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.

LM7372: looking for a part that has faster settling

Part Number: LM7372
Other Parts Discussed in Thread: THS3001, THS3491, THS4271, THS4211, LM7171

Currently I'm using LM7372 in a design but the settling time is only specified to 0.1% for a 10V step, also noise is higher than I'd ideally like. Does TI have a suitable replacement for this part that has settling to 0.01% step in less than 70nS specified ?  Also preferably with noise down in the 1 - 3 nV/RtHz region. I know ADI make such a part but wondered if TI had anything newer that would be worth evaluating ?

  • Hello,

    Which part from ADI are you considering that has this capability? Are you comparing their settling time to the same test conditions mentioned in the LM7372 datasheet?

    The LM7372 might meet your requirements for .01% settling in 70nS. It depends on your configuration. Can you share with me your schematic/application, as well as the gain and input signal? With a 10V step it is possible that the LM7372 will settle in 70ns to .01% step.

    Thanks!
    -Karan
  • Hi Karan, The AD811 is specified to settle in 65nS to 0.01% for unity gain driving a 150 Ohm load. The LM7372 is specified driving a 500 Ohm load. We are using it with a gain of x-1.5 where Rfb = 1.5K however we are driving a coaxial cable so there is 150pF in series with 50 Ohms capacitive load also. We are seeing good settling to 0.02 % but at 0.01% the part has a long settling tail (see attachment) the magenta trace shows settling to 0.01% for a 6.5v step (-6.5V to 0V) with +/-16V supply rails. Where the trace intersects with the center line is where is has decayed to 100ppm. As you can see increase the limit to say 0.015% and the settling would be really fast. In our case we actually need good settling at 0.0025% so the thermal tail becomes problematic. This looks more like a CFA than a VFA for thermal settling. So the part we need the settling has to be well behaved at low levels to give us the settling that we need.

  • Forgot to say circuit is a standard inverting opamp Rin = 1K, Rfb = 1.5K, load = 50 Ohms + 150pF (high quality coax cable). Power decoupling is 10 Ohms and 3uF to ground on each regulated 16V supply. Also I did try bypassing the 10 Ohm resistors but it had no effect on low level settling.
  • Hey Steve, I see you are still on this quest for no thermal tail where we had this THS3001 discussion.

    e2e.ti.com/.../2919142

    The AD811 is actually the first +/-15V CFA part from ADI and would not have very good thermal tail either. I think they did work on lower tail parts but more in the lower supply voltages.

    So the LM7372 was done before National bought Comlinear, so they too would have no concept of thermal tail. (you can tell that by the LM instead of the LMH prefix)

    The THS +/-15V parts were being done in Dallas before they bought BurrBrown - so again, no concept of thermal tail.

    I can't think of any +/-15V CFA where there was an effort to zero out thermal tail. Elantec might have tried but I think most of their parts are gone now.

    However, just looking at your requirements (and your load is confusing, a TINA file would help - are you driving a only source terminated coax - nothing at the far end?), you are putting a big step power shift in the output stage NPN side. To reduce thermal tail in a device, reduce the power shift in the output transistor. So, if you really only need to go 0V to 6.5V, solve for the +/-supplies with a pull down R to the negative rail that will give the same power in the output NPN at 0V and 6.5V. That would be using say a -5V and +8V supply, then adding a resistor to the -5V at the output - I have not set that problem up yet to solve, but might be promising to try. Of course usually, it is more common to have a range of output swings required where that simple approach will not work.

    So the elements of good large signal step response are many.
    1. Do not slew limit, this can be assessed by the SSBW shape to get the peak dV/dT and then compare to the device slew rate - and yes, running inverting in CFA will give you a bit higher slew rate - usually not modelled.
    2. actual 2nd order response ringing -this is a phase margin thing, easily tuned
    3. internal thermal tail shifting the input offset voltage
    4.

    It might be some other type of op amp or composite circuit will be necessary. What you are looking for is a high slew rate device with better than average input error terms and low drift .
  • Hi Mike, Yes I'm afraid so ! Also there are multiple amplifiers in the signal chain, our conversation was addressing the intermediate amplifier, this one is regarding the output stage which causes significant degradation to the settling tail performance I was surprised because the LM7372 is a VFA but the results look more like I'd expect from a CFA !!! To answer your question the cable capacitance is the dominant load because at the far end the load is only 1Meg 20pF (very similar to a standard scope input).  The device needs to be bipolar ie drive at least between +12V and -12V and for any 10V step within that range able to slew/settle to <0.01% in tens of nanoseconds. I don't use Tina, we use the industry standard LTSpice for compatibility with existing models - unfortunately modelling won't help because most devices use approximations or behavioral models these days so fine for gain/phase/frequency response but not much help for low level settling !

    I did try the THS3491 which is a newer faster part for which I had high hopes, but it turns out that which it is faster on slew rate the settling tail is actually worse so no luck there either.

    Regarding the AD811 yes it is a CFA (so not ideal for my application) but from the datasheet  the specified settling is 65nS to 0.01% for a 10v step with 1.7nV/RtHz, presumably the thermal tail kicks in below 0.01% which isn't ideal but may be acceptable. Also this would still be better than the LM7372 that kicks in at around 0.015 - 0.02%, with14nV/RtHz. noise.

    Yes it would be nice to have a device where the thermal tail has settled out but in the absence of this if settling time to 0.01% <70ns  is published on the datasheet this would be an excellent second choice ! Also I'm open to designing my own hybrid if needs be but obviously an off-shelf solution that comes close would be more cost-effective.

    So going back to my question are there any TI parts that are specified for +/-15v or higher supplies with a settling time of less than say 65nS to 0.01% for a 10V step ?

  • Hello,

    Have you looked at the THS4271? Would that work for you?

    Thanks!
    -Karan
  • Hi Karan, the THS4271 looks like a nice part, unfortunately it only works on +/-5V supply rails and we need +/-15V. Do you have anything that runs on +/-15V that has a similar or better specification ? (noise/settling time)

  • Also forgot to say the THS3491 is better than I gave it credit for, settling time for a 5V step to 0.01% is only a few hundred nanoseconds so better than the THS3001 and probably the best TI part I've measured so far. Unfortunately it really runs out of gas on a 10V step so 0.01% settling goes up into the microseconds range so can't really use for my application.
  • Hello Steve,

    There is also the THS4211, however it doesn't meet your voltage noise requirement. If that is okay with your system, that's another part to consider.

    Thanks!
    -Karan
  • Hi Karan, using this to replace the LM7372 then 7nV/Rt Hz would be an improvement, compared to the current 14nV/Rt Hz. The issue is that it's only specified for +/-7.5V operation. We need +/-15v as a minimum, preferably +/-18V. Any other suggestions, new parts etc that could work ?

  • Steve, you are obviously the most active designer in settling time in the current epoch.

    It would be hard to imagine with the THS3491, but if your 10V step slipped over into slew limiting that will vastly extend your settling time. Inverting is faster, and it has a lot of slew rate on demand. I was surprised at the big shift from 5V to 10V settlting. You might try slowing your input edge just a bit to see if it pop back in.

  • Hi Mike, maybe at this moment, I have a client with bleeding edge requirements. Possibly could be exceeding the slew rate because I have some very fast parts on the front end. Having said that 10V step is more of a benchmark because I figure if a part has reasonable settling to 0.01% for a 10V step then likely it will be reasonable for a 5V step to 0.0025% which is the parameter I actually care about. So the THS3491 isn't out of the running but I'm comparing the best of the best to see which is the best fit to my requirements. It seems like it's easy to find good parts that will run on +/-5V but +/-15V i much more of a challenge.
  • well yes - if you go into slew limiting your linear settling time calculation goes out the window,

    As for parts, there is a fundamental speed vs. breakdown tradeoff that is moving you across different process nodes. The THS3491 is the first new high voltage/speed process in maybe 10yrs. 

    If you have the small signal F-3dB on the THS3491, the approximate peak dV/dT for an output step of Vstep will be 2.85*Vstep*F-3dB

    That is a pretty good approximation for 0.6<Q<1.5 from this pair of articles, look at figure 5 part 1. 

    https://www.edn.com/design/analog/4443250/What-is-op-amp-slew-rate-in-a-slew-enhanced-world--Part-1

    https://www.edn.com/design/analog/4443345/What-is-op-amp-slew-rate-in-a-slew-enhanced-world--Part-2

  • Hi Karan, any updates, or new/existing parts that might be a good replacement. I understand this was an old National part and one of their earlier ventures into high speed opamps so I'm sure it must be possible to better this given the technological advances and greater levels of expertise at TI ?

  • Hi Steve,

    It seems that THS3491 is your best option. I have searched through all the parts to see if anything meets these requirements. It may be possible that we never had the need to develop a high slew rate part with such high supply rails. I will try to find more information.

    Thanks!
    -Karan
  • Hi Karan,

    The THS3491 is definitely a great part and I plan to use it in my new design at x5 where it has better settling to 0.01% specified on the datasheet than most VFA's which is surprising for a CFA, But for replacing the LM7372 I need to use lower gain and larger step sizes than shown on the datasheet. Do you have any other settling time data say to 0.01% for a gain of x1/x2 and a 5V/10V step? The THS3491 iteslf is a high slew rate part on higher rails, what may be the case though is that the emphasis was on slew rate rather than settling time given that there is usually a trade-off between DC precision and high speed performance.

  • Hi Karan, another question, is there a single version of the LM7372 ? ie a single device with a standard SO-8 package op-amp pinout ?

  • Hi Steve,

    We have the LM7171. This is very similar to the LM7372 and available in an 8 pin D package. We have nothing like that in the DDA package you are asking about.

    Thanks!
    -Karan
  • Thanks Karan, looks like it may possibly be an even better part from the slew rate and settling specs. Of course the datasheet doesn't really tell me what happens after 0.1% so I'd have to measure. I'm assuming this is a similar part but not an identical circuit on the same process as the LM7372 ?
  • Hi Steve,

    They are both built on TI's complementary bipolar process and released in the same year. They are not internally identical as their specs are different however these parts are very similar.

    Thanks!
    -Karan
  • Thanks Karan, I may have found a more suitable part but I will keep these in my back pocket because they could make a good alternative

    Best regards,

    Steve