Maker Pro
Maker Pro

current-mode opamps

J

John Larkin

Jan 1, 1970
0
I'm still trying to figure out current-mode opamps.

In a conventional opamp, the emitters of an input diff pair are fed by
a current source, and the diff input voltage steers that current. In a
current-mode opamp, the complementary input transistors have a small
quiescent bias, but the non-inverting (bases) input signal invokes
large collector currents, which are amplified all the way to the
output. So in a sense we are getting information from the signal
applied to a regular opamp, but we're extracting power from the input
to a current-mode opamp. Which is why slew rates can be so high: more
error loads the input signal harder, and drives the outputs harder,
without running out of current.

If you look at this,

http://focus.ti.com/lit/ds/symlink/ths3201.pdf

you'll see that the ni input impedance is spec'd as 780K + 1 pF. They
also provide an S11 plot, which, after it's untangled, models as a
high resistance shunted by about 2 pF or so, presumably because of
extra C on the horribly-laid-out eval board.

OK, I have an open-drain (a phemt) with a 50 ohm pullup to, say, +2
volts. Unloaded, risetime is 100 ps or so. If it's then coupled to the
THS3201 as a gain=+2 amp, the output falling edge is screaming fast,
but rise is slow. Looking at the fet drain, it looks as if the pullup
resistor is loaded by about 10 pF. Pig city.

So at large-signal swings, the input transistors are running out of
beta as frequency goes up. That makes the input appear to be
capacitive. That's not suggested by the input specs nor by the
small-signal S parameters.

Tricky.

John
 
J

Joerg

Jan 1, 1970
0
John said:
I'm still trying to figure out current-mode opamps.

In a conventional opamp, the emitters of an input diff pair are fed by
a current source, and the diff input voltage steers that current. In a
current-mode opamp, the complementary input transistors have a small
quiescent bias, but the non-inverting (bases) input signal invokes
large collector currents, which are amplified all the way to the
output. So in a sense we are getting information from the signal
applied to a regular opamp, but we're extracting power from the input
to a current-mode opamp. Which is why slew rates can be so high: more
error loads the input signal harder, and drives the outputs harder,
without running out of current.

If you look at this,

http://focus.ti.com/lit/ds/symlink/ths3201.pdf

you'll see that the ni input impedance is spec'd as 780K + 1 pF. They
also provide an S11 plot, which, after it's untangled, models as a
high resistance shunted by about 2 pF or so, presumably because of
extra C on the horribly-laid-out eval board.

Sez 1pF in the data but they only give a typical value, no max. But it
depends on what you have hanging on the output.

OK, I have an open-drain (a phemt) with a 50 ohm pullup to, say, +2
volts. Unloaded, risetime is 100 ps or so. If it's then coupled to the
THS3201 as a gain=+2 amp, the output falling edge is screaming fast,
but rise is slow. Looking at the fet drain, it looks as if the pullup
resistor is loaded by about 10 pF. Pig city.

So at large-signal swings, the input transistors are running out of
beta as frequency goes up. That makes the input appear to be
capacitive. That's not suggested by the input specs nor by the
small-signal S parameters.

Maybe it didn't look good on the datasheet? But seriously, what's on the
output?


I've never really warmed up to the CFB concept. Kept things discrete,
mostly. One of our profs told us that using a FET in common gate
configuration would be outright stupid. I laughed so hard I had to leave
the auditorium.
 
W

Winfield Hill

Jan 1, 1970
0
John said:
I'm still trying to figure out current-mode opamps.

In a conventional opamp, the emitters of an input diff pair are fed by
a current source, and the diff input voltage steers that current. In a
current-mode opamp, the complementary input transistors have a small
quiescent bias, but the non-inverting (bases) input signal invokes
large collector currents, which are amplified all the way to the
output. So in a sense we are getting information from the signal
applied to a regular opamp, but we're extracting power from the input
to a current-mode opamp. Which is why slew rates can be so high: more
error loads the input signal harder, and drives the outputs harder,
without running out of current.

If you look at this,
http://focus.ti.com/lit/ds/symlink/ths3201.pdf

you'll see that the ni input impedance is spec'd as 780K + 1 pF.
They> also provide an S11 plot, which, after it's untangled, models
as a high resistance shunted by about 2 pF or so, presumably
because of extra C on the horribly-laid-out eval board.

OK, I have an open-drain (a phemt) with a 50 ohm pullup to, say,
+2 volts. Unloaded, risetime is 100 ps or so. If it's then coupled
to the THS3201 as a gain=+2 amp, the output falling edge is screaming
fast, but rise is slow. Looking at the fet drain, it looks as if
the pullup resistor is loaded by about 10 pF. Pig city.

So at large-signal swings, the input transistors are running out
of beta as frequency goes up. That makes the input appear to be
capacitive. That's not suggested by the input specs nor by the
small-signal S parameters.

Tricky.

You've got Vin = 2V and G = +2. What are your
two CFB "-input" resistor values?
 
J

John Larkin

Jan 1, 1970
0
You've got Vin = 2V and G = +2. What are your
two CFB "-input" resistor values?


At G=2, 100 ohm load, they suggest that both resistors be 715 ohms for
good large-signal performance. Playing with the resistors changes
transient response but doesn't much affect the apparent 10 pF input
capacitance.

The eval board, and the S-params setup, effectively drive the ni input
from a 25 ohm source. The frequency response curves are unclear on
what the generator source impedance may be. They give inverting and ni
curves small-signal, but the large-signal stuff is all inverting. Are
they trying to hide the actual performance?

TI is beginning to disappoint me.

John
 
J

Joerg

Jan 1, 1970
0
John said:
At G=2, 100 ohm load, they suggest that both resistors be 715 ohms for
good large-signal performance. Playing with the resistors changes
transient response but doesn't much affect the apparent 10 pF input
capacitance.

The eval board, and the S-params setup, effectively drive the ni input
from a 25 ohm source. The frequency response curves are unclear on
what the generator source impedance may be. They give inverting and ni
curves small-signal, but the large-signal stuff is all inverting. Are
they trying to hide the actual performance?

TI is beginning to disappoint me.

Ask them. But don't hold your breath, I had a case a while ago where
they could not figure out why a TPS regulator went "exotherm" despite
everything around it being well within specs. They didn't want to give
me the SPICE model either so that kind of tied my hands. It was a LDO,
by the way ...
 
J

John Larkin

Jan 1, 1970
0
Ask them. But don't hold your breath, I had a case a while ago where
they could not figure out why a TPS regulator went "exotherm" despite
everything around it being well within specs. They didn't want to give
me the SPICE model either so that kind of tied my hands. It was a LDO,
by the way ...

I commented earlier on our experience THS3062's frying. It took them
about 8 weeks to respond to our many requests for help. Finally
prodded by TI's VP of Ethics, they admitted that the amp is
fundamentally broken, and that the datasheet effectively lies.

It's sad that they have about the fastest linear process on the
planet, and they use it so badly.

My finger is much better. I may not have a permanent SO-8 shaped scar.

John
 
J

Joerg

Jan 1, 1970
0
John said:
I commented earlier on our experience THS3062's frying. It took them
about 8 weeks to respond to our many requests for help. Finally
prodded by TI's VP of Ethics, they admitted that the amp is
fundamentally broken, and that the datasheet effectively lies.

They have a VP of Ethics? Dang, I could have forwarded the TPS regulator
issue to him because IMHO the handling was ethically, ahem, let's say
borderline. Do they give an "ombudsman" link somewhere? But it's water
under the bridge now, I designed it out for good. No more LDOs.

It's sad that they have about the fastest linear process on the
planet, and they use it so badly.

Their non-CFB THS amps are fantastic. CFB I don't know, I tried those
out years ago from another manufacturer. Back then CFB was hailed as the
ultimate cat's meouw. To me it sure wasn't ...

My finger is much better. I may not have a permanent SO-8 shaped scar.

A friend does have a permanent scar. 16 dots. He leaned on a table and
failed to notice that 16 pin DIP chip laying there, upside down ...
 
T

Tom2000

Jan 1, 1970
0
At G=2, 100 ohm load, they suggest that both resistors be 715 ohms for
good large-signal performance. Playing with the resistors changes
transient response but doesn't much affect the apparent 10 pF input
capacitance.

The eval board, and the S-params setup, effectively drive the ni input
from a 25 ohm source. The frequency response curves are unclear on
what the generator source impedance may be. They give inverting and ni
curves small-signal, but the large-signal stuff is all inverting. Are
they trying to hide the actual performance?

TI is beginning to disappoint me.

John

John, have you had a chance to compare the TI CFBs with the AD
versions?

(Great thread, by the way...)

Thanks,

Tom
 
J

Jim Thompson

Jan 1, 1970
0
John Larkin wrote: [snip]
My finger is much better. I may not have a permanent SO-8 shaped scar.

A friend does have a permanent scar. 16 dots. He leaned on a table and
failed to notice that 16 pin DIP chip laying there, upside down ...

I stepped on one once... in my socks :-(

Completely cured me of walking around with no shoes in the lab.

...Jim Thompson
 
J

John Larkin

Jan 1, 1970
0
John Larkin wrote: [snip]
My finger is much better. I may not have a permanent SO-8 shaped scar.

A friend does have a permanent scar. 16 dots. He leaned on a table and
failed to notice that 16 pin DIP chip laying there, upside down ...

I stepped on one once... in my socks :-(

Completely cured me of walking around with no shoes in the lab.

...Jim Thompson


Me too, dip24 barefoot. And once I knocked a soldering iron off the
bench and, superb athelete that I am, caught it in mid-air.

John
 
J

John Larkin

Jan 1, 1970
0
John, have you had a chance to compare the TI CFBs with the AD
versions?

(Great thread, by the way...)

Thanks,

Tom

The ADI parts are great if you don't need a lot of voltage swing. We
use AD8009's, 8014's, 8001's, all great parts without quirks.

The TI's have insane slew rates and bandwidth with +-7.5 or even +-15
rails, like nothing else I know of.

TI also makes some insanely fast fixed-gain amps, THS4302 and 4303.

John
 
T

Tom2000

Jan 1, 1970
0
The ADI parts are great if you don't need a lot of voltage swing. We
use AD8009's, 8014's, 8001's, all great parts without quirks.

The TI's have insane slew rates and bandwidth with +-7.5 or even +-15
rails, like nothing else I know of.

TI also makes some insanely fast fixed-gain amps, THS4302 and 4303.

John

Thanks, John. Great info. I'm keeping this whole thread for
reference.

Tom
 
J

Joerg

Jan 1, 1970
0
John said:
John Larkin wrote: [snip]
My finger is much better. I may not have a permanent SO-8 shaped scar.

A friend does have a permanent scar. 16 dots. He leaned on a table and
failed to notice that 16 pin DIP chip laying there, upside down ...
I stepped on one once... in my socks :-(

Completely cured me of walking around with no shoes in the lab.

...Jim Thompson


Me too, dip24 barefoot. And once I knocked a soldering iron off the
bench and, superb athelete that I am, caught it in mid-air.

The ultimate nightmare happened to a telco engineer. He wore those
climber's sandals where it takes a while to get them off, with thick
wool socks. A blob of molten solder rolled off a board, onto the toe
area, then phsssss... through the sock.
 
J

Jim Thompson

Jan 1, 1970
0
John Larkin wrote: [snip]

My finger is much better. I may not have a permanent SO-8 shaped scar.


A friend does have a permanent scar. 16 dots. He leaned on a table and
failed to notice that 16 pin DIP chip laying there, upside down ...

I stepped on one once... in my socks :-(

Completely cured me of walking around with no shoes in the lab.

...Jim Thompson


Me too, dip24 barefoot. And once I knocked a soldering iron off the
bench and, superb athelete that I am, caught it in mid-air.

John

I reached for the iron without looking up, ONCE ;-)

...Jim Thompson
 
J

Joerg

Jan 1, 1970
0
Jim said:
On Mon, 25 Feb 2008 17:22:07 GMT, Joerg

John Larkin wrote:
[snip]
My finger is much better. I may not have a permanent SO-8 shaped scar.

A friend does have a permanent scar. 16 dots. He leaned on a table and
failed to notice that 16 pin DIP chip laying there, upside down ...
I stepped on one once... in my socks :-(

Completely cured me of walking around with no shoes in the lab.

...Jim Thompson

Me too, dip24 barefoot. And once I knocked a soldering iron off the
bench and, superb athelete that I am, caught it in mid-air.

John

I reached for the iron without looking up, ONCE ;-)

An engineer in The Netherlands (where they drink a lot of coffee) wanted
to place his back without looking up. Instead of the Weller spring he
landed it in a full styrofoam coffee cup.
 
C

Clifford Heath

Jan 1, 1970
0
John said:
Me too, dip24 barefoot. And once I knocked a soldering iron off the
bench and, superb athelete that I am, caught it in mid-air.

Surely a superb athlete would have caught the correct end though?
I've caught a dropped iron a couple of times - with the presence of
mind to catch it by the *cord*.

I left the soldering iron on the ground while working on the car's
tail-lights, *once*. My three-year-old stepped on it barefoot, and
didn't realize for several seconds. No idea why it didn't do more
harm, but he was fine after a couple of days, though it was a painful
burn. Still makes me feel queasy 18 years later...
 
J

John Larkin

Jan 1, 1970
0
Thanks, John. Great info. I'm keeping this whole thread for
reference.

Tom

Today's experiments suggest that the "fake capacitance" at the THS3201
input is different depending on the polarity you happen to be slewing.
Even more fun!

We're looking into output stages that use discrete GaAs fets.

John
 
M

MooseFET

Jan 1, 1970
0
On Feb 25, 9:22 am, Joerg <[email protected]>
wrote:
[...]
They have a VP of Ethics?

If someone has a plumber on staff, I tend to think they have or at
least had a problem with pipes leaking.


[... TI ...]
Their non-CFB THS amps are fantastic. CFB I don't know, I tried those
out years ago from another manufacturer. Back then CFB was hailed as the
ultimate cat's meouw. To me it sure wasn't ...

CFB op-amps were the "new and improved" product back then. Today the
dust has settled and they have found use in a few places. I like the
LT1206 because you can "overcompensate the heck out of it" with just
resistor values.
A friend does have a permanent scar. 16 dots. He leaned on a table and
failed to notice that 16 pin DIP chip laying there, upside down ...

For a while I had a nifty pattern on the bottom of one foot. I worked
bare footed at home back then.
 
M

MooseFET

Jan 1, 1970
0
John Larkin wrote: [snip]
My finger is much better. I may not have a permanent SO-8 shaped scar.
A friend does have a permanent scar. 16 dots. He leaned on a table and
failed to notice that 16 pin DIP chip laying there, upside down ...

I stepped on one once... in my socks :-(

Completely cured me of walking around with no shoes in the lab.
I think it is the copper in the leadframe that makes it hurt out of
all proportion to the real damage.
 
J

John Popelish

Jan 1, 1970
0
MooseFET said:
I think it is the copper in the leadframe that makes it hurt out of
all proportion to the real damage.

Could be, but the time I stepped on one and some of the the
pins hit bone and folded over like staples, hurt for another
reason.
 
Top