So the RFM95 is using 75uA when the MCU is in deep sleep (current for the Arduino board not included)?
What current does your Arduino board draw in deep sleep (RFM95 not included)?
What board(s) are you using?
I recall some older post about deep sleep and using external pull-up resistors for (part? of) the RFM95’s interface lines to prevent them from floating during deep sleep (the floating caused additional current draw on the MCU or RFM95). But cannot find it back and do not remember if this was related to AVR (ATmega328xxx) and it may as well have been related to STM32 (blue pill board) instead.
There are issues with the ESP32, when in deep sleep some of the IO pins are disconnected and float, which does increase sleep current of the LoRa device.
An ATMega 328/1284 keeps the pins active in deep sleep and no additional pullups are needed. Although a pullup on the NSS pin is always a good idea.
With the LoRa device configured for sleep mode, current is normally circa 0.1uA, data sheet quotes 0.2uA typical to 1uA max.
I would add, and this applies to ATMega boards too, it can be very difficult indeed to troubleshoot high sleep current on a finished design.
Far far better to start with bare bones devices on a breadboard so that when you add stuff, such as a regulator or LoRa device, you can see exactly what affect adding a component does have on sleep current.
Putting a ‘bare’ (3.3V) ATmega, ESP8266 or STM32 on a breadboard is doable (although for ESP8266 there are very few options) because there are rather basic boards available for these MCU’s that are breadboard friendly and require only the voltage regulator and one or more LED’s to be removed for low power / reliable measurements.
For ESP32 I still have not been able to find a suitable board that is breadboard friendly and does not have a USB to serial chip and not have battery charging support integrated on the board. For boards that have USB to serial and battery charging support integrated, these components are (very) difficult to remove and often require a microscope to do the job.
For (3.3V) ATmega328 an Arduino Pro Mini is the most practical.
For STM32 a basic ‘blue pill’ board (STM32F103C8) can be used.
For ESP8266 there are very few options (such boards) available. I use this one: ESP8266 SMD Adapter Board R2 [version] - ElectroDragon (it’s voltage regulator has a quescent current of up to 90uA so worth removing for low power).
For ESP32 I have not found any similar option which fits a normal breadboard (and which has pull-ups/downs and 2 switches required for normal operation already included).
That is a real nice charger with an informative display. In addition to Liitokala Lii-402 the Folomov-A4 also has 250mA and 3A charging modes and it can charge 4 cells with 2A at the same time (and costs almost 3 times as much).
Good reviews of the LiFeSO4 suitable A4 and Lii-402 chargers:
Folomov-A4: Review of Charger Folomov A4
Liitokala Lii-402: Review of Charger LiitoKala Lii-402
Before you ask, the DCDC converter is not the reason. I also make measure on the 3.3V directly with a 3.3V source.`
So the ‘DCDC converter’ (regulator ?) was removed or otherwise disconnected from the 3.3V supply to the processor and RFM95 when you measured the current ?
I would also suggest that unless you want to re-work that board, by removing or replacing the regulator, your never going to get sleep current below (typically) 90uA.