ADA4817 differential probe

It is my first project with a high speed op amp, so I would like to ask a few questions about using the ADA4817 before starting the prototype, to avoid basic errors.

My project consists of reproducing a differential probe with characteristics similar to Keysight's N2792A tip (10:1, 200 MHz, output ± 2 V, 1 MΩ // 3.5 pF between inputs, differential ± 20 V, common mode ± 60 V).

This design I believe consists of three stages:

1) Input attenuator;
2) Non inverters for buffering;
3) Subtractor, converting the differential inputs into a single output;

To be able to go ahead and make the first prototype are, I wish to know:

1) Can I remain with unit gain on the preamps and on the output subtractor, and perform all necessary attenuation only on the RC input attenuator? I see that ADA4817 its unity gain stable, my -3dB operating frequency its on the range for the signal's amplitude, feedback resistor and crossover frequency checked.

2) Can I have ADA4817 non-inverters with unit gain, using 200R feedback resistors and without the gain resistor Rg (like datasheet states at page 27, just after equation 16)?

3) Considering that the maximum output amplitude happens only when a signal at +20 V and another at -20 V, so the subtractor will perform output 1 - (- 1) = 2 V, can I consider that the circuit operates most of the time with small signal characteristics? In addition, the datasheet states that the Input Common-Mode Voltage Range from −VS to (+ VS - 2.8). My source will be ± 5 V, so I think its ok in that point.

4) Is there a problem with using low values ​​like 350R in the subtractor resistors?

5) The feedback pin FB on the ADA4817 packages it's only a clever way to route feedback signal from the output? It's a good idea to use it for feedback traces?


    •  Analog Employees 
    on Aug 19, 2020 7:42 AM 3 months ago


    Thank you for  having interest on ADA4817 for your project.

    Here is my response to your queries and hoping this will help.

    1. Yes you can have your signal conditioning at the Voltage Divider Network (VDN). Your target attenuation is 1/20, and you have used 500k:25k. To attenuate the signal 20X, the R values should be 1/19 since you will use the formula (R2/R1 + R2) as an input to your unity amplifier. Based on Figure 11 of the TPCs, 200MHz is at the -3dB but it is recommended to have your application at flat point and usually 1 decade below the -3dB point as a rule of thumb.

    2. The Rg in the data sheet of figure 66 which have labels Rs1 and Rs2 may serve two purposes: a) to serve as current limiting in the event
    of Input Over Voltage and b) as compensation resistor to have balance Offset Current between two input pins. From your circuit,
    you have this equivalent R as the parallel combination of 500K and 25K. The Bias current of the ADA4817 is relatively small, in picoamps level.
    It will introduce small DC error but if your design permits this DC error, then you can proceed.

    3. Your input Voltage to the amplifier will be 1V only as attenuated in your VDN and that is within specs as you have mentioned. However,
    having input frequency of 200MHz and 1Vp, that will require a Slew Rate(SR) of 1256V/uS which is quite high compared to the SR of ADA4817.

    4. The parameter that will look into this is if the amplifier can drive this load. Based on the Spec table, for RL of 100 ohms,
    your Output Voltage High is Vs – 1.5V and therefore, we can still achieve your target output voltage of 2V.
    Another thing you need to consider on your subtractor stage is the matching of these resistors to achieve high CMRR.
    Here is a good read in this link using composite amplifiers, and using discrete Diff Amp on the second stage.

    5. The FB pin of the ADA4817 is an additional feature to ease the layout of feedback. This will help to remove additional parasitic that the
    layout trace may induce. 

    Thanks and regards.

  • Thanks for your answer Fdelaram. 

    I've thought on what you wrote.  I've seen a lot of 200 MHz differential probes with -3 dB point at this frequency. I understand what you mean. I would like to have a that in the flat region too. But have the max frequency 1 decade before the -3 dB point was the hard to read. Really a great advice. 

    Besides all, the slew rate limitation was the worst to me. I can change my Voltage Divider Network to work at small signals level in pre amps, staying within ADA4817 Slew Rate limits (870 V/us and 200 MHz limits to theoretical 0,69 V) . But at the last stage, I'll need to diff the signals and apply some gain on the result to achieve ±2 Vp to drive the oscilloscope input. The idea of a discrete diff amp is amazing (with discrete resistors), but I can't find one part that can drive the signal ±2 Vp at 200 MHz with the gain. Do you know if there's other solution to make this output driving?

    Again I'm referencing the design on comercial probes specifications, where ±2 Vp its typical. Maybe if a reduce it to small signal and tell the scope that its a 100:1 probe resolves the question. But I still want to know how they do it on professional probes. What do you think they do?

    I'll study using LTC6268-10 at diff stage too. 

    •  Analog Employees 
    on Aug 20, 2020 7:47 AM 3 months ago in reply to forrequi


    Yes it is quite hard to find a discrete diff amp with wideband as 200MHz.

    You can opt to use ADA4817 but I suggest, you have to use matched resistors network rather than discrete one.
    You can check LT5400 for quad matched resistor network.

    The LT5400-7 has 1:4 ratio, you can optimized this part to have a gain of 4.

    Hope this helps.