TLV767: Inaccurate Spice Model

Hi Ethan,

How are you doing? Let me check this tomorrow. Please note that these simulations are not really accurate when it comes down to temperature specs of the device itself. The modeling includes component variance, but we recommend testing an EVM on a bench for an accurate test. 

Thank you,

Josh Nachassi 

Hi Ethan,

Our models are behavioral models, so they won't represent true transient behavior. The temperature variance is not modeled within the simulation. Please let me know if you have any further questions and I would be happy to help.

Thank you,
Josh Nachassi 

Hi Josh,

I understand that the model has limitations and incorporates changes due to temperature. The issue I have with the model is that accuracy across ALL temperature is supposed to be 1% (stated in the datasheet). The accuracy of the model is off by 10% of setpoint! This is huge and violates the datasheet spec.

Thanks,

Ethan

Hi Josh,

Sorry for the late response, we had some design thrash which resulted in us removing this part from our design. Now we have it added back. While I understand the the SPICE model is just one specific state of the all the different potential part variations. The issue is that, even factoring in the temperature, silicon skew, minor revisions of the die, etc... It still doesn't account for the massive difference observed at the FB pin. The voltage error is supposed to be 1% across PVT, but the SPICE model has it at 10%!!!!!!!! That's why I think this is an error in the model.

Thanks,

Ethan

Hi Josh,

I think we are talking past each other. The only reason I brought up temperature is to point out the fact that even if we factor in the effects of temperature, the accuracy of the device is wrong by over 10%.  What I suspect is happening, is that the SPICE model has the wrong default voltage programmed for the FB pin. Please check the simulation and see that this voltage on the FB pin is wrong by other 10%.

Thank You,

Ethan

Hi Ethan,

Yes, we were talking past each other it seems like. Ill check the model and let you know what I see.

Thank you,

Josh Nachassi

Hi Ethan,

We tested the PSPICE model on our end and I see that the VFB is 0.8-V as shown in this scope shot. I put the feedback in unity gain so Vout=Vfb and I also tested vfb with a probe with feedback resistors and it shows 0.8-V. Can I see how you are getting this value of 0.72? What is your test case so we can try to replicate it. 

Thank you,
Josh Nachassi

Here is the simulation I ran in LTSpice. AS you can see, it regulates to 0.72V at the FB node. Maybe the Unencrypted model is different than the one you used?

Hey Ethan,

Thanks for sharing. I used a model that I found in an archive because the simulation on ti.com was not working for me and was showing a different device. This is very interesting and although you are reading this in the Spice model as 0.72-V, we can assure you that our VFB does not deviate like this in normal condition. 

Thank you,

Josh Nachassi 

Hi Josh,

So we are using different SPICE models, and we get different results! That being said, can TI provide me with an unencrypted SPICE model so I can run my simulations with the expected values?

Thanks,

Ethan

Hi Ethan,

Yes give me some time let me see if I can get it for you.

Thank you,
Josh Nachassi

Hi Ethan,

I checked back with my colleague, and we used the unencrypted .lib file and imported into the encrypted design for testing and we got the 0.8-V. I am still curious how you are getting the 0.72-V in the simulation. Are you using the unencrypted file in ti.com?

Thank you,

Josh Nachassi 

Hi Josh,

Yes, I am using the unencrypted file.

Thanks,

Ethan

Hi Ethan,

My colleague and I are unsure why your model is representing this value when we are using the same file as you to run the test. Was there anything that was modified in your test from the original file?

Thank you,

Josh Nachassi

There was nothing I modified in the file.

Hi Ethan,

Understood. Give me some time to look into this case of the simulation. The Vfb of this part does not deviate by 10% so obviously there is something with the simulation. 

I will try to help you solve this issue further, let me speak with some more people on my team. Have a great weekend.

Thank you,

Josh Nachassi  

Hey Josh,

Thanks for looking into it. I'm going on vacation and will be back in 3 weeks.

Thanks,

EThan

Hi Josh,

I'm back. Is there an updated SPICE model?

Thanks,

Ethan

Hi Josh,

I'm trying to simulate my circuit before I build it. This would require the model to regulate to the correct voltage. I can try importing the SPICE model again. Can you send me the exact link for the SPICE model I should be using for LTSpice?

Thanks,

Ethan

Hi Josh,

I downloaded the file from the link and imported it into LTSpice, and unfortunately, I do not see the Vfb regulating to 0.8V. I talked to my team and agreed that is was okay to share the simulation file / models with the TI Forum. If you could open the simulation up and take a look at the feedback node, maybe that can help us make some headway into this problem.

Thanks,

Ethan

SD Card Level Translator System.zip