open-source-silicon.dev

Possibly silly question. I do not see it in the PDK, but I think jfets on SKY130 may be possible? Has anyone looked into this? <https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6338421/>

image.png

I'm currently working on a paper using this kind of technique. It not new. Bulk driven differential amplifiers use PMOS devices that behaves as JFETs.

<https://ieeexplore.ieee.org/document/700924>

My paper uses a variation of this circuit. I'm using the Skywater PDK also.

The problem with this JFET, at least with Skywater PDK, is this kind of gate layout, with poly gate corners inside the diffusion, are forbidden.

Ah. Thanks for the explanation. Seems pretty interesting!

My true interest in this is zero/negative threshold devices for low startup voltage power conversion. It looks like the native nmos/pmos devices might work? <https://skywater-pdk.readthedocs.io/en/main/rules/device-details.html#id55>

I am a bit unclear on the nomenclature used in the table but I think that the table implies that the threshold voltage is ~ zero? I cant seem to find a legend for the table but the figures that I think are threshold voltage seem to be close to zero

For Dickinson charge pumps, the problem can be the bulk connection. Even if you use native transistors in diode configuration, there will be reverse body bias, and it will increase VT. But, yes, the native zero-VT transistors are a good choice.

<@U01QTMG2K8R>:  The poly gate corners are not forbidden by manufacturing rules;  that is, you can put them in the layout and the layout won't be rejected.  There are SkyWater devices with 45 degree angles on gates, so my gut feeling is that the rule is an unnecessary restriction;  nevertheless, if you can make the layouts so that there are no 90 degree poly corners inside the diffusion (that is, use 45 degree bends only), it's probably safer.

<@U017UPJEGKZ>:  Yes, the native Vt devices have thresholds very close to zero, although it's a good idea to run the device model through SPICE and get an I-V curve.

Is there any decoding for the device table parameter values as presented in the readthedocs documentation? I can guess which ones are the threshold voltage from context, but its pretty unclear

<@U017UPJEGKZ>:  The best way is to look at the SPICE model.  For example, in `$PDKROOT/sky130A/libs.ref/sky130_fd_pr/spice/sky130_fd_pr__nfet_05v0_nvt.pm3.spice` , see the expression for `vth0`:
```+ vth0 = {0.053+sky130_fd_pr__nfet_05v0_nvt__vth0_diff_0+MC_MM_SWITCH*AGAUSS(0,1.0,1)*(sky130_fd_pr__nfet_05v0_nvt__vth0_slope/sqrt(l*w*mult))}```
The first value (0.053) is the nominal threshold voltage, and the rest is just adjustments.