high-side current sense circuit - jelly bean parts only

[rob77]

Hi Guys,

i ran into this very nice document by Intersil
http://www.intersil.com/content/dam/Intersil/collateral/en/cookbooks/st-091.pdf

that current sensing circuit is extremely clever and simple ! it does the high side sensing amplification and level shifting in one pass.. thing of beauty !  and it doeas it with a low voltage opamp... ideal circuit for higher voltage rail (it's really hard to find a opamp for 40V+ supply and common mode reaching to the top rail).

so i was immediately thinking of building it with jely bean parts and see how it went
i decided to have a rather high shunt/burden voltage to make the opamp's input offset and thermal drift insignificant (the original design was in the 10 mV range) and changed the supply for the opamp - used a 12V zener (going to use this in 30V rail supply) and also the opamp is powered from "before" the shunt because the power consumption of TL081 is higher than th eoriginal opamp used.

the schematic and the board built:  (sorry for the mixed SMD+THT construction , but i built it immediately from parts i had at home )








btw... decoupling is missing by purpose - to test stability without decoupling.

going to do some measurements and share the results later , but the first tests are better than expected thermal stability is more than acceptable despite the standard 200ppm resistors used.


EDIT:  forgot to mention.. the goal was to have 5V output for 1Amp (to match the 5V ADC of arduino/atmega).

[trevwhite]

Thanks for sharing, really interesting article. Be good to see your results.

[Yansi]

R1  10 kilos? Are u serious? Or is that a value left from copying parts in the editor?

[miguelvp]

R1  10 kilos? Are u serious? Or is that a value left from copying parts in the editor?

He did populate it in the board (1002 = 100*10^2 = 10K)

[void_error]

I'm using the same jelly-bean solution for high side sensing in my bench PSU, although slightly modified. In theory it can accept input voltages between roughly 5 & 50V. Not tested yet, only simulated, but it should work. Used BJTs as I wanted to keep the design largely unchanged for higher output voltages and high-voltage logic level MOSFETs are not that common, especially P-type ones. There is some beta-related error but it's so small it can be neglected - 0.01% for transistors with a gain of 100. Used a constant current sink in place of the (10k?) resistor to make sure the voltage across the zener (LMV431 here) remains constant across the input voltage range.

I hope you don't mind me posting the whole power supply schematic but I was too lazy to edit/upload only the high-side sense part.

[rob77]

R1  10 kilos? Are u serious? Or is that a value left from copying parts in the editor?

He did populate it in the board (1002 = 100*10^2 = 10K)

yes i did

and what's wrong with 10K ? or should i burn the production of the local power plant in that little poor zener ?

[void_error]

and what's wrong with 10K ? or should i burn the production of the local power plant in that little poor zener ?

Nothing wrong with 10k as long as there's enough current through the zener diode to have 12V across it - double check with the datasheet.

[rob77]

I'm using the same jelly-bean solution for high side sensing in my bench PSU, although slightly modified. In theory it can accept input voltages between roughly 5 & 50V. Not tested yet, only simulated, but it should work. Used BJTs as I wanted to keep the design largely unchanged for higher output voltages and high-voltage logic level MOSFETs are not that common, especially P-type ones. There is some beta-related error but it's so small it can be neglected - 0.01% for transistors with a gain of 100. Used a constant current sink in place of the (10k?) resistor to make sure the voltage across the zener (LMV431 here) remains constant across the input voltage range.

I hope you don't mind me posting the whole power supply schematic but I was too lazy to edit/upload only the high-side sense part.

don't mind it at all nice supply , i like it   
the 431 shunt regulator instead of plain zener is a good idea - makes the thing more versatile.

[TerraHertz]

That's a neat circuit Rob77, thanks.

[Yansi]

I've mystified myself - I have overlooked that you use 30V supply. So there IS enough current flow through the 10K resistor to power the OPamp and zener. My fault, sorry!

[moffy]

Isn't the input voltage outside the minimum Common Mode Input Range for the TL081?

[Yansi]

Interesting note, but the classic JFET opamp like TL0xx should typicaly work when the input commonmode voltage is near or at V+ level. But I think it is not guaranteed, it just might work. I've never tried to measure and verify that, so I wouldn't  rely on this behaviour.

[toli]

It might be the early morning hours speaking here, but unless I'm wrong seems like you guys forgot about something

When calculating the value of the resistor that sets the current (in both posted designs), don't forget it will pass the current of the zener/431 as well as the current of the OPAmp. The TL081 for example has Icc of up to 2.8mA (1.4mA typical). That alone is capable of turning the zener off completely even from 30V supply.
Same for the design void_error has posted, the MCP6021 can have Icc as high as 1.35mA, and the CCS is set to just 1mA. So it might be best to set the CCS somewhat higher - 2mA or so will leave plenty of current for the voltage divider (~350uA) + LMV431 to operate (will have at least 300uA).

[rob77]

It might be the early morning hours speaking here, but unless I'm wrong seems like you guys forgot about something

When calculating the value of the resistor that sets the current (in both posted designs), don't forget it will pass the current of the zener/431 as well as the current of the OPAmp. The TL081 for example has Icc of up to 2.8mA (1.4mA typical). That alone is capable of turning the zener off completely even from 30V supply.
Same for the design void_error has posted, the MCP6021 can have Icc as high as 1.35mA, and the CCS is set to just 1mA. So it might be best to set the CCS somewhat higher - 2mA or so will leave plenty of current for the voltage divider (~350uA) + LMV431 to operate (will have at least 300uA).

thanks for the comments, but i have built the circuit and is working fine... including the voltage drop across the zener
TL081 has a typical suppply current of 1.8mA - that's 18V drop @ 10k resistor... with a 30V supply rail and 12V zener it's just the right value, isn't it ?

why should someone burn more power than needed in a zener ? you should always burn just the bare minimum in a zener - just enough to cover the variations in input voltage and variations in the current through the load.

i haven't measurent the supply current of those TL081s i have , but i would expect it will be less than the typical 1.8mA at 12V single supply. i assume the 1.8mA is @ the higher supply voltages +-15V (that's 30V single supply).
 
regarding the power consumption of the tL081 - it's supposed to be pretty static, the inputs are sitting pretty much at the same potential all the time and the output is driving a small mosfet in it's linear region , and it's a mosfet with very little gate capacitance anyway.

i would rather expect that someone will raise an objection against the TL081 it's not guaranteed that the common mode of tl081 will reach the top rail... but if you have a look at the datasheet... it says: typical common mode input +15/-12 @ supply +-15V => therefore it should reach out to the top rail. btw... tested a couple of TL081s and they're all reaching the top rail.

EDIT:

btw... we can safely assume the typical supply current , because we are relying on the typical common mode range. if the TL081 in question has a higher supply current , then it's not a typical chip and has some deviations including other parameters - like the common mode input range we are relying on

[toli]

rob, a couple of things:
1- the bias current doesn't vary much with supply voltage, most biasing circuit are designed in such a way that the current doesn't vary much with supply voltage, otherwise performance would vary significantly with voltage and this is something designers are trying to avoid.
2- the fact its working well at the time you have tested doesn't guarantee it will keep on doing so over time and temperature variations. Its very common for biasing circuit to be PTAT to minimize BW variations as temperature changes so chances are the current will increase as the device heats up. Good practice is to design for worst case scenario unless you are going to put some sort of detection for this situation. If the datasheet states 2.8mA as max Icc, it would be best to design accordingly.
2.8mA across 10K is 28V, turning the zener off completely. chances are another 2mA won't make much difference (if any) in terms of power consumption

Edit:
as for what you've added, the two things are unrelated. The fact Icc isn't at its typical value, doesn't mean input CM range is other than typical. As a rule of thumb, it probably better to draw an extra 2mA than count on the amplifier to be typical. Statistically speaking, 50% of these will be >typical.

[rob77]

rob, a couple of things:
1- the bias current doesn't vary much with supply voltage, most biasing circuit are designed in such a way that the current doesn't vary much with supply voltage, otherwise performance would vary significantly with voltage and this is something designers are trying to avoid.
2- the fact its working well at the time you have tested doesn't guarantee it will keep on doing so over time and temperature variations. Its very common for biasing circuit to be PTAT to minimize BW variations as temperature changes so chances are the current will increase as the device heats up. Good practice is to design for worst case scenario unless you are going to put some sort of detection for this situation. If the datasheet states 2.8mA as max Icc, it would be best to design accordingly.
2.8mA across 10K is 28V, turning the zener off completely. chances are another 2mA won't make much difference (if any) in terms of power consumption

Edit:
as for what you've added, the two things are unrelated. The fact Icc isn't at its typical value, doesn't mean input CM range is other than typical. As a rule of thumb, it probably better to draw an extra 2mA than count on the amplifier to be typical. Statistically speaking, 50% of these will be >typical.

fully agree  wouldn't do that for mass production but for one-off designs which are hand built and hand tested it's just ok to rely on the typical values because you can hand select the chip for your one-off product;)

regarding the variations - agree as well ... i should have written - if the supply current is not typical, then there is a high chance of not having a typical CM range. (btw... the typical values are not 100% unrelated)

[toli]

Hi rob,

sure you can do this for a one off, but as temperature rises its more than likely that the value of Icc will rise - so unless you test at the worst conditions the circuit is expected to operate under, it would be best to leave a margin even for a one off.
As for mass production, since you've posted it on the forum, keep in mind that many of the people who'll use it, don't understand it/have no idea what assumptions were taken when designing the circuit. So as a schematic posted for common use it will be best to put an according value of resistor or included an appropriate note under which conditions the resistor was calculated/how to calculate it for each use.

[dannyf]

Isn't the input voltage outside the minimum Common Mode Input Range for the TL081?

Bingo! Right on the money.

[void_error]

It might be the early morning hours speaking here, but unless I'm wrong seems like you guys forgot about something

When calculating the value of the resistor that sets the current (in both posted designs), don't forget it will pass the current of the zener/431 as well as the current of the OPAmp. The TL081 for example has Icc of up to 2.8mA (1.4mA typical). That alone is capable of turning the zener off completely even from 30V supply.
Same for the design void_error has posted, the MCP6021 can have Icc as high as 1.35mA, and the CCS is set to just 1mA. So it might be best to set the CCS somewhat higher - 2mA or so will leave plenty of current for the voltage divider (~350uA) + LMV431 to operate (will have at least 300uA).

Oops, I probably did the math for a LMV321 for some unknown reason instead the MCP6021, I'll probably set the CCS to 3mA and replace Q8 for a BJT which can dissipate more than 100mW. That makes R37 33ohms. Thanks for pointing that out