High-side high-current source independent of Vcc

[tom66]

I want to source a current into a variable load, around 100-200mA. The load sinks current, and is grounded at one end, so I cannot use a low-side current sink. The current needs to be precise and independent of the load impedance up until the source saturates up to a volt or so. The current is too high to use devices like LM334. Linear Tech has a device that can source 200mA but it's about £4 in onesies from Farnell, so I'd like to avoid it.

I thought about using a high side op-amp driven current sink (like this

), but with a dedicated Vref instead of the divider resistors... but this is dependent on the difference between Vcc and Vref, effectively this means my Vcc must also be a precision voltage source which increases the complexity as well as that precision source needs to be able to supply 200mA. I do have a precision Vref @ 2.048V, but it can't source much more than a few mA.

I came up with what's attached. It works, but it requires two opamps plus two power devices.  The first opamp essentially generates a 2xVref output, and the second op-amp regulates into a current using a P-channel FET (I did not want to use a PNP as the base current would create an error.) The source current is set by Vref/Rset.

I'm thinking that there's a simpler option. Does anyone have any suggestions?

[macboy]

A very simple solution is a LM317 with a sense resistor of 12 ohm for 100 mA, or 6 ohm for 200 mA. Dropout voltage will be 1.25 plus the dropout of the LM317 (<2 V).

You can also build a current source from a Vref, op-amp, transistor, and sense resistor in exactly the same manner as the classic constant current DC load; you just move the Vref, sense resistor and transistor up to the positive rail. Don't forget to use a R2R I/O op-amp. Something like attached (no guarantees!).

[Marco]

A TLV431 current source can do it with reasonable accuracy ... cathode current is an error, but not significantly greater than opamp offset voltage will cause in your circuit if you don't use choppers.

PS. LM317 has same level of error current, so guess that makes more sense.

[tom66]

Accuracy is the main design characteristic. Second is drift over temperature and time as it warms up - the op-amp is ideal in that respect as that can be thermally isolated from the hot transistor although any IC with a low tempco would also work well.

The '431 solution looks interesting. I'm not a fan of the LM317 due to the Vref being spec'd as 1.2-1.3V, that's +/-4% error...

[macboy]

Accuracy is the main design characteristic. Second is drift over temperature and time as it warms up - the op-amp is ideal in that respect as that can be thermally isolated from the hot transistor although any IC with a low tempco would also work well.

The '431 solution looks interesting. I'm not a fan of the LM317 due to the Vref being spec'd as 1.2-1.3V, that's +/-4% error...

LT1083 is better at 1%, still not as good as TL431 at 0.4%, but you don't need to consider additional error from op-amp offset etc.
I'd you don't specify such constraints on your initial question, how can we give the answers you want?

[macboy]

In my quick and dirty circuit above you will notice the Darlington output. This is because any base current pulled by the op-amp will flow through the sense resistor but not through the load, so is a source of error. The Darlington output reduces this error due to it's very high current gain.

[tom66]

Precision: within 0.5% with well toleranced components.

The darlington output is interesting, I'll have to model it. One of my concerns is relying on the transistor hfe too much, as that has poor tolerance and temp coefficient. But, if it had an error <0.5% that could work.

[macboy]

Precision: within 0.5% with well toleranced components.

The darlington output is interesting, I'll have to model it. One of my concerns is relying on the transistor hfe too much, as that has poor tolerance and temp coefficient. But, if it had an error <0.5% that could work.

It can also work well with a PMOS, with the added benefit of even lower error current.

[Marco]

LT1083 is better at 1%, still not as good as TL431 at 0.4%, but you don't need to consider additional error from op-amp offset etc.
I'd you don't specify such constraints on your initial question, how can we give the answers you want?

You don't need to consider it with the TL431 either, the current source works essentially the same it just takes an extra resistor and transistor to turn the TL431 into a three terminal voltage regulator.