Current Reference - Beta Multiplier

[AMSA84]

Hi guys,

I have to design a current reference. I have decided to design he Beta Multiplier Current Reference Source.

This kind of topology is power supply independent.

While searching for other topologies, I saw that there is too a topology called Delta Vbe Current Source.

Can someone tell me what are the main differences between the beta multiplier topology and the Delta Vbe topology?

Kind regards.

EDIT: Taking the advantage of this post, allow me to ask another thing:

I would like to know how temperature will affect the Beta Multiplier Current Reference. I was wondering if that it is possible to be done with the DC analysis selecting the temperature field and varying it from -25ºC up to 125ºC. It is a reliable way or this must be done using the PVT analysis?

 

[beetwee]

What do you mean with "Delta vbe current source". Please provide a schematic and/or more information.
E.g. in a beta multiplier with the transistors biased in subthreshold, the voltage across the resistor (so the reference current) is proportional to the thermal voltage kt/q, just like in case of 2 BJT's with a different current density.

Yes, you can simulate the temperature dependency with a DC analysis with the temperature option. Don't forget to add a start-up circuit!

 

[AMSA84]

Hi,

Thanks for the reply. In facatI don't remember the page on Bakers book, but you can put on google images this: delta vbe current source and you'll find the schematic. It's based on two transistor - left branch with one PNP right branch 8 PNP (8 because ln(8)=2 and it's more easy to do the math). The current mirrors are cascode.

Regarding the temperature, I asked that because I thought that when doing corners (PVT analysis) the models used or so could be different and doing the DC Analysis varying the temperature would mislead the results.

So the results that I will get, thinking on typical (@ 25ºC), using the DC analysis varying the temperature will be the same as if I was doing the corners but only varying the Temperature there?

About the comment you made of threshold, that means having the transistors Vdsat around what, 100mV? 50mV?

Apart from this, allow me to ask another question.

What would be a suitable or a good result regarding the current variation function of temperature? That is, for example, for a 10uA current reference a good spec would be 10uA +- 10% from -40ºC to 120ºC or the spec is more relaxed? I am not very familiar with this. The specs here are flexible.

For the beta multiplier topology, considering a 20uA reference current, I got + 7uA @ 125ºC and -5uA @ -40ºC. This is good result? It is something like 20uA +35% and 20uA - 25%.

About the other question I did, the difference between the delta vbe current source and the beta multiplier, is that the beta multiplier is power supply independent and the delta vbe current source is power supply independent and temperature independent? If it is temperature independent it will have a smaller variation of the current reference when compared with the beta multiplier?

Thank you very much for your attention, in advance.

Kind regards.

 

[D.A.(Tony)Stewart]

To design "anything" , 1st you define all input and output constraints and environmental range and supply range.

Then you define stability, tolerance and start/stop behavior.

Then you can compare solutions.

Using custom L/W ratio bandgap voltage reference you can choose e. g. 1.5 and Vds max > Vdd.

The Beta Multiplier Current Reference Source uses such custom L/W ratios with a current mirror and folded cascode output.

 

[beetwee]

Hi,
Regarding the temperature, I asked that because I thought that when doing corners (PVT analysis) the models used or so could be different and doing the DC Analysis varying the temperature would mislead the results.

So the results that I will get, thinking on typical (@ 25ºC), using the DC analysis varying the temperature will be the same as if I was doing the corners but only varying the Temperature there?

In case of a corner simulation you can choose between fast,slow and typical transistor (and resistor) models. In case of a DC temperature sweep you use the typical model. That's good, because you design with a typical transistor and determine the minimum/maximum performance using corners and MonteCarlo simulations.

About the comment you made of threshold, that means having the transistors Vdsat around what, 100mV? 50mV?

When Vgs<VTh, the voltage across the resistor is n* kT/q *ln (K), with K the ratio of the mosfets. You still want the transistors to be in saturation, so vds must be >100mV! If I remember correct, you can also find something about the beta multiplier in subthreshold in the book of Baker.

What would be a suitable or a good result regarding the current variation function of temperature? That is, for example, for a 10uA current reference a good spec would be 10uA +- 10% from -40ºC to 120ºC or the spec is more relaxed? I am not very familiar with this. The specs here are flexible.

For the beta multiplier topology, considering a 20uA reference current, I got + 7uA @ 125ºC and -5uA @ -40ºC. This is good result? It is something like 20uA +35% and 20uA - 25%.

I think you cannot do it better than that. +-25-30% is normal. In a beta multiplier you can optimize the dependency of VDD with a better design (e.g. a cascode current mirror), but you cannot make the temperature dependency better because it's determined by the resistor and mosfet models.

You can make a more accurate current source using a bandgap, resistor and opamp, but then the accuracy of the current is then still limited by a resistor, which has about +-15% accuracy.

 

[AMSA84]

But beetwee, when you say using a band gap, that has nothing to do with that beta multiple. Those are tão different topologies. I don't know if you have searched in the google images, but I believe that the delta vbe current source don't have nothing to do with the band gap you are referring to. However it uses BJT transistors.

What about that delta vbe current source? It migh allow a better accuracy?

 

[FvM]

Is it asking too much if we want to see a circuit or a link?

There are apparently some misunderstandings, beta multiplier is usually referring to MOSFET reference circuits. "Delta vbe" is a common operation priciple of most bipolar references and in so far rather unspecific. It may be that you have more specific circuits in mind, but it's no clear which.

 

[AMSA84]

There is the schematic.

 

[FvM]

The circuit shows a PTAT current source, output current proportional to absolute temperature. See e.g. Razavi, Design of Analog CMOS Integrated Circuits, Chapter 11.4 PTAT Current Generation.

Generation of a (approximately) temperature independent current need an additional compensation means.

 

[AMSA84]

So that means that this circuit it is not totally temperature independent. However it is more accurately than the previews one? (beta multiplier)

Do you know any reference where I can find a temperature independent (and at same time power supply independent) current source?

Thanks in advance.

 

[beetwee]

A 1rst order temperature independent bias requires more complexity than the relative simple beta multiplier and often has the following parts:
- a PTAT reference
- a CTAT reference
- an adding circuit

How complex may it be and what are the specs regarding absolute accuracy? If you use a bandgap, an opamp and a resistor you can get a good temperature independent current source. But the absolute accuracy of the current is then still determent by the accuracy of the resistor (+-15%).

Regarding whether a beta multiplier or the delta vbe circuit is better depends on other specs. The BJT diodes requires 0.6V voltage headroom. If you design for 1.2V you don't have that voltage headroom. For 2.5V on the other hand, it won't be a problem.
I don't know exactly how good a beta multiplier is strong inversion regarding temperature dependency. But I know that a beta multiplier in subthreshold is comparable with the delta vbe circuit. I advise you to simulate both circuits and test it yourself (and share it with us so everyone can learn from it).

 

[AMSA84]

Thanks for the help beetwee (and others).

That's what I'll do. I'll try both circuits and see the results. I found some references on how to "test" those circuits (nothing special though).

But from what you said, I think that I can conclude that the effort on using the bandgap, opamp and a resistor does not compensate for the gain that one would have (+-15%) when comparing to the beta multiplier or delta vbe current source (+-30%).

My power supply voltage is 3.3V.

Regards.

 

[AMSA84]

EDIT: Guys, I forgot to ask another thing.

I have several bias current around 500uA. It is a better practice to design a current source with an output current of 1uA or 10uA, and then, when biasing the other blocks use transistors with W ~ 1000um or it is better to have a output current of 100uA and small transistors? I know that you might say that will depend on power consumption and area. Both are open, nothing is defined.

 

[beetwee]

In my opinion it's a waste of power to make the reference 100uA or 500uA. What's best depends on the specs of your circuit and the layout. E.g. a ratio of 1:100 is possible, but in the layout it will require more work and mismatch can become a problem (again, depending on the application). I would use a ratio around 1:10. So 50u bias current.

When biasing is only 5-10% of your current consumption, you don't gain much by optimizing the current mirror ratio and i would easily consider it "good enough".

 

[AMSA84]

Ok beetwee, I see. So a ratio of 1:10 is a good choice. I will try first to get a new comparator and a new OTA that consume less current.

Taking the advantage of this post, can anyone tell me a nice a nice temperature independent current source topology?

I have seen several topology, PTAT, CTAT, others PTAT with OPAMP, and other one with an OPAM, resistor and MOST (has someone mentioned earlier).

You can see then here: **broken link removed**

As you can see in that picture you have all the types of current sources I mentioned.

Now the question is, we can make an almost temperature independent current source using the PTAT and CTAT. I can imagine that it is putting together circuit 1 and circuit 4 (or 3). But How can I do that? Don't know if it is possble.

Another solution would be using a bandgap voltage reference in the Vref part of circuit 5, and then make the current reference using the transistor with the resistor. Being the bandgap voltage reference almost independent of temperature, I wonder if the current would be too. One must take into account that there are process variations with temperature that might influence the final result (current) because of the OPAMP and the resistor.

Does any one can help me out here? How to get an almost current source independent of temperature?

Kind regards

 

[beetwee]

I can't open your attachment. Can you upload it again?

 

[AMSA84]

Yes I can. Wait a second.

Here you go:



- - - Updated - - -

Basically, my idea was to put together a PTAT current source and a CTAT, to compensate the temperature influence. I think that using two current mirror into a node and then another current mirror to get the compensated current is possible.

 

[beetwee]

Adding currents together isn't that hard: e.g. mirror the PTAT and CTAT current to a PMOS transistor. Then connect the drains of these 2 PMOS together. Connect that to a nmos mirror and you have the sum of 2 currents...
The challenge is to make sure that the PTAT and CTAT current have the same (opposite) temperature coefficient.

It is also possible to make 2 PTAT voltages and substract those voltage from each other. That leaves you with a temperature independent voltage (which can be converted to a current with a resistor).

Another option is to look at the paper of Banba about his bandgap reference. You can find it on google or IEEE Xplore. His output is in fact a temperature compensated current.

 

[AMSA84]

Yes, that's right. However, I think that it is not required 100% that the PTAT and CTAT current have the same temperature coefficient. To achieve that the output current reference should be a straight line, correct?

Normally, from what I have seen, what is possible to achieve is something like an hiperbola, right? One can allow 1% or 2% of error, or that is too much? Of course it will depend on the application, but generally speaking or from what you have seen?

 

[beetwee]

From what I remember you could get an accuracy of ~2% over temperature when you compensate with a PTAT and CTAT current. Then you get indeed an inversed parabola.