Post Go back to editing

LT3094 Spice Model Problem

I'm trying to build a +/-16.5V power supply using either two LT8364 (which I already have) or one LT8582 and a pair of LT3045-1 / LT3094 to post-regulate the output of the DC/DC converters. As I'm trying to make use of the VIOC feature, I'd like to run some simulations in LTspice.

However, I've run into an issue where the LT3094 is not behaving as expected when supplied with input voltages below -15V.

To reproduce, I've modified the LT3094 demo circuit very slightly to show the issue:

  • I've removed the current limiting resistor (doesn't make a difference)
  • I've changed the PGFB resistor divider for a threshold of -13.59V
  • I've changed Rset to 140k for an output voltage of -14V
  • I've changed the load resistor to 50 ohms
  • I've changed the input voltage to -14.9V

This works as expected:

However, changing the input voltage to exactly -15V makes the circuit behave very strangely:

Further decreasing the input voltage, e.g. to -15.5V, will simply make the output of the LT3094 clamp to just above the input voltage, which is also unexpected:

Now the real funny thing is: if I replace the LT3094 with an LT3093, the circuit behaves as expected again:

This looks to me like a bug in the LT3094 model (I've a a handful of LT3094's on my bench and can confirm that they do indeed work with input voltages of -19V).

Unfortunately, there are a couple of differences between the LT3093 and LT3094, and I need around 400mA in my power supply. So while I can certainly do some simulations with the LT3093, I'd much prefer to use the LT3094 model.

I'm using LTspice XVII for OS X, build May 27 2020, 18:51:13.

  • Hi mhx,

    I'm so glad to find someone else having a suspicion about bugs in the LT3094 model

    I started from the LTspice example circuit for the LT8582 and gradually walked it towards the +/- supply example shown in the LT3094 datasheet. The positive rail works fine, but the LT3094 output just tracks its negative rail input, during the startup ramp and still after it has gone beyond (more negative than) the voltage on the set pin. (Molly Zhu said that it sounds like the internal opamp is saturated)

    I substituted the LT3093 in place of the LT3094 in the simulation circuit, and increased the load resistors to reduce current draw 150mA. I changed nothing else and the circuit then started to behaved as it should. Each 10ms simulation run takes an hour or so, but I can live with that. 

    Incidently, I have previously run all my nominally "+/-15V" circuits at +/-16.5V. This time I was going to stick to +/-15V, because the absolute voltage limits on the LT3045-1 output and set pins are 16V. You might know something I don't or you might want to double check.

    I was using LTSpiceXVII from Jun 17, 2020, on Windows7.

  • Hi psupine,

    Funnily enough, a few minutes after posting this I had another look at the 3045-1 datasheet and saw the +15V output limit. I've read the datasheets pretty extensively, but never noticed this before. Maybe I was just assuming the devices to be more symmetrical than they are. I've had no issues running my prototypes at +/-16.5V, but I'll be dropping them down to +/-15V.

    But yeah, something's definitely funky with the 3094 model. You can actually run the LT3093 model way beyond the specified current limit of 200mA, but not close to the 450mA that I actually need.

    WRT simulation time: one thing I found speeding up LTspice quite significantly is turning off the graphs during simulation. I usually end up with simulation speeds of around 40us/s with the LT3045/LT3094, so 10ms will usually take around 5 minutes to finish.

  • I'd really appreciate if someone from AD/LT could at least confirm that something doesn't look right here. I'd have a look at the model myself, but as far as I can tell there's no unencrypted version of the LT3094 model available.

  • I reported this issue to our modeling group. Thank you for your feedback.