4W from 2W standard Vref buffer

[iMo]

This is a more general stuff, thus I start a new thread, if I may.

One of the thoughts was to be able to support up to 10 mA output current without major problems.

with 10 mA I would think about 4 Wire output.
with best regards
Andreas

This is a quick and dirty idea how to make 4W out of the current "2W standard design". It requires negative voltage and 2 additional opamps, that is a complication, indeed, but it works fine 4W or 2W. The major issues I still see are:
1. the inverter's gain should be the same as the 7->10V ratio divider,
2. the zener's bias changes a little bit with the amount of the drop on the cables.
Perhaps it could be done differently, simpler??

PS: the 4W/2W leads are 1m long 0.3mm dia copper wires..
The non-inverting input of the U7 should go directly to the STAR2 point, to be precise..

[Kleinstein]

There is a simpler way to do it.

The positive side is kind of OK, as one option to do it. There is an alternative way, that is not simpler but without adding much noise:
The sense goes directly to the FB divider, without the extra buffer. To reduce the current needed to drive the FB divider (in the cicuit 10 V / 30 K) one can provide that current from an auxiliary supply. E.g. amplify the 7 V to an auxiliary 14 V and than have a resistor (some 7.5 K in the example) to the feedback input. So instead of an extra input a direct connection and current compensation is used.

A similar alternative way can be used for the low side:
 The low side dirve is connected to the power supply and the actual reference and 7 to 10 V divider is powered from the low side sense. An auxiliary supply will provide the current needed.
 This only works well, if the reference part consumes a constant current, so a separated oven part.
If the reference current is variable I would not use the way with current compensation, but use a buffer to drive the refrence low side, a little like the high side of the original circuit.
Alterntive one could also have the reference at the supply ground and drive the low side input, so that the low side sense is equal to ground.

[iMo]

..The sense goes directly to the FB divider, without the extra buffer. To reduce the current needed to drive the FB divider (in the cicuit 10 V / 30 K) one can provide that current from an auxiliary supply. E.g. amplify the 7 V to an auxiliary 14 V and than have a resistor (some 7.5 K in the example) to the feedback input. So instead of an extra input a direct connection and current compensation is used.
..

Ok, that high side compensation may replace the upper opamp, for example a 350uA current source will limit the SENSE_POS input current to max 1uA at 33mA..
But you have to short externally the FORCE_POS with SENSE_POS in order to use 2W in this case.

[iMo]

This is with the floating zener, 2 opamps and negative -5V.
The graph is in ppm for various loads.

[branadic]

There is one solution given on xdevs:



-branadic-

[iMo]

4W ala xdev..

[Kleinstein]

The version from Xdev would ideally have an extra current source (e.g. resistor of some 12 K to +14 V) to compensate the typical current for the FB divider. This would reduce the effect of resistance in the FB path.

[DavidKo]

EDC 106 have the adjustable compensation on negative rail. Trimmer 2a do the job. To adjust the decade is set to zero and sense link on negative terminals is removed. On 10V range the value should be adjusted to <1mV.

[branadic]

There is also Data Precision 8200 schematic with 4W

http://www.ko4bb.com/manuals/129.69.132.157/8200_20_BIT_DAC_A1_PN_60-1090_REV_0AD_Schematic.jpg

-branadic-

[iMo]

4W ala XDEV with a buffer for SENS_POS..

Note: mind the zener is simulated as a voltage source with 1.5ohm internal resistance (see the schemtics).

There is still an issue with the zener's bias - for example the FORCE_POS (I've been biasing it from) changes by aprox 9mV with the load (changes depend on the voltage drops at the leads)..

[iMo]

The version from Xdev would ideally have an extra current source (e.g. resistor of some 12 K to +14 V) to compensate the typical current for the FB divider. This would reduce the effect of resistance in the FB path.

But that would require some setting with the resistor or current source based on the divider values.. Would not be a buffer opamp a better solution? No need to adjust the compensation current..

PS: and as I wrote above the zener's bias cannot be taken from the FORCE_POS as the voltage there changes with load and leads resistances.. It seems to me we need either a complete separate Vref and then a 4W buffer, or the zener will be biased by a simple resistor from a stabilized voltage source (and not bootstrapped)..

[Kleinstein]

For the sense input one can use a buffer, but that buffer adds noise and possibly drift, even of the not much compared to the ADR1399. An AZ buffer has some spikes at the input and to a small degree it can seen an effect of the input capacitance. So the amplifiers offset can change with the input impedance, which is not good. So ideally one would have a little filtering a the buffer input, that additionally slows down the loop.

The current source on the other side does not add voltage noise or an offset voltage. The accuracy of the resistors in the auxiliary current source sets how much of the input current is compensated. Even just 1% accuracy there would give a 99 % reduction of the input current, and thus quite some improvement over no compensation.

It depends on the priorites if the extra buffer or current compensation is more suitable.

One could still bootstrap the zener reference - with the auxiliary voltage for current compensation it can be from that votlage part directly, so that even a not connected feedback would not screw things up.
Instead of an extra LT1021 ref. more like an extra amplifier (e.g. x 2 with an OP07 or sililar).

[iMo]

..One could still bootstrap the zener reference - with the auxiliary voltage for current compensation it can be from that votlage part directly, so that even a not connected feedback would not screw things up.
Instead of an extra LT1021 ref. more like an extra amplifier (e.g. x 2 with an OP07 or sililar).

It gets more and more complicated

[Conrad Hoffman]

Maybe of interest how Data Precision did it with the 8200-
http://w.ko4bb.com/manuals/72.43.255.174/8200_20_BIT_DAC_A1_PN_60-1090_REV_0AD_Schematic.jpg

[Kleinstein]

They use a seprate isolated supply for the reference and frist part of the DAC.  So the buffers in the DAC effectively help with the low side sense.  This is clever, but somewhat limited to a limited number of DAC types.

[iMo]

And the last version above (added the current limit to the low side driver) - the voltage change in ppm with the load resistances..

[Kleinstein]

It there is no other low impedance link to the supply / ground of the circuit, there is no real need to have extra protection/current limit also for the low side.

[iMo]

A variant for the upcoming ADR1001..

[GigaJoe]

so what happens if sense wire will not connected by error ? (my guess 100ohm resistors should be between sense and output , not in line of sense)
if an output shorted it would be 1W on transistor dissipation
is really necessary to use Darlington,  trade to risk of oscillation, and solve unknown problem?
R1 C1 - voltage divider with uncertainties
C5 0.1u + 1k resistor and 1-3Mohm in parallel seems would be better  (my guess)
sorry didn't get about bav199 diodes

with all this stars are you going to 4 layer board ?

[iMo]

so what happens if sense wire will not connected by error ?

[Kleinstein]

If the sense wires are not connected, the diode would take over the job. The BAV199 however have a rather high votlage drop. The 200 mV shown in the similation look rather low and would depend on the op-amp bias current. So it could also be 500 mV or so that the voltage is high (or low). A more normal diode (e.g. BAV99) would give less difference and still little leakage at the usually small difference between drive and sense.

[iMo]

..and the 10Meg load with 2W (sense and force shorted externally)..