Circuits between Banggap and DAC core circuit(current steering)

[turtlewang]

hello everyone,

Now I am studying the current steering DAC. Since what the DAC core required is a precision current. Now I have one bandgap on chip which can provide a 1.2V voltage. I want to use the temperature-independent voltage to the required current. How can I achive that?

Are there any conversion circuits that commonly used in the literature to convert this temperature-independent voltage to the required current?

what's the design requirements of this conversion circuit?

Are there any papers that I can refer to?

Thanks

Charles

 

[palmeiras]

Hi Charles,
A very common way is to use a operational amplifier to force the voltage across some resistor equal to the bandgap voltage (~1.2), therefore, generating a current almost independent of the temperature. However, due to the resistor process variations, this current can vary more or less 20%. I don’t think this is enough for you. Is it?
There are other current sources in literature that the output current doesn’t depend on resistors values.

What is the specification of your current?
Best regards,

 

[turtlewang]

Thanks!
I think a external resistor will be OK, is that right?

But I don't understand the meaning "What is the specification of your current". What I want is to generate the corresponding current that needed by the current steering DAC core.

For example, I want to design a 12-bit current steering DAC with INL<=2LSB, DNL<=1.5LSB; SFDR>=55dB in the Nyquist range while the clock frequency>500MHz.

I use the current steering architecture which is 6bit binary+6bit unitary. The current source in the DAC core is a cascoded current source. So the full range output current is the sum of all the current sources.

So i need convert the bandgap voltage to a precision current. Then mirror the current to the DAC core.

So ,another question is how to know the specification of the current from the overall specification (they are INL/DNL,SFDR/SNR).

Thanks very much!

Hi Charles,
A very common way is to use a operational amplifier to force the voltage across some resistor equal to the bandgap voltage (~1.2), therefore, generating a current almost independent of the temperature. However, due to the resistor process variations, this current can vary more or less 20%. I don’t think this is enough for you. Is it?
There are other current sources in literature that the output current doesn’t depend on resistors values.

What is the specification of your current?
Best regards,

 

[palmeiras]

HI Charles,

Perfect! If your design permits to use an external resistor, in such way you can adjust it or choose the proper value, this solution is good for you. I have seen current steering DAC that uses this approach.
My above question was not proper made. The approach we discussed is valid. Regarding the temperature behavior of this current, it would be almost equal to the temperature coefficient of your external resistor when I = VBR/Resistor. Therefore, if you know an estimation of this value, you can simulate the impact of this variation. Sometimes the designers try to configure the temperature coefficient of the BGR voltage to have the some TC of the resistor. In such way, your current will present “zero” TC.
But anyway, your simulations will say that if you need to do this or no. I don’t think so.

Regards,

 

[turtlewang]

Hi palmeiras,

Thanks!

I am wandering are there some other methods that convert the BGR voltage to current. Since now, I want to compare some methods, find the pros and cons of different methods. Then I will choose one according to my design. Any paper or materials can i resort to ?

HI Charles,

Perfect! If your design permits to use an external resistor, in such way you can adjust it or choose the proper value, this solution is good for you. I have seen current steering DAC that uses this approach.
My above question was not proper made. The approach we discussed is valid. Regarding the temperature behavior of this current, it would be almost equal to the temperature coefficient of your external resistor when I = VBR/Resistor. Therefore, if you know an estimation of this value, you can simulate the impact of this variation. Sometimes the designers try to configure the temperature coefficient of the BGR voltage to have the some TC of the resistor. In such way, your current will present “zero” TC.
But anyway, your simulations will say that if you need to do this or no. I don’t think so.

Regards,

 

[palmeiras]

Hi Charles,

The discussion above can be found in CMOS Analog Circuit Design - Allen and Holberg (section Bandgap Voltage references).
What you can also do is to design other current reference that does not use resistors. There are several works in literature. For instance:

Low Voltage CMOS Current and Voltage References without Resistors - Christian Falconi, Arnaldo D’Amico.

If you find a better way, please tell me, and we can discuss more about it.

Regards,

 

[turtlewang]

Hi palmeiras,

Thanks, I will read these materials. Then we will discuss later

Regards