PGA305: PGA305 EVM GUI calculations not matching 'PGAcoeffgen24.dll' calculations

Mohammad Waqas Imtiaz

Part Number: PGA305

Hi Ti,

We have successfully executed and understood the PGA306 EVM GUI calibration process.
The calculated values function as expected.

However, there is an issue with the TI-provided "PGAcoeffgen24.dll" as it does not yield the anticipated results.
The calculated values, derived from the same input, differ, and the plot does not display the expected behavior.

This matter is a follow-up to the discussion in this thread:
https://e2e.ti.com/support/sensors-group/sensors/f/sensors-forum/837458/pga305-calculating-coefficients-without-evm-gui

PGA305 GUI - Guided calibration:

We employ the 3-pressure-points and 1-temperature-point method.
Additionally, we simulate the bridge voltage using an external DC power supply.

Reading Padc values and converting them to float:

Bar	DC voltage on INP+ and INP-	Read Padc	Integer	/2^22
0	0.01v	0x24A57	150103	0.03578734397
50	0.06v	0xEEA45	977477	0.23304867744
100	0.11v	0x1BE0C8	1827016	0.43559455871

Measured/calculated DAC values:

Dac	Hex	Integer	/2^14
4 mA	0x7B2	1970	0.12023925781
12 mA	0x1714	5908	0.36059570312
20 mA	0x267B	9851	0.60125732421

Calculated coefficients from the wizard:

Parameter	Hex	Integer	/2^22
h0	0x174817	1525783	0.36377501487
g0	0x4D0000	5046272	1.203125
n0	0xFB26FA	-317702	-0.07574605941
Padc gain	1
Padc offset	0xF0EA71	-988559	-0.23569083213

Validating the coefficients:

D(Padc) = h0 + g0*(Padc + Poffset) + n0*(Padc + Poffset)^2

D(0.03578734397) = 0.1202392117522 ~ 0.12023925781: Good
D(0.23304867744) = 0.3605956437268 ~ 0.36059570312: Good
D(0.43559455871) = 0.601257259766 ~ 0.60125732421: Good

Concluding that the calculated coefficients from the guided calibration are good

PGAcoeffgen24.dll:

LabView 2023 Block Diagram and Front Panel:

Validating the coefficients:

D(Padc) = h0 + g0*(Padc + Poffset) + n0*(Padc + Poffset)^2

D(150103) = -2046.756952413414 != 1970, Wrong

D(977477) = 2954.0007042884827 != 5908, Wrong

D(1827016) = 8088.726546431059 != 9851, Wrong

Concluding that the calculated coefficients are not correct, or i am missing something

Tried many different combinations, but no success:

Any help would be appreciated.
I can also jump on a call if that is easier.

over 1 year ago

0 Bob Benjamin over 1 year ago

TI__Guru** 113765 points

Hi Mohammad,

It is not totally clear to me how the GUI is actually calculating the coefficients, but the DLL should correlate closely. One thing that I think may be an issue is you have a blank or zeroes for the temperature data. I think that is probably your biggest issue as to why the two calculations are not matching. Unfortunately I can't tell you how the temperature of 25 deg C gets computed for the GUI computation.

The other thing I noticed is the GUI is show pressure gain of 1, but the DLL is computing to 2.

Best regards,

Bob B

0 Mohammad Waqas Imtiaz over 1 year ago in reply to Bob Benjamin

Prodigy 40 points

Hi Bob,

Thank you for your quick response.

So far, we have experimented with hundreds of different combinations and values, both read and simulated, in an attempt to see patterns and reverse engineer the inner workings of the function. Without a clear understanding of the function, it is challenging to effectively use and certify it. This is why we seek TI's assistance to move forward.

I posted the wrong image before where Tadc is set as 0, the correct image includes values for Tadc obtained from the GUI. Despite this, the calculated coefficients remain identical, suggesting that Tadc at 3P1T has no impact.

The DLL file calculates a gain of 2, while the GUI calculates a gain of 1 with identical Padc values. Why is there a difference?

Am i correct to understand that there are two different Pgains, one which is used before Padc values are read (5 - 400), and the other in P = Pgain * Padc + Poffset?

I believe the key to understanding this function lies in understanding the Pgain and Poffset calculation in the polynomial. Why is it necessary to have a Padc offset in the first place? Given the Padc values above, I can successfully calculate a polynomial that passes through all three points without needing any offset. The coefficients fall within the 6 digit hex range and can be stored on the EEPROM. However, when I do that, the compensation engine stops working.

Like this for example: This didn't work

Do you have any insights into why Poffset is used? The datasheet doesn't provide any information beyond the formula itself.

Any help would be appreciated

0 Bob Benjamin over 1 year ago in reply to Mohammad Waqas Imtiaz

TI__Guru** 113765 points

Hi Mohammad,

I truly feel your pain as well. I've have spent a great deal of time trying to understand the nuances myself between the GUI and the DLL. The GUI does not use the DLL in its computation and I am still trying to figure that one out too as to the differences. Pretty much everything I know is on the E2E forum already. I sent you a couple of links in the other thread you posted before, but I will put all of the pertinent information that I have been able to come up with in the list below:

If you look in ti.com under your myTI account for your Secure resources, look under PGA305-SOURCE you should see the zip file for PGA305 Coefficient Calculation DLL Source v1.0 which contains the matlab script used to compile the DLL.

Best regards,

Bob B

0 Mohammad Waqas Imtiaz over 1 year ago in reply to Bob Benjamin

Prodigy 40 points

Thank you once again,

I have gone through all these links multiple times, but have not yet found a complete solution.
It's reassuring to know that Pgain and Poffset don't have any constants affecting them, but what remains unclear is how they are calculated.

Thank you for trying to share the MATLAB file; it might be the key to success.
However, when I checked under "My secure resources," it indicated: "You do not have any secure resources."

Could you please attempt to share the file again?

If this information is insufficient, then we might not have a way to proceed further. Is there no one at TI who might have relevant insights? Whether it's the datasheet writer, firmware coder, DLL compiler, GUI maker, sensor developer, or anyone else?
The PGA30* series contains multiple IC's that might have similar calculation.
Any additional assistance would be greatly appreciated.

Thank you for responding in the thanksgiving week.
We appreciate it.

0 Bryan Lizon86 over 1 year ago in reply to Mohammad Waqas Imtiaz

TI__Guru 52156 points

Hi Mohammad Waqas Imtiaz,

This a holiday week in the US, we will respond to you next week. Thanks for your patience

-Bryan

0 Bob Benjamin over 1 year ago in reply to Bryan Lizon86

TI__Guru** 113765 points

Hi Mohammad Waqas Imtiaz,

You should be able to access the download now.

Mohammad Waqas Imtiaz said:
Is there no one at TI who might have relevant insights? Whether it's the datasheet writer, firmware coder, DLL compiler, GUI maker, sensor developer, or anyone else?

Unfortunately the answer to the question is no. Much of the work was contracted outside of TI and not well documented. As the device was developed some number of years ago many of those involved are no longer with TI or have moved on to other projects. So myself and a few others are basically doing what you are doing to understand more about the GUI and what differs from the DLL. This is an ongoing process and once better understood we will document accordingly, but at this time what we know is what you know.

Best regards,

Bob B

Sensors

Sensors forum

PGA305: PGA305 EVM GUI calculations not matching 'PGAcoeffgen24.dll' calculations