[SOLVED] Current reference, is it possible without using resistors and having a 5% tolerance?

[jgk2004]

Hello all,

Are there any possible architectures which can produce an accurate current reference without using a resistor? I have a set temperature and Supply since it is for an implement. I need a 15uA current with a +-5% tolerance. Unfortunately, all of my resistors in my technology have a +-15% so thus I can not use them.... Additionally, I can not do a resistor trim because I am pad limited in the design....

Does anyone have any ideas?

JGK

 

[Junus2012]

Do you know I have the same problem,
even during the corner analyses give a huge change on the resistor value that effect the current

I didint hear about a current reference circuit without a resistor, it will be interesting for me as well if i found one like this

 

[dick_freebird]

Some foundries' e-trim can be accessed by serial inputs
with the programming ckty embedded (not pad-per-bit).
Maybe not yours, but you could ask.

Some technologies may support laser trimming if the
resistor is thin-film and the oxides don't mess you up
optically. There were also laser-interconnect schemes
purported to be fairly process-portable (creating
line-line bridging when zapped under bias) at one
time, but this seems to have not caught on.

I haven't seen a process that runs better than 5%
tolerance -on anything-, and I've worked with a
bunch.

 

[BradtheRad]

Here is something which might be suitable. I don't know much else about it.

https://en.wikipedia.org/wiki/Bandgap_voltage_reference

"A bandgap voltage reference is a temperature independent voltage reference circuit widely used in integrated circuits, usually with an output voltage around 1.25 V, close to the theoretical 1.22 eV bandgap of silicon at 0 K."

 

[Junus2012]

still it use the resistor , he wants without resistors

Here is something which might be suitable. I don't know much else about it.

https://en.wikipedia.org/wiki/Bandgap_voltage_reference

"A bandgap voltage reference is a temperature independent voltage reference circuit widely used in integrated circuits, usually with an output voltage around 1.25 V, close to the theoretical 1.22 eV bandgap of silicon at 0 K."

- - - Updated - - -

it is right, 5% change is pretty good. you may not find better tolerance than this value

Some foundries' e-trim can be accessed by serial inputs
with the programming ckty embedded (not pad-per-bit).
Maybe not yours, but you could ask.

Some technologies may support laser trimming if the
resistor is thin-film and the oxides don't mess you up
optically. There were also laser-interconnect schemes
purported to be fairly process-portable (creating
line-line bridging when zapped under bias) at one
time, but this seems to have not caught on.

I haven't seen a process that runs better than 5%
tolerance -on anything-, and I've worked with a
bunch.

 

[erikl]

Are there any possible architectures which can produce an accurate current reference without using a resistor?
Does anyone have any ideas?

Hello John,

here's the earliest resistor-less CC source (from E.A. Vittoz) which I found:

I tried it in a .18µm process for a 5µA source, the temperature stability was excellent, and the spread over the comm. temp range and a single wafer was below 5%, AFAIR. MC simulations incl. full PVT ranges resulted in current changes <10% within 1σ .

Below is another paper on a rather complex resistor-free CC source - unfortunately I got only the first 2 pages from the article (A_Resistor-free_temp-comp_CMOS_current_reference_p1-2.pdf). I didn't try that one.

Finally one more on a very low current (9nA) resistor-less CC source from De Vita/Iannacone (~2006 , 0.35µm) :

erikl

 

[jgk2004]

Hello all,

Thanks for the help! Its funny but I just build that Vittoz reference two nights ago Erikl, nice suggestion. It works great and looks like it meets the specs I need. I get 5% tolerance over a 200mV change is supply and that meets my requirements. Last night I also built the A_Resistor-free_temp-comp_CMOS_current_reference_p1-2.pdf. That one had around 10% tolerance over MC and thus I did not use. Of course it has a better temperature robustness since it has a PTAT and CTAT, but since my application is implements it only has to work at body temperature, thus 37C!

Thanks for the suggestions

Jgk

- - - Updated - - -

Hello all,

Thanks for the help! Its funny but I just build that Vittoz reference two nights ago Erikl, nice suggestion. It works great and looks like it meets the specs I need. I get 5% tolerance over a 200mV change is supply and that meets my requirements. Last night I also built the A_Resistor-free_temp-comp_CMOS_current_reference_p1-2.pdf. That one had around 10% tolerance over MC and thus I did not use. Of course it has a better temperature robustness since it has a PTAT and CTAT, but since my application is implements it only has to work at body temperature, thus 37C!

Thanks for the suggestions

Jgk

 

[Junus2012]

do you have an article for the circuit you used ?

thank you

Hello all,

Thanks for the help! Its funny but I just build that Vittoz reference two nights ago Erikl, nice suggestion. It works great and looks like it meets the specs I need. I get 5% tolerance over a 200mV change is supply and that meets my requirements. Last night I also built the A_Resistor-free_temp-comp_CMOS_current_reference_p1-2.pdf. That one had around 10% tolerance over MC and thus I did not use. Of course it has a better temperature robustness since it has a PTAT and CTAT, but since my application is implements it only has to work at body temperature, thus 37C!

Thanks for the suggestions

Jgk

- - - Updated - - -

Hello all,

Thanks for the help! Its funny but I just build that Vittoz reference two nights ago Erikl, nice suggestion. It works great and looks like it meets the specs I need. I get 5% tolerance over a 200mV change is supply and that meets my requirements. Last night I also built the A_Resistor-free_temp-comp_CMOS_current_reference_p1-2.pdf. That one had around 10% tolerance over MC and thus I did not use. Of course it has a better temperature robustness since it has a PTAT and CTAT, but since my application is implements it only has to work at body temperature, thus 37C!

Thanks for the suggestions

Jgk

 

[erikl]

do you have an article for the circuit you used ?

Ciao Junus,

the above shown slide I found in a copy of E. Vittoz' lecture during a
CCCD Workshop in Lund, Oct. 2-3, 2003
WEAK INVERSION IN ANALOG AND DIGITAL CIRCUITS, slide 15

I found the same circuit also in this book:
Alice Wang, Benton Highsmith Calhoun, Anantha P. Chandrakasan
Sub-threshold Design for Ultra Low-Power Systems
© 2006 Springer Science+Business Media, LLC, 233 Spring Street NY 10013
Chap. 8: Analog Circuits in Weak Inversion by Eric A. Vittoz
8.4 Exponential Characteristics
8.4.1 Voltage and Current Reference, p. 156

Saluti, erikl