how to read the datasheet correct ( 2uV op amp ? huh )

[eevblogfan]

well . I don't know if that's the correct forum to discuss that subject ,

I'm working on high power electronic load , but first . I decided I deeded 0-1A load , so I've build it ,

I used lm324 and just random 13007 ( I thought it was a Nmos , but it turned to be NPN transistor ... )


work's , the shunt is just 1 ohm probably 1W 1% resistor from cheap Chinese (damm ) multimeter that blew in my hand while measuring mains volts , stupid meter ! ,

anyway , the shunt is +- couple of milivolts out ( at 1A setting ) .



when I received my 5X units of the mcp6V27's , I decided I'll protect the VDD of the op amp via connecting it throe the 0-1A (also have another mode-> 0-100mA ) , I've set the current max to be approx 3mA , and it work's quiet fine ( I by mistake over loaded the VoutA of the op amp ) and the current jumped from 1.2mA to 3.3 or so ,

then 2 days later , I plugged the op amp opposite (    )  and the load wasn't fast at all , the current rose to few hundred of mA's and then settled at 3.3mA , the op amp smoked and I was crying ( it had cost me a lot ! )

so I have 4 more units , I decided to do a faster 0-1A load ( just changing the shunt and the op amp to the mcp6V27 ) ,

I've changed the shunt to 50mohm one ( lower voltage operation ,+ lower power on shunt = more stable )

it turns out that 2uV op amp can be out by quiet a lot ! ,

0-100mV and it can be anyway from 9uV to 20uV , lower than 1.278 (or so ) mV and it remain in that voltage ,

is that's related to the fact that it's been powered by 1X( 4.5V) power supply ( can not reach negative rail ? )

and what about the +-  9-20uV ?  I'll mention that above 100mV it goes to 2.5-3uV as quoted on datasheet (I think )

NOTE(1): by putting some cap between the VoutA and GND , I get +- 10uV at the range of 0-100mV

NOTE(2): the circuit is half on variboard and half on bread board , with like 10C'M long , approx 20AWG wiers

thank you in advance and sorry for the long post

NOTE : I'm not shure if the title is fine either , please tell me if it doesn't , thank you !

[blackdog]

Hi eevblogfan

Can you post a picture and a schematic?
Working with voltage below 1mv is difficult.
Be very careful with the opamps specially MCP6V27 associated with static electricity.
I did this to => I plugged the op amp opposite, even after more than 30 years of experience 

Kind regarts,
Bram

[amspire]

I would be surprised if the opamp offset was more then 2uV between 0V in and the supply volts on the inputs.

If you seem to be getting a bigger offset, one of three things are probably happening. The opamp could be oscillating. Secondly, the way you are connecting the power to load circuit is adding some extra offset to the amplifier. Are you using two supplies, or are you powering the opamp and the load from one supply?  Last it could be the method you are using to measure the offset is causing a problem.

Ideally the load supply connects to the sense resistor and the mosfet drain. There should be a second supply for the opamp circuit. The ground of the opamp circuit connects to the low sense input of the sense resistor. Other then this one connection, the load ground and the opamp circuit ground should not be sharing the same ground track on the PCB. The resistor back to the opamp connects to the high sense input of the sense resistor.

How are you detecting the offset voltage of the opamp?

Richard.

[eevblogfan]

it will take me quiet allot of time to draw the schem ,

it's basic negative feedback (accros the shunt ) and this time , because of low voltage ( 4.5V ) and low current out of op amp , I decided I needed Pmos , so basically it's just VoutA control the gate of the Pmos which control the B of the PNP transistor ,

there's trimmer accros the 1ohm resistor in order to "calibrate" the thing ,

also 105.0 nF cap accros the VoutA and GND ( and about 500ohm resistor from VoutA and Gate of mosfet )

well , I'll draw a persise shem , in order to clearfy the circuit for you , it will take me some time ( I'll draw it on paper and photo it in order to make it faster )

offset voltage is meassured by comparing both ( the 287 and 3478A ) one another , then offset my 287 ( has constant 3uV offset  ( ? ) )  then meassure the inverting voltage and the non inverting voltage ,  the GND leads are as tide as I can to each other , and so as the + and _ one's

thank you in advance ! 

[amspire]

Any chance of a photo of the test setup? Just want to see how the regulator circuit connects to the MOSFET and current sense resistor circuit.

[eevblogfan]

well.  there is a chanse , but first , I'll say that it's messy and Not ideal set up , so please promis me you woun't kill me for that ( ><" )


I'll upload both together when I'll finish the drawing

[amspire]

well.  there is a chanse , but first , I'll say that it's messy and Not ideal set up , so please promis me you woun't kill me for that ( ><" )


I'll upload both together when I'll finish the drawing

Doesn't matter in the slightest if it is messy.

[codeboy2k]

There are many sources of error in an opamp circuit. yes, please post the full circuit so we can look at it.

The Vos in a datasheet is the raw offset measured with the chip just sitting there on a lab bench doing nothing. usually the inputs are tied together and grounded, and the offset is read directly at the output.  That's the old way. A new way to measure low offsets in the uV ranges is to tie one input to ground, build a gain of about 1000 or so, and tie that gained up input to ground too. Now measure the output directly, the offset voltage is the output divided by the gain setting (and you need precision resisters for this too).

Input bias currents can cause your offset to go up, make sure your input resistors are low values, like a few hundred ohms only.  Also, remember all the offset and noise at the inputs will show up at the outputs, amplified. 

you said you have a capacitor at the output, then a resistor to the gate of the FET ?  I hope you actually have a resistor first, then a cap to ground then another resistor to the gate; the first RC is a low pass filter, you need this to clean up the output of any amplified noise from the inputs. The second resistor is for the gate. You may not need it, but you should at least put a low value one there, to further isolate the gate capacitance from the opamp output. The first resistor of the RC circuit will do that, but it may not be enough . and on a PCB, especially the first cut, I always put space for one, and sometimes just fill it with 0-ohms. I also put pads as placeholders around the op amps of my circuit, in all the usual places for compensations and fill them later. That's just good advice for you when you finally get around to making a PCB.

and are you sending your feedback loop from one breadboard 20cm over to the veroboard and back ? I'm thinking you breadboard has the opamps and your veroboard has the transistors and current sense resistor.

[codeboy2k]

My first thought... that 2k input resistor on the + input to the opamp is way too high.

Consider this:  ohms law -- E = I*R. 

If the opamp has 20nA of bias current on your + input, and you have a 2k ohm resistor that the current has to be sourced from, then there is a voltage drop of 40 uV.  That means your input pin is going to be 40uV different from your setpoint at the junction of the 2k/22k/222k, and whatever you set by the 22k pot will always appear different (by an offset of 40, 50, 60 uV, depending on the bias currents).

If you were to change that 2k into 100 ohms or 50 ohms then the voltage drop from the bias current is way, way less.

100-ohm : E=IR = 20nA * 100-ohms = 2uV drop across the resistor.
50 -ohm  : half that -> 1uV drop
5 - ohm : 1/10 more -> 100nV drop. now you're talking... you've contributed less offset than your opamp spec.  That's good right there!

Even after that if you want more precision, you can still add circuitry at the inputs to offset those bias currents and effectively null them out. 

Lesson of the day here for all players:  Keep your input resistors low low low to reduce offset errors.

point #2:  I suggest you include a low pass filter at the op amp output, it will always help to stabilize things.



point #3: I am not sure why you have that diode at the output of the 7805. I think you are trying to get reverse polarity protection. But that diode drop might hurt you. If you need 5V back, you can raise the 7805 ground pin with another diode in the 7805's connection to ground. Put a diode in the standard forward biased configuration to the ground rail. This will raise your ground for the 7805 so when the voltage drops across the output diode, it will be 5V again.

In this case it's probably not needed, the opamp you chose is specified for 2.3V to 5.5V.

point #4: I don't like the trim pot over the 1-ohm sense resistor. Your going to have high currents through your trimmer, and that can heat it up, and then it's off calibration. And trimmers don't have good tempcos anyways.  Normally you would do this sort of thing by using a second op amp in the feedback loop, to feed the (-) input of the first opamp. One of the inputs of that opamp will have an offset trim adjustment, so you can trim the voltage from the sense resistor to match the setpoint voltage.  The problem is now you will need negative voltages or another opamp to invert it again if you want to stay with only positive voltages.



point 5: and this is for everyone, don't be afraid of negative voltages, they are easy to generate, and you can do alot with opamps when you have voltages available below 0. opamps love large positive and negative voltages and they get their best performance from higher voltages, not lower voltages.  There is an industry push to make lower voltage opamps.. it sort of works, but I still see the best specs are when the opamps are running with +/- 15V.

Actually with your power supply design using the 7810 and 7805 , you can easily make a small negative voltage rail that is 2 diode drops down, from the GND refererence, and float everything else above that.

OK so those are my suggestions. Pick and choose what you like and give some a try.

Cheers!

[eevblogfan]

hey

here is the "correct" schema :

besides , why is that constant 1.27X mV when the (-) connected to GND > ?

[os40la]

point 5: and this is for everyone, don't be afraid of negative voltages, they are easy to generate, and you can do alot with opamps when you have voltages available below 0. opamps love large positive and negative voltages and they get their best performance from higher voltages, not lower voltages.  There is an industry push to make lower voltage opamps.. it sort of works, but I still see the best specs are when the opamps are running with +/- 15V.

Could you give us some examples. I have seen examples that use  555's, MAX232's, LM79XX's. Not sure what would be the best. 

[eevblogfan]

hey

the best is "the best" for the requirments , ie , for low power and low noise , use resistor devider (like dave done on the uCurrent ) then buffer the center tab , there you have it ! , low noise low power split power supply ,

you hae the switching caps the 555 timer etc ( alo if AC voltage is avalible , you can use voltage dubler and then split , assuming the amplitude is too low for the rails you need , eg , 3Vppk , you can build  the voltage up to +- 17V , regulate and then split it into +- 15V )

blablabla (yes , I'm tierd , so ? )

[free_electron]

sigh ...

That's all i can muster for now. Stuff like this depresses me.

"we need stellar performance" but it's too much to ask to built a correct, properly specified, stable supply. Let alone properly bias the opamps...
Instead we'll muck about with two cascaded piece-of-shit regulators , slap in a diode and write down that we see 4.406 ( note the three trailing digits ! ) volts.. for -2mV per degree C ? if your room changes 5 degrees your second trailing digit is already off.... by the time this thing sits in a box the first trailing digit will be wobbling as it is running... raise the temp 50 degree over ambient and you are wrong in the 100mV range !

we also use a 4.654 ( not the three trailing digits ) ohms resistor to send a current through two of the crappiest diodes we could find coming from an 11.27 volt supply ... you are sending roughly 2 ampere through these diodes ? ... wow ... what's Tj ? 500 degrees ?

But i have a good opamp 2uV  . surely that will comensate no ?

you have to be stark-raving mad to design like this... this whole thing makes as much sense as a screen door on a submarine ...

Throw in this stinker : ' i put in a transsitor ... i though it was a mos but turned out to be a bipolar....'..

How does that song go ? They're coming to take me away  ...  i need to go for a little walk..... this is too much.
when i'm back maybe i'll post some tips ...

[eevblogfan]

hhhhhh , you've killed me , I can't stop laughing , ( realy )

as I've said , I'm sorry for that Not apropriate building ....

the 0-1A was just athinkering and It worked ! , but the LM324 is too slow , so I wierd it to the mcp6V27 just to see if there is any other problem ( ofcorse that when I'll finish I'll replace the KIA7810 into the KIA7805 , and another few changes )

but my question is why there is constant 1.27XmV when the (-) connected to VoutA and the (+) beed fed with 0.8mV  , why ?

If I had to build that from nothing , I'll buy *anything* new ( which I can't ) and make my own PCB . put some shiny binding posts as well ,

oh and as for the 13007 , I haven't serched for any datasheet , it seemd to work no problem with the Vout of the LM324 directly ,  (compare to another transistor's who killed my op amp ! - 10ohm is not enough ... )  sorry for that if it makes you mad on me ...

really , I'd like to learn electronics , not poetry , but I liked you're song

well , I'll be here tommorow , I'm going to sleep, hope to hear from you !

my bad ! . I'll be here within 30 min's , after that , I'll go to asleep

[blackdog]

Hi eevblogfan,

I tested this schematic (yes, i have built it) for you this afternone and its works :-)
And i think free_electron has a good point about the powersupply...
My schematic is not perfect but its works without osscilation.

Schematic:
www.bramcam.nl/eevblogfan-01.JPG

Test Setup
www.bramcam.nl/eevblogfan-02.JPG

Voltage @ + input opamp
www.bramcam.nl/eevblogfan-04.JPG

Current
www.bramcam.nl/eevblogfan-03.JPG


Kind regarts,
Bram

[eevblogfan]

lol , I've pulled almost everything out of non working devises ! dont blame me for that ! , I know KIA7805 doesnt have that stable output, it doestnt ment to be at first place ,

the bridg rectefier was there for AC transformer testing "under load " result "after *filtering*  him " . the 400V cap and the 13007 is there vor more then 100V operating ( ofcorse at lower currents ) but ther's not need in building new unit just for that , note , that CPU heat sink is bearly warm at higher the 30W continusly ( with fan voltage of about 6-7V and the fan noise is not annoying ) ie, 0-1A continusly at 30V without a dout (it can go higher but I hadn't tested it yet ... )

waiting to hear from free_electorn !

[codeboy2k]

LOL's at free_electron's reply ... I was more restrained 

Yeah, I was wondering about that 4.654 ohm precision resistor too

but I had my vent for the day... by the time I replied, I already vented at someone else in another thread, and was feeling helpful again 

[eevblogfan]

hey .

  he isn't precision , I've just go ahead and posted the measured value ,there is anything I can do now in order to reduce your non helping approach ? . really . I'd like to understand ,  when the mcp6V27 is on the breadboard , shorting (-) to the VoutA , putting quite stable 5mV to (+) , and compare the fluke 287 VS the HP-3478A , offset the fluke 287 according to the HP-3478A , and the HP-3478A measured the VoutA while the fluke 287 measured the 5mV input to (+) , the difference between them is anywhere between 9uV and 20uV , when (+) is shorted to GND , the VoutA give me 1.27XmV reading . can someone explain me only that fact ?  please ?

thank you in advance !

[blackdog]

Hi eevblogvan,

Your problem is the offset voltage, look at the datasheet, the offset change if your input and/or output is near the powersupply voltage.
If you look @ my schematic you wil see a resistor in series with the 22K current setting potmeter.
The +input of the opamp cant reach "0" and the opamp remains stable.
Also C4 and R6 are also required to remain maintain stable.

I have built for your schematic yesterday and it worked well.
Even with a 7805 as a reference :-)

Sorry for the English language, I am a dyslexic Dutchman

Kind regarts,
Bram

[amspire]

hey .

  he isn't precision , I've just go ahead and posted the measured value ,there is anything I can do now in order to reduce your non helping approach ? . really . I'd like to understand ,  when the mcp6V27 is on the breadboard , shorting (-) to the VoutA , putting quite stable 5mV to (+) , and compare the fluke 287 VS the HP-3478A , offset the fluke 287 according to the HP-3478A , and the HP-3478A measured the VoutA while the fluke 287 measured the 5mV input to (+) , the difference between them is anywhere between 9uV and 20uV , when (+) is shorted to GND , the VoutA give me 1.27XmV reading . can someone explain me only that fact ?  please ?

thank you in advance !

The minimum output voltage is specified at less then 15mV (5mV typical), so your problem is probably that for this test, the output cannot go low enough. This is not a problem for your constant current load circuit at all since the opamp output is never anywhere near zero volts. It is usually well above 1 volt. If you add a voltage divider to the output - 1k from the output to a 10 ohm resistor, with the other end of the 10 ohm resistor to ground, and you connect the negative input to the 1K/10ohm junction, then the voltage across the 10 ohm resistor should be much more accurate.

The inputs of the opamp should have less then a 2uV offset all the way down to 0 volts and all the way up to the positive rail voltage.

Richard.

[eevblogfan]

have fun :







I must say . as for that 7805 , that's really stable (I dont have any scope for further info )

it has constant 0.9~0.6mV offset . I have no idea why , but with 1 ohm resistor it's ok , as for the 50mohm . I dont think so .

I've noted that the panel meter has low input resistanse , and he dropped the voltage set , so now I'm sensing the voltage via the unused VoutB as simple follower ( if I'll lower the shun't resistance , it will help me , I'll put some gain in order to read direct the current )

I've tried few transistor's ( with gate mosfet ofcorse ) and they didn't have the preformance I get when using the IRFP064N , so maybe that was the 13007's fault !

anyway , when I'll have the time .I'll put everithing togeter and I'll leave some place for 10X 10ohm resistors (better accuarcy + lower power on each one leads for better preformance , enough for my usage )

I decided to put about 1M , accros 5K pot , the range switch will disconnect the leg of the 5K pot and the range will be lower by order of 10's

thank you blackdog  for the schematic !

but I haven't yet solved the "problem" , thank you amspire , I'll try you're idea and post the result(s) in here when I'll be able to

thank you in advance and have a nice day !

[T4P]

Where did you get those classic intel heatsinks?

[eevblogfan]

oh ,, good question

my father is computer technician , he had a-lot of quiet old computers , and they taks a-lot of plase either . I told him I can make a use of the motherboards, ATX's ,CPU's heat sink's ,

there's a pic of the wier's I took from many ATX's :

 

thank you in advance !

[T4P]

Wow ... My father has nearly the same position but here everyone throws out their still usable PC and not do anything ... i can't even get hold of a old heatsink 

[eevblogfan]

you can get these heat sinks for quiet cheap , serch on ebay ( if you need )