RAK3172 versus RAP3172-SIP (power consumption and low-level differences)

I’m evaluating both the RAK3172 and RAK3172-SIP for what RAKwireless refers to as “low-level development”.

There is a “low-level” document, but at the time of writing it seems to be specific to the RAK3172 (non-SIP). I’d link to it, but the forum prevents me.

On that page, there is a “Low Level Development zip file” link that was clearly written only for crystal (not TCXO) usage (e.g. for the non-SIP module).

That is straightforward to change, but given the power consumption measurements that I am seeing, I am suspicious that there are other changes that need to be made.

Using the “RUI” firmware that RAKwireless provides, it is possible to operate the module in receive and transmit to measure power consumption:

For receive:

AT+TRX=100
(followed by an ATZ to end receive)

For transmit:

AT+CW=923000000:0:60
(followed by an ATZ to end transmit)

where the second argument “0” can be changed for different power levels.

Here are the current measurements all obtained with a calibrated 6.5 digit bench multimeter in statistics average mode (30 samples with high TX power, 50 samples with medium TX power, and 100 samples for RX):

RUI RAK3172 (non-SIP)
RX 8.417 mA
[email protected] 46.589 mA
[email protected] 52.455 mA
[email protected] 114.647mA

RUI RAK3172-SIP
RX 9.784 mA
[email protected] 47.057 mA
[email protected] 52.578 mA
[email protected] 98.645 mA

For comparison, here are the numbers I get doing the equivalent with the Semtech library:

low-level RAK3172 (non-SIP)
RX 5.372 mA
[email protected] 43.641 mA
[email protected] 49.531 mA
[email protected] 109.647 mA

low-level RAK3172-SIP
RX 6.928 mA
[email protected] 50.398 mA
[email protected] 57.078 mA
[email protected] 128.515 mA

Note that:

  1. My implementation is more efficient in receive than RUI (probably just because of a lower CPU clock speed); the same improvement can also be seen for transmit for non-SIP only
  2. The SIP module draws more current in receive than the non-SIP module (extra consumption due to TCXO?)
  3. There does not seem to be big difference with RUI for transmit, but yet my Semtech-based code draw substantially more current.
  4. The 20dBm numbers are higher than the RAKwireless datasheet, but those are only quoted for 868MHz.

#2 makes me wonder if the RAKwireless datasheet “Electrical Characteristics” are accurate, since they claim identical receive performance between the non-SIP and SIP:

#3 makes me suspect that there are other changes between the RAK3272 (non-SIP) and RAK3272-SIP that need to be documented.

For comparison, here are the same numbers using the ST NUCLEO WL55JC1:

low-level ST NUCLEO WL55JC1
RX 6.707 mA
[email protected] 10.637 mA
[email protected] 11.538 mA
[email protected] 79.972 mA

The lower current draw at lower transmit power is likely due to ST supporting the low-power PA in addition to the high-power PA. The RAK3272 (non-SIP or SIP) design is only using the high-power PA.

It is not clear to me why the RAK3272 is less efficient at 20dBm; this is likely related to #4 above.

The low-level document that I referred to above is here: