What's the normal value for opamp bias current mirrors

[vistapoint]

And how to decide? The current mirrors seems only provide a reference current. Why not make it as small as possible? How about 1uA or 100nA? I have no idea how they decided that.

 

[ambreesh]

dear vistapoint
their are issues rtelated to noise.
Dr Boser has discussed it in his lectures.

 

[always@smart]

dear vistapoint
their are issues rtelated to noise.
Dr Boser has discussed it in his lectures.

-------------------------------------------------------------

oh..whre are Dr. BOser lecture note? where can we get it?

regards,
samrt

 

[kwkam]

usually, I will choose around 50~100uA except special requirement.

 

[vistapoint]

I saw a lot of designs by Johan H. Huijsing and his students. They used less than 10uA as reference current source for their opamp cells. Sometimes it's even less. What's the effect if small current is used? Will the crcuit still be biased properly? Is noise significant? I am always not very certain about noise. It seems intangible.

 

[jiangwp]

I think it is considered based on matching to reduce the erro of current mirror.

In the meanwhile , the noise and psrr are the key factors effecting the performance of opamp due to the bias importance.

 

[alvays]

that/s a good question. I also meet this situation, but i havent thinked about it.

what Jiangwp said is about the w/l ratio and length so on. but what the author said is about the bias reference current

 

[jordan76]

Hi alvays,

I agree with jiangwp. Suppose the error of a certain current mirror due to mismatch is 0.1uA, if the bias reference current is 1uA, the error is 10%;
but if the bias reference current is 10uA,then the error is only 1%.

Hope it helps.

regards,
jordan76

 

[nEuron]

i did read some papers on bandgap circuits. some of the designs are having very low current consumption. for example, current consumption less than 20uA.
if that is the case, what kind of actions we need take to make sure the error is small enough when a small current is used?

 

[alvays]

jordan, do you mean that the absolute current error has no dependance on the current volume?

However, i would like to consider it as the percentage if the current is not too small. that is, if the error is 0.1uA when volume is 1uA,then it will be 1uA when volume turns to be 10uA

the error comes from random error and systematic error, the former comes from technology mismatch, the latter comes from design method, for example, difference of vds between input voltage and output voltage

 

[jiangwp]

alvays, you are right. I agree with you.

What you say is a part of what i mean.

First, The bias current is limited by the system power dissipation.

Second, for a good performance amplifier, it must be robust operation over wide process variations and termperature. The bias determines the operation point for a amplifier, so good bias is a precondition for a good amplifier.

Third, for the low noise and precision amplifier, the noise of transistor and the fluctuate of power supply of bias will effect the performance of amplifier. so the size and ratio of transistor and the psrr are also factors for bias.

 

[paladinzlp]

I think bias current is decided by the whole circuit's
power(such as OP-AMP,BANDGAP)and other design SPEC(such as gain(f=0)).

 

[vistapoint]

I raised the question because I saw some designs with amaizingly small current. They use 200nA for current sources, 0.8uA for the input pair, 100nA for CMFB amplifier.

I am now making my own design. I doubt the above bias currents will make the circuit work, even though the power consumption is attractive. The specs are not tight on gain and noise.

Another factor is that I am a newbie. I want my chip work first. What's the safest way to bias the current? or what kind of bias makes the circuit not robust?

 

[paladinzlp]

What is your circuit's vcc and vss?

 

[vistapoint]

3.3V and 0? Any suggestions?

 

[songsong]

It depends on the specified slew rate.