Current mirroring a DAC output, instead of resistor loading

[sharkies]

When designing a current-steering DAC, your load is usually a 50Ohm load.
If you look at slide6 of the attached PPT slide(by Baker), this particular design uses a current mirror load instead of the resistor.
Does anybody have more reference on this method? I found the original paper for the ppt slide, but it doesn't give any further details.

I have a feeling that there must be DAC designs out there with current mirror loads instead of the normal resistor load, but I'm not able to find one yet.



Please let me know. Thank you

 

[erikl]

Re: Current mirroring a DAC output, instead of resistor load

... this particular design uses a current mirror load instead of the resistor.

I think this has 3 advantages:

• Independent of the current, the voltage at the op & om nodes is (nearly) constant - in contrary to a resistor load. This helps for accuracy.
• The current mirror multiplies the current (by a factor of 10 in this case), so the output resistors can be a factor of 10 lower.
• The output resistors/voltages now are referenced to GND

 

[sharkies]

Re: Current mirroring a DAC output, instead of resistor load

-> op & om nodes is (nearly) constant
I'm not so sure on this, In fact op & om will change as the current on each branch changes. For example when current is all steered on one side, than the diode connected transistor will have high Vgs, but when the current is near zero, Vgs will have be near zero as well.
I'm not convinced that this will particularly suit higher accuracy, and that's why I'm looking for a paper on it.

 

[dick_freebird]

For best fidelity a current DAC should be operating into a
constant voltage (typ virtual ground). This eliminates output
common mode (Early or lambda) effects.

A current mirror is likely to be more bias-invariant than a
resistor, as far as outut voltage deflection vs current in
total, but more nonlinear. If you servoed the output voltage
point to remain fixed you could get better DC accuracy
but probably at the expense of bandwidth and settling time.

Of course the last DAC I worked on was built with NJFETs,
single ended output, and about 15 years ago.

 

[sharkies]

Re: Current mirroring a DAC output, instead of resistor load

dick_freebird, I can agree with you on what you said.
However, my concern is with the 1/gm of the current-mirror load. When we have a 50ohm load, as long as the finite output impedance of the current sources in each cell is high, we can guarantee linearity up to a certain degree. Formula can be derived, and it's in literature.
When we have a current-mirror instead, we have 1/gm and not 50ohm. 1/gm is a lot bigger than 50ohm. And therefore, to guarantee to same linearity, we need higher output impedance of the current mirror.... I think... right?