https://open-source-silicon.dev logo
Title
p

Pritesh Ps

03/31/2023, 11:33 AM
Hi, I was trying to simulate a circuit2 from 'MOSFETs-only sub-1-V voltage references for ultra low power applications (DOI: https://doi.org/10.1007/s10470-020-01619-8)". However, a error appear as could not find a valid model name for M4, when units are used like (8u). But, when units are not used [like 8] same error but for 'could not find a valid model name for M8' appears. MOSFET M8 has w=0.24um. and i also awre of to use numbers between 0.5 to 99 without unit. why is that? what's the default unit (i think in spice manual say default is um (micrometer)). also what could be the possible solution to resolve any one of the error please, i need this circuit working and as per paper producing Vref around 750mV.
l

Luis Henrique Rodovalho

03/31/2023, 11:38 AM
Maybe the problem is the transistors with 50 um of length. It is just too large
https://github.com/google/skywater-pdk-libs-sky130_fd_pr/blob/f62031a1be9aefe902d6[…]d6f59b57627436/cells/nfet_01v8/sky130_fd_pr__nfet_01v8.bins.csv Try to use transistors with those dimensions. Any other dimension has a high chance of having non accurate models.
If you really need other aspect ratios, try to use transistor arrays https://lci.ufsc.br/pdf/_00309905.pdf
p

Pritesh Ps

03/31/2023, 11:50 AM
@Luis Henrique Rodovalho One thing i just tried, without units mentioned and made M8 pfet 2x size and it simulated without any problem. and got Vref half of the expected value. i know this is not correct way to change in aspect ratio. but, i was wondering whether 50 units is the problem or what? without units only M8 becomes the problem as it's w=0.24
and if I put units infront of all aspect ratio then only 50um makes sense to have error
l

Luis Henrique Rodovalho

03/31/2023, 11:52 AM
Sometimes the problem is using fingers or multipliers, as the model try to use a fraction of W instead of using transistors in parallel. This way, the transistor model uses less than the minimum width.
Anyway, M8 is too small. Any mismatch would make your reference fail your specs. The bulk terminal is connected with forward body biasing. The models aren't accurate for large VBS, so, be careful.
p

Pritesh Ps

03/31/2023, 11:55 AM
oh! Thanks for pointing it out.
l

Luis Henrique Rodovalho

03/31/2023, 11:59 AM
You could replace the transistor M8 with a diode. Source and drain terminals are short circuited, so, there is only a diode made between the bulk terminal and the other terminals. The most accurate model is for this parasitic diode, so, what really matters is the area of their diffusions, not exactly the channel dimensions.
s

Stefan Schippers

03/31/2023, 4:00 PM
@Pritesh Ps The big problem here is that you are using dimensions in micro-meters. sky130 has a global SCALE=1e-6, so a 50um width transistor should have W=50 You need to remove the 'u' suffix from all transistor dimensions.
If you still get errors it means the width is too big. In this case reduce the width and use 2 or more parallel devices, setting mult=2 or more.