On This Page
- Power Supply Rejection Ratio
- Mismatch Accuracy
- Absolute Accuracy
- Power Supply Rejection Ratio
- Startup/Shutdown time
- Startup with slow and fast supply
- Supply Glitch recovery
- Opamp biasing should be done with a PTAT current (See note in Opamp Design tips document)
- Biasing Network should be designed so that it tracks the circuit you are trying to bias. Therefore if a circuit contains a MOS and then a BJT stacked, then the biasing circuit should also contain the same model MOS and BJT stacked to generate the bias current or voltage for the stack so that they track each other with process and temperature variations.
- Biasing Network should be designed keeping in mind a lot of supply variation and should switch on at low supplies (before the rest of the circuit is expected to start, this can be done by pumping some extra current during the startup procedure through the biasing circuit. This is especially important if the biasing circuit is part of a startup circuit.)
- To bias particular nodes which do not have a well defined voltages like voltage being set by Vds of transistors, the Vgs/Vth or Vbe of a transistor can be used to set the voltage of that node.
- To improve the PSRR put a capacitor across the Gate-Source or Emitter Base of the mirroring transistor so that it passes off the high frequency disturbances into the biasing string without changing the Vgs/Vbe of the current source transistors. See the PSRR of CS analysis in notes for more intuition on good PSRR design.
- Mirroring device should have poor transconductance for low offset, better matching(since overdrive will be higher) and noise contributions.
- Output resistance of the mirroring device should be high.
- Mirroring device should have Low input compliance (Low Input voltage requirement for a given current) and large output compliance (Low headroom)
- Should be a Low noise device.
- Operable for a wide range of currents.
- Device matching should be good
- The circuit should occupy least area
- Some things to take care in Bipolar mirrors:
- When is it going to saturation?
- When supply goes to 0 does the base bias become high impedance and retain its voltage? Compare to MOS. PNP Bipolar mirrors can cause high impedance point at base if the supply is ramped down suddenly.
- Beta and base current errors and how to reduce them
- Early voltage - direct calculation of ro
|Transconductance can be made smaller for a given current by making W/L smaller.(ADV)||Transconductance fixed for a given current unless a emitter resistance added to reduce gm|
|MOS Matching can be improved by increasing the channel area of the device.(ADV)||Emitter resistance needed to have better Bipolar matching, would increase area, and requires good resistor matching which may require large dimensions for the resistor|
|MOS generally has a lower output compliance voltage (Larger headroom) than bipolar. However, a good MOS current mirror (with small W/L) may have an even larger compliance voltage than bipolar.||Bipolar generally has a lower input compliance voltage (Low input voltage requirement for the same current) than MOS and a higher output compliance voltage. Degeneration resistors for the bipolar will tend to increase both compliance voltages. (ADV)|
|The output resistance can be increased by increasing the L. Can be increased by Cascoding. MOS cascoding generally easier than Bipolar. (ADV)||Bipolar output resistance is fixed for a given current. Degeneration resistor improves it. Cascoding will improve it.|
|The drain current noise of a MOSFET is proportional to the square root of the gm hence can be reduced by reducing gm. (ADV)||The collector current noise of a BJT is proportional to the square root of the current and hence cannot be changed by changing the gm.|
Bipolars generally have smaller flicker noise component. (ADV)
|MOS will hit subthreshold for low current values.||Bipolar operation is ideal over many decades of current. (ADV)|
|MOS devices will be smaller so better packing on chip and hence better matching. Generally to achieve very accurate matching comparable to Bipolars the size has to be made very large.||Bipolar devices are generally larger. But Bipolars are generally better matched devices, i.e. dI/I is better, and is improved a lot by using degeneration resistors (ADV)|
|A MOS Device has lower output impedance over frequency (at higher frequencies). This can be advantageous for situations where the PSRR needs to be improved at higher frequencies. (ADV)||A BJT device has higher output impedance over frequency(at higher frequencies) makes it useful if we want to bias the tail current of a diffamp operating at high frequencies. (ADV)|
|MOS devices have zero gate current hence no mirroring error. (ADV)||Bipolar transistors can have a MOS transistor to act as a Beta helper to remove the mirroring error.|
|MOS devices don't have the problem of leaking more current to the gate when they go to the triode region and hence causing loading or other effects to connected circuitry. (ADV)||For Bipolar mirrors care has to be taken to think about when they go into saturation at low supply voltages especially during startup and may cause some undesirable startup problems in silicon if not done carefully.|