Post Go back to editing

LTC2945 supply setup for negative voltages

I have several negative supply voltages I need to monitor (-2V, -5V, and -30V).

Looking at the block diagram (p. 9)... if I'm connecting the negative supply to the GND pin, am I correct in assuming I'll need optical isolators for I2C in all cases since the internal "ground" on those pins will now be <0V (zeners tripping), or is the control logic somehow isolated?

The few negative-supply examples show INTVCC with only a bypass cap to the GND pin (i.e., no separate supply on INTVCC) for cases where the supply voltage is better than -4V.  For my low-voltage case (-2V), do I need to provide a separate -5V supply to INTVCC for the internal logic, or is there another way around?  I assume the same rules apply to the I2C lines (if I use a -5V INTVCC supply, I'll need isolators pulled to -5V, as well).

Thinking in negative voltage does something horrible to your brain ;-)

  • On a related note... is PAD connected to the GND pin internally?  If so, should it be tied to the negative supply being monitored?

    What about ADIN?  If unused, should that be tied to the negative supply being monitored, as well?

  • Hi Dan,

        You are correct. Signals below GND (pin) voltage are not allowed for the LTC2945. You can tie the GND pin (and optionally PAD, which is not connected internally, but does provide thermal advantages) to the lowest potential (your negative supply). So you can power the LTC2945 by hooking-up the GND pin to any negative voltage more than -4V. In this case you will need isolators for the I2C bus and any other signals that go to the ground-referenced world.

        The INTVCC pin has a regulator that steps down from VDD, and you must ensure a minimum of 2.7V to INTVCC. So you can simply hook-up VDD to any voltage larger than 4V OR apply a voltage > 2.7V to INTVCC (keep the capacitor). In the case of your -2V supply you will be forced to provide an alternative supply to achieve > 2.7V at INTVCC. Of course, you will still need isolators for the I2C lines.

       As you pointed out, dealing with negative supplies does something to your brain. The ADIN pin is no exception. Voltages on this pin will be referenced to the GND pin of the IC, so, for example, if you apply -5V at GND, and -3V at ADIN, the ADC will tell you that it measured +2V. Don't let the ADIN pin go more than 7V (abs max) above the GND pin, and never below the GND pin.

  • I should explicitly state that when you tie the GND pin to -5V or -4V you will tie the VDD pin and SENSE+ pin to the system ground (0V).

    When you tie the GND pin to -2.7V you will tie both the VDD and INTVCC pins to system ground, as well as SENSE+.

    When you need to monitor -2V you will have to tie GND to a well controlled -5V supply, and VDD and SENSE+ to system ground. Remember that ADC readings are referenced to the GND pin (-5V), so your -2V supply on SENSE+ will register as (-2V) - (-5V) = +3V.

    Clear as mud, right?

  • Thanks for the help so far, nenger.  So, if I'm understanding all of this correctly, the attached pic shows a rough outline of what I'm designing towards (ignoring SCL, addressing lines, etc.):

    The -30V and -5V monitors should be essentially identical in format and are the "easiest" to design as they're above the minimum INTVCC requirements.

    Your second reply did muddy my understanding a bit. Tieing the VDD pin to system ground for monitoring negative voltages makes sense, but you said for -4V to -80V monitoring I should also tie SENSE+ to system ground... this doesn't match up with Figure 3d (p. 14), and I'm not sure how the monitor would be able to determine current/power in such a configuration.

    For the -2V monitor, you said the GND pin needs to connect to a -5V supply... you lost me on that one.  Why doesn't the incoming -2V supply on the GND pin work?

    Incidentally, the >=2.7V supply to the INTVCC pin for sub-4V monitoring... is that relative to the GND pin, or absolute?  In other words, If the GND pin is monitoring -2V and VDD is connected to system ground, does INTVCC need only a +0.7V supply (or greater).  Or perhaps it's relative to the VDD pin, so if VDD is at system ground, then INTVCC is VDD+2.7V?


  • Surprisingly, I do have both -5V and +2V supplies already on the board, so either solution will work for the -2V monitoring.  Here's my current design based upon using the -5V supply...

    ... though I think I may go with your final suggestion and use the +2V supply to VDD instead of system ground.

    Due to design criteria outside of my control, I do not have access to a supply-segregated "return" line before it reaches system ground, only the "outgoing" negative supply line... hence, low-side sensing is my only real option, as far as I can tell.

    I need a few good days of "deep think" on this one before it sinks in.  I'm monitoring 32 supplies ranging from -30V to +20V at currents of anywhere from 100mA to 30A.  The eye twitch is real ;-)

    Many thanks, nenger.  Unless other questions come up in the meantime, I'll update in a few months when I get a few boards made.

  • Hi Dan,

        As you have drawn it you are using "low-side sensing", which is fine, and will work for the -5V case in your drawing (and looks like the datasheet drawing). I recommended the "high side sense" in my suggestions. High side sensing places the sense resistor between the highest potential and the load. Low-side places the sense resistor between the load the the lowest potential. The reason that I recommended that you use high side current sense is because the LTC2945 can sense the voltage between SENSE+ and GND. In high-side sensing that voltage is the input voltage, and can be used to calculate power (I x V). In low-side current sense the SENSE+ voltage is near the GND pin voltage, which does not reflect the input voltage. You can still read the input voltage at the VDD pin, though. Either the low or high sense approach will work, and the way you have drawn it is fine for the voltages larger than |-5V|.

        The way that you have drawn your -2V circuit will not work because you will force the INTVCC pin above the VDD pin, which should be avoided. Note that the datasheet says when you apply a voltage to INTVCC from the outside you should short INTVCC to VDD. My earlier recommendation that you tie the GND pin to -5V causes the VDD pin to be the most positive voltage at the IC. When you do that INTVCC will regulate itself, and only need a capacitor.

        Instead, you could also tie the VDD pin to a voltage above system ground (like +2.7V or +5V) and keep the GND pin tied to -2V. In that case don't drive INTVCC. Do it like this:

    Hopefully this clears-up the confusion. I blame it on the negative voltages!