Comparator - fried due to excessive output current?

Discussion in 'Analog & Mixed-Signal Design' started by Col John Matrix, Sep 6, 2016.

  1. Col John Matrix

    Thread Starter Member

    Apr 16, 2009

    I'm using a TIA followed by a comparator in order to convert an optical signal received by a photodiode (a sequence of 0s and 1s) into a corresponding electrical signal that's 0 to 5 V. This is to allow an optical link to be used in place of the previous electrical link.

    The circuit seems to function perfectly well, converting the optical signal into an electrical signal. However, a couple of the comparators have failed - I observed it suddenly trying to draw a large current from the power supply (200 mA or more), followed by a drop in that current and a subsequent inability to output 0 to 5 V signals.

    My guess is that the way I had things set up meant the comparator output was delivering a much higher current than it was rated for, causing the part to heat up and eventually fail. The max. output current is given as +/- 50 mA, together with a note that at such a current "a heat sink may be required to keep the junction temperature below the absolute maximum rating". I had things set up so that the comparator output was 0 to 5 V, delivered to a 50 Ohm load of an oscilloscope - so would the comparator output current therefore be nominally far too high (e.g. up to a max 5V / 50 Ohm = 100 mA?) and fried the part?

    The part in question is the Linear Circuits LTC6752, in the MS8 package. The circuit is very similar to that shown in Figure 10 on the data sheet. The input voltages to the comparator inputs were about 1 V.

    I'm just looking to get confirmation (or not!) that I've identified the cause of failure, and some pointers on how to avoid it (I'm thinking - change the scope load to 1 MOhm, put on a heat sink maybe...)

  2. AlbertHall

    Well-Known Member

    Jun 4, 2014
    I would not expect a max 50mA output to put up with 100mA for very long.
    If your application requires this signal to feed a 50Ω load then you will need to use a higher current amplifier as well as/instead of the LTC6752.
    If your application requires this signal to feed a much higher load impedance then just don't use a 50Ω oscilloscope probe.
  3. Alec_t

    AAC Fanatic!

    Sep 17, 2013
    Your suspicions confirmed. You could use the comparator output to drive the 50Ω load via a current-boosting emitter-follower stage.
  4. crutschow


    Mar 14, 2008
    Why are you using the 50Ω impedance setting for your oscilloscope. :confused:
    That's normally only used if you have a 50Ω source and 50Ω coax cable from the source to the oscilloscope to minimize high frequency reflections in the cable.

    For your measurements you should be using a 10:1 10MΩ probe with the oscilloscope set to 1MΩ.
  5. Col John Matrix

    Thread Starter Member

    Apr 16, 2009
    Simple carelessness! It was set to 50 Ohm to look at something else and I then connected the comparator input into it without bothering to think!

    I tried a fresh part. When connecting the comparator output to 1 MOhm scope, the current being drawn by the comparator was fairly low (< 10 mA). When switching the scope to 50 Ohm, the part attempted to draw a much larger current (prevented from doing so by the current limit set on the supply). That's consistent with an excessive current draw with a 50 Ohm output.

    Thanks for confirming my suspicions! I'll try to be more careful in future ;)
  6. Kjeldgaard


    Apr 7, 2016
    There may be a further problem:
    Absolute maximum supply for the output circuit (Vdd-Vee) is 3.6 Volt.
  7. Col John Matrix

    Thread Starter Member

    Apr 16, 2009
    Thanks for the heads up. Ultimately I'll be using 0 - 3.3 V, which wouldn't be an issue, but at the moment I'm using 0-5 V simply to avoid having to use an extra supply.
  8. Col John Matrix

    Thread Starter Member

    Apr 16, 2009
    Hello everyone,

    I'm rebooting this thread with a follow-up question. The comparator output is connected using a 50 Ohm SMA coax cable to the scope. The scope input impedance is set to 1 MOhm, so as not to fry the comparator as discussed above.

    With this configuration, the comparator runs fine, no more problems with frying it due to excess output current as was the case when setting the scope to 50 Ohm. However, on the rising/falling edges of the comparator output there is quite a lot of ringing when the scope impedance is set to 1MOhm. When I (briefly!) switch back to 50 Ohm the edges are nice and clean.

    The signal is ultimately to be read in by an FPGA. From what I can tell the input impedance of the FPGA is 50 Ohm - states "The characteristic impedance of all routes from the FPGA to the expansion connector is approximately 50-Ω."

    - I guess if that's the case, the FPGA should receive a "clean edge" signal rather than one with ringing? I'm trying to get my head around which signal seen on the scope, using either 50 Ohm or 1 MOhm, is representative of what the FPGA is going to see.

    Thanks in advance for any input!
  9. crutschow


    Mar 14, 2008
    The capacitance and inductance of the coax cable, when not connected to its characteristic impedance, are causing the ringing you see.
    Any digital signal will cause such ringing into that configuration.
    You should use a 10:1 probe.

    The FPGA itself is likely a high impedance, but the line to it's pins is apparently a 50Ω transmission so the signals should be driven from a source capable of driving 50Ω.
  10. tindel

    Active Member

    Sep 16, 2012
    I agree with crutschow. It's all about your transmission lines. You're ringing like a bell and not showing the true characteristics of the circuit! I've attached a transmission line simulation that shows some different options.

    • The first is a 50 ohm coax and load. This works great if you can source the 100mA required. But you can't.
    • The second is what you're finding to be a horrible solution - and rightly so. The reflections and ringing are caused by a mismatch in the transmission line with respect to the load. I see people do this all of the time for some reason. And it works reasonably well at low speeds (<~1MHz). But at the high speeds of a typical comparator this will not work.
    • The third is my favorite option. Adding probe resistance to your circuit board reduces the amount of current the driver needs to source to 10mA. You get at 10:1 divided down voltage, but most scopes can accommodate for this without having to think at all. This also allows the convenience of connecting directly to your board with an SMA or BNC connector. You can still measure higher frequencies than a typical passive probe. You can change your resistance to different load values to see how (or if) that changes the characteristics of the circuit.
    • The last option is a standard passive probe. This is a terrific option. Most scope and probe combinations go to about 100MHz. It has the lightest loading effect of all of the options eliminating loading as a impact on the circuit. It does require slightly more thought during layout to optimize probing techniques and locations, but this is not always required. Every good lab has many of these probes laying around for your use.
    I purposely had the rising and falling edges be 350ps. This equates to 1GHz. This is to illustrate the passive probe rolling off the edges, while the 50ohm does not. Really the only time you have to go to a 50ohm scope impedance is in high speed applications >100MHz. A passive scope probe does not do too well in these situations. But it's terrific in about 90% of applications. Occasionally you'll need to go to a 50 ohm impedance for low signal applications as well (<~10mV).