Mismatch in power measurements with BCU and a standard logic analyzer Saleae

[Vikas7Bal]

I am trying to perform power measurements on the iMX93EVK board using the host based BCU utility.

I am trying to validate the "SOC Power" by comparing the "Group SOC power" reported by the BCU and compare it with the data collected from a standard logic analyzer ("Saleae"), by probing the sense resistor connected to SOC rail (Vdd_SOC)

The schematic of the sense resistor I probed is as follows :

I tried to capture the SOC power for some " time window" (in which a piece of code executes ) by toggling a GPIO pin at the start and end of execution of code

Also, from the SOM board schematics, I came to know about how BCU measures power, which is shown in the below image:

Now as per the schematic diagram for the sense resistor, I can observe the following :

Rs = 1 + 0.005 = 1.005 ohm

BCU data:
I_sense = 99.5 mA
Vsense = I_sense * Rs = 99.99 mV

Please find below the waveforms observed from Saleae for the same SOC rail

With Saleae I am observing the sense voltage to be around 16 mV (whereas it is reported as 99.99 mV by BCU)

Saleae data:
V_sense+ = 880 mV
V_sense- = 864 mV
V_sense = “V_sense+” - “V_sense-” = 16 mV

So, I am not sure if there is any issue in the procedure that I followed as "Saleae" is a standard logic analyzer and ideally there shouldn't be any such mismtaches.

So could some one help me in resolving this issue?

[JorgeCas]

Hello,

That is the correct way to measure it but you need to consider full-scale current and sense current equations of PAC19314 datasheet.

It is difficult to compare BCU with your measurement since the tool has instantaneous and average measurements.

Best regards.

[Vikas7Bal]

@JorgeCas , thanks for your reply.
I have few questions, (please bare with me as I have a long post with multiple questions)

1) I have gone through the PAC1934 current equations, I think those computations are required to be done on the "VSENSE results registers", to get the actual voltage and current drawn by sense resistor i.e., can you confirm that the probed voltages through Saleae must match with the final computed voltages using PAC equations?

2) Also it looks like PAC1934 driver is part of the BCU utility itself, so I think it is not possible to access the PAC registers when my code executes on Cortex-M33. Can you confirm?

2) From the SOM schematics, I observed the "VDD_SOC_0V85" is the power rail that goes close enough to SOC, is there any other rail that can go even closer to measure power at Cortex-M33 level or at the Ethosu NPU level?

3) Assuming the probed voltage is not having any errors, can you validate the following procedure to compute the SOC power?

- Probe Vsense+ and Vsense- voltages
- Calculate sense resistor voltage (Vsense = Vsense+ - Vsense-)
- Calculate the sense current (Isense = Vsense / Rs). In my case, Rs = 1 + 0.005 = 1.005 ohm
- Calculate SOC voltage which will be same as Vsense- (since they are in series)
- Calculate SOC power (Power = Vsoc * Isense) (since they are in series)

With this procedure, I am getting Power = 13,755.22 uW

Can you provide me your insights on these assumptions?
Thanks

[JorgeCas]

Hello,

1) That is correct, the Vsense value is stored in registers but it should match with voltage drop in sense resistor.

2) Yes, it's not possible access to PCA1934 registers from Cortex-M33.

3) Your procedure is correct but, it is difficult to compare BCU with your measurement since the tool has instantaneous and average measurements.

Best regards.

[Vikas7Bal]

@JorgeCas, I understand that, Logic Analyzer and BCU can't be compared because the latter has average and instantaneous measurements.
The only reason why we are trying to go with Logic Analyzer is that, the sampling rate of BCU is coarse (it is working at 10 ms sampling rate) and my piece of code executes much faster (< 1 ms). So since we couldn't find a way to increase the sampling rate we tried logic analyzer.

1) So if there is a way to increase the sampling rate of BCU, then that would be better?

Also, in the SOM schematics, I found below table which provides details about voltages and currents drawn by each rail :

I tried to measure them by using both Logic analyzer and BCU  and these are my findings :

Power rail	Expected current from schematics (mA)	Measured current through Saleae (mA)	Measured current through BCU (mA)
NVCC_BBSM_1P8	2	2.3	0.090942
VDD_ANA_0P8	186	50	20.06531
VDD_ANA_1V8	389	72	13.22022
VDD2_DDR	676	360	15.625
VDDQ_DDR	360	280	34.60693

2) As you can see, the BCU values are not even matching with the expected currents from schematics which should be the ideal case.

So can you provide some help regarding these two queries?

[JorgeCas]

Hello,

1) You may need to modify the BCU tool source code and build it. We do not have references to modify it.

The reported result in schematic is just a reference and is not guaranteed because the data presented is based on empirical measurements performed on a small sample size.

2) For power consumption measurements I suggest you use power measurement application note since the test conditions changes the power consumption.

Best regards.