ADA4096-2 & the Filter Wizard

Hi,

After using the filter wizard for a design, I am finding some ambiguity in the results of the software, vs data sheet specifications. Or probably a misunderstanding on the results on my part.

The ADA4096 data sheet indicates an input overload swing of +/-32 Vpp. However, when selecting this part at the front end of an active filter design, when viewing "Voltage Range", it indicates a maximum input of 50mVpp, and 33mVpp nominal.

Are these numbers with respect to boundaries of filter performance, or input voltage that may damage the amplifier? The filter is a 40dB gain, 7th order Chebyshev. A, B stage = ADA4096, C,D stage = ADA4051.

Regards and thanks.

  • 0
    •  Analog Employees 
    on Mar 9, 2014 6:04 PM

    groger,

    I'd like to attempt to clarify some of the Filter Wizard results, using a simpler design as an illustration.  If you still have questions about your specific design, please feel free to follow-up with those questions.

    Let's look at the following design in Filter Wizard: a low-pass filter, leaving all of the fields in the "Specifications" tab at their defaults, except changing gain to "40 dB."

    Now, looking at the "Stages" drop-down view in the "Component Selection" tab:

    Here, you can see the targeted specifications for each stage of your design.  So, in this example, we have two stages - Stage A targets a gain of 6.44 V/V, and stage B targets a gain of 15.54 V/V.

    Now, take a look at the "Voltage Range" drop down view (still in the "Component Selection" tab):

    The "Amplifier Specs" section shows you the datasheet specs for the selected amplifiers (with the voltage supplies you're using - in this example, +/- 5V) - slew rate, max input range, max output range.

    Two things to notice:

    1. The "warning" icons indicate limiting factor in your design, with regards to signal range.  In this example design, the output of stage B is maxed out.
    2. At the top of this view is "Maximum Input Signal Possible:", which is 0.10Vpp in this example design.

    So, simply put: Harry is correct.  You've set up a circuit with a gain of 100, so assuming you're running on rails of say, +/- 10V, the output of your circuit can never exceed 20Vpp.  Which means your circuit won't operate correctly if you drive it with anything larger than approx. 200mVpp.

    The specs for staying within the operation range of your circuit is different then the specs in the datasheet that indicate the voltage range that will do irreversible damage to your device.  Filter Wizard doesn't report any "smoke point" specs. 

    As busy as the "Voltage Range" view looks, it gets more busy (and more informative) when you move to the "Component Tolerances" tab.

    There is now lots of information about "Worst Case."  What is that?

    Simply put, Worst Case specifications take into account the variations in your design due to device tolerances.  You can adjust the tolerances, but in the example, I'm showing the 5% capacitor, 0.5% resistor, and 20% GBW tolerances.  These variations will cause your design to vary in it's intended parameters, including gain, Q, cutoff frequency, etc.

    In this example, the gain of the circuit varies pretty significantly due to tolerance:

    So, even though the target gain is 40dB, there are variations in the components that could result in 45dB gain (178 V/V).

    In the worst case scenario, your output swing on the last stage is still 9.6 Vpp.  But, now considering that it's possible that your circuit could have a gain as high as 45dB, you run the risk of overdriving your design if you drive with a input signal any greater than 52 mVpp (9.6 Vpp / 178 = approx 52 mVpp).

    If you are still having trouble understanding the Voltage Range information reported by Filter Wizard, start by trying to understand it in the Component Selection tab first.  Adding the variation due to tolerances is a very powerful way to see "real world" design issues, but is certainly more information to digest.

    I hope this example was helpful in understanding your specific design.  Good luck with your filter design!

    -Anne

  • Thanks Anne, harryH.

    Your replies were insightful and helped in better understanding the filter tool.

    Regards,

    Gary

  • Hi,

    After a week or 2 to think about the overall design a bit more, another question arises - this one is a bit more to do with the supporting cast around the filter.

    The incoming signals could be quite weak, as low as 60uV. In addition, an AD8307 log amp will be included in the processing chain, so I do not want to have any gain in front of that (for obvious reasons), but would like to use a BP filter to ensure the log amp is responding just to frequencies of interest.

    So the question is: would it be better to amplify the signal, then filter, or the other way around (keeping in mind the additional use of the AD8307 and it's need for a no pre-gain signal)

    BTW - the "amplify" is referring to amplifying in preparation to drive an ADC. The AD8307 is a separate process for the microcontroller.

    Hope this question fits here - if not, I could move it to another forum.

    Thanks!

  • 0
    •  Analog Employees 
    on Apr 9, 2014 10:16 PM

    Hi groger,

    The low end of the signal range before any amplification is 60 uV (is this RMS?) or about -84 dBV or -71 dBm.  What is the expected high end of the signal range?  The reason I ask is the -71 dBm is on the low end of the AD8307's dynamic range, if you need to amplifier then filter, and you tapped off the signal right after the amp and put it into the AD8307, as long as you knew the gain and the biggest signal was with the dynamic range of the AD8307, you could calibrate out the gain. 

    For example, if you had a 20 dB gain block, to figure out what the input power was, all you would have to do is subtract 0.5 V from whatever the voltage on the OUT pin is reporting (the slope is 25 mV/dB, times 20 dB yields 0.5V). 

    Also, as i think you alluded to, this log amp will detect any signal within its bandwidth.  So if you have unwanted signals within the part's bandwidth, they will need to be filtered out for proper measurement.

    Hope this helps,

    Joel

  • Ok, thanks for the helpful reply.

    I think the AD8307 datasheet indicates the range is to -75dBm (56uV) which would work for the intended app. it can be shifted with a matching network to -88dBm.

    I guess there were 2 distinct questions which I should have delineated. The one I really need to know is regarding the performance of a non-amplified bandpass filter working at these levels. Because of the addition of the AD8307, I changed my filter design to 0dB gain, with maximal stop bands (40dB) I used the filter wizard to design it - but no indication of just what the sensitivity is. I selected a low noise option, and with good board layout and careful interfacing, I'm assuming I should get the specified noise of 700nV RMS. I will still use a unity gain buffer at the front end to the sensor. So, is it still prudent to amplify the signal for optimal filter performance, or should it perform just as well with a (i.e.,) 60uV signal as it would a 60mV signal?

    Thanks

    Gary