Differential Current Sensing

Discussion in 'General Electronics Chat' started by scubasteve_911, Jan 8, 2008.

  1. scubasteve_911

    Thread Starter Senior Member

    Dec 27, 2007
    1,202
    1
    Hi Everyone!,

    I'm trying to implement the circuit attached which was found in an older US patent.

    http://www.google.com/patents?id=4uwnAAAAEBAJ&dq=5703490

    I'm using a dual-hbridge in parallel so that I can access each bottom FET, so that is taken care of. I'm very worried about the bottom end of the circuit, which is the dual precision half-wave rectifiers along with the differential amplifier.

    I am modifying the precision half-wave rectifier with an additional diode in the feedback path to prevent any stability problems, most will know what I am referring to. Apparently, this circuit will be able to sense both the direction and magnitude of current flow during fast-decay and slow-decay modes of operation.

    My question is, will this circuit require a bipolar supply? I am thinking 'yes' because the current flow can induce a negative voltage across either sense resistor, which is bad news for a single supply opamp. Secondly, should I stick to the low-offset and drift opamps for this? What sort of diodes should I use for the best performance?

    Apparently, I should use the most gain at the rectifier stage for better noise performance. This was mentioned to me by a professor and in the patent. My low-drift opamps have limited capabilities with 2.5MHz GBW. I want the highest frequency range possible that I can get. Should I split up the desired gain between the rectifier and differential amplifier stages?

    Also, Will I get an overall positive output from this circuit? I would think so, if the current is overall positive in a single direction (which I need).

    I am planning on using this opamp, since it has the lowest offset and stability. http://www.analog.com/en/prod/0,2877,AD8628,00.html
    *** NOT ANYMORE , JUST REALIZED SINGLE SUPPLY OPERATION :( ***

    I'm really stuck and would appreciate your input!

    Kindest regards,

    Stephen
     
  2. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    Why not have a look at Linear Technology's LT1007/LT1037 series opamps? They have around a 60MHz bandwidth product, are low noise, low drift, high slew rate (11V/uS for 1037) and a 117dB CMRR, Vgain of 20M into a 2K Ohm load.

    Zero offset and drift would be nice, but if it can't handle the BW, it's pretty useless.

    If you're really concerned about temp drift (ie: if 0.6uV/°C isn't good enough) then you could consider using Peltier devices driven by an op-amp to keep the temperature stable.

    I don't see any diodes across the bridge. If the load is inductive, it'll blow the lids right off of your FETs. Speaking of which, have you made a choice there yet? What's your supply V? If it's 30V to 70V, take a look at IRF1404, IRF1405, IRF1407.

    You could use clamping diodes on the inputs of your op amps to limit Vsense to your supply rails.

    Keep Rsense (21A,21B) low in value to minimize wasted power. They must also be strictly resistive; ie: not wirewound.

    You probably are already aware of the hazards of having the upper and lower half of the same side of the bridge conducting at the same time. Many power MOSFETS require much more time to turn off than to turn on.

    As far as the polarity of the output - have you modeled it yet? I don't see a single resistor or capacitor value in that image; only the existance of them. It's going to be a bit difficult to predict circuit performance without that.
     
  3. scubasteve_911

    Thread Starter Senior Member

    Dec 27, 2007
    1,202
    1
    Hi Steve,

    Thanks very much for your reply! I'm looking into other opamps, I have decided to use a AD8599 for the precision rectifiers and stick with the single supply AD8628 for the differential amplifier. I'm currently doing some simulations in hope of a bit of confidence when I build it, I don't like throwing money away :)

    The new OA has 10MHz GBW, 16V/uS slew, 10uV offset, ~1nV/sqrt(Hz) noise. I'm a really big fan of Analog Devices products, which is what I am using for a lot of it.

    I'm very sorry, but I am not using a discrete H-bridge. It's a fully integrated dual h-bridge IC L6205PD. I'm paralleling both bridges to form two half-bridges, then using individual 0.1 Ohm precision sense resistors. I had to go the integrated route, since I am using 8 of these circuits! Also, these bridges include nice features such as overcurrent protection, intrinsic diodes, overtemperature, cross-conduction protection, etc.

    I'm using a 42V @ 14A supply for the drivers, since there are 8. Their current demands are related differentially, so it is close to enough.

    You're right about my output, I am very concerned about what I may be facing. The problem is that the parameters are dynamic. I'm expecting about 820 milliohms in the coil, capacitance I am completely uncertain about, and inductance is looking to be around 50uH. The inductance changes with my highly dynamic bias current and position of a metallic object which is moving. I expect about a 20uH variation which is non-linear with these two parameters. This was determined with Finite element modeling, so I am unsure until I build it. Do you think I will need a snubber?

    Steve
     
  4. scubasteve_911

    Thread Starter Senior Member

    Dec 27, 2007
    1,202
    1
    Simulation has exposed my ignorance once again :p

    My precision rectifier OA isn't rail to rail, so I might give in and check out LT products. Secondly, there's some diode switching noise present on one output. I am thinking of scrapping the precision rectification and just go fully differential. I will have double the current rating during recirculation though, but I can compensate for that.

    Steve
     
  5. beenthere

    Retired Moderator

    Apr 20, 2004
    15,815
    282
    You might want to look at the sense lines in the L6250 again. Both low side devices share the same sense line. Paralleling the two circuits on a chip merely places four low side FET's together on the sense line. They do not appear to be separable into half bridges, either. As one side is off while the other conducts, the point of the differential current sensing is hard to understand.
     
  6. scubasteve_911

    Thread Starter Senior Member

    Dec 27, 2007
    1,202
    1
    beenthere,

    I think you may have misunderstood or I was not clear. I am using two H-bridges individually as half-bridges by paralleling the individual bridge's outputs. Then, I am using the bridge's individual sense pin for current sensing.

    And as an update, I'm using a programmable gain instrumentation amplifier for my current sensing. The theory is to have a constant gain of 10V, then switch to 5V during recirculation, thus giving me the correct output. I plan on switching up to 100KHz, and the opamp AD8250 can switch gains at 10MHz! I think this is a solid plan :)

    Steve
     
  7. beenthere

    Retired Moderator

    Apr 20, 2004
    15,815
    282
    Just came back to correct myself, scubasteve - I was looking at the spec sheet & at the arrangement for paralleled operation to double current capability.
     
  8. scubasteve_911

    Thread Starter Senior Member

    Dec 27, 2007
    1,202
    1
    It's okay, very easy to miss! The worst part about paralleling each bridge as a half bridge is that they do not double the current limitation of the device. In actuality, the Vs and Sense pins are wire-bonded to the die, these limit the current. If I were to parallel a half-bridge from one bridge with another half-bridge from the other bridge, then you can get twice the current. I cannot do this because of the shared current paths through the sensing resistor :( I do get a lower RDSon though, which is a good thing :D

    Steve
     
Loading...