AD620 as a voltage to current converter

Originally posted by peter@siemens in the EngineerZone feedback forum (before the Amplifier forum was live).


We are trying to build avoltage to  current output stage for a new product with the above device. Input -5v  to +5v required output current -12.5mA to +12.5mA

A  colleague built a prototype unit based on the circuit in the data sheet  and the performance was very good. I laid out a board using SOIC  devices and the +ve performance was good but the -ve would only remain  stable to about 12.3mA after that, if set to 12.5, it would slowly decay  on what appears to be a temerature related factor as cooling the device  restores the output. I laid out a further board for PDIP parts and  tested these - again, negative output decays.

The  gain pins on the device are Open Circuit giving a gain of 1 and the  input scaling resistors are pretty low - 2k7 in series and 1k shunt.

One  theory I had - the input comes into pin 3 (non inverting) - Pin 2 is  grounded - should this go via a series resistor or shouldn't that make  any difference?

When I used the device from my colleagues prototype, it worked fine - swap back and the problem is still there.

Do you have any thoughts?



  • 0
    •  Analog Employees 
    on Feb 15, 2011 4:12 AM

    Hi Peter,

    Depending on your load, it may be that you are getting a little too close to the AD620's output voltage compliance limit.  Let's say you have a 1 kohm load.  This generates a -12.5V drop across the load resistor.  Another -1.25V drop occurs across the current setting resistor.  This means the AD620 should supply -13.75V at its output:  In other words, 1.25V from negative rail when using +/-15V supplies.

    The AD620's output swing is specified at 1.2V from the negative supply at room temperature.  So it initially looks like the AD620 should meet the spec.  However this spec is with a 10 kohm load:  i.e. with about 1.4 mA flowing.  Like any amplifier, the AD620 can't get as close to the supplies with heavier loads.

    We have two typical performance curves in the datasheet that apply:  figures 22 and 23.  Your load of -12.5 mA at -13.75V is equivalent to a 1.1 kohm load.  We can see from figure 22 that a typical part with a 2 k ohm load is already pretty much at the limit your circuit requires.  A 1.1 kohm load will be worse.  We can also see from figure 23 that their is a noticable degradation in output swing from 2 kohm to 1 kohm load.

    If the problem goes away when you use your current source to drive a 0 ohm impedance this is likely the problem.  You could get around this issue by using larger supplies (say +/-16V).  Decreasing the value of your current setting resistor will also help.

    If the problem is still there even when you drive a 0 ohm impedance, let me know.


  • Thanks for that Matt

    The problem isn't load dependant - even with the output connected directly to the DVM so pretty much a zero ohm load, the output sags. I've tried varying the supply and it appears to be worse at higher voltages leading me to suspect maybe a resistance in the negative drive transistor within the device?

    I think that the output approaching the rail for 1k load is going to be a different issue (the spec for 1K is pretty much an arbitary figure) which I think the customer may well conceed is a little pessamistic (in real life, its likely to be closer to a couple of hundred ohms or so).


  • 0
    •  Analog Employees 
    on Feb 15, 2011 7:48 PM

    Hi Peter,

    I talked to an in amp designer about this question.  He mentioned that it might be possible that the short circuit current limit is engaging, although he would be surprised.  The short circuit current limit is something we spec as a typical number, and not something we test.  We would expect this number to normally fall within +/-20% of the 18 mA we give in the datasheet.  However the AD620 was released about 20 years ago.  It is possible in the years since it was released, this short circuit current value has slowly drifted so that on some parts it might be reaching the 12 1/2 mA area you are now seeing.  Since we do not test this parameter, we might not have noticed.  We use a feature of the process in the short circuit current limiting circuit that we don't use elsewhere and wouldn't be caught by other tests.  However if a typical parameter has changed substantially we often hear feedback from our customers -  and this is the first I have heard of this issue .

    If it was the short circuit limit, it would also make sense that increasing the supply voltage would make the issue a little worse.  A larger supply voltage would cause the AD620 to dissipate more heat (both because quiescient power will go up as well as the power dissipation due to the current you are sinking).  A rise in temperature will cause the short circuit limit to kick in slightly earlier.

    One simple way to check:   short both input and the reference of the AD620 in question to ground.  Hook up a resistor (say 500 ohms) between the output of the AD620 and a variable voltage source.  Measure the output of the AD620 while varying the voltage source.  If the AD620 can't maintain the 0V output with -12.5 mA running out of it, then you know you are indeed running into a short circuit issue.

    You may want to consider our AD8270 through AD8274 line of difference amplifiers.  These amplifiers have much more output current capability (60 mA) so you won't need to operate close to the edge.  Because it is a difference amplifier, your voltage divider becomes a little more tricky - you now have to compute the effect of the input impedance of the difference amplifier.  You may be able to use one of the gains these amplifies provide, since some of them attenuate.  You could get around this by buffering the diff amp with a dual op amp.  Or if you really just need single ended input you could use table 9 on page 17 of the AD8271 datasheet (or just use a high power output op amp.)

    Matt Duff

    (Instrumentation Amplifier Applications Engineer)

  • Thanks Matt

    I performed the test as you described and can confirm that the suspect devices are unable to maintain the output at zero volts with -12.5mA output - the 2 good devices I have are able to operate at this output current and must therefore conclude that as you suggested, the spec on the current limit circuit has changed over the years. I guess this varies from batch to batch or, mask set to mask set - the good devices have one code on them after the part number and the bad devices have a different code but without characterising devices before purchase, it could work out as a very expensive exercise!

    do you know what the number means? Good are #1008 and the bad ones are #1018 (I'm guessing its the week number of 2010 when they were manufactured?).

    Thanks for your help - looking like its going to have to be a modified Howland design after all :-(


  • 0
    •  Analog Employees 
    on Feb 16, 2011 2:07 AM

    Hi Peter,

    Well at least we know what is going on.

    You are correct about the date codes.  #1008 is the 8th week of 2010, #1018 is the 18th week of 2010.