poor op-amp settling time

[Winfield Hill]

I selected a Maxim MAX4239 opamp for a low-power design
project, for its 2.5uV offset, its Iq = 0.6mA, and for
G>10, its 6.5MHz GBW. For G=10, that's a 650kHz -3dB
bandwidth. Slew 1.6V/us. Response, t = 1/pi 650kHz =
0.5us. But the datasheet's 0.1% settling time spec is
500us, 1000x times slower! Makes little sense. Sad.
Dunno whether to believe a very poor datasheet number.


--
Thanks,
- Win

 

On Sun, 12 Apr 2020 22:24:36 -0400, bitrex <user@example.net> wrote:

On 4/12/2020 10:20 PM, jlarkin@highlandsniptechnology.com wrote:
On 12 Apr 2020 18:41:59 -0700, Winfield Hill <winfieldhill@yahoo.com
wrote:


I selected a Maxim MAX4239 opamp for a low-power design
project, for its 2.5uV offset, its Iq = 0.6mA, and for
G>10, its 6.5MHz GBW. For G=10, that's a 650kHz -3dB
bandwidth. Slew 1.6V/us. Response, t = 1/pi 650kHz =
0.5us. But the datasheet's 0.1% settling time spec is
500us, 1000x times slower! Makes little sense. Sad.
Dunno whether to believe a very poor datasheet number.

Never Buy Maxim!

I use the DS1077 i2c clock oscillator sometimes it's cheap and does the
business. But that's a Dallas Semi part to be fair.

Yikes, they measure frequency errors in per cent, not parts per
million. You can get a programmable crystal oscillator for less than
that.

Maxim deliberately uses nonstandard pinouts. And then they discontinue
parts without notice.



--

John Larkin Highland Technology, Inc

Science teaches us to doubt.

Claude Bernard

 

[Phil Hobbs]

On 2020-04-12 21:41, Winfield Hill wrote:

I selected a Maxim MAX4239 opamp for a low-power design project, for
its 2.5uV offset, its Iq = 0.6mA, and for G>10, its 6.5MHz GBW. For
G=10, that's a 650kHz -3dB bandwidth. Slew 1.6V/us. Response, t =
1/pi 650kHz = 0.5us. But the datasheet's 0.1% settling time spec is
500us, 1000x times slower! Makes little sense. Sad. Dunno whether
to believe a very poor datasheet number.

That often happens when the vendor does some funny stuff with the
compensation. You wind up with a pole/zero pair that's closely spaced
but not coincident. The ancient trick for speeding up an LM101 by
capacitively coupling to the offset null pins to bypass the lateral PNP
level shifters is a classic example.

It gives you a little whoopdedoo in the transfer function that dominates
the settling behaviour at late times when other contributions have died
out. The LM318 was famous for that as well. (Come to think of it,
National never did get over the idea that 400 ns was a fast settling
time. Or that 1 mV was a low offset voltage.)

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

[Phil Hobbs]

On 2020-04-12 21:59, Phil Hobbs wrote:

On 2020-04-12 21:41, Winfield Hill wrote:
I selected a Maxim MAX4239 opamp for a low-power design project,
for its 2.5uV offset, its Iq = 0.6mA, and for G>10, its 6.5MHz GBW.
For G=10, that's a 650kHz -3dB bandwidth. Slew 1.6V/us. Response,
t = 1/pi 650kHz = 0.5us. But the datasheet's 0.1% settling time
spec is 500us, 1000x times slower! Makes little sense. Sad. Dunno
whether to believe a very poor datasheet number.

That often happens when the vendor does some funny stuff with the
compensation. You wind up with a

low-frequency

pole/zero pair that's closely spaced but not coincident. The ancient
trick for speeding up an LM101 by capacitively coupling to the offset
null pins to bypass the lateral PNP level shifters is a classic
example.

It gives you a little whoopdedoo in the transfer function that
dominates the settling behaviour at late times when other
contributions have died out. The LM318 was famous for that as well.
(Come to think of it, National never did get over the idea that 400
ns was a fast settling time. Or that 1 mV was a low offset
voltage.)

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

On 12 Apr 2020 18:41:59 -0700, Winfield Hill <winfieldhill@yahoo.com>
wrote:

I selected a Maxim MAX4239 opamp for a low-power design
project, for its 2.5uV offset, its Iq = 0.6mA, and for
G>10, its 6.5MHz GBW. For G=10, that's a 650kHz -3dB
bandwidth. Slew 1.6V/us. Response, t = 1/pi 650kHz =
0.5us. But the datasheet's 0.1% settling time spec is
500us, 1000x times slower! Makes little sense. Sad.
Dunno whether to believe a very poor datasheet number.

Never Buy Maxim!

The overload recovery time is milliseconds, so there's probably some
node inside that winds up when it rails or slews. It's an autozero,
and they can get messy. Intersil made one autozero opamp that took
*seconds* to recover.

The overload recovery graph is suggestive. Look at that time constant!




--

John Larkin Highland Technology, Inc

Science teaches us to doubt.

Claude Bernard

 

[keith wright]

On Sunday, 12 April 2020 18:42:16 UTC-7, Winfield Hill wrote:

I selected a Maxim MAX4239 opamp for a low-power design
project, for its 2.5uV offset, its Iq = 0.6mA, and for
G>10, its 6.5MHz GBW. For G=10, that's a 650kHz -3dB
bandwidth. Slew 1.6V/us. Response, t = 1/pi 650kHz =
0.5us. But the datasheet's 0.1% settling time spec is
500us, 1000x times slower! Makes little sense. Sad.
Dunno whether to believe a very poor datasheet number.


--
Thanks,
- Win

It is a chopper stabilized opamp - I've been bitten by the long settling time of those types when recovering from overdrive.

kw

 

[bitrex]

On 4/12/2020 10:20 PM, jlarkin@highlandsniptechnology.com wrote:

On 12 Apr 2020 18:41:59 -0700, Winfield Hill <winfieldhill@yahoo.com
wrote:


I selected a Maxim MAX4239 opamp for a low-power design
project, for its 2.5uV offset, its Iq = 0.6mA, and for
G>10, its 6.5MHz GBW. For G=10, that's a 650kHz -3dB
bandwidth. Slew 1.6V/us. Response, t = 1/pi 650kHz =
0.5us. But the datasheet's 0.1% settling time spec is
500us, 1000x times slower! Makes little sense. Sad.
Dunno whether to believe a very poor datasheet number.

Never Buy Maxim!

I use the DS1077 i2c clock oscillator sometimes it's cheap and does the
business. But that's a Dallas Semi part to be fair.

The overload recovery time is milliseconds, so there's probably some
node inside that winds up when it rails or slews. It's an autozero,
and they can get messy. Intersil made one autozero opamp that took
*seconds* to recover.

The overload recovery graph is suggestive. Look at that time constant!

 

[bitrex]

On 4/12/2020 10:40 PM, jlarkin@highlandsniptechnology.com wrote:

On Sun, 12 Apr 2020 22:24:36 -0400, bitrex <user@example.net> wrote:

On 4/12/2020 10:20 PM, jlarkin@highlandsniptechnology.com wrote:
On 12 Apr 2020 18:41:59 -0700, Winfield Hill <winfieldhill@yahoo.com
wrote:


I selected a Maxim MAX4239 opamp for a low-power design
project, for its 2.5uV offset, its Iq = 0.6mA, and for
G>10, its 6.5MHz GBW. For G=10, that's a 650kHz -3dB
bandwidth. Slew 1.6V/us. Response, t = 1/pi 650kHz =
0.5us. But the datasheet's 0.1% settling time spec is
500us, 1000x times slower! Makes little sense. Sad.
Dunno whether to believe a very poor datasheet number.

Never Buy Maxim!

I use the DS1077 i2c clock oscillator sometimes it's cheap and does the
business. But that's a Dallas Semi part to be fair.

Yikes, they measure frequency errors in per cent, not parts per
million. You can get a programmable crystal oscillator for less than
that.

It has a very long divider chain, I don't know what programmable xtal
oscillator for less money would be a good substitute with that
granularity that can be adjusted on the fly.

Maxim deliberately uses nonstandard pinouts. And then they discontinue
parts without notice.

 

[Ricky C]

On Sunday, April 12, 2020 at 11:19:09 PM UTC-4, bitrex wrote:

On 4/12/2020 10:40 PM, jlarkin@highlandsniptechnology.com wrote:
On Sun, 12 Apr 2020 22:24:36 -0400, bitrex <user@example.net> wrote:

On 4/12/2020 10:20 PM, jlarkin@highlandsniptechnology.com wrote:
On 12 Apr 2020 18:41:59 -0700, Winfield Hill <winfieldhill@yahoo.com
wrote:


I selected a Maxim MAX4239 opamp for a low-power design
project, for its 2.5uV offset, its Iq = 0.6mA, and for
G>10, its 6.5MHz GBW. For G=10, that's a 650kHz -3dB
bandwidth. Slew 1.6V/us. Response, t = 1/pi 650kHz > >>>> 0.5us. But the datasheet's 0.1% settling time spec is
500us, 1000x times slower! Makes little sense. Sad.
Dunno whether to believe a very poor datasheet number.

Never Buy Maxim!

I use the DS1077 i2c clock oscillator sometimes it's cheap and does the
business. But that's a Dallas Semi part to be fair.

Yikes, they measure frequency errors in per cent, not parts per
million. You can get a programmable crystal oscillator for less than
that.

It has a very long divider chain, I don't know what programmable xtal
oscillator for less money would be a good substitute with that
granularity that can be adjusted on the fly.

Maxim deliberately uses nonstandard pinouts. And then they discontinue
parts without notice.

I don't know for sure about packages with the crystal in them, but they have PLL type devices that will give you pretty much whatever frequency you want with a pretty fine granularity. I know there are programmable oscillators with dividers, so I assume they come with the PLL as well. Seems silly not to.

--

Rick C.

- Get 1,000 miles of free Supercharging
- Tesla referral code - https://ts.la/richard11209

 

[bitrex]

On 4/13/2020 12:52 AM, Ricky C wrote:

On Sunday, April 12, 2020 at 11:19:09 PM UTC-4, bitrex wrote:
On 4/12/2020 10:40 PM, jlarkin@highlandsniptechnology.com wrote:
On Sun, 12 Apr 2020 22:24:36 -0400, bitrex <user@example.net> wrote:

On 4/12/2020 10:20 PM, jlarkin@highlandsniptechnology.com wrote:
On 12 Apr 2020 18:41:59 -0700, Winfield Hill <winfieldhill@yahoo.com
wrote:


I selected a Maxim MAX4239 opamp for a low-power design
project, for its 2.5uV offset, its Iq = 0.6mA, and for
G>10, its 6.5MHz GBW. For G=10, that's a 650kHz -3dB
bandwidth. Slew 1.6V/us. Response, t = 1/pi 650kHz =
0.5us. But the datasheet's 0.1% settling time spec is
500us, 1000x times slower! Makes little sense. Sad.
Dunno whether to believe a very poor datasheet number.

Never Buy Maxim!

I use the DS1077 i2c clock oscillator sometimes it's cheap and does the
business. But that's a Dallas Semi part to be fair.

Yikes, they measure frequency errors in per cent, not parts per
million. You can get a programmable crystal oscillator for less than
that.

It has a very long divider chain, I don't know what programmable xtal
oscillator for less money would be a good substitute with that
granularity that can be adjusted on the fly.

Maxim deliberately uses nonstandard pinouts. And then they discontinue
parts without notice.

I don't know for sure about packages with the crystal in them, but they have PLL type devices that will give you pretty much whatever frequency you want with a pretty fine granularity. I know there are programmable oscillators with dividers, so I assume they come with the PLL as well. Seems silly not to.

It works great for clocking a bucket-brigade delay!

<https://www.emusician.com/.image/t_share/MTUxNzQ3MDEzNDQ2NzM5MDcw/image-placeholder-title.jpg>

 

[Uwe Bonnes]

jlarkin@highlandsniptechnology.com wrote:

On 12 Apr 2020 18:41:59 -0700, Winfield Hill <winfieldhill@yahoo.com
wrote:


I selected a Maxim MAX4239 opamp for a low-power design
project, for its 2.5uV offset, its Iq = 0.6mA, and for
G>10, its 6.5MHz GBW. For G=10, that's a 650kHz -3dB
bandwidth. Slew 1.6V/us. Response, t = 1/pi 650kHz =
0.5us. But the datasheet's 0.1% settling time spec is
500us, 1000x times slower! Makes little sense. Sad.
Dunno whether to believe a very poor datasheet number.

Never Buy Maxim!

The overload recovery time is milliseconds, so there's probably some
node inside that winds up when it rails or slews. It's an autozero,
and they can get messy. Intersil made one autozero opamp that took
*seconds* to recover.

I looked at AD8628 and LTC1052 and did not find any specification for
settling time neither....
--
Uwe Bonnes bon@elektron.ikp.physik.tu-darmstadt.de

Institut fuer Kernphysik Schlossgartenstrasse 9 64289 Darmstadt
--------- Tel. 06151 1623569 ------- Fax. 06151 1623305 ---------

 

[Winfield Hill]

Phil Hobbs wrote...

On 2020-04-12 21:41, Winfield Hill wrote:
I selected a Maxim MAX4239 opamp for a low-power design project, for
its 2.5uV offset, its Iq = 0.6mA, and for G>10, its 6.5MHz GBW. For
G=10, that's a 650kHz -3dB bandwidth. Slew 1.6V/us. Response, t =
1/pi 650kHz = 0.5us. But the datasheet's 0.1% settling time spec is
500us, 1000x times slower! Makes little sense. Sad. Dunno whether
to believe a very poor datasheet number.

That often happens when the vendor does some funny stuff
with the compensation. You wind up with a [low-frequency]
pole/zero pair that's closely spaced but not coincident.

Exactly. I solved the problem by adding a range switch,
which relaxed the Vos requirement by 10x. Then I selected
an LMP7711 opamp. 17MHz GBW, 150uV Vos, and an enable pin.
TI doesn't provide a settling-time spec for that part, but
it has a well-behaved GBW plot; I'm sure it'll be just fine.


--
Thanks,
- Win

 

[John Larkin]

On Sun, 12 Apr 2020 23:19:07 -0400, bitrex <user@example.net> wrote:

On 4/12/2020 10:40 PM, jlarkin@highlandsniptechnology.com wrote:
On Sun, 12 Apr 2020 22:24:36 -0400, bitrex <user@example.net> wrote:

On 4/12/2020 10:20 PM, jlarkin@highlandsniptechnology.com wrote:
On 12 Apr 2020 18:41:59 -0700, Winfield Hill <winfieldhill@yahoo.com
wrote:


I selected a Maxim MAX4239 opamp for a low-power design
project, for its 2.5uV offset, its Iq = 0.6mA, and for
G>10, its 6.5MHz GBW. For G=10, that's a 650kHz -3dB
bandwidth. Slew 1.6V/us. Response, t = 1/pi 650kHz =
0.5us. But the datasheet's 0.1% settling time spec is
500us, 1000x times slower! Makes little sense. Sad.
Dunno whether to believe a very poor datasheet number.

Never Buy Maxim!

I use the DS1077 i2c clock oscillator sometimes it's cheap and does the
business. But that's a Dallas Semi part to be fair.

Yikes, they measure frequency errors in per cent, not parts per
million. You can get a programmable crystal oscillator for less than
that.

It has a very long divider chain, I don't know what programmable xtal
oscillator for less money would be a good substitute with that
granularity that can be adjusted on the fly.

A heap of companies make them, PPB programmable over a huge range.
Check Digikey.

--

John Larkin Highland Technology, Inc
picosecond timing precision measurement

jlarkin att highlandtechnology dott com
http://www.highlandtechnology.com

 

[bitrex]

On 4/13/2020 3:37 PM, John Larkin wrote:

On Sun, 12 Apr 2020 23:19:07 -0400, bitrex <user@example.net> wrote:

On 4/12/2020 10:40 PM, jlarkin@highlandsniptechnology.com wrote:
On Sun, 12 Apr 2020 22:24:36 -0400, bitrex <user@example.net> wrote:

On 4/12/2020 10:20 PM, jlarkin@highlandsniptechnology.com wrote:
On 12 Apr 2020 18:41:59 -0700, Winfield Hill <winfieldhill@yahoo.com
wrote:


I selected a Maxim MAX4239 opamp for a low-power design
project, for its 2.5uV offset, its Iq = 0.6mA, and for
G>10, its 6.5MHz GBW. For G=10, that's a 650kHz -3dB
bandwidth. Slew 1.6V/us. Response, t = 1/pi 650kHz =
0.5us. But the datasheet's 0.1% settling time spec is
500us, 1000x times slower! Makes little sense. Sad.
Dunno whether to believe a very poor datasheet number.

Never Buy Maxim!

I use the DS1077 i2c clock oscillator sometimes it's cheap and does the
business. But that's a Dallas Semi part to be fair.

Yikes, they measure frequency errors in per cent, not parts per
million. You can get a programmable crystal oscillator for less than
that.

It has a very long divider chain, I don't know what programmable xtal
oscillator for less money would be a good substitute with that
granularity that can be adjusted on the fly.


A heap of companies make them, PPB programmable over a huge range.
Check Digikey.

and i2c bus? Not SPI or some bespoke serial interface?

I like my library code for quick turn-arounds. I2C, button-debouncing,
LED display driver, even rotary encoders have their own libs

 

Looks like a STM32G0 will be cheaper and have simelar specs wrt jitter and accuracy

You the get the rest of the micro for free

Cheers

Klaus

 

On Mon, 13 Apr 2020 16:03:53 -0400, bitrex <user@example.net> wrote:

On 4/13/2020 3:37 PM, John Larkin wrote:
On Sun, 12 Apr 2020 23:19:07 -0400, bitrex <user@example.net> wrote:

On 4/12/2020 10:40 PM, jlarkin@highlandsniptechnology.com wrote:
On Sun, 12 Apr 2020 22:24:36 -0400, bitrex <user@example.net> wrote:

On 4/12/2020 10:20 PM, jlarkin@highlandsniptechnology.com wrote:
On 12 Apr 2020 18:41:59 -0700, Winfield Hill <winfieldhill@yahoo.com
wrote:


I selected a Maxim MAX4239 opamp for a low-power design
project, for its 2.5uV offset, its Iq = 0.6mA, and for
G>10, its 6.5MHz GBW. For G=10, that's a 650kHz -3dB
bandwidth. Slew 1.6V/us. Response, t = 1/pi 650kHz =
0.5us. But the datasheet's 0.1% settling time spec is
500us, 1000x times slower! Makes little sense. Sad.
Dunno whether to believe a very poor datasheet number.

Never Buy Maxim!

I use the DS1077 i2c clock oscillator sometimes it's cheap and does the
business. But that's a Dallas Semi part to be fair.

Yikes, they measure frequency errors in per cent, not parts per
million. You can get a programmable crystal oscillator for less than
that.

It has a very long divider chain, I don't know what programmable xtal
oscillator for less money would be a good substitute with that
granularity that can be adjusted on the fly.


A heap of companies make them, PPB programmable over a huge range.
Check Digikey.


and i2c bus? Not SPI or some bespoke serial interface?

I have an idea: check Digikey!



--

John Larkin Highland Technology, Inc

Science teaches us to doubt.

Claude Bernard

 

[Michael Kellett]

On 13/04/2020 02:41, Winfield Hill wrote:

I selected a Maxim MAX4239 opamp for a low-power design
project, for its 2.5uV offset, its Iq = 0.6mA, and for
G>10, its 6.5MHz GBW. For G=10, that's a 650kHz -3dB
bandwidth. Slew 1.6V/us. Response, t = 1/pi 650kHz =
0.5us. But the datasheet's 0.1% settling time spec is
500us, 1000x times slower! Makes little sense. Sad.
Dunno whether to believe a very poor datasheet number.

Is the Microchip MCP6V91 any use to you - the specs don't compare very
directly but the Microchip part looks a lot better on recovery time.

And Microchip are much nicer to buy from

MK

 

[Winfield Hill]

Michael Kellett wrote...

On 13/04/2020 02:41, Winfield Hill wrote:
I selected a Maxim MAX4239 opamp for a low-power design
project, for its 2.5uV offset, its Iq = 0.6mA, and for
G>10, its 6.5MHz GBW. For G=10, that's a 650kHz -3dB
bandwidth. Slew 1.6V/us. Response, t = 1/pi 650kHz =
0.5us. But the datasheet's 0.1% settling time spec is
500us, 1000x times slower! Makes little sense. Sad.
Dunno whether to believe a very poor datasheet number.

Is the Microchip MCP6V91 any use to you - the specs don't
compare very directly but the Microchip part looks a lot
better on recovery time.

Hey, nice part, I hadn't found that one! Oops, no Enable
pin. :-( Also, while the 65us settling/step-recovery is
much better than the Maxim part, it's still rather long.
Anyway, I solved the problem by adding a range switch,
allowing a higher Vos spec, and avoiding a chopper amp.

> And Microchip are much nicer to buy from

I agree, I'm a huge Microchip fan.


--
Thanks,
- Win