AFE4404EVM: Optimal timing settings for PPG with AFE4404EVM

Hi Alexandra,

Interference between the channels can be reduced by increasing the separation between sampling signals.
However this also means that it will reduce the ambient light rejection. So you should play with settings to find which one works best for you.

Regards,
Prabin

Hello Pravin,

My apologies for the late reply. Thank you for the information, I will do as you suggest.
I have done some measurements to optimize the SNR of the signal, and I am not sure how to interpret the results. Ideally, in a system limited by photon noise, I would expect an increase of the SNR as a square root of the increase in averaged samples or supply current. However, so far, I did not observe this trend with my AFE4404EVM and the NJL5513R sensor. In order for me to find the optimal settings to achieve the highest SNR, could you please tell me what is the limiting noise in this system?

Thank you very much in advance.

Best regards,

Alexandra

Hi Alexandra,

Can you share some data such as DC, standard deviation for the case with low and high LED current?
ADC in the AFE has some noise and it is dominant for the signal with DC close to 0. As you increase the LED current , the noise from the external source ( including LED driver) will start dominating the total noise. This phenomenon is described in the figure 4 of the AFE4404's datasheet.

Regards,
Prabin

Hello Prabin,

Thank you for your detailed reply. I have repeated my measurements with the AFE4405EVM and the same optical sensor (NJL5513R): I have acquired a few signals sampled at 50Hz, with a Duty Cycle of 0.5%, over 10s.

I am extracting the mean DC value of each signal,, and computing the energy of the bandpassed signal (between 30 and 240bpm) as a measure for SNR.

Please find below some of the experimental values I measured: please note the units are "codes".

For 1 averaged sample, and for the Green channel: DC = 7.4x10^5 ; Energy = 1.4x10^5

For 4 averaged samples (Green channel): DC = 7.6x10^5 ; Energy = 1.1x10^5

For 16 averaged samples (Green channel): DC = 7.7x10^5 ; Energy = 1.1x10^5

For a current of 5 mA on the Green LED: DC = 1.6x10^5 ; Energy = 2.8x10^4

For a current of 20 mA on the Green LED: DC = 4.8x10^5 ; Energy = 6.6x10^4

For a current of 40 mA on the Green LED: DC = 7.6x10^5 ; Energy = 1.2x10^5

As you can see, there is a trend for Energy when it is increasing with the current, but there does not seem to be any improvement when using averaging... Would you have any idea of the reason behind this obervation?

On my side I will repeat these measurements a few times to have more statistical relevance.

Thank you very much in advance.

Best regards,

Alexandra

Hi Alexandra,

Are you using AFE4404EVM for these experiments?
When you say you are changing he averaging , I assume you are changing NUMAV. Also while changing NUMAV, you have to change the conversion width too ( or fix the conversion widths for the maximum NUMAV).
Regarding the data you have posted, is it for LED phase or LED-AMB phase. We should look for LED-AMB phase for better performance.
Can you also report the standard deviation of LED-AMB data ?

I suggest you try following experiment to see the effect the averaging.
1) Disconnect the PD using PD_DISCONNECT bit.
2) Program the conversion widths ( for all 4 phases) for the NUMAV of 15. See table 7 in the datasheet
3) Set NUMAV = 0 ( 1 ADC averaging) , capture some data and note down STD DEV of LED-AMB phase.
4) Repeat the step for different values of NUMAV.

Regards,
Prabin

Hello Prabin,

I have recently received the AFE4405EVM evaluation kit and therefore decided to run some tests on it. As it has some interesting features for my application, I will use it from now on. I understand that this post does no longer belong in this thread: should I create a new post with the AFE4405EVM tag for further questions?

Regarding the AFE4404EVM, I realized (after posting my first message) that I did not set properly the conversion width according to the NUMAV. This is obviously a source of error in my measurements, as you pointed out.

However, I did change the conversion width using the default settings provided in the GUI of the AFE4405EVM: the results I posted were obtained with the AFE4405EVM and LED phase only. For the LED-AMB, do I perform a simple subtraction of the signals (LED1, 2, 3 - AMB) or is the operation more complex?

I will now perform the operations you described: could you just tell me why disconnecting the photodiode is important?

Thank you very much for your precise and detailed answers and your availability!

Kind regards,

Alexandra

Hi Alexandra,

Since AFE4405 is still under NDA, we can take this discussion offline at yadav@ti.com.
You can either digitally subtract the LED and the AMB values or AFE subtracts and give it as 0x2E and 0x2F register.
The main reason to disconnect the PD is to eliminate noise coming from input ( such as LED, photodiode etc). By doing this we can measure the noise only due to AFE.

Regards,
Prabin