Problems with LM358 Current Sense Differential Amplifier

[hfletch]

Hello all,

I'm a bit stuck troubleshooting the current sense differential amplifier on Dave's uSupply. http://www.eevblog.com/files/uSupplyBenchRevC.pdf
Now, I'm aware that the current sense/feedback loop on the uSupply is known to be unreliable as the LM358 has a low offset voltage of around 5mV above the negative rail. Which, as mentioned here: https://www.eevblog.com/forum/beginners/current-sense-on-usupply-rev-c/msg996002/#msg996002, is too low to sense small currents.

However, the issue I'm experiencing seems to be unrelated to this (I think).
I have a PCB with the exact same circuit displayed in the first link built up.
I've set V+ to 10V, set VSET to give an output voltage 5V and tied a 100R resistor across the supply output to achieve a 50 mA load current.


With a measured 50 mA load on the output a 80mV drop is measured across the 1R shunt resistor (I havent figured out where this extra 30mA is going). This 80mV appears across pin 3 of U12A (the input to the voltage follower) and the high side of R23.
The voltage follower (U12A) output is measured to be 1.15V less than its input. I believe, this is because the LM358 cannot output close to the V+ rail, which of course is near the input to voltage follower!  .

Measuring the voltage across the inputs of U12B (the differential amp) gives a voltage which oscillates around 4.6mV  wrt pin 6, the negative input. This result is present with and without output current flowing through the shunt resistor. The output of U12B is then measured at a constant 1.2V and limits the supply output down unless the current reference voltage (ISET) is driven above 1.2V. When the ISET is set above 1.2V voltage regulation works as expected, with 5V measured on the output.

My 0.1% 18K resistors have been measured to be within 5 Ohm of each other. Could this difference in resistance cause the negative input of the differential amp to be consistently higher than the positive input?

Any ideas? Please let me know if you require clarification/more detail.

Thanks

[David Hess]

There must be more discussion about that circuit somewhere but U12A could never work correctly as shown.  U12A's input common mode range is violated.  U12B will also have problems because it can only pull its output to ground with 50 microamps which only allows a common mode input of less than 1.8 volts across its feedback and input resistor.

U12A and U12B should be replaced with a circuit like this or this.  Note however that this requires an operational amplifier with a positive common mode input range to its positive supply.  The old LM301A is suitable as are many JFET input operational amplifiers.

[hfletch]

Thank you for the reply David,

U12A's input common mode range is violated

I did find a couple more threads on the same subject. All which boiled down to exactly that.

I'll dive into the suggested devices next chance I get. It's looking like I'll be drawing up a patch board to recover the current limit function of the supply  . All in all a great learning experience!

Cheers

[David Hess]

I'll dive into the suggested devices next chance I get. It's looking like I'll be drawing up a patch board to recover the current limit function of the supply  . All in all a great learning experience!

If you want to keep the same exact circuit, than a rail-to-rail input and output operational amplifier can replace the LM358/LM324 but it will not be as good as my suggestion.  The LT1366 is could work in this application in place of the LM358.