Post Go back to editing

LTC1668 Gain Error

In a circuit with a LTC1668 driving a differential buffer stage, and using precision (0.1%) resistors, I'm seeing an output gain error of -4% (there's small variation from unit to unit, but the error is always close to -4%).  I've seen the same error across three different revisions of the PCB, on 15 different individual assemblies, both hand built and machine built.

The error isn't a problem - I can very easily compensate for it - but I'd like to get to the bottom of its source.

This is the buffer stage:

Not shown is the LTC1668 reference input, which is driven externally by a LTC6655-4.096 (Vref) through a 0.1% 3.32k resistor (Rref).

The buffer stage is the same topology as the buffer in figure 7 of the LTC1668 data sheet, but the component values are different.

The amplifier is an ADA4637-1.

Given all that, with a DAC code of D:

The factor of 8 is taken from the LTC1668 datasheet, the rest is from basic circuit analysis.  RV = 49.9R, RI = 1k, RF = 6.81k at the moment, but the exact values aren't important.  I've used a couple of values for RF during testing, but the error is always -4%.

Anne know what the explanation is, or seen the same issue?

  • What is the frequency content of the signal are you outputting?  When you say -4% you mean that the voltage you are measuring is always 4% lower that what you expect?  If you change the amplifier to a different amplifier, does the error stay with the DAC or with the amplifier?  Does the new amplifier also produce a -4% error, or is it a different number.   What is the output of the amplifier driving?  All of the resistors are 0.1% resistors?  How tightly are the power supplies regulated?  Can I see a layout and schematic?

  • Hi Clarence,

    Thanks for the response.

    Yes, the output is 0.96 times what I expect, across all the frequencies at which I've tested it (from dc to about 20kHz in this application).

    I haven't tried with a different amplifier, but if I take voltage readings at VA and VB, the output of the amplifier is what you'd expect, and it looks as though the currents at IOUTA and IOUTB are lower than they're expected to be. (And I assume that you mean a different TYPE of amplifier; as I pointed out in the post, I've seen the effect on numerous different prototype boards).

    I've also used LTSpice to verify that the amplifier operates as expected with the component values I'm using and the voltages (and therefore currents) I'm seeing at VA and VB.

    The amplifier drives a high impedance load but, as I've described in the previous paragraph, the output of the amplifier is consistent with the voltages at VA and VB.

    The supplies are very well regulated; the positive supply is regulated by a LT3065 and the negative supply by a LT3090.  The board is four-layer, and very comprehensively decoupled.

    No, I'm not going to publish the layout and schematic on a public forum :o)

    My post here has followed a long running and fruitless attempt to get to the bottom of the issue through technical support, who HAVE had a copy of the schematic (but not the layout).  The support ticket number was SR-91665-N2Y6W5, if that helps.

  • How are you measuring this 4% error?  Is it that you are getting 1.036V and the datasheet says it should be 1V?  

  • The output I measure is not the same as the output I predict using the transfer function I showed in the first part of the post.  So, using your arbitrary levels, where I would expect 1.00V out, I actually see 0.96V.

    There is, of course, a distribution of gain values, since there's a tolerance of 1% on the gain of the LTC1668 (most other error sources in the circuit are much smaller than that, so I'd expect the DAC gain error to be the dominant factor).  But all the units I have are consistently producing an output that's _about_ 0.96 times the expected output.

    The transfer function is calculated using equation (1) in the LTC1668 datasheet to determine the DAC's differential output current, and then straightforward circuit analysis to scale that with the differential buffer stage.

  • Hi guys,

    Just a quick check up on this. What is the output of LTC6655 showing and what is the voltage at the "other" side of the Rref resistor ?



  • Hi Rob,

    The LTC6655 output, measured at the end of the Rref resistor, is 4.0967V.  At the 'other' end of the resistor (pin 16 of the LTC1668), the voltage is -3.7mV.  Both readings taken wrt 0V at pins 17 and 18.



  • I believe the error is given by the internal 1.1k resistor that is at each output of the DAC, that changes the RV actual value.

  • I believe the error is given by the internal 1.1k resistor that is at each output of the DAC, that changes the RV actual value.

  • That looks about right.  I hadn't re-visited this since the last previous post.

    The relevant section of the datasheet (for the benefit of anyone reading the thread and assessing LTC1668 performance) is figure 3 and the "LADCOM" description beneath it.