Are there any unity gain stable fast op-amps for single rail

[Lostgallifreyan]

If anyone knows a part that fits the subject description (GBP at least 50
MHz), please let me know what it is. Digikey's filters don't reliably limit
to single-rail choices, and say nothing about unity gain stability, so there
seems to be no easy way to find any.

 

[Lostgallifreyan]

Lostgallifreyan <no-one@nowhere.net> wrote in
news:Xns9F9DAC92FAB66zoodlewurdle@216.196.109.145:

If anyone knows a part that fits the subject description (GBP at least
50 MHz), please let me know what it is. Digikey's filters don't reliably
limit to single-rail choices, and say nothing about unity gain
stability, so there seems to be no easy way to find any.

As an example, I found LT1364 on Digikey, but while their filter implies
operation on a single supply (5 V ~ 36 V, ą2.5 V ~ 18 V), the data sheet
makes it clear that the output swing DOES NOT include the negative rail!

Please help me if you can, because it's clear I can't easily get any guidance
I can trust, if I do what I'm 'supposed to do'.

 

[Rich Webb]

On Mon, 14 Nov 2011 11:29:48 -0600, Lostgallifreyan <no-one@nowhere.net>
wrote:

Lostgallifreyan <no-one@nowhere.net> wrote in
news:Xns9F9DAC92FAB66zoodlewurdle@216.196.109.145:

If anyone knows a part that fits the subject description (GBP at least
50 MHz), please let me know what it is. Digikey's filters don't reliably
limit to single-rail choices, and say nothing about unity gain
stability, so there seems to be no easy way to find any.


As an example, I found LT1364 on Digikey, but while their filter implies
operation on a single supply (5 V ~ 36 V, ą2.5 V ~ 18 V), the data sheet
makes it clear that the output swing DOES NOT include the negative rail!

Please help me if you can, because it's clear I can't easily get any guidance
I can trust, if I do what I'm 'supposed to do'.

If you can go with a smaller GBP (10 MHz typical, 7 MHz spec'd minimum)
National's LM6132/6134 has an input CM range that includes the rails and
can swing to within a few mV of both output rails (depending on how hard
it's driving, of course). Unity gain stable.

--
Rich Webb Norfolk, VA

 

[Lostgallifreyan]

Rich Webb <bbew.ar@mapson.nozirev.ten> wrote in
news:b5n2c79q6ins2ccgesbv0shteatkk3tfjh@4ax.com:

If you can go with a smaller GBP (10 MHz typical, 7 MHz spec'd minimum)
National's LM6132/6134 has an input CM range that includes the rails and
can swing to within a few mV of both output rails (depending on how hard
it's driving, of course). Unity gain stable.

I almost certainly need a lot more GBP but thankyou, it's a nice suggestion
to check out. It seems to have all the other virtues I need.

(As an aside, I'll mention that I sent a message to Digikey about their
search filters returning dodgy results. I got a reply that tuned up just one
amp, but it still wouldn't take output to negative rail. For whatever reason,
Digikey do not seem to realise that this requirement is what 'signal rail'
means, it's not just a different of perspective like stating +5V to +36V
instead of ą2.5V to ą18V. The reply I got actually inferred that I'd asked
for single stage, too... Ň^O)

 

[Lostgallifreyan]

Lostgallifreyan <no-one@nowhere.net> wrote in
news:Xns9F9DD18B54BFEzoodlewurdle@216.196.109.145:

'signal rail'

SINGLE...

 

[Jamie]

Lostgallifreyan wrote:

Lostgallifreyan <no-one@nowhere.net> wrote in
news:Xns9F9DAC92FAB66zoodlewurdle@216.196.109.145:


If anyone knows a part that fits the subject description (GBP at least
50 MHz), please let me know what it is. Digikey's filters don't reliably
limit to single-rail choices, and say nothing about unity gain
stability, so there seems to be no easy way to find any.



As an example, I found LT1364 on Digikey, but while their filter implies
operation on a single supply (5 V ~ 36 V, ą2.5 V ~ 18 V), the data sheet
makes it clear that the output swing DOES NOT include the negative rail!

Please help me if you can, because it's clear I can't easily get any guidance
I can trust, if I do what I'm 'supposed to do'.
Well, I don't know who you can trust? How ever, I looked at that spec

sheet for the quoted part#. The problem isn't so much that is can't
include the (+) rail, it does, it just does not get all the way down to
the absolute common side. You'll see ~ .7 or more when the output is in
the low swing. This also falls true for the high side with that amp.

Single rail circuits of this nature is tricky to get it right, have
you thought about operating a small on board (-) voltage generator for the
(-) rail of the op-amp or any op-amp not having rail abilities to the
common? I mean, you only need like -1 volt or so.

Jamie

 

Lostgallifreyan schrieb:

Rich Webb <bbew.ar@mapson.nozirev.ten> wrote in
news:b5n2c79q6ins2ccgesbv0shteatkk3tfjh@4ax.com:

If you can go with a smaller GBP (10 MHz typical, 7 MHz spec'd minimum)
National's LM6132/6134 has an input CM range that includes the rails and
can swing to within a few mV of both output rails (depending on how hard
it's driving, of course). Unity gain stable.


I almost certainly need a lot more GBP but thankyou, it's a nice suggestion
to check out. It seems to have all the other virtues I need.

Unfortunately, there is an antagonism between high speed and high supply
voltage, and modern products tend to be low-voltage anyway. Nor do the
tricks needed for full rail-to-rail inputs increase speed.

If you can dispense with the high supply voltage: TSH72 by STM. Input
seems to include the negative rail, output incudes both rails, max.
recommended supply 12V.

If you can dispense with the rail-to-rail property: One of the
current-feedback amps dressed up as voltage-feedback types, like the
LT1361 and LT1364 by LT, or the LM6171 by National. Or true
voltage-feedback types like the LM6161 (unity-gain stable) or LM6164
(unstable) by National, or the THS4001 by TI, all with max. supply
around 32V.

These are often labeled video amplifiers, so this should make a good
search term.

Martin.

 

[Lostgallifreyan]

Jamie <jamie_ka1lpa_not_valid_after_ka1lpa_@charter.net> wrote in
news:8ehwq.29352$v_4.29305@newsfe21.iad:

Lostgallifreyan wrote:

Lostgallifreyan <no-one@nowhere.net> wrote in
news:Xns9F9DAC92FAB66zoodlewurdle@216.196.109.145:


If anyone knows a part that fits the subject description (GBP at least
50 MHz), please let me know what it is. Digikey's filters don't
reliably limit to single-rail choices, and say nothing about unity gain
stability, so there seems to be no easy way to find any.



As an example, I found LT1364 on Digikey, but while their filter
implies operation on a single supply (5 V ~ 36 V, ą2.5 V ~ 18 V), the
data sheet makes it clear that the output swing DOES NOT include the
negative rail!

Please help me if you can, because it's clear I can't easily get any
guidance I can trust, if I do what I'm 'supposed to do'.

Well, I don't know who you can trust? How ever, I looked at that spec
sheet for the quoted part#. The problem isn't so much that is can't
include the (+) rail, it does, it just does not get all the way down to
the absolute common side. You'll see ~ .7 or more when the output is in
the low swing. This also falls true for the high side with that amp.

Single rail circuits of this nature is tricky to get it right, have
you thought about operating a small on board (-) voltage generator for
the (-) rail of the op-amp or any op-amp not having rail abilities to
the common? I mean, you only need like -1 volt or so.

Not this time. I did that in another circuit where it had the extra
advantage of some huge headroom (laser power meter, up to over 20 watts on a
single scale, to be read by any voltmeter on a 20V range). In this case
though, I want clipping at ground as part of protection for a laser diode. So
the output (and input) must be able to reach the negative rail. To do
anything else increases circuit complexity a lot, compared to what I have
now.

Re trust, I'm gettting good answers here (yours too), a sense of what
compromises make these particular amp specs hard to find. I knew they were
rare, but not so much about why. (LT specifically mention LT1215 as unusual,
I guess I was lucky to find it so easily).

 

[Lostgallifreyan]

clicliclic@freenet.de wrote in news:4EC1A742.E758B2EF@freenet.de:

Unfortunately, there is an antagonism between high speed and high supply
voltage, and modern products tend to be low-voltage anyway. Nor do the
tricks needed for full rail-to-rail inputs increase speed.

If you can dispense with the high supply voltage: TSH72 by STM. Input
seems to include the negative rail, output incudes both rails, max.
recommended supply 12V.

Definitely possible. One use I had in mind could use up to 36V supply, but
that scheme didn't need speed, was just a neat way to use the existing
circuit. The real need that made me build it was for a laser diode, not an
LED array, and I can use a supply as low as 6V, but I will try for more
because some people like to drive multiple diodes in series, so the more
headroom the better..

TSH72 looks nice, but has very small differential input max, 4V effectively,
and it seems limited to symmetry about ground. I don't need very much
difference, but I do need at least 5V. (Also, once I settle on a really good
one, I know I might reach for it more often if it has some generous allowance
all round. Ideal fantasy: CA3140 on steroids, GBP around 70 MHz

If you can dispense with the rail-to-rail property: One of the
current-feedback amps dressed up as voltage-feedback types, like the
LT1361 and LT1364 by LT, or the LM6171 by National. Or true
voltage-feedback types like the LM6161 (unity-gain stable) or LM6164
(unstable) by National, or the THS4001 by TI, all with max. supply
around 32V.

I found an LM6142 (after the suggestion given to me, of LM6132), which looks
like a proper contender, so I'll certainly look at these others you named.
Thankyou. (I don't need to-rail on the postive supply, just the negative..)

These are often labeled video amplifiers, so this should make a good
search term.

Definitely. I'll try that too. Ideally I'll end up with one that is as
futureproof as the CA3140 has turned out to be, but I guess no-one knows the
form book for new devices, so maybe all bets are off.

 

Lostgallifreyan schrieb:

clicliclic@freenet.de wrote in news:4EC1A742.E758B2EF@freenet.de:

If you can dispense with the high supply voltage: TSH72 by STM. Input
seems to include the negative rail, output incudes both rails, max.
recommended supply 12V.

TSH72 looks nice, but has very small differential input max, 4V effectively,
and it seems limited to symmetry about ground. I don't need very much
difference, but I do need at least 5V.

Unfortunately ST datasheets are hardly complete. The GHz input
transistors very likely need protection since transients of the order of
the supply voltage always have to be reckoned with. This is probably
supplied on-chip using diodes in combination with series resistors. The
max. admissible +-2V (not 4V as far as I can see) between inputs
presumably refers to a steady-state stress. If you want to exceed this
value, external series resistors should keep the junctions happy. To
determine the resistor values, measure the current actually drawn at 2V.
As not to compromise speed, tiny capacitancies may be added in parallel.

You may also want to look at Linear's LT1632: rail-to rail input and
output, 45MHz, 45V/us, max. admissible supply 36V. The center figure on
page 9 of my datasheet (last change appears to have been 2007) is
incompatible with a GBP of 45MHz; I suppose the gain scale is off by
20dB. Here, the on-chip input protection is directly shown on page 12.

Martin.

 

[Lostgallifreyan]

clicliclic@freenet.de wrote in news:4EC2C21A.87DE1AD2@freenet.de:

Unfortunately ST datasheets are hardly complete. The GHz input
transistors very likely need protection since transients of the order of
the supply voltage always have to be reckoned with. This is probably
supplied on-chip using diodes in combination with series resistors. The
max. admissible +-2V (not 4V as far as I can see) between inputs
presumably refers to a steady-state stress. If you want to exceed this
value, external series resistors should keep the junctions happy. To
determine the resistor values, measure the current actually drawn at 2V.
As not to compromise speed, tiny capacitancies may be added in parallel.

I'll pass on that one. I like the idea of the current limiting resistors, but
in my current circuit I have already made boards, and need to keep their size
down while using through-hole parts. (I'll aim for a second edition using SMT
some time but first things first.

Btw, isn't 4V centred about zero just another way to say ą2V? I'm ok with
saying ąnV if it removes ambiguity though.

You may also want to look at Linear's LT1632: rail-to rail input and
output, 45MHz, 45V/us, max. admissible supply 36V. The center figure on
page 9 of my datasheet (last change appears to have been 2007) is
incompatible with a GBP of 45MHz; I suppose the gain scale is off by
20dB. Here, the on-chip input protection is directly shown on page 12.

Now that DOES look nice. Thankyou. Wasn't part of my last LTspice update,
but I swapped it in for LT1215 in my circuit model and it works well enough
that I'll try to get some to try in the real one soon.

The circuit as it is now is here:
http://repairfaq.cis.upenn.edu/Misc/laserdps.htm#dpsldd317

Currently, when I crank the input frequency (using square waves) to above
about 350 MHz I get some nasty spiking that would threaten the laser diode
that the circuit is driving. That ONLY happens if the dual pot that mixes
between DC 5V internal reference, and 5V peak input, were to fail in such a
way as to make the total difference of input voltage greater than 5V. With a
good pot this is unlikely, but if I can reduce the risk further, I will.
Spice didn't reveal this problem, so I'm not yet sure what the cause is,
other than it seems related to cliping at zero volts. THis cliiping is a
clear advantage at up around 300 Hz but apparently a major risk above this.
Ň^O

The first thing to test seems to be a different amp, one that is more stable
with unity gain (the first stage in that circuit has LESS gain, as it has to
divide by four, whatever total results from the mixed input). The CA3240
never showed this behaviour, but then it wouldn't, it's not a fast amp. So
I'm still left wondering if changing the amp will fix this, or whether I need
to look elsewhere. Figuring out what other IC's work well here seemed like a
good first test...

 

[Lostgallifreyan]

Lostgallifreyan <no-one@nowhere.net> wrote in
news:Xns9FA19959150AAzoodlewurdle@216.196.109.145:

cliping at zero volts. THis cliiping

Clipping.
Just so people know.... >

 

[Lostgallifreyan]

Lostgallifreyan <no-one@nowhere.net> wrote in
news:Xns9FA19959150AAzoodlewurdle@216.196.109.145:

300 Hz

KHz. This typo actually mattered...

 

Lostgallifreyan schrieb:

Unfortunately ST datasheets are hardly complete. [...]

[...] Btw, isn't 4V centred about zero just another way to say ą2V?
I'm ok with saying ąnV if it removes ambiguity though.

My TSH7x datasheet says on page 2:

ABSOLUTE MAXIMUM RATINGS

Symbol | Parameter | Value | Unit
----------------------------------------------------------------
Vid | Differential Input Voltage (2) | +- 2 | V

with the footnote:

(2) Differential voltages are non-inverting input terminal with
respect to the inverting terminal

Now that DOES look nice. Thankyou.

It is listed by ST as being replaceable by their TSH72! They fail to
mention the much higher max. supply of the LT1632 though.

Martin.

 

[Lostgallifreyan]

clicliclic@freenet.de wrote in news:4EC6C1EC.205FF440@freenet.de:

Btw, isn't 4V centred about zero just another way to say ą2V?
I'm ok with saying ąnV if it removes ambiguity though.

My TSH7x datasheet says on page 2:

ABSOLUTE MAXIMUM RATINGS

Symbol | Parameter | Value | Unit
----------------------------------------------------------------
Vid | Differential Input Voltage (2) | +- 2 | V

with the footnote:

(2) Differential voltages are non-inverting input terminal with
respect to the inverting terminal

Oh. So literally ONLY 2V, from each other. I didn't notive that restriction,
I assumed they meant ground, therefore total peak difference of 4 when both
were were at opposite maximum. (Normally when I see ą anything, that's what
is meant).

Anyway, if you looked at the thing I linked to, you'll see I have weird ideas
about versatility and use of widely available parts with a good degree of
futureproofing. So a wide-range differential input is a must.

Now that DOES look nice. Thankyou.

It is listed by ST as being replaceable by their TSH72! They fail to
mention the much higher max. supply of the LT1632 though.

I can see that recommending parts can be more than awkward. We're expected to
specify minutiae to limit choices, but that's a raw call when makers are
failing to do what we must do, when we can't if they don't. Again, this is
ahy I try to choose parts carefully. I like to make choices I can stand by
for years. I have no clue what LT's record is of pulling IC's out from under
people's feet, but surely as times get harder, they might want to limit the
number of different devices to get reliable sales. I don't know how to figure
out likely stocks that last, it seems that a non-technical approach can work
just as well, like seeing what gets sold widely enough to end up on eBay and
lots of small hobbyist supplies. Apart from "if it works and people want it
enough" I can't see how to make choices.

Btw, the CA3240 works a damn sight better than I said it does in that laser
driver page I linked to. I just coped it again today, and was more than
impressed. It's actually better at 500 KHz that some are at 100 or 200 KHz
when using video bandwidth amps, if it doesn't have to do any gain. It seems
to have enough slew rate and stability at unity gain to make me wonder why
I'm searching so hard to better it at all!