Parallel DACs: How many I/V resistors ?

[Bernhard]

Preparing a prototype 16x chip per channel with passive I/V.
Don't want to stack the chips, so there are traces of a certain length and resistance between the chips.

Any opinions if I should use a 500 ohm resistor from each Iout to ground, or just put a single 30 ohm resistor in the middle of the trace ?

The trace resistance ( mohm ? ) could have an impact on the whole thing with that low ohm I/V resistor ???

 

[dddac]

îf you use dacs with current output, in my opinion, it is the best to lead all current lines to one resistor, where the currents add up. The current is not changing depending on the length of the wire (mOhms), so you should take the voltgae directly from the two legs from the I/V resistor ...

best regards
doede

ps which chips ?

 

[hagtech]

I would worry more about crosstalk and ground noise. Having a single resistor sort of forces you into thinking about traces going to the same place. That one ground at the single resistor has to go to each dac. Make sure digital side noise isn't getting coupled in.

Other than that, I don't see a problem with separate resistors for each. The one benefit would be that if the dacs were socketed, then you could play with any number of them installed at a time, without having to change resistors. Output voltage would be constant, regardless.

jh

 

[triode_al]

If I am correct, the currentsource resistance is somewhat like 1.000 to 1.500 ohms (for instance the PCM56 appl note states 1k2 for Output Impedance).
So comparing this with the sampling resistor (like 500 ohm/n where n=number, for the PCM56 series; but 100/n can be used as well) should not influence the output of one over the other.

To me there is no difference of one 30 ohm or several 500 ohms resistors: they are connected at the top (output) anyway.

But, depending on the output topology (is is not a resistor!) there might be an influence of interference. Now my reasoning is as follows. By adding an output resistor of 5 ohms you will still have the same current value (the current never dies), from each chip, which sum up. The extra stopping resistor stops any far end interference in the mesh between all output pins.
regards
albert

 

[dddac]

xxxx
wrong post 😀

 

[Zoran]

Note
In PDF datasheets for the PCM56,
they used 200 ohm external IV res.
(to compare sett. time against OP inside... )
Maybe it will be better to use IV res. close to the DAC chip,
because of the IV conn.
thats led us to use N#DAC IVres.

 

[Bernhard]

Note
In PDF datasheets for the PCM56,
they used 200 ohm external IV res.
(to compare sett. time against OP inside... )
Maybe it will be better to use IV res. close to the DAC chip,
because of the IV conn.
thats led us to use N#DAC IVres.

If it is a low inductance path, IMHO a max. distance of 10cm from Iout pin via Ri/v to ground could not do any harm ?

--> But if no deglitcher is used...

PCM56 settling to 0,001% in Vout mode : 4µs
PCM56 settling to 0,001% in Iout mode : 0,5µs
OPA627 settling to 0,001% : 1,5µs
PCM56 0,5µs + OPA627 1,5µs = 2µs 😕

Best to use passive I/V and non os IMHO.

--> transformer coupled passive I/V :

If as usual the transformer's primary is parallel to the I/V resistor:

The load resistor on the secondary transforms back to the primary and acts like an inductive resistor parallel to the Vishay bulk foil non inductive I/V resistor.

And to avoid junk on the secondary, it must be properly terminated.
That can well be something like 3kohm which will be transformed into a primary inductive 300 ohm when 1:10 is used.

Avoid or use very wide band transformer 😕

 

[Bernhard]

I will try 500 ohm each chip, with and without stopping resistors, and single 30 ohm.