[TEST] Testing PD monitor #19
Loading…
Reference in New Issue
There is no content yet.
Delete Branch "%!s(<nil>)"
Deleting a branch is permanent. Although the deleted branch may exist for a short time before cleaning up, in most cases it CANNOT be undone. Continue?
Frequency response of kirdy photo diode input.
Setup:
Stimulus : 408 uA RMS
DC bias : 1.44 mA
Injected current measured through ZCP30 50MHz AC/DC current probe.
Output signal measured at PD_MON net with x1 probe
Results:
-3dB bandwidth : 911 Hz
After thinking more about it, we probably want to use a better op-amp there in the end.
I saw that at the end of the AFE front end, there is a LPF with a 1.5923566879kHz cutoff frequency and R100 output resistor at a large value(1k Ohm), which should be the contributing factor to the low bandwidth. They should be the contributing factor to the low bandwidth.
Why do we need such low cutoff frequency?
If the target cutoff frequency is such low, using a better opamp(with better GBW, bias current etc) may not improve performance I think.
TLV2172, Dual Channel CMOS Amplifier should be a reasonable opamp that we are looking for.
(Cost a HKD10.52 per pcs at 100 order quantity)
Transimpedance Amplifier Consideration:
With some calculations and Simulations based on TI Application Notes, the Transimpedance AFE stage can handle 5MHz PD Signal.
Specification Used to Design Transimpedance Amplifier (Followed TI Application Notes)
TINA-TI SPICE Simulation Circuit
Simulation Result:
The cursors measure the frequency response of the PD_MON net output
Might not really be a low-pass filter, could be mostly a capacitor to suppress ADC input current spikes. Also the 1k/100nF values could be just placeholders.
Biasing a photodiode also reduces its capacitance.
I looked through different types of laser driver(LD) and concluded that the integrated photodiode(PD) capacitance can have a maximum 20pF. But, not all manufactures or types of LDs have their integrated PD fully specified. If there is tracking error specs for the integrated PD, it is in range of +-0.5dB to +-1.5dB.
Thus, the following design specification for the PD monitor(TIA+ADC+Voltage Reference) is revised:
PD Monitor Specs: (Ref: TI SBOA220A)
I_in Range = 0-2.5mA
V_out Rnage = 0-2.5V
Maximum Photodiode Capacitance = C_pd_max = 20pF
Precision = 0.25dB -> Error of +- 3%
Bandwidth(-3dB) = Depends on ADC Sampling Rate(Max sampling rate: 2.4M Sample/s)
Note: Initial Accuracy is not taken into account
OPA381: Transimpedance Amplifier
Note1: 2nd Pole for the LPF can be added using existing R and C pads for better cutoff and noise.
Note2: R102 improves the capacitive load capability of OPA381 without adding any DC error.
Analog Front End Simulated Result:
Note: The following types of error is taken into account. Initial Accuracy of Rf, Temperature Drift of Rf, Max input bias current, Max input offset voltage, AFE total noise at 600kHz.
Frequency Response:
Noise Measurement:
REF3033 Voltage Reference:
STM32F407 ADC:
Note: STM32 ADC is missing a lot of specification like noise, temperature coefficient etc.
Total Amount of Error combined = +-2.06% < +-3%
I will proceed with this design in the next, but should I add a trimming network with potentiometer to trim the feedback resistor in TIA to reduce the error even further?
Looks like a pretty old and niche part without exceptional performance. Any particular reason why you picked that one?
Generally I find the Burr-Brown photodiode-related products not to be all that great compared to their other ICs. We can probably get better results with a regular op-amp wired up as a TIA.
This could be trimmed in firmware in any case, no?
OPA381 has lower input offset current and input offset voltage than the other opamp at similar price range and in the input voltage range that accept(+15V to -6V). It also achieve this performance in reasonably low power(0.8mA).
(I only looked through TI's catalog)
I think improving Analog Front End further may not yield better result unless a discrete ADC is used. Now, the bottleneck comes from STM32 ADC. I think OPA381 already meets the specification requirement and so it should be good enough?
Ar yes, it should not have any differences if it does not hit the upper bound of the input voltage range.
Maybe but the requirement isn't high and we can probably find a IC that is cheaper and/or already used on other Sinara boards.
Okay. Will look into other options later.
TL082 is used in Driver Sinara boards to monitor the output current of its laser driver. It is a industry standard OpAmp with other manufacturers it in the same pinout and P/N.
TI TL082: Dual JFET OpAmp
Frequency Response(@PD_MON Output):
Total Noise Output(@PD_MON Output):
Max Error (AFE + ADC + ADC) = 2.91% < 3%
Root Sum Square Error = 1.75108% < 3%
This should be cheap and good enough for our design goal.