Op Amp replacement

beenthere

Joined Apr 20, 2004
15,819
There are a great number of instrumentation amps available. Can you give some parameters to work from? If it does what you want, $9.48 isn't too bad.
 

Wendy

Joined Mar 24, 2008
23,415
Unless the circuit takes 20 or so...

The fact you want a drop in replacement suggests you are working with a comercial or finished circuit. You may want to install sockets and use direct replacements if this is the case.
 

SgtWookie

Joined Jul 17, 2007
22,230
They're under $10 each - I don't call that extremely expensive.

An Ebay seller I buy from occasionally has a lot of 4 for $25+$3.25 shipping, or $6.25 ($7.06 w/shipping) each. If you need more, ask for a custom listing. They currently have around 80 of them.
http://cgi.ebay.com/INA114AP-INA114...VQQcmdZViewItemQQ_trksidZp1638Q2em118Q2el1247
If you use a lot of 'em, better stock up. TI currently has none in inventory, and vendors are starting to run low on stock.
 

SgtWookie

Joined Jul 17, 2007
22,230
I beg to differ with you, Bill.

Using sockets with instrumentation amps is not recommended. Even a very slight difference in resistances of the socket connections will result in very degraded CMRR.
 

Thread Starter

DC_Kid

Joined Feb 25, 2008
1,072
the board is a new design and is about to be sent off to my board maker for a proto run of 12 boards (roughly 6"x3" each). this rev of design uses six op-amps with about 100 components total on the board. it will be sold in the $150-$200 price range when finished so i need to keep costs down to a minimum (op-amps alone are roughly half that already), hence why i was wondering about a less expensive op-amp. i can easily make changes to my board design to accomodate a new op-amp, etc. since this is a automotive application the CMR is important. the 115db CMR of INA114 is a nice spec, but 100+db would suffice.

i'm also glad SgtWookie mentioned sockets (or issues with). for proto purposes i was gonna use dip sockets for the amp, but i need to re-evaluate this idea. my circuit might not be that sensitive so using dip sockets may be ok. since i will most likely move some stuff around on the board before sending it off for a production run so using sockets on proto boards will allow me to re-use the amps on the production boards thus saving $$$.
 

beenthere

Joined Apr 20, 2004
15,819
Have you gone to sites and searched for IA's using the CMRR as a parameter? Through-hole stuff is getting a bit scarce these days, too.
 

SgtWookie

Joined Jul 17, 2007
22,230
What's your gain?
What's your CMRR requirement?

You might take a look at INA128/INA129, which are largely pin for pin compatible. CMRR is better than the INA114.
Linear Technology's LT1167, LT1168, LT1920 have attractive specs; with higher gain CMRR gets up to 140.
http://www.linear.com/pc/viewCategory.jsp?navId=H0,C1,C1154,C1009,C1045
Analog Devices has a slew of instrumentation amps.
http://www.analog.com/en/

Everything's a trade-off. But using sockets for an instrumentation amplifier is sort of like putting truck tires on a Ferarri.
 

Thread Starter

DC_Kid

Joined Feb 25, 2008
1,072
thanks SgtWookie, i dunno why i missed the INA128/129 units. 128 is a few $ cheaper than 114 and does the same thing. my gain is only 100, which in my case will cause the output to saturate to supply voltage (+5v) and is ok in my application because i am using the output as a master reset to a 4 bit counter IC. amp is being used to sense a current pulse (approx 1amp) across a 1% 0.05ohm "air" resistor.

i'm also glad you guys had me looking at the data sheets again as i found my circuit did not account for the required input bias return path. i'm not a op-amp genius but i think i can follow whats required for the bias path. since i have a 0.05ohm resistor sitting across the inputs of the amp the source impedance is considered to be low, so i can use just one resistor to bias the + input to ground. the resistor value is what i'm not so clear on. is a 47k ok?
 

SgtWookie

Joined Jul 17, 2007
22,230
The input impedance of the 128/129 is so high that you could use just about any value.

But just to play the devil's advocate... :D

Using a bias resistor to ground on just one side of the sensor will affect your CMRR by the ratio of sensor resistance / bias resistance.

Bias current is in the vicinity of 2nA, so you have a good bit of room to finagle. A 1 meg bias resistor would result in a 2mV offset, 100k = 0.2mV, etc. Basically, you're just looking to keep the inputs from floating so that the amp doesn't saturate.

You may want to experiment with different values to find the best trade-off.
 

Thread Starter

DC_Kid

Joined Feb 25, 2008
1,072
in the datasheet they show 3 examples of biasing. my circuit is very close to the thermocouple example. .05 ohm sitting across the amp inputs looks almost like a piece of copper wire, very low impedance. they only use one resistor to bias the inputs in that tc example. a few mV offset on the output is not going to make a diff in my case as the master reset is TTL level and anything around 5v will trip the reset. you still think i need a resistor to ground on each amp input?
 

SgtWookie

Joined Jul 17, 2007
22,230
Not really.

It wouldn't hurt to provide a pad for an SMT resistor on each side as a just-in-case, particularly on a proto board - unless you're really tight for space. Sometimes such extra touches are the difference between "usable" and "scrap".

I hope your analog and TTL sides are well isolated from each other, or your life will get unpleasant very quickly - sort of like having a punk band perform in a library. :eek:
 

Thread Starter

DC_Kid

Joined Feb 25, 2008
1,072
I hope your analog and TTL sides are well isolated from each other, or your life will get unpleasant very quickly - sort of like having a punk band perform in a library. :eek:
since my design has already been put down on a board space is very tight, but i can make some minor adjustments to accommodate two resistors per amp.

as for the isolation, the amp output gets tied directly to the MR of a 4bit counter IC. the amp is powered by a single supply (0v, +5v) so the max output of the amp cant go beyond 5v. my expected current through the current sense resistor will be approx 1amp and is driven by varying PWM. the PWM varies between 0-85% duty cycle between 20-60Hz. a 1amp pulse should result in a 0.05v drop across the current sense resistor and the amp should then multiply this diff by 100 yielding a short ~5v output which will reset the 4bit clock.

i am trying to minimize in-between isolation IC's for timing purposes. the current pulse is controlled by a external NPN transistor. i monitor the collector of this NPN, and when it goes low i flip the signal via a pFET which causes my clock to advance. if current is sensed then the clock gets reset. if NPN gets fired but no current is detected then i know something is wrong. i actually wait for bit "1" (numerical 2) goes high to call it a fault. i do this because of the possible time delay between NPN getting turned on and the current pulse. its possible for the clock to advance before the master reset occurs.

are you suggesting a opto-isolator of some sort between amp output and clock MR?
 

SgtWookie

Joined Jul 17, 2007
22,230
Opto-isolator = yes, that was precisely what I was thinking of.
Along with separation of supplies, ground planes being joined at only one point, and plenty of bypass caps.
 

Thread Starter

DC_Kid

Joined Feb 25, 2008
1,072
every amp has a .1uF poly and a 10kpF ceramic within 1mm of supply pin. the clocks get filtered once with the same pair. power is automotive voltage, comes through a muRata bnx002-01 filter which feeds LM7805. i only have one 7805. a ceramic and 220uF lytic hangs directly on the 7805 output. the 7805 feeds the amps and clocks by way of multiple traces. to divide up currents on the ground side i have many traces that terminate at a central ground pad.
 

Attachments

Top