| Author |
Message |
liushangpiao
Joined: 27 Mar 2008
Posts: 17
|
 03 Oct 2008 12:58 Bandgap Problem!! |
|
|
|
Hi,everybody!
I have some problems about BANDGAP circuit design here.
Those are below:
1、The ration of the two transistor area is usually 8 ,not 4 ,16 ,why?
2、Why is the temperature coefficient of Vbe -1.5mv/k to -2.3mv/k?;
what is it related to?
3、what type of BJT is ususlly used ?why?
4、What type of resistor is used,polyresistor,P/N diffusion resistor or others? why?
who can give me some answers,thank u very much!
|
|
| Back to top |
|
 |
mpig09
Joined: 26 Aug 2005
Posts: 123
Helped: 6
|
| 04 Oct 2008 14:50 Re: Bandgap Problem!! |
|
|
|
Dear liushangpiao:
1、The ration of the two transistor area is usually 8 ,not 4 ,16 ,why?
==> for layout matching, 1:8 is the minimum ratio.
2、Why is the temperature coefficient of Vbe -1.5mv/k to -2.3mv/k?;
what is it related to?
==>please explain clearly.
3、what type of BJT is ususlly used ?why?
==>I usually the smallest size BJT for the smallest layout area.
4、What type of resistor is used,polyresistor,P/N diffusion resistor or others? why?
==>if you want to design bandgap, Temp. Coef. is a important point, so you need to know which Temp. Coef. of resister type is suit for your circuit.
I hope these can help you.
mpig
|
|
| Back to top |
|
|
haff99
Joined: 10 Mar 2007
Posts: 36
Helped: 4
|
| 04 Oct 2008 15:50 Re: Bandgap Problem!! |
|
|
|
Most of the questions were already answered in the first post but I'll add a few comments.
1. 8 to 1 is typically used so you can layout the device with the 1X device in the center and the 8 devices built around the 1X device. That is why you typically see 8 to 1. You can use 4 to 1, 16 to 1 or anything else you want.
2. It doesn't really matter what the TC of the VBE is because you end up adjusting your (r2/R1)KT/q voltage to cancel it out. You sort of have to rely on the fact that your models are correct in the simulator to this exactly. VBE's should, in most cases, be modeled extremely accurately. THe typical book example of a VBE is -2mV/C. Obviously there will be some differences with your particular process.
3. Again you can use whatever size BJT you want. Typically bigger emitter areas=better matching. If size is criticaly go smaller. Also, depending on your conditions you may want to plug your BJT's to decrease the collecter resistance. Again, it all depends on your design.
4. REsistors are the same as well. YOu typically will want to use resistors that match the best since the R2/R1 ratio is critical. If size is important you use the resistors with highest sheet resistance. TC of the resistors shouldn't be an issue since the TC's should cancel out. It becomes an issue if you have some sort of current spec over tempeature or something like that. So again it all depends on what your design calls for.
|
|
| Back to top |
|
|
fmgay
Joined: 19 Jun 2008
Posts: 7
|
| 05 Oct 2008 14:02 Bandgap Problem!! |
|
|
|
The answers of above two posts are very well, I want to say a little more:
for issue 4, I always use POLY resistors, for the sake of small Voltage Coef.
as diffusion resistors or well resistors, you must connect substrate to proper bias, that will intruduce some mismatch as results of different voltage.
|
|
| Back to top |
|
|
liushangpiao
Joined: 27 Mar 2008
Posts: 17
|
| 12 Oct 2008 5:28 Re: Bandgap Problem!! |
|
|
|
Thanks very much for your kindness...
I believe I have understaned these questions....
thanks again..
|
|
| Back to top |
|
|
endru
Joined: 10 Nov 2007
Posts: 21
Helped: 2
|
| 12 Oct 2008 10:19 Re: Bandgap Problem!! |
|
|
|
Just want to add, for question 1: the higher the ratio, the better it is (to minimize VBG variation due to mismatch).
Because if there is an offset within your opamp or any device which make sure that the voltages between the two branches are the same (if you use that structure), your PTAT current is no longer (delta VBE)/R, but rather (delta VBE + offset)/R. Hence, if your delta VBE is big, the effect of offset can be minimized.
However, the area and/or the current become the trade off.
For example, with ratio of 8, you have delta VBE = 54 mV.
To increase it to twice, you need the ratio of 64, which is pretty large.
One of the ways to solve this is to stack the BJTs. Two BJTs with area of 8 in one branch, and two BJTs with area of 1 in one branch. Or, you can make 4 branches, one branch is for PNP with area of 8, one branch is for another PNP with area of 8-but with its base connected to the emitter of the first branch, and two branches for BJTs with area of 1 with the same configuration (this is useful if you make use of vertical PNP out of CMOS process). With this, your delta VBE is simply multiplied by two, and your total area only increases twice. You can refer this to this book: analysis and design of analog integrated circuits.
Hope it is useful. =)
|
|
| Back to top |
|
|
liushangpiao
Joined: 27 Mar 2008
Posts: 17
|
| 14 Oct 2008 4:16 Re: Bandgap Problem!! |
|
|
|
Thanks again.
Your advices are very useful for me..
|
|
| Back to top |
|
|
