How High a Source Impedance can be to Opamp Input?

[Mechatrommer]

is it wise to do thing like in the picture? a 1Mohm and 1Kohm divider to "1Mohm input" type opamp? i believe if R2 is changed to near opamp input impedance, things are going to be nasty? how to specify the source (divider's) impedance seen by the opamp? is it 1Kohm or 1Mohm? noobs

[ejeffrey]

Yeah, that is fine.  First off, the opamp input sees the two resistors in parallel, so the opamp sees 1 kohm input impedance.  If you increased R2 to near 1 megaohm, you would want to use a FET input opamp.

[free_electron]

what do you mean with '1 megaohm type input ' opamp... that is not how opamps are specified.
There is a parameter called 'input bias current'. that is what you are interested in. both the value AND the sign !!! some opamps give current , some opamps take current !
lets say you cave an opamp with a 1 uA input bias current. a current flowing IN the opamp.

you apply 1 volt to your input. you have 1 megaohm. the total current in the system is now limited to 1 uampere.. but the opamp will take that current.... So now you have a problem. you actually have a very low impedance in paralle with the 1K. your voltage divider is completely out of whack ... you were expecting (a ratio of roughly 1/1000 ) and you end up up with something completely arbitray because of current flowing in the opamp.

Let's say the bias current comes OUT of the opamp.... the situation worsens. you have 1 microamp coming from the 1 meg resistor + 1 microamp from the opamp. gives you 2 microamp through the 1 K.. you now measure DOUBLE what you expect !

It gets worse .... the other input of the opamp ALSO draws or delivers current. so you are actually creating a voltage differential between the two inputs.. this delta will get amplified by whatever gain you have.

in precise systems you will see that the designer keeps the perceived impedance for both inputs identical ( by putting a resistor in series with the inputs )

Lesson to be learned : if you are dealing with very high impedance circuitry around an opamp : check the bias current specification for your chosen opamp. A bipolar opamp ( like and LM741 LM324 etc ) has a large input bias current.
JFET opamps also have a bias current ( TL072 TL082 )
Only CMOS input opamps have extremely low input bias current. an LMC6603 goes down to half a femtoampere .. )

So : for high impedance work : use cmos input opamps.

[ejeffrey]

I was rather assuming that he wasn't using a 1000:1 voltage divider for a signal of 1 volt.  Even so, 1 microamp is a huge bias current.  10-100 nanoamps is more common for bipolar input opamps.  In any case, a jfet or mosfet amplifier would be a poor choice here as offset voltage is likely to be a bigger problem than bias current due to the low impedance _as seen by the amplifier_ of 1000 ohms.  If the DC accuracy really needs to be in the microvolt range, the only solution is a chopper stabilized amplifier which have embarrassingly low offset voltage and bias current.

One more tip: some opamps have a specified "input offset current" which is less than the input bias current.  This means that both inputs have approximately the same current flow.  If your op-amp is like this you can improve matters by putting a resistor between the output and the (-) input instead of the wire link equal to the parallel impedance R1//R2.  Both inputs will then see the same impedance, and the voltage offsets caused by the bias current will partially cancel.  Make sure to check the data sheet, if the offset current isn't markedly less than the bias current, don't bother, and again this is more a thing you do with higher impedances than 1kohm.  It won't help fix the offset voltage.

[Mechatrommer]

the planned opamp is THS3092:
http://www.datasheetcatalog.com/datasheets_pdf/T/H/S/3/THS3092.shtml

bias current: 20uA (±20uA/°C)
offset current: 15uA (±20uA/°C)
offset voltage: 3mV (±10uV/°C)
gain to be used: 10X

now how am i suppose to calculate that? i can dig again the formula in book, but what about the drift? how do i ensure my circuit/signal jump up and down at reasonable percentage accuracy? afaik, i havent read this kind of calculation, and the temperature swing can be very broad.

and the reason for the selected opamp is because its the only i can find with ±15V supply BW up to 100MHz. yes the drift/offset/noise spec is not that interesting, but not much selection from TI's parametric. the reason for the 1/1000X divider so that the opamp can take sort of mains voltage stuff as its input, and why it has to be 1Mohm there, so it has sort of high impedance input. think like a fet probe here that can measure mains. i know the buzz of OPA656, but its expensive, low power supply ±6V iirc. and i'm not sure it can swing for my 10X gain. and i'm using 2 opamp for this setup. 2xOPA656 is a fortune... dilemma

about the offset voltage, cant i just offset/bias the opamp output so it can null again? and the problem is... if its possible, i'm trying if the 1/1000X divider can take the 1volt signal , might change the spec later? meaning i have to reduce the divider by increasing R2, might also worsen the situation... dillema i've been "merry go rounding" for weeks.

[free_electron]

If your op-amp is like this you can improve matters by putting a resistor between the output and the (-) input instead of the wire link equal to the parallel impedance R1//R2.

exactly. that's what i said 'by putting a resistor in series with the inputs '. in case of a unity gain stage you put this between output and input

planned opamp is THS3092

WHOA . HOLD YOUR HORSES BOY ! that is a current feedback opamp ! and totally unsuitable for what you are doing ! you can't even put that thing in unity gain like you have drawn it there. you cannot connect the output straight back to the inverting input. you will kill the opamp if you do that. there needs to be a resistor there + one from inverting input to ground.
See figure 52 in the datasheet.

I happen to know those THS3092 opamps. I was working on ADSl and VDSL and we used those as line drivers. that is what they are made for.

throw that thing out. look for an LMC6xxx series. oh wait... 100 MHz bandwidth with +-15 volt rails ? you are solving the wrong problem...
Explain first why you need 100MHz bandwidth.
Then explain why you want to use +-15 volt rails... a full swing , lets say 30 volts ( you won't get anywhere near that due to output limitations , but let's use 30 volts )
100 Mhz is 10 nanoseconds.... this means you want an opamp that can do 3000 volt per microsecond ? Let me know when you find one ...

[free_electron]

right. 2 minutes of scouring.

go to www.digikey.com
type in the search box 'opamp' and hit enter
in the list click on 'Linear - Amplifiers - Instrumentation, OP Amps, Buffer Amps (10,768 items)'

you will get a filter page
in the 5th column you will find slew rate. Select anytinhg that is 3000v/us and faster. ( ctrl-click and shift-click , you know the drill )
underneath the left column there is a little checkbox that says 'in stock' click that one too.
click the button 'apply filters'

you will get the same filter view but with all non-relevant opamps removed.

now, in the third column under 'amplifier type' select 'voltage feedback'
and click the 'apply filters buffer again.

you end up with 3 or four ..

LM7171
LM6172
LMH6550 and one or 2 others.

if you want the datasheets of these things, scroll all the way to the right of the page. there is in the last column a little symbol with the letter 'D'. click and the pdf files open.
if you wan tthe cheapest one : second to last column hs a price filter. click on the little blue arrow that points upwards and it will sort by price. if you get a popup : type how many pieces you are going to buy and it will find the cheapest part for that quantity.

and you can drill down on device package, tape and reel or stick and many other parameters ....

isn;t life great if you know where to look.... no need to run around in circles for 3 weeks.
Oh, and forget those garabge sites like alldatashit .... go to the manufacturers website , or use distributers websites like digikey and mouser to find what you need. the parametric engines there are waay better than what the manufacturers have.
besides , alldatashit and others often have outdated datasheets with errors.

digikeyt and mouser link tot he manufacturers website directy so you are guaranteed to have the latest revision.

[Mechatrommer]

man, i know the digikey drill allright, except i'm not master in all opamp specs such as those offset and bias, and i still dont get why i have to match effective impedance between +ve and -ve opamp input, since i only connecting to +ve input and -ve input get its current from the opamp output?

whats in my mind is only 30Vpp supply, 100MHz, and thats right, resulting in thousands of V per microsesond, you'll be lucky finding vfb/fet opamp with that spec. why 30Vpp? well because i want greater dynamic range? and i have no problem supplying it. and why 100MHz? because i want it to measure 100MHz signal? of course i'll need to reduce this if necessary later (if not feasible). and why opamp? because discrete bjt/fet linear amplifier is not my playground.

and i know cfb we cant short output and inverting inp and i'm planning 20Vpp swing from 30Vpp supply, i know its no rail2rail. i skipped the detail for simplicity, my problem lies in the divider (still )

LM7171
LM6172
LMH6550

i thought i've came across them, if they are vfb, no, i havent found something that can go 1000+V/us slewrate if its vfb (i may be wrong i'll check again when i'm home). and if i have to reduce the gain (for lower opamp output swing) i'm afraid i will deal great amount in noise later, thats my fear

edit: and i used http://www.datasheetcatalog.com/ not other datashit. other datashit are really shit. i can almost find 99% of the datashit i'm looking for from datasheetcatalog, if i cant then i'll go to another datashit until i find it. and it loads pretty fast to, digikey is kind of slow from my place here, thats unfortunate for me

[Mechatrommer]

ok maybe the divider is one term. but there's another... attenuator, i havent any luck finding circuit example in the net.

[free_electron]

i skipped the detail for simplicity,

Well, there is your problem. DO NOT SKIP details. I don't know you and i don't know your proficiency level in all this stuff. So i assume the worst case scenario : this guy i susing CFB opamps and shows schematic for a VFB opamp ... ouch ...
You also mentioned TI's parametric search so i pointed you to Digikey's parametric as it is manufacturer independent.... since i judged your profiency level low ( becasue you did not feed me all info / false info ) i unrolled the digikey walkthrough. .. So don't get pissed at me if i'm trying to help . OK ?

As for the impedance matching between the inputs : Opamps generate a biascurrent on their input pins. this current can flow in or out of the opamp depending on the internal technology used ( bimos , cmos , bipolar , jfet and where the current pump is for the longtail. ( if the pump sits to the ground side the current will flow in the opamp , if the pump sits against the rail the current will come out of the inputs ).

This current will create a voltage drop across the impedance of your source. Since and opamp is a DIFFERENCE amplifier : you want to create the same drop on BOTH inputs , so it cancels itself out. You are creating a common mode voltage made by the bias current of the inputs . if you the inputs see a different impedance you are no longer creating a common mode but a differential error. this will get amplified by whatever gain you have in your system already... 1 microampere may not be much , but into 1 megaohm it is 1 volt. if your opamp has a gain of 20 ... the output will be stuck against positive rail no matter what you try ...

So , yes , impedance matching is important. especially when dealing with large ohmic values.

Now, back to the problem at hand.
THe THS3092 is not suitable for what you want to do due to its large input bias current that thing is typcially 20uA and it drifts all over the place.
So here is your dilemma.

Let's say you measure 10 volts.. 10 volts into 1 megaohm is 10 uA... but, the opamp also gives 20ua ... that is 30 uA flowing through the 1K resistor. you just injected a 200% error into your measurement.. Send that through whatever gain you have and it all goes to snot. Let's say you can calibrate that away .... Heat it up or cool it down and it still drifts all over the place.

in other words you need something that is CMOS input based , or has an input bias current in the order of picoampere. A quick look found nothing beneath 75nA and those are all 5 volt max..

the only one i can find that fits the 30 volt rail criteria is the LM6172... and that has already a 1uA input bias...
so if you measure 1 volt at the input you already have a 100% error ...

You may have to drop that 1 megaohm down .... and live with a lower input impedance.

besides that 1 megaohm is going to be annoying when you start going above a few MHz.... you will have to put a small ( few pf ) cap in parallel.. that resistor is going to become inductive... unless you use thin film types. and even then.... the stray capacitance on your board is going to kill you..

1pF at 10 Mhz is roughly 15K... in parallel with your 1k ... thats a 7% error. and that isassuming your stray capacitance is 1pf ...
it'll be more in the order of 5 pf... and then you have rouglhy 3K ... in parallel with 1k... that's a big error ....
so you will need to compensate that with a capacitor across the 1 Mohm  to balance it out.

[Mechatrommer]

now dont get me wrong my friend. i throw "pissfullness to forum members" away long time ago. the only thing left for me is "pissed off to myself". and assuming the worst scenario that i'm incompetent is just the right decision for you and i'm glad you replied

and now to be more accurate, apologize for my sloppiness, i thought it doesnt matter, it turned out to be the bias current problem here, so here the more accurate.. latest ammendment "revision F" , including compensation capacitors, i will thoroughly study the loading effect on compensation later, if i'm stuck i sure will beg for help in another thread refer to below picture, i actually did reduced the impedance across the signal input and opamp to 330Kohm. and schema showing ths3091 instead ths3092, i dont have the part in that sim, they ar close brother 1amp vs 2amp chip. so... lets say measuring 1V signal into 330Kohm -> 3uA + 20uA opamp current into 1Kohm =  0.023V, 1/331x divider should give 0.003V. thats 666% error?! well thats too bad! am i calculating it correctly?

about the feedback resistor, i dont have much choice since its cfb, i read i need to comply with datasheet value to make the best compromise between bandwidth and stability. datasheet says 608ohm iirc, but i'm trying 1Kohm Rf in hoping for better stability, and not so low Rg to ground. and on the output signal, there's another stage not shown. more on that later.

[free_electron]

ah, now we have a complete picture. good.
since you are using a current feedback opamp you de not need to match the impedances (that is the advantage of a current feedback system) , but you still have the input bias current that is going to annoy you...

brr. 680pF across the 1k... lets see. at 1Mhz ( you are going for 100Mhz bandwidht right ) 1Mhz would give us 1/(2 PI 1000000 680e-12) or 234 ohms...
234 ohms in parallel with 1K .. is roughly 190 ohms... a far cry from the 1k wou'd expect...
you expect a 1/330 ration. Your 1 vpp will be 190 volts pp...

At 10Mhz you'll have 23 ohms in that capacitor... you'll end up with 23 millivolts instead of 1 volt ...
See what i'm getting at ... you are never going to get 100MHz bandwidth. your -3db point ( 0.707) sits in the kilohertz range ... get rid of that 680 pf cap. it is killing your bandwidth.

Now, lets analyze this in current mode.

Code: [Select]


Vin (330volt) -----[330k]------+---<--(20uA)--- GND
                               |
                              1K
                               |
                             GND


grr. i wish the text editor would use a fixed width font. it messes up  my ascii art.

Send the above schematic through the simulator. the (20uA) is a current source. the < symbol denotes the direction of the current. in pspice they have current sources.
plot Vin versus the voltage across the 1K and you will see the distortion coming..

330 volt in 330k is 1 milliampere. you you get 1.02 milliampere through the 1k resistor. your induced error is about 2%. the bias current is of no concern in this case.
if you try to measure 0.33 volt ... you have 1uA coming from the 330k side plus 20uA from the opamp. that is a catastrofical error...
If you apply 6.6 volts at the input your system would report 13.2 volts... becasue now both branches add 20uA . so you read the voltage for 40uA as opposed to the 20 uA you would expect.

So the question now is : what voltage are we dealing with... are you always going to measure in the 300 volt range , or do you sometimes need to measure in the 2 volt range.
if your sinewave is 300 vpp you will get some crossover distortion when the voltage approaches 0.... because at that point the bias current starts messing with you.

Now, why do you wan tto attenuate so strong and then amplify it up...
if you make R1 33k instead of 330k and set the opamp in unity gain.. you also end up with 10 vpp. and the impact of the bias current is much lower !

[Mechatrommer]

if you make R1 33k instead of 330k and set the opamp in unity gain.. you also end up with 10 vpp. and the impact of the bias current is much lower !

this option is reserved for last, it will need extra intermediate opamp stage in between. that 10x gain is critical for the next stage, but it seems i failed even before the opamp stage. revision back to drawing board, and will take this bias current into consideration. thanks free_e for the advice, lets see what i'll come out with. the fault is traced back to my neuron synapses but not sure why 1KHz square wave compensation (similar to probe compensation) giving me good result with 680pF and still showing nice (well... acceptable) square wave at 10MHz at output.

[Mechatrommer]

right. my mistake. they are vfb and thousands of V/us slewrate.
LM7171
LM6172
still, uA range of Io and Ib. havent come across LMH6550. but it seems i need to change priority in parametric search, ie Io and Ib.

[ejeffrey]

I am still highly skeptical of your goal.  I don't really believe you have a 100 MHz source at 230 volts RMS, and if you do, you are going to cook your circuit with it.

So what might be the case is that you are trying to measure a couple of volts RF on top of a 230 VAC.  The problem here is that when you atttenuate the AC voltage by a factor of 300, you attenuate the RF as well, and you end up with a few mV of RF.  Those CFB opamps are very low noise, so you can actually amplify that back up to what you want, but you are just unnecessarily handicapping yourself.

Is there any reason you can't AC couple /high pass filter your signal to block the high voltage and only amplify the RF component? If you need to see the 230 VAC component for some reason you can add a second output with a low bandwidth amplifier and an attenuator, or you can use a shelving filter that attenuates the low frequency part by ~1000:1 while passing the RF unattenuated to keep them on the same signal.

If your goal is to make a general purpose amplifier like a scope front-end, then you are going to need programmable attenuators, level shifters, and AC coupling just like a scope input.

[Mechatrommer]

adding 20uA current source with or without the opamp failed to simulate (zero opamp output 1st picture). without 20uA, showing nice curve (2nd picture), i'm not sure why its nice even though i've included compensation caps, and now its increased to 2.2nF for flat respond (VM3 vs VM4 or c2 vs c3). loading to 100ohm source also non existence (VM1 vs VM2 or c1). i'm not sure where i'm doing wrong here. i've done the diy pcb'ing. i'll capture when i'm back this night. iirc i've got quite usable attenuation, but noisy

I am still highly skeptical of your goal
So what might be the case is that you are trying to measure a couple of volts RF on top of a 230 VAC

you getting close there my friend. now it seems it hard for me to conseal it... its a differential amplifier, 3 opamp config, available in text everywhere. the next stage not shown is just another differencing opamp but i'm still struggling at the input stage so far it seems, its pretty usable to measure smps offline side. i believe still far from perfection, i need a sound design but i feel like walking blind here

[Mechatrommer]

maybe the 20uA offset/bias current is not going into the 1Kohm divider? but instead provided by input through 1Mohm? and ground into the opamp?
anyway. here's my capture of current circuit. 1KHz square compensation and 1K-30MHz sine signal. yellow is input, blue is ins-amp output. using 1/100X divider (130K+1.3Kohm, i'm merry go rounding remember?, this is whats currently on there) still crude setup but anyway... the signal start to roll off at 20MHz, and 30MHz and above, things start to go tricky (even square wave at 1-10MHz), but i didnt get any 600%(recalculating with 3Vpeak... there should be 200%) error as calculation exercised before, so..

ps: i've used 24V LA batt to reinforce my crappy psu for anybody who wonder, the last pic is the schematic (simplified, removing psu, switches etc), FWIW.

[free_electron]

growl. this is just unbelievable. I asked you to SHOW THE REAL THING  and not to hold back any information ! you keep misleading me and then you end up with problems that are no problems. Also : do yourself a favore and chuck that simulator you have in the garbage can, it is bullshitting you.

An instrumentation amp is a totally different beast than a single ended animal.

- bias current matching . done , not a problem since both sides inject the same error they cancel each other out.
- common mode rejection . done , not a problem that's what instrumentation amps are all about. the common mode is created in both brances and the diff amp backstage cancels that one out. any imbalance you get in the frontend is self-canceling in the backend. that is the beauty of an instrumentation amplifier. the damn thing eliminates everything.

if you buy a diff probe from tektronix for high voltages it does 100MHz without problems. they use a stupid quad opamp from maxim to make the instrumentation amp, followed by an op27 output buffer. they have come cermet input resistors with long body to prevent flashover. you can throw 1000 volts in these probres , they don't even blink.

See, the problem is that you are not explaining what you are trying to do, then you give half a schematic. everything i said is only applicable to single ended stages.
once you go the instrumentation amp route it cancels out.

[BravoV]

Serious yet funny thread you got there Mech. 

[robrenz]

Mech, FWIW Tell free_electron conceptually what you are trying to accomplish not what you are doing.  He obviously has the chops to help you get to your destination as do many others on this forum.  I think we will all learn more in less time that way.  I realize you can sometimes learn more by the mistakes on the way but I think you compained about projects eating into spending time with your wife. BTW I am enjoying your electronic travels.

[Mechatrommer]

sorry free_elec. i was trying to solve this particular issue. your advice is not for nothing dont worry about that. if i reveal the whole thing in 1st place, how could i learn about Ibias effect and the intention is not to hold back anything, but trying to solve one thing at a time. everything cancel out you said? not that easy... not that easy... just to get -60dB CMRR at 1MHz is a miracle

and this one is a big problem, even though i can compensate the signal, but many other peculiar aspects of this, there's too much to be solved. how do i know the attenuation curve at each stage? should i compensate after the first opamp stage? etc etc, and peculiar loading/ringing effect. now Robrenz seems following me from the start "silently", he should know how much i want to cry on this thing! if just men are allowed to cry

and in another thread, i asked too much question (about 50ohm termination), not even half are answered. maybe i was overwhelming the viewers? or maybe they are "too interested"? so i prefer showing a simple thing one at a time. easier chances for readers to answer my question... focus, think focus now free_elec, do you want to solve all the problems for me? tek diff probe? how much does it cost? i saw the Pintek Diff Probe that also can do 1KV 100MHz too many variant at $100+ how poor i am. i've set it as the benchmark, think i will reach there? not sure, maybe i should just call it OFF!

[Mechatrommer]

now... the swmbo asked for a divorce again, what should i do? if i reveal the whole thing 1st, then what would the advice be? buy a tex diff probe? furthermore, the knowledge here may be applied if one to build just a simple "fet like probe" get the idea?

[robrenz]

Its easy to get caught up in the thrill of learning, designing, building. Trust me I know.  It generally seems like family on this forum and forum participation can become an obsession also. Trust me I know again. This stuff can seem more important than it is. But in the end, when my body or mind starts to fail, Nobody on this forum is going to change my diaper or feed me or be there for me.

Mech, I thought I was writing this for you, but I think I am speaking to myself.

robrenz

[Mechatrommer]

It generally seems like family on this forum and forum participation can become an obsession also

so maybe if nobody answering your question, that can be a good thing i guess.

Mech, I thought I was writing this for you, but I think I am speaking to myself.

so it seems we have something in common? nobody here in my place that have something in common with me. i'm the special one or the most unspecial one

Nobody on this forum is going to change my diaper or feed me or be there for me

absolute! except... i dont wear diaper

i keep thinking many times, read somewhere... once you're married, you're doomed, no more hobby, or at least we cant be so obsessed like before. some truth to it. time management... thats what i'm lacking most. urghh! 10yrs passed by, i feel like i'm still a child. i should re-consider/re-think my purpose in life...

[ejeffrey]

you getting close there my friend. now it seems it hard for me to conseal it... its a differential amplifier, 3 opamp config, available in text everywhere. the next stage not shown is just another differencing opamp but i'm still struggling at the input stage so far it seems, its pretty usable to measure smps offline side.

Is the intent here to measure the high side current by measuring the voltage drop across a sense resistor?  If so, an instrumentation amplifier like you have shown will work, but is not necessarily the best option.  A current transformer will have better CMMR than even an excellent implementation of a differencing amp, and way better than a half-assed implementation.  The obvious downside is that you lose the DC component.

Also, keep in mind that no matter how good your instrumentation amplifier, the CMMR will be limited by the accuracy of your attenuator network.  Your divider needs to be made of high precision resistors (.01% is not too good) or you need to trim it. You can trim it into tune for DC with a pot, but keeping the impedances balanced at AC is more difficult.

I still don't like the fact that you have a big attenuator to handle the DC voltage that attenuates your signal just as much.  Your plan will work if implemented properly, and it is great if what you need is general purpose differential amplifier with a huge common mode range. It will have a poor SNR compared to a circuit that doesn't have the attenuator network in front.