Post Go back to editing

Weird overshoot LTC2641

Category: Hardware
Product Number: LTC2641

Hi, 

I am testing a new prototype board using the LTC2641 - 16 bit DAC. I am sending the DAC a single command asking for 0.5v for 50ms, then a clear signal is sent. The pulse is repeated every second for initial testing  

The problem I am having is that the output is not a square wave with a 1x probe or ez clips/bnc cable. See attached image. The output has some overshoot but it is atypical overshoot IMO. With a 10x probe, the output looks more like a square wave. I have tested several different leads and several different oscilloscopes/ADCs and the results are the same. 

To isolate the problem, U1 and C2 are not populated. I have tried shorting R2 and various other values with no change. 

1x 
10x 


(Cropped) Board schematic:
Parents
  • Hi  , 

    Not sure what you mean by the output is not a square wave. Is it because of the oscillations on the first plot? Those could be caused by loading effect of the probes. Different probe type and configurations have varying load effects. One comment on your schematic is that you should ut 0.1uF bypass caps near the Vdd and Ref pins. 

    I agree that is just some typical overshoot you're seeing. You may want to try using a digitizer instead of an oscilloscope for a lesser loading effect. 

    And does this overshoot propagate to the output of your amplifier?

    Best regards,

    Ian

  • I have also used a 32bit dev board to measure the signal and I get the same waveform as the 1x probe on the oscilloscope. 

    I thought it was a loading issue caused by the measurement of the dac output, but the waveform is the same when measured on the right hand side of the circuit (other side of the op amp). The op amp alone should not be enough to over load the dac and the waveform is the same with and without the op amp in its socket. 

    I also tried keeping the Dac off and connecting the equivalent square wave pulse with an arbitrary waveform generator to the dac test point. The other side of the op amp circuit then reads a very clean square wave which is what is what pointed me to the Dac as the root cause. 

    One more piece of data: as I increase the dac output, the oscillation magnitude stays the same but the noise increases, to the point that the 2.5v looks like it doesn’t have the oscillation but the noise is really just the amplitude of the oscillation and drowns out the oscillation. 

    I believe I already have the bypass caps you are talking about at Vdd and VRef. I am hoping the fix is something like this though. 

  • Interesting. I had a .1uf and a 11uf cap on the input of both the Vdd and Vref for the dac. I removed the 11uf and that removed quite a bit of the oscillating. I still have a single large dip. It looks like the initial rise time is very good, then the dip lasts an additional 100us. 



    I also still have the noise issue where the higher the Dac output voltage, the higher the noise I see. 

  • Removing the bulk capacitor 11uF, you lose a lot of energy on the transient response. Hence the large dip that lasts for some time. I was looking the reference datasheet and 1uF in parallel with the 0.1uF is recommended. 

    One more thing that could contribute to the noise is the ground return path. There may be other circuitry that couples the noise into the ground. You could try measuring to a point near the GND pin of the DAC. 

    Does this happen on other units as well? 

  • The cap(s) on the Vref line are the driver of some, if not all, the oscillation. I ran through a bunch of values and it looks like the lower the value, the higher the amplitude of the single dip down to ~10nF (where things get weird) and then up to 100nF where above that you see overshoot after the first dip. I’d love to eliminate the initial dip as well though…

    The noise issue was my fault. Each time I increased the dac output, my oscilloscope was changing ranges, which in turn looked like more noise on the signal but was actually just more noise in the measurement. 






  • Hi,

    This is still unresolved. I have monitored the DAC output vs all of the other points in the circuit. The issue does not align with any digital signals (CS, etc). The power and ground lines are stable. The only other location where this issue is apparent is on the Vref line. See image - orange is DAC output, Blue is Vref.

    This issue is apparent over a large number of tested probes, oscilloscopes, ADCs, and power sources.

  • Hi, 

    Looks to me that the issue is with the transient response of the reference. One way to solve this is as you've tried, experimenting with values of the capacitors and also adding damping resistor in series to try and dampen the step response. you could try it between the vref output and the ref input of the DAC or on the output of the DAC. 10-100 ohms should suffice depending on how much damping is allowed on your output square wave's edge. 

    Best regards,

    Ian

  • Hi Ian,


    Thanks for the response. I did some more testing today and I added a big cap to the Vref line (1000uF aluminum). This stabilized the Vref line as well as the DAC output. I had previously tried bigger cap but nothing this big. Although this fixes the issue, I feel like I am putting a band-aid on an unknown issue.

    I have maximized the output rise time allowable, but I will try a resistor between the DAC and Vref IC.

    I have been using ADR4525 as my voltage reference. Maybe this is incompatible with LTC2641. To test this, I have ordered LT1019, the reference voltage IC mentioned in the LTC2641 datasheet. Luckily the pin out of the two references are the same.

    Here is the vref line (blue 2.5v) and the DAC output with the 1000uF + 1uF + .1uF capacitors on Vref.

  • That is quite huge for a vref capacitor. There must be some kind of loading effect introduced to the ref output by the ref input of the DAC but I can't find that in the datasheet. You could check with a current probe to see that. In any case, other form of solution could be a buffer in between ref and the DAC.

    Your transient response has been significantly dampened, hope this is still OK in your application. 

  • I changed out the reference voltage ADR4525 for LTC1019 and the problem went away. I can now use the recommended 1uF and .1uF caps and get datasheet performance from the LTC2641.

    Thanks for you help.

Reply Children
No Data