The signal in question is highlighted pink in the below schematic:
INA826 has long settling times and slow slew rate.
This bears out in simulation and with real parts.
The red trace corresponds to the highlighted wire in the schematic above. There is significant delay between the red trace's rising edge and the blue trace's rising edge (which in this screenshot best represents the input to the INA826), especially at the largest 10V input step where it can take over ten microseconds to settle.
This is what it looks like in the o-scope. Ch1/yellow trace is used to trigger the oscope. Ch2/green trace is a stimulus delivered to a 1k resistive load. Ch3/orange trace is the highlighted node in the schematic.
For some reason, this delay appears even worse with real parts than in simulation; this 5us delay between the green trace's 3V step and the orange trace's rising edge instead would take about 1-2us in simulation.
This latency is particularly problematic when using short pulse trains (20us) which entirely precludes providing real-time feedback:
The INA849 has improved transient performance according to the datasheet.
The simulation seems to show this as well.
I propose testing a simple part swap, the INA849 is a pin-for-pin drop-in replacement for the INA826.
The signal in question is highlighted pink in the below schematic:
INA826 has long settling times and slow slew rate.
This bears out in simulation and with real parts.
The red trace corresponds to the highlighted wire in the schematic above. There is significant delay between the red trace's rising edge and the blue trace's rising edge (which in this screenshot best represents the input to the INA826), especially at the largest 10V input step where it can take over ten microseconds to settle.
This is what it looks like in the o-scope. Ch1/yellow trace is used to trigger the oscope. Ch2/green trace is a stimulus delivered to a 1k resistive load. Ch3/orange trace is the highlighted node in the schematic.
For some reason, this delay appears even worse with real parts than in simulation; this 5us delay between the green trace's 3V step and the orange trace's rising edge instead would take about 1-2us in simulation.
This latency is particularly problematic when using short pulse trains (20us) which entirely precludes providing real-time feedback:
The INA849 has improved transient performance according to the datasheet.
The simulation seems to show this as well.
I propose testing a simple part swap, the INA849 is a pin-for-pin drop-in replacement for the INA826.