Author Topic: Interpolating a curve from 2 Curves  (Read 1006 times)

0 Members and 1 Guest are viewing this topic.

Offline Glenn0010Topic starter

  • Regular Contributor
  • *
  • Posts: 226
  • Country: mt
Interpolating a curve from 2 Curves
« on: April 28, 2021, 10:48:37 am »
Hi,

I have this curve below, I'd like to create a new curve for Vge at 14 V. The strategy I am going for is to take data points from the 13 V and 15 V curves and split the difference to create a new 14 V curve. I think this is a simple and quick method that will get relatively accurate results.

Saying that, are there any better way of approaching this?



Cheers

GG
 

Offline MadTux

  • Frequent Contributor
  • **
  • Posts: 809
Re: Interpolating a curve from 2 Curves
« Reply #1 on: April 28, 2021, 11:04:00 am »
If you rely on the analog properties of a switching transistor, your design is wrong/bad!
Thereby interpolating curves is quite pointless......
 

Offline Conrad Hoffman

  • Super Contributor
  • ***
  • Posts: 2139
  • Country: us
    • The Messy Basement
Re: Interpolating a curve from 2 Curves
« Reply #2 on: April 28, 2021, 11:13:55 am »
I'm sure there's some method that involves cutting across the curves with a line, plotting the intersections and using that curve to interpolate, but I'd just pencil in the new curve by hand and eyeball. As for the usefulness, they do publish this stuff for a reason. Sure, designs shouldn't be sensitive to transistor parameters but you still need to work things out to know.
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22435
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: Interpolating a curve from 2 Curves
« Reply #3 on: April 28, 2021, 11:17:47 am »
This.  The curve for any given device varies plus or minus a volt or so -- see the min/max spread in Vge(th).

Also, this appears to be an IGBT, which probably isn't rated for operation in this regime (high Vce and Ic), at least not for very long, and certainly won't last very long even if it is so rated.

If you need a controlled current under pulsed or fault conditions, consider a MOSFET (will have wider SOA), and an analog circuit to limit or regulate drain current.

By "fault conditions", I mean a normally-saturated switch, entering this regime, which is detected by the control circuit, and turned off within a reasonable amount of time (typically a few microseconds).

Vge(on) is still a valid question, but not to any precision -- normal design application is to set Vge(on) at least high enough to sink the desired maximum load current, without exceeding the maximum rated short-circuit / fault current, or SOA, at the same voltage.  For example, given the plot above, and a load current of 5-10A, you might choose Vge(on) >= 12V, which should draw a fault current over 20A, so a peak current detector of 15-20A should be adequate, or a desat detector can be used.  The device will need a peak current rating over 50A (corresponding to the current drawn at an equivalent 13-14V Vge(on), because of the spread in Vge(th)), and I'm guessing the maximum fault duration will need to be under 10us or so; confirm with the SOA.

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22435
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: Interpolating a curve from 2 Curves
« Reply #4 on: April 28, 2021, 11:21:38 am »
I'm sure there's some method that involves cutting across the curves with a line, plotting the intersections and using that curve to interpolate, but I'd just pencil in the new curve by hand and eyeball. As for the usefulness, they do publish this stuff for a reason. Sure, designs shouldn't be sensitive to transistor parameters but you still need to work things out to know.

As for numerical methods, a quadratic interpolation with the neighboring curves (3 points in e.g. a vertical line) should give very good results -- the transfer curve is somewhere between quadratic and linear in this regime.  In general, a more complete model will have many higher order terms than just quadratic -- it's just that, in this area, almost all of them are negligible.  If you can find a SPICE model for the device, it may give a better fit.

Again, simply not knowing where that curve happens to be, on a real random part, let alone at whatever junction temperature it happens to have in the moment -- makes it a moot point.

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline Glenn0010Topic starter

  • Regular Contributor
  • *
  • Posts: 226
  • Country: mt
Re: Interpolating a curve from 2 Curves
« Reply #5 on: April 28, 2021, 11:55:36 am »
Hi All,

Thanks for the feedback, I understand that it is difficult to draw any accuracy from this due to tempereture and device to device variations.

However, my main question was more out of curiosity in case I might need something similar in the future ratehr than out of need.

Cheers
 

Online graybeard

  • Frequent Contributor
  • **
  • Posts: 482
  • Country: us
  • Retired III-V RF/mixed signal/device engineer
    • Chris Grossman
Re: Interpolating a curve from 2 Curves
« Reply #6 on: April 30, 2021, 05:12:17 am »
In most cases transistor to transistor variation will probably be bigger than any error you get from just eyeballing the difference.

Hopefully whatever you are doing does not rely on your device exactly duplicating the spec sheet curves since device to device variation and temperature variations will practically ensure that your device will not exactly match the specified curves.


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf