Investigate RF load switch malfunction in v0.3 assembly run 1
Description
Even when the 5V_RF_S and 5V_RF_U load swiches are closed via the EN pins, a voltage of about 1.1V is observed at their outputs.
Test log
| Date | Changes | Results |
|---|---|---|
| 30/11/2023 | Removed the 18 Ohm resistor on the FLAG line of the 5V_RF_S load switch, essentially isolating the 5V_RF_S and 3V3_RF_S load switches. | No particular results. Most of the measurements were made after this change. |
| 01/12/2023 | Set up a breadboard circuit with the same load switch and a 47k Ohm load. Connected a 15 Ohm load to the output and an STM32 NUCLEO board pin to the EN pin. The EN pin was toggled every 5 seconds and the output of the load switch was measured. | The breadboard circuit showed normal operation of the load switch. When the EN pin was LOW, the device fully shut off and the output was 0V. |
| 01/12/2023 | On the breadboard setup, measured the voltage on a floating EN pin when the input voltage is 5V and 3.3V. | The voltage on the EN line was not big (~300mV). |
| 01/12/2023 | On satnogs comms, we took measurements of the FLAG_S and FLAG_U lines. These are pulled-up to 3V3_RF_x. | Although a max voltage of 3V3 is to be expected, the max voltage on the FLAG lines was 5V, when the power system opened. When it was closed, the voltage on the FLAG lines was 1.1V. This was true both for S-band (where the 18 Ohm R has been removed) and for UHF. |
| 01/12/2023 | We removed the 10k pullup resistors on the FLAG lines of the 5V load switches (EN lines of the 3v3 ones) - R608, R612 - to remove a possible connection of 5V_RF_x and 3V3_RF through the 3V3 load switches. | The 5V_RF_S output was relatively clean - 5V open, 300mV closed. The 5V_RF_U was cleaner - 5V open, ~0V closed. We note here that the 18 Ohm resistor (R616) is present on the UHF circuit. |
| 7/12/2023 | Reconstructed the load switch circuit on a breadboard | Confirmed that there is no weird inherent behavior of the load switches |
| 7/12/2023 | Removed the 18 Ohm resistor on UHF to have identical circuits in the 2 bands. | Same behavior |
| 7/12/2023 | Observed that when the 5V_RF efuse closed, it did not open again. Observed a 1.7V drop on the D502 zenner diode. We removed it. | After removing the Zenner, the efuse operated as expected. |
| 7/12/2023 | Refering to UHF: Noticed that there might be some path from a signal with 3.3V to the 5V_RF_x rails. Because we do not TX, this path can only be throught the !EN_TX_U. We removed the NOR gate (U1011). | The problem went away, both the 5V_RF_U and 3V3_RF_U were operating fine. |
| 7/12/2023 | Reconnected the 18 Ohm resistor on UHF. Reconnected the 18 Ohm resistor on S-band | Same behavior on UHF. |
| 7/12/2023 | The capacitors C616 and C617 were replaced with 4.7k Ohm and 4.87k Ohm resistors respectively. | This formed a voltage divider that dropped the 1.1V on the 3V3_RF_x EN pin to a voltage below the enabling threshold, thus closing the respective load switch. Once the load switch closes, the power source of the NOR gate is removed and the path causing the 1.1V voltage to appear on the -closed- 5V_RF_x line vanishes. Still, we could not get a TX on S-band and we are suspecting that a fault with the RF-Mixer is the cause. |
List of possible causes
Considering the board after the removal of the 18 Ohm resistor on the FLAG line
-
Inherent load switch behavior -
In the condition where the EN pin of the 3V3_RF_x load switch is floating, its voltage is 1.1V and because it is pulled-up to the 5V_RF_x rail (which is closed from its load switch), the voltage there becomes 1.1V as well. -
One of the devices in the RF chain leaking to a pin connected with the 5V_RF_x rail when it is in an undefined state. The reason for the undefined state is because the 3V3_RF_x EN pin is in an undefined state and thus we should not consider that the 3V3_RF_x load switch operated normally. Also for the components powered by 5V_RF_x, we already know that they do not operate normally. - The issue is common on both UHF and S-band so we can start by looking at components present in both chains.
- The AGC and VGA are common between the 2 chains. Those are enabled by the output of a NOT gate which is powered from 3V3_RF_x. We should note that the EN_TX_S signal is a 3.3V signal from the AT86 which is HIGH when TXing, while the 3V3_RX_x signal comes from the aforementioned NOT gate.
The third possible case was proven to be the issue. There was current originating from the NOT gate though the VGA.
Edited by Dimitrios Moustroufis