Operational Amplifier Design

[Jim Thompson]

Jim Thompson wrote...

[snip]

The "simplified schematic" is flawed... it can't be as-drawn... there
is a fundamental error. I tried to tease Win into _stepping_in_it_,
but he won't take the bait ;-)

Some of us have work to do!

Me too, but I hang out here when long simulations are running.

I assume you're referring to Q7's collector-
base diode in parallel with Q13's base-emitter (NSC diagram, Q12 and Q11
for Motorola), during pulldown. We don't know the relative I_s values...

But we do. From that era we know that the B-C junction of the NPN is
the same doping (same process step) as the E-B junction of the PNP.
And the current limit resistor is probably 35 ohms. So, as-drawn, we
would have serious (effective) beta reduction when the B-C of the NPN
forward biased. My guess is that there is at least one diode (and
probably two) between the collector of Q7 and the base of Q13

Do you have access to a proper full schematic?

No I don't. The LM324 is one of the few devices that I didn't trace
looking for patent infringement issues ;-)

He's referring to the compensation cap connected to the base of a pullup
output transistor, rather than to the output, forcing a slew time delay
when traversing from pullup to pulldown, and back.

OK, Now I understand what he meant. The slew time thru the dead-band.

No doubt the chosen
cap connection is more stable,

Dramatically so, I've been down that kind of design path.

and perhaps protects the cap from static
discharge damage too.

Also true. Capacitor plates on pins are to be avoided... they tend to
respond to ESD with pin-hole shorts.

But it exacerbates the crossover distortion.

Yes. But I don't know of a single commercial amplifier that isn't set
up that way.

...Jim Thompson

 

[Winfield Hill]

Jim Thompson wrote...

But I don't know of a single commercial amplifier that isn't set
up that way.

Doesn't adding a modest series resistor (along with an ESD network
after the resistor) give one a chance to both stabilize the loop,
and avoid ESD failure?

 

[Rich Grise]

While I never dissected an LM324, I suspect that some versions derived
the 50uA from mirroring via a replica of Q12 (National data sheet
reference designators).

Might be an interesting exercise to trace one out under a microscope:
If you look at the published schematic, the output stage cannot
function properly, it has a major problem... can you tell us what it
is?

The short across Q5, Q6, Q7/Rsc, and Q13?

Thanks,
Rich

 

[Rich Grise]

While I never dissected an LM324, I suspect that some versions derived
the 50uA from mirroring via a replica of Q12 (National data sheet
reference designators).

Might be an interesting exercise to trace one out under a microscope:
If you look at the published schematic, the output stage cannot
function properly, it has a major problem... can you tell us what it
is?

I posted asking, The short across Q5,6,7,13, and Rsc? with no
link to the schem.
http://perso.wanadoo.es/chyryes/componentes/LM324.pdfThanks,
Rich

 

[Winfield Hill]

Jim Thompson wrote...

But we do. From that era we know that the B-C junction of the NPN is
the same doping (same process step) as the E-B junction of the PNP.
And the current limit resistor is probably 35 ohms. So, as-drawn, we
would have serious (effective) beta reduction when the B-C of the NPN
forward biased. My guess is that there is at least one diode (and
probably two) between the collector of Q7 and the base of Q13

I always put a diode in the current-limit transistor's collector (as
is standard in the power-amp crowd). If we assume that Q7's area is
dramatically smaller than Q13's area, does that make a difference?

What are you thinking of when you say more than one diode? Do you
think there's a class AB diode or two in there afterall?

 

[Jim Thompson]

I posted asking, The short across Q5,6,7,13, and Rsc? with no
link to the schem.
http://perso.wanadoo.es/chyryes/componentes/LM324.pdf
Thanks,
Rich

What "short"?

...Jim Thompson

 

[Jim Thompson]

Jim Thompson wrote...

Doesn't adding a modest series resistor (along with an ESD network
after the resistor) give one a chance to both stabilize the loop,
and avoid ESD failure?

Actually, most compensation caps _do_ have a series resistor (to
introduce a zero).

...Jim Thompson

 

[Jim Thompson]

Jim Thompson wrote...

I always put a diode in the current-limit transistor's collector (as
is standard in the power-amp crowd). If we assume that Q7's area is
dramatically smaller than Q13's area, does that make a difference?

What are you thinking of when you say more than one diode? Do you
think there's a class AB diode or two in there afterall?

Judging from the dramatic performances differences between National
and Motorola parts, I'd say so.

Are you aware of the infamous "let's copy a 741 episode", where the
Fairchild published schematic was copied into silicon by (IIRC)
Raytheon, except that 'tain't the circuit ;-)

...Jim Thompson

 

[Winfield Hill]

Jim Thompson wrote...

Are you aware of the infamous "let's copy a 741 episode", where the
Fairchild published schematic was copied into silicon by (IIRC)
Raytheon, except that 'tain't the circuit ;-)

Do you know what the correct circuit is?

 

[Rich The Philosophizer]

What "short"?

In the schematic on the top of page 6,
The black line from the collector of Q12 and base of Q13, straight
up to the collector of Q7, up to the base of Q5 and the 100 uA C.S.
The base of Q5 is shorted to the base of Q13, so what does anything
in between them do?

Obviously, I'm missing something, but what?

Thanks,
Rich

 

[Tony Williams]

The biggest problem was brand variability... Motorola's copy had
such severe cross-over distortion as to make the part unusable
for active filters... which is why I put the Motorola LM324 on
the GenRad (OmniComp) do-not-purchase list.

Plus a tendency for a little burst of hf oscillation
from the output transistors at swapover. It was best
with the LM324 to think of it as delivering positive
output current only, with the negative current being
supplied by an external pulldown.

 

[Kevin Aylward]

Actually, most compensation caps _do_ have a series resistor (to
introduce a zero).

Ahmmm. I disagree. If they do, its often a poor design. This particular
resister usually fails to do what it is idealised to do. It can only
work if the stage is not running out of steam at its zero frequency.
This is not often the case when you want to use such a trick. The cap
does a pole split, i.e. moves other poles further away, e.g low output
impedance driving the load. As soon as the resister starts coming in,
the poles start to migrate inwards again getting you back to where you
were without the Miller cap at all. i.e. loads of poles all over the
place.

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

 

[Winfield Hill]

Jim Thompson wrote...

OK, Now I understand what he meant. The slew time thru the dead-band.


Dramatically so, I've been down that kind of design path.


Also true. Capacitor plates on pins are to be avoided... they tend to
respond to ESD with pin-hole shorts.


Yes. But I don't know of a single commercial amplifier that isn't set
up that way.

I'm sure I've seen several with capacitors directly from the output
pin, at least so far as the drawing can be trusted. I'll try to
locate a few specific examples.

 

[Winfield Hill]

Jim Thompson wrote...

But we do. From that era we know that the B-C junction of the NPN is
the same doping (same process step) as the E-B junction of the PNP.

Jim, are you sure that must be so, given the LM324 output pulldown
transistor is a substrate PNP?

And the current limit resistor is probably 35 ohms. So, as-drawn, we
would have serious (effective) beta reduction when the B-C of the NPN
forward biased. My guess is that there is at least one diode (and
probably two) between the collector of Q7 and the base of Q13

I got up early and spent a few hours updating my computer with old
LM324 datasheets from various manufacturers, so I have now checked 48
of them. Many have very detailed schematics, for example Fairchild
and Samsung, etc. One of the oldest parts, and most complete drawings,
the Fairchild uA124QB, has 95 transistors (16 have multiple collectors),
11 resistors, 4 caps and one JFET. BTW, this happens to exactly match
the number of parts claimed by the TI datasheets to be in their version.

Despite the high degree of detail (one even shows which resistors are
pinch resistors), none of the schematics shows class-AB biasing diodes,
or a collector diode associated with the current-limit transistor.

Even the abbreviated drawings do show considerable difference between
various manufacturer's parts. For example, the Fairchild types use two
current sources to increase bias current for the input emitter-follower
of the input Darlington pairs, while the Motorola types use a collector
tap on the differential transistors to get additional current for the
input EF transistors. This creates a defined beta for the differential
transistors, and either approach reduces the offset-voltage error from
the input EF transistors, by more accurately defining and matching their
two collector currents. The Fairchild types also have collector taps on
the differential transistors, going to ground. Both approaches rob much
of the differential transistor pair's collector current, thereby reducing
the transconductance of the input pair, and allowing the use of a smaller
comp cap (5pF) than would otherwise be possible, according to Motorola.

 

[Jim Thompson]

Jim Thompson wrote...

Do you know what the correct circuit is?

Yes ;-) At least I _think_ I remember it still.

...Jim Thompson

 

[Jim Thompson]

Jim Thompson wrote...

Jim, are you sure that must be so, given the LM324 output pulldown
transistor is a substrate PNP?

Yes. See.....

http://www.analog-innovations.com/SED/NPN-VPNP.pdf

the _diffused_ structures used in LM324, LM339, etc.

In later years, when _implants_ came into use, the VPNP emitter could
be more richly doped, upping its beta.

I got up early and spent a few hours updating my computer with old
LM324 datasheets from various manufacturers, so I have now checked 48
of them. Many have very detailed schematics, for example Fairchild
and Samsung, etc. One of the oldest parts, and most complete drawings,
the Fairchild uA124QB, has 95 transistors (16 have multiple collectors),

I couldn't locate the uA124QB. Could you send a copy? Thanks!

11 resistors, 4 caps and one JFET. BTW, this happens to exactly match
the number of parts claimed by the TI datasheets to be in their version.

Despite the high degree of detail (one even shows which resistors are
pinch resistors), none of the schematics shows class-AB biasing diodes,
or a collector diode associated with the current-limit transistor.

IIRC this is one of Widlar's jewels. He was fond of leaving critical
components off of schematics, or using diffused structures that just
happened to create additional, invisible components.

Even the abbreviated drawings do show considerable difference between
various manufacturer's parts. For example, the Fairchild types use two
current sources to increase bias current for the input emitter-follower
of the input Darlington pairs, while the Motorola types use a collector
tap on the differential transistors to get additional current for the
input EF transistors. This creates a defined beta for the differential
transistors, and either approach reduces the offset-voltage error from
the input EF transistors, by more accurately defining and matching their
two collector currents. The Fairchild types also have collector taps on
the differential transistors, going to ground. Both approaches rob much
of the differential transistor pair's collector current, thereby reducing
the transconductance of the input pair, and allowing the use of a smaller
comp cap (5pF) than would otherwise be possible, according to Motorola.

Yes, that always surprises everyone... low gm is desirable in the
frontend for best GBW product _and_ slew rate.

...Jim Thompson

 

[Ken Smith]

The output section has about 3 diodes worth of crossover distortion and
the compensation does not enclose that distortion so at high frequencies,
it appears in the output. It also makes the output impedance rise quite a
bit for increasing frequencies and vary a lot with load current. This can
lead to trouble when you use them to drive a transistor buffer.

The "simplified schematic" is flawed... it can't be as-drawn... there
is a fundamental error. I tried to tease Win into _stepping_in_it_,
but he won't take the bait ;-)[/QUOTE]

I had assumed that the schematic was simplified but that the 3 diode cross
over issue was real because of what I had seen in in actual use.

And what's this statement, "...compensation does not enclose that
distortion..."?

Consider these oversimplifications:

ASCII art:

Not enclosed:
-----------
! !
-------+--------! Distorty !------ output
! ! ! stuff !
--- ! ! !
---Cc ! -----------
! !/
---+----!
!\ Q1
!
GND



Enclosed:

-------------------------------
! !
! ----------- !
! ! ! !
! --------! Distorty !---+-- output
! ! ! stuff !
--- ! ! !
---Cc ! -----------
! !/
---+----!
!\ Q1
!
GND


When the compensation capacitor takes the feedback from the far side of
the distortion, the gain of Q1 will be acting to make the whole section
act more linear. ie: its collector will slew more quickly through the
cross over distortion. This greatly lowers the cross overdistortion of
the circuit.

This method only works for about 1 transistor followed by something like a
complimentry emitter follower. If you try to enclose too many devices,
the local feedback system can become unstable for some loops gains.
Transistor stages running right near the cut off point have a much lower
frequency responce. As the output switches from the NPN to the PNP, the
frequency responce of that part sweeps down and then back up. The local
feedback loop must be made stable for all of the gain/pahse values the
circuit sweeps through. I assume that this problem is why the LM324 does
not enclose the non-linearities in the compensation loop.

 

[Winfield Hill]

Jim Thompson wrote...

Yes ;-) At least I _think_ I remember it still.

...Jim Thompson

OK, we're all eyes and ears!

 

[Ken Smith]

Yes. But I don't know of a single commercial amplifier that isn't set
up that way.

I assume you are still talking of IC. amplifiers with the compensation cap
as an internal part. IIRC most op-amps that have external
(over)compensation provisions the external cap runs to the output.

 

[Jim Thompson]

Jim Thompson wrote...

OK, we're all eyes and ears!

I'll look thru my archives. I think that's one of the problem areas
for which I wrote an ICE report.

...Jim Thompson