesh!

I know what you're saying but it works out in practice rather better than the
numbers alone might suggest.


This also creates a high output noise
(10x 18nV = 180nV, which is referred back to the input as
180nV/gain, which is sizable at low gains). Need we go on?

If the 680R value has no effect, imagine making it 10 ohms?

Addressed above. With hindsight I realised they could have their value increased
somewhat to decent effect.


Note, you can substantially increase this resistor without
affecting the preamp, using up a little common-mode range.
Nevertheless, the collector resistors are the bane of this
design approach.

See my later "improved mic amp".

I should perhaps point out that the "better mic amp" was one that went into my
'cheapest' products. The higher end ones get the "improved mic amp". Also that
the history of this topology and its variants goes back over 30 years in which
time various derivatives have been developed. There's others I haven't yet
posted here.

---
Blah, blah, blah, blah, ****ing blah.

LOL, and I bet you wired up all of your crap with special low THD
wire, huh?


--
JF

A better mic preamp
 

John Fields wrote:

LOL, and I bet you wired up all of your crap with special low THD
wire, huh?

Don't be ridiculous.

A better mic preamp
 

The IC mic amps that are around are internally quite like their discrete
counterparts but just not quite as good and vastly more expensive.

They're a boon for 'designers' who can't actually really design though.

Graham


Guys like me.....

lol

Actually Graham that does bring up an interesting question though. The
SSM2019 is only a couple bucks apiece in for 100 or more. I'm sure you must
have close to that in parts in your design. There seems to be a preference
for discrete designs - what makes them better than something like the 2019?
Or are they actually better?

Thanks

A better mic preamp
 

On Sat, 17 Mar 2007 14:32:01 +0000, Eeyore
wrote:



John Fields wrote:

LOL, and I bet you wired up all of your crap with special low THD
wire, huh?

Don't be ridiculous.

---
You mean you _didn't_???


--
JF

A better mic preamp
 

"Winfield Hill" wrote in message
...

But come on Graham, it's elementary. The output opamp has
to work harder with a fixed gain of 10 you've created with
the 680-ohm resistors. This creates distortion, because
only 1/10 of the already-poor TL072 loop gain is left for
the preamp to use, as the 3MHz TL072 bandwidth is reduced
to 300kHz. Sheesh! This also creates a high output noise
(10x 18nV = 180nV, which is referred back to the input as
180nV/gain, which is sizable at low gains). Need we go on?

If the 680R value has no effect, imagine making it 10 ohms?

Note, you can substantially increase this resistor without
affecting the preamp, using up a little common-mode range.
Nevertheless, the collector resistors are the bane of this
design approach.


--
Thanks,
- Win

Errr, Nope...... As a neophyte I found this one when LTspicing SMPS stuff
and therefore I am a believer. There are probably equations that deal with
this stuff so don't excuse me while I get it wrong.

Take your inverting amplifier with it's feedback from the output of your
thing and then you have optimised its gain to do the business. You think
that that part of the sum works out to be two so you divide that into the
GBW of the op-amp and think you are sitting pretty at 500KHz.

Unfortunately you had to stuff another resistor up its bum from the
reference to set the output in the first place. So what you should have done
is work out the gain based on that resistor and the one from the output in
parallel with it and the sum works out to be fifty.......

Which drops the steam in your op-amp down in the place where you thought you
were in pretty the first place and you are in trouble.

It is left as an excercise for the lecturer(s) to explain.

DNA

Have you tried out my IGBT thing yet?

A better mic preamp
 

John Fields wrote:
On Sat, 17 Mar 2007 09:42:57 +0000, Eeyore
wrote:



MassiveProng wrote:


Eeyore Gave us:

MassiveProng wrote:

Eeyore Gave us:

John Larkin wrote:

Eeyore wrote:


What do you reckon the 4R7s are for ?

More noise?

LOL. No not that.

Clue: it turns out that in 0805 size they can sometimes fail open.


Nearly all metfilm small form factor resistors "fail open".

But *WHY* would a 4R7 fail ?

Any number of reasons. Poor affixing can cause the termination(s)
to fail, an ESD event before assembly, A current surge that far
exceeds its power capacity.

I meant in operation. I know what causes it to fail. I was actually asking you.



The film is very thin, you know. And we all know that the
termination is the weakest point on an SMD resistor. Have you tried a
thick film version?

Actually, it was a mistake to use SMD for that part. I'm going to revert to a more
physically substantial leaded part.


---
Good idea. Something like this:

http://www.ohmite.com/cgi-bin/showpa...duct=89_series

should keep you out of trouble.


:-)

A better mic preamp
 

John Larkin wrote:

Eeyore wrote:
John Fields wrote:

LOL, and I bet you wired up all of your crap with special low THD
wire, huh?

Don't be ridiculous.

I did once use some brass-plated steel parts I thought where solid
brass. They caused a huge and confusing harmonic distortion at around
200 amps.

Do elaborate.

I've never run a load over about 50 amps.

Graham

A better mic preamp
 

On Sat, 17 Mar 2007 14:32:01 +0000, Eeyore
wrote:



John Fields wrote:

LOL, and I bet you wired up all of your crap with special low THD
wire, huh?

Don't be ridiculous.

I did once use some brass-plated steel parts I thought where solid
brass. They caused a huge and confusing harmonic distortion at around
200 amps.

John

A better mic preamp
 

"tempus fugit"

The IC mic amps that are around are internally quite like their discrete
counterparts but just not quite as good and vastly more expensive.

They're a boon for 'designers' who can't actually really design though.


Actually Graham that does bring up an interesting question though. The
SSM2019 is only a couple bucks apiece in for 100 or more.


** Only true for the SMD version.


I'm sure you must
have close to that in parts in your design. There seems to be a preference
for discrete designs - what makes them better than something like the
2019?
Or are they actually better?


** The SSM2019 is relatively new ( released in 2003 ) - it replaces the
SSM2017 that simply disappeared off the market about 8 years ago. Other SSM
ICs were similarly discontinued ( ie SSM2142 line drive) leaving makers,
gear owners and service people completely in the lurch.

That is a the problem with any unique IC that is not multi-sourced. Pro
audio gear has a very long life cycle - mixing desks and power amps are in
use up to 20 years after sale - so it is foolish to use any single sourced
components that have no substitute.

Plenty of examples exist of big name manufactures making this dumb mistake
and paying for it.

EG:

US amp maker " BGW" produced a series of amps using NEC output devices in
unique FOUR leg, 2 bolt flat packs.

The devices proved to have a serious inherent flaw causing early failures,
NEC dropped the part with no replacement and so all those BGW amps are now
scrap.




....... Phil

A better mic preamp
 

"tempus fugit" ...

Phil Allison


** The SSM2019 is relatively new ( released in 2003 ) - it replaces the
SSM2017 that simply disappeared off the market about 8 years ago. Other SSM
ICs were similarly discontinued ( ie SSM2142 line drive) leaving makers,
gear owners and service people completely in the lurch.

That is a the problem with any unique IC that is not multi-sourced. Pro
audio gear has a very long life cycle - mixing desks and power amps are in
use up to 20 years after sale - so it is foolish to use any single sourced
components that have no substitute.

Plenty of examples exist of big name manufactures making this dumb mistake
and paying for it.

EG:

US amp maker " BGW" produced a series of amps using NEC output devices in
unique FOUR leg, 2 bolt flat packs.

The devices proved to have a serious inherent flaw causing early failures,
NEC dropped the part with no replacement and so all those BGW amps are now
scrap.


....... Phil



Thanks for that Phil. I didn't realize that they had stopped making the
2017
(that's what I'm using) I just thought the 2019 was kind of an upgrade.
What
would you say about the sound/performance though? Did they manage to get
as
good (or close enough to it) a design and sound into the IC as could be
done
with a discrete design like yours or Graham's? Might the tendency to stop
producing ICs like this be a factor in using discretes, or is it more the
sound? I'm sure you've heard that argument - "Oh, I hate the sound of
opamps
so I want something discrete".



** Read my post again ( I have re-posted it clean to make that easy. )

Cos you missed a whole lot the first time.





....... Phil

A better mic preamp
 

"Phil Allison" wrote in message
...

"tempus fugit"

The IC mic amps that are around are internally quite like their
discrete
counterparts but just not quite as good and vastly more expensive.

They're a boon for 'designers' who can't actually really design though.


Actually Graham that does bring up an interesting question though. The
SSM2019 is only a couple bucks apiece in for 100 or more.


** Only true for the SMD version.


I'm sure you must
have close to that in parts in your design. There seems to be a
preference
for discrete designs - what makes them better than something like the
2019?
Or are they actually better?


** The SSM2019 is relatively new ( released in 2003 ) - it replaces the
SSM2017 that simply disappeared off the market about 8 years ago. Other
SSM
ICs were similarly discontinued ( ie SSM2142 line drive) leaving makers,
gear owners and service people completely in the lurch.

That is a the problem with any unique IC that is not multi-sourced. Pro
audio gear has a very long life cycle - mixing desks and power amps are in
use up to 20 years after sale - so it is foolish to use any single
sourced
components that have no substitute.

Plenty of examples exist of big name manufactures making this dumb mistake
and paying for it.

EG:

US amp maker " BGW" produced a series of amps using NEC output devices in
unique FOUR leg, 2 bolt flat packs.

The devices proved to have a serious inherent flaw causing early failures,
NEC dropped the part with no replacement and so all those BGW amps are now
scrap.




...... Phil

Thanks for that Phil. I didn't realize that they had stopped making the 2017
(that's what I'm using) I just thought the 2019 was kind of an upgrade. What
would you say about the sound/performance though? Did they manage to get as
good (or close enough to it) a design and sound into the IC as could be done
with a discrete design like yours or Graham's? Might the tendency to stop
producing ICs like this be a factor in using discretes, or is it more the
sound? I'm sure you've heard that argument - "Oh, I hate the sound of opamps
so I want something discrete".