• State Resolved
  • Locked Locked
  • Replies 11 replies
  • Subscribers 50 subscribers
  • Views 1768 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

[High-Speed Amplifier Circuits]

Xuan Truong Nguyen
Prodigy 120 points
Other Parts Discussed in Thread: THS4001, THS3001, OPA847, LMH5401, THS4302, OPA695, OPA2695, LMH6702, LMH6629, VCA821, THS4303

Dear, 

We would like to design an amplifier circuit under conditions:

- Gain: +20

- Input signal: 50mV~1V with pulse width is ~10ns. Input can include a single pulse or some consecutive pulse.

Considering the problem, we have chosen some high-speed amplifier models:

THS4001  - www.ti.com/.../getliterature.tsp

THS3001 - www.ti.com/.../THS3001

OPA847 - www.ti.com/.../opa847.pdf

However, the shape of output signal did not maintain the input one as the falling edge of output slowly changed while falling edge of input is stiff. 

My questions are:

- What is the amplifier chip that is suite to our application?

- Is there any reference circuit that satisfies the given conditions?

Thank you very much. 

Regards, 

Truong

  • 0
    TI__Genius 12631 points
    Truong,
    Assuming a 1ns rise/fall time for a 1V signal, the required slew rate is 1000V/us. Also, because of the high gain that the part is in the required bandwidth in G=20 is 0.35/1ns = 350MHz. This equates to a gain bandwidth product (GBW) of 350MHz * 20 = 7GHz. We do not have a single-ended voltage feedback amplifier (VFA) that meets this requirements.

    1. One option is to use a fully-differential amplifier like the LMH5401.
    2. Use two cascaded VFAs. The THS4302 is an option here if you are willing to use 2 amplifiers.
    3. A current feedback amplifier (CFA) does not follow the GBW relationship like a VFA, however the LMH6702 and OPA695, which are our fastest amplifiers do not have sufficient BW at that gain so again you will need to use a cascaded set of amplifiers. The OPA2695 is a dual channel amplifier in a single package so that may work.

    You mentioned that your input signal is 1V which in a gain of 20 will be 20V output. However you also said that you used the OPA847 which can only work on 12V max supplies, so Im not sure if you meant 50mV input * 20 = 1V. Can you please clarify?

    If you do need a 20V output signal then the required slew-rate is 20V/1ns = 20000V/us. We unfortunately do not have any amplifier that will meet this spec.

    -Samir
  • 0 in reply to Samir Cherian
    Prodigy 120 points
    Samir,

    Thank you very much for your answer.
    Input Range is: 50mV-1V
    Output Range is: 1V-3.3V.
    Thus the Gain G = 20 actually is the maximum of gain.
    Now we have several THS4001 and OPAMP847, do you think we can make a cascaded VFAs?

    Thanks.
    --Truong
  • 0 in reply to Xuan Truong Nguyen
    TI__Genius 13935 points
    Hello Truong,

    The LMH6629 has a bandwidth of 900MHz at a gain of 10V/V. This may be suitable for your application. You could use an amplifier like the LMH6702 at a gain of 2 and then an LMH6629 at a gain of 10 for 20V/V.

    Keep in mind that any termination losses from the source or load will reduce overall gain. If you need more than 20V/V you can run the LMH6629 at a gain of 4 as the first stage. THe LMH6629 is not unity gain stable, it is best if used at gains of 4V/V or higher.

    Regards,
    Loren
  • 0 in reply to Loren Siebert 1
    TI__Genius 12631 points
    Truong,
    Do you need some kind of gain programmability? I am just looking at your numbers:

    Input Range is: 50mV-1V
    Output Range is: 1V-3.3V.

    This would suggest gains ranging from 20V/V to 3.3V/V. How do you plan on implementing this? The VCA821 is a programmable gain amplifier that could be used in case that is what you need. Of course you could always use switches/relays in your feedback path to program the gain but you would need to take great care as multiple switches and high-speed amplifiers don't always mix very well.

    -Samir
  • 0 in reply to Samir Cherian
    Prodigy 120 points
    Samir,

    Based on the specs, my plan is also to design in a cascaded manner: Amplifier + Discriminator (Comparator). And we got the problem with the amplifier level as I mentioned before. If I use VCA821, how do you think about the pulse-width factor (our input signal has very narrow width (10ns))? Actually, as far as I know, the OPAMP847 has G = +20 at 100MHz so I thought it satisfy our requirement. But the actual circuit did not work as what we expected.

    Thanks.
    --Truong
  • 0 in reply to Loren Siebert 1
    Prodigy 120 points
    Hello Loren,

    Thanks for your answer.
    Based on your analysis, the solution is to use two cascaded VFAs.
    Assuming that we have some options:
    A: LMH6702 (2V/V) => LHM6629 (10V/V)
    B: LMH6629(4V/V) => LHM6629 (5V/V)
    C: LMH6629(4V/V) => THS4302 (5V/V)
    Could you give me a comparison among them?

    Regards.
    Truong
  • 0 in reply to Xuan Truong Nguyen
    TI__Genius 12631 points
    Truong,
    The VCA821 has sufficient slew-rate to give you what you need in a gain of 10V/V. You will need to send me a schematic and results in order for me to understand and debug why you are seeing issues.
    -Samir
  • 0 in reply to Samir Cherian
    TI__Mastermind 22825 points

    Hi Truong,

    You asked about comparison between the combinations which give you the required response:

    "A: LMH6702 (2V/V) => LHM6629 (10V/V)
    B: LMH6629(4V/V) => LHM6629 (5V/V)
    C: LMH6629(4V/V) => THS4302 (5V/V)"

    For slew rate reasons, the 2nd stage needs the higher slew rate device (i.e. LMH6702 as opposed to the LMH6629). The lowest noise device (LMH6629) should appear as the 1st stage to minimize noise impact.

    If you are not getting the response from OPA847 (as you seem to be stating) with response being "slow" or unexpected, I am suspecting that you are being limited by the OPA847 large signal -3dB response limited to just above 200MHz (Vo = 2Vpp):

    OPA847:

    The LMH6629 large signal behavior is also likely to limit you similarly. The LMH6702 large signal bandwidth (Av= 4V/V, datasheet figure 1) will be around 300MHz.

    If you don't need to adjust the gain (like a normal OPA with external resistor values), you will have a faster response from the THS4302 you have mentioned, to close to 1GHz:

    THS4302:

    THS4303 is another candidate.

    Regards,

    Hooman

  • 0 in reply to Hooman Hashemi
    Prodigy 120 points

    Hooman, 

    Thank you very much. 

    Assume that I would like to use LMH6629 at the first stage, do you think it is possible to amplify our signal 10 times?

    --Truong

  • 0 in reply to Xuan Truong Nguyen
    Prodigy 120 points
    Hooman,

    For example, we use AN-1897 LMH6629 LLP-8 evaluation board to amplifier our signal

    Thanks.
  • 0 in reply to Xuan Truong Nguyen
    TI__Mastermind 22825 points

    Hi Troung,

    Yes, there is a populated board for the LMH6629 set for a gain of 10V/V (but can be changed by changing board components) that you can use:

    I have copied your original conditions below for reference:

    "We would like to design an amplifier circuit under conditions:

    - Gain: +20

    - Input signal: 50mV~1V with pulse width is ~10ns. Input can include a single pulse or some consecutive pulse."

     

    I'm not sure how you will handle when your input is large (1V). This will saturate the 1st stage of your design. Is that going to be an issue in your system? Otherwise, you may have to clamp the signal or somehow prevent this large input condition.

    Keep in mind that with large signals, you may be limited by slew rate. Here is the highest sinusoidal frequency you can "handle" based on the LMH6629 slew rate (SR= 1,600 V/us) and the peak-to-peak output swing (Vpp):

    f_max (Hz) = SR / { pi() * Vpp}

    So, with 3Vpp output swing (assumed), you can handle sinusoids up to 170MHz, and so on:

    f_max (Hz) = 1.6e9 (V/s) / { pi() * 3Vpp} = 170MHz

     

    In terms of square waves / pulses, your output rise / fall times are:

    Rise_Fall_time (seconds) =~ Vpp / SR

    Rise_Fall_time (seconds) =~ 3Vpp / 1.6e9 = 1.9ns

    So, out of the 10ns pulse width you've specified, 2x1.9ns (= 3.8ns) would be taken to traverse low to high and then high to low transitions (leaving you with 6.3ns of "valid" output. You'd have to review and see if this is fast enough for your application?

     

    Regards,

    Hooman