Implementing current mirrors with MOS or BJTs in a BiCMOS ?

Current Mirrors

Is it better to implement Current Mirrors using MOS or BJTs in a BiCMOS process? One reason I can think in favor for MOS is that there is no gate current so the mirroring will be more accurate compared to the Base current error of BJTs. Are there any other important reasons?

Re: Current Mirrors

Current mirrors implemented in CMOS has a smaller area than that in bipolar.

Current Mirrors

offcourse CMOS due to low power and better matching

Current Mirrors

it takes care of area or even can be more densely packed transistors,

reduces the cost of the chip

Re: Current Mirrors

aryajur said:

Is it better to implement Current Mirrors using MOS or BJTs in a BiCMOS process? One reason I can think in favor for MOS is that there is no gate current so the mirroring will be more accurate compared to the Base current error of BJTs. Are there any other important reasons?

Click to expand...

But if u want your current mirrors can get better performance in practice, I think BJTs is indispensable.

Added after 6 minutes:

There are some examples in practical circuits.

Re: Current Mirrors

qiushidaren said:

But if u want your current mirrors can get better performance in practice, I think BJTs is indispensable.

Click to expand...

Sorry, I didn't understand the reason, why are BJTs (used as mirrors) better in 1st circuit below?

Re: Current Mirrors

Also, I am doubt what obi1 said. Why the bipolar has better matching than CMOS?

Re: Current Mirrors

sorry,I can't understand.Can you explain why BJTs is better

Re: Current Mirrors

I think that, in general, MOS transistors are better current mirrors due to better matching and zero gate current. Also, to match BJT's you usually need to degenerate their emitters for better matching performance. This can be area and headroom consuming. ( Grey and Meyer is a valuable source for data in this regard )

However, BJT's can offer some advantages as current mirrors/sources:
1. They usually have larger output impedance
2. They have very small flicker noise components. ( However, they usually have larger thermal noise )

Re: Current Mirrors

just go through Behzad razavi..

Re: Current Mirrors

aryajur said:

Is it better to implement Current Mirrors using MOS or BJTs in a BiCMOS process? One reason I can think in favor for MOS is that there is no gate current so the mirroring will be more accurate compared to the Base current error of BJTs. Are there any other important reasons?

Click to expand...

One other important reason is that more compact cmos layout can have better matching property.

Re: Current Mirrors

aryajur said:

qiushidaren said:

But if u want your current mirrors can get better performance in practice, I think BJTs is indispensable.

Click to expand...

Sorry, I didn't understand the reason, why are BJTs (used as mirrors) better in 1st circuit below?

Click to expand...

They are used to get a low temperature coefficient.
VREF=VBE+kVT, can get a low tempco.

Added after 2 minutes:

VBE has a negative tempco, and VT has a positive one, so by choosing k correctly, we can get a low even zero tempco.

Current Mirrors

the mismatch of mos is large

Re: Current Mirrors

montage3000 said:

the mismatch of mos is large

Click to expand...

Regarding this, I also noticed that in one of the BiCMOS process I am using, the mismatch of Vth between 2 identical MOSFETs can be 15mV maximum, while the mismatch of Vbe's of 2 identical BJTs has been reported to be 2mV maximum. Doesn't this data suggest that the matching of the Bipolar devices is better?
If this is the case should the configuration shown below be the best mirroring configuration? The emitter resistor is to reduce effect of Is mismatch.

Re: Current Mirrors

Here is my take on bipolar vs MOS for current mirrors:

1- To make a good current mirror, you want POOR transconductance in the devices for low offsets and noise contributions. With poor transconductance, a small change in input current represents a large change in gate or base voltage. It also takes a larger base or gate voltage change to cause an equal current change at the output current.

For a given current, the bipolar transconductance is constant, which means that the use of degeneration resistors is the only way to reduce transconductance. This means large area, and your matching is now dependant not only on the bipolar, but the resistor matching, too!

For a given current, the MOS transconductance can be changed using W/L. Making a small W/L will make a small transconductance. (Using long L will also help with output resistance of the mirror)

Advantage - MOS.

2- A good current mirror has a low input compliance voltage and a wide output compliance voltage.

Bipolar generally has a lower input compliance voltage than MOS.

MOS generally has a lower output compliance voltage than bipolar. However, a good MOS current mirror (with small W/L) may have an even larger compliance voltage than the bipolar. Degeneration resistors for the bipolar will tend to increase both compliance voltages.

Advantage - possibly slight advantage to bipolar.

3- A good current mirror has a high output resistance.

Bipolar output resistance is fixed for a given current. Again, the degeneration resistors will improve this.

With a MOS device, the output resistance can be changed by changing L.

Both bipolars and MOS can be cascoded for improved output resistance. MOS cascodes are generally easier to make, especially when many current mirrors are needed.

Advantage - MOS.

4- A good current mirror has low noise.

Bipolar current noise is sqrt(2QIB), MOS current noise is sqrt(8/3KTGmB). For a given current, the noise from the bipolar is fixed, unless degeneration resistance is added. Again, low tranconductance (Gm) improves MOS performance for noise.

Advantage - MOS.

5- A good current mirror operates well over a wide range of currents.

Bipolar operation is ideal over many decades of current. MOS hits subthreshold operation at low currents and velocity saturation at high currents.

Advantage - bipolar.

As you can see, the MOS tends to win in most comparisons. However, it can be seen that when certain aspects are critical, there are cases where the bipolar will win.

Re: Current Mirrors

Hello JPR,
thanks for your reply. I had a couple of questions. By the smaller input compliance on Bipolar transistors do you mean that the input voltage variation will be less due to the exponential dependance of current? If this is the case why is this property desirable for current mirrors? Also why are MOS cascodes easier to make?

Re: Current Mirrors

I too agree with JPR on his observations and facts listed

Current Mirrors

I THINK IF YOU WANT TO GET A GOOD CURRENT MIRROR ,YOU SHOULD USE A BANDGAP OR JUST A PATA CURRENT

Re: Current Mirrors

By compliance voltage, I was talking about the voltage overhead needed in order to operate the current mirror. Vbe will typically be smaller than Vgs. Vce_sat and Vds_sat will probably be similar for similar performing current mirrors.

In terms of the statements about using a bandgap current reference or a PTAT current reference, they are good if what you are looking for is a specific current. However, if you want to mirror a current that already exists (in the load of a differential pair, for example), you would want a current mirror rather than a reference.

Current Mirrors

Hmm, there seems to a lot of reasoning but still hard to tell which one is better. As far as I heard, BJT is much better in terms of matching. One question here, does CMOS have poor matching in subthreshold? if so, any reason?