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.

LMH6624: LMH6624MA low capacitance input overvoltage protection

Part Number: LMH6624
Other Parts Discussed in Thread: LM7171, OPA690, OPA820, OPA656

Hi, 

I am using the LMH6624MA with an ADP, which is biased at 500V, can you please suggest a low capacitacnce protection method in case the 500V ends up on the input? i was considering a singal diode (BAV99) but they are only rated to 100V.

Thanks

  • Morning Andrew, 

    The BAV99 is normally what I use, but not clear on your comment. Normally those diodes would be on the inverting input node and seeing low volage across it, then if asked to suppress a fault, there would need to be a limiting resistor as the diode conducts to keep it from being destroyed. Can you attach a schematic - or a TINA file is better. 

  • Yes, i am curranty using a BAV99.D3 is the ADP and and D1 and D2 is the BAV99. The worst failure mode would be the APD going short circuit and 500V reaching BAV99.

    APD.TSC

  • I am working on your TINA file, your detector diode is leaking quite a lot pushing the output voltage to the rail - do you have a model number or a better TINA model for the APD - eventually would need its capacitance. 

  • The TINA file is just to show the schematic not for simulation, unfortunately I don't have a model for D1, D2 or D3 (just used standard diodes). The actual circuit works fine.

    just wondering if there is a better input protection idea

  • Ok, here I reduced your bias to 20V to just get things going, you probably already have them, but a couple of caps will help your noise, That noise peaking is very normal and dependent on the source C

    The BAV99's add a little bit of C. If the APD shorts out, the 10kohm will limit the current to 50mA, well in range for the BAV99. Their voltage with 50mA might increase beyond the internal input stage protect diodes (if any) that would conduct back to the V+ node.,  -That current split would be limited by the 5kohm on the V+ input to ground. So, with 50mA into the BAV99, only say about 0.7V on the V- node -overdriven to the output but not beyond the supplies - I would think fine. and the file, 

    LMH6624 Zt input protect with BAV99.TSC

  • Thanks for the simulation.

    I have two 100nF high voltage ceramic caps very close to the APD pins.

  • As long as I have the file, lets look at the 2nd stage where the feedback cap risks instability as it is shaping the noise gain to 1 for this decomp part. Usually easier to see in the output spot noise, yep - looks kind of chancy but easy to improve, 

    We can shape the noise gain to stay higher with another cap on the inverting input, hold flat noise gain (you will still get the gain filtering). 

    yes, this is a lot better phase margin with the noise peaking with a broad peak, 

    And the signal response curve, 

    More than you would ever want to know about this inverting comp thing I published way back in the BurrBrown days. This was recently republished. 

    https://www.edn.com/unique-compensation-technique-tames-high-bandwidth-voltage-feedback-op-amps-2/

  • Hi Michael,

    Thanks for the response and linked article (I will give it a read, always open to learning opportunities).

    Yes, the second stage amplifier gave instability (~450Mhz noise / oscillation on the signal), this was resolved by replacing the 9pf feedback cap with a 2.2pf. But given the second stage does not require the bandwidth provided by the LMH6624 it has been swapped for the LM7171AIM.

    The final circuit is below (only missing the 100nF and 10uF caps on all the supply lines), the 10k pot is to give some very small offset control.

    C1 maybe changed to 0.3pF and C2 to 2.2pF to provide more bandwidth (~10MEG instead of ~6MEG)

    Final.TSC

  • Hi Michael,

    Thanks for the response and linked article (I will give it a read, always open to learning opportunities).

    Yes, the second stage amplifier gave instability (~450Mhz noise / oscillation on the signal), this was resolved by replacing the 9pf feedback cap with a 2.2pf. But given the second stage does not require the bandwidth provided by the LMH6624 it has been swapped for the LM7171AIM.

    The final circuit is below (only missing the 100nF and 10uF caps on all the supply lines), the 10k pot is to give some very small offset control.

    C1 maybe changed to 0.3pF and C2 to 2.2pF to provide more bandwidth (~10MEG instead of ~6MEG)

    Final.TSC

    File with APD and caps

    Final_APD.TSC

  • Well Andrew, 

    It is kind of encouraging that the LMH6624 model basically predicted that oscillation - little mismatched on frequency probably due to not having the correct loading in sim vs bench. You have been showing only a 75 series out - a lot of this higher speed part will move towards less phase margin with no load so I have been adding a terminating 75 in sim. 

    The LM7171 is kind of an odd choice, did you have stock in it or something? 

    It falls into a class of VFA that has a high tranconductance input stage to offer CFA type slew rate (the OPA690 is the part we did at BurrBrown like this) - that comes with much higher input voltage noise, 

    Your source in the model is not correct for an APD (even though they often specify a 50ohm out, that is for their characterization I think - Hammamatsu I assume)

    Anyway, a similar GBP lower noise 2nd stage might be the OPA820. Here is an output noise comparison, 

    And this file, 

    LMH6624 to OPA820 stage.TSC

  • I did have the OPA656U but due to stock and storage issues (MSL 3 for most of the OPA parts) the LM7171 (MSL 1) was chosen.

    In lab testing the noise is very similar with either output stage (OPA656U or the LM7171)

  • Well I did the OPA656 as well, unity gain stable JFET input which is not needed in your design. Relatively pricey to get the JFET input device. 

    MSL3 is kind of odd, I had standardized on MSL2 with the thinking the MSL1 allows a customer to completely abuse a part in storage and the company will back that up. 

    The National guys in that late 90s timeframe where pretty aggressive in a number of ways. MSL1 might be one of those, the LM7171 was going after similar parts down the road at LTC - the LTC families of that type were kind of overdone - they got real enamoured with that topology as they had kind of missed the boat on CFA. 

    Yes the SO-8 OPA656 is level 3, but the sot23 is level 2. Maybe there was some actual problem in those days getting MSL2 for the SO-8. I don't think that continued over time. I think maybe the low leakage input bias current for the JFET might have perhaps failed MSL2 testing in the SO8?

    Yes, the OPA820 is more like what I recalled, all packages MSL2

  • Thanks for your feedback.

    The OPA656U was used in a previous working design, I agree the OPA820 is a much better part.

    I have a working design / layout so did not want to change the design from SO-8. I will have a look at the OPA820.

    I am keen to keep to MSL 1 if possible, as I am only making a few devices and don't have a humidity controlled room for part storage and from my testing the LM7171 seems to work fine.

    Edit: Looks like I can get the OPA820, so will order some for testing and comparison.

  • Do you know why the SO-8 is MSL-2? as most SO-8 packages are MSL-1.

    Is this due to possible lead corrosion if kept at the wrong humidity? if not i could hand solder the OPA820 and store in a sealed bag.

  • Well there is both a physical consideration and a strategic component. While a part might be quailifiable as MSL-1, it might not be prudent to do so. I looked into at one time pretty closely, and decided to target MSL-2 as a good balance between risk and market acceptance. Other marketing managers made different decisions. 

  • Many thanks for your comments and feedback it is much appreciated. 

    I should be able to test the OPA820 early next week and will probably move to that part instead of the LM7171.