Voltage follower output voltage

Thread Starter

Oussama Zaidi

Joined Mar 1, 2016
53
Hi
I use TLE2024 in the circuit in the picture below as a voltage follower, I choose it for its low offset voltage (850uV max), the problem is when I simulate it in proteus with an input very close to zero the output is much bigger when Vee is connected to GND, only when Vee= -Vdd I have an output close to 0 too, I can't figure the reason, does anyone has an idea?follower.PNG follower_1.PNG
 
Last edited:

Kjeldgaard

Joined Apr 7, 2016
476
The selected operational amplifier has the correct data for common mode voltages at the input (includes the negative rail).
But the output can, according to the data sheet, with 10K Ohm load and 5V supply, just go down to approximately 0.8V.
So there must be found a type of amplifier with rail to rail output.
 

crutschow

Joined Mar 14, 2008
34,285
Try putting one or two diodes in series with the feedback loop (anodes to output) and take the output from the diode cathode (connected to the minus op amp input) with a 10kΩ resistor to ground at the minus input.
That should add enough offset to give zero volts across the 10kΩ with 0V in.

Here's the circuit:
upload_2016-5-27_10-10-6.png
 

JMac3108

Joined Aug 16, 2010
348
Like Kjeldgaard said, you need a rail-to-rail type op-amp. The TLE20204 will only go to within about 0.8V of its rails (positive and negative supplies). Look around on the TI site for op-amps designated as "rail to rail".

@crutschow
Cool circuit ... never thought of that!
 

Phil-S

Joined Dec 4, 2015
238
Just tried the diode trick - one and two 1N4148's in series.
Improved the output on a TLE2141 with IN+ grounded from 68-mV out to 11-mV out. Two 1N4148's made no difference to one.
Close, but not quite
 

danadak

Joined Mar 10, 2018
4,057
Keep in mind when you use the diodes your output Z goes form
being ~ = a voltage source to much higher impedance, basically
the dynamic diode resistance.



Regards, Dana.
 

Phil-S

Joined Dec 4, 2015
238
Thanks for that.
A tad above my op amp knowledge/capabilities.
Getting to realise rail to rail output ain't quite what it says. Staggered to find TLE2021 even worse - pushing 700-mV voltage follower with input grounded.
Only one op amp that I have ever used comes really close (within approximately 1 or 2 mV to true grounded input at output) and that's MAX407. Expensive, but dependable.
I'm going to try an LM7705 charge pump with a -ve 0.23-V output to see if I can get the elusive zero
 

crutschow

Joined Mar 14, 2008
34,285
Keep in mind when you use the diodes your output Z goes form
being ~ = a voltage source to much higher impedance, basically
the dynamic diode resistance.
But that's the open loop impedance.
The closed-loop impedance is approximately the diode impedance divided by the open-loop gain, making it very small.
 

danadak

Joined Mar 10, 2018
4,057
@crutschow

The sim shows, at 1 Khz, a node Z of ~ 12K. I thought you were correct once I read your
post, that (2 X Zdiode) / AOL, but then I kept thinking I am looking into a 10K in parallel
with diode dynamic R. But that would have produced < 10K, whereas the sim shows the Z
up, at 1 Khz. I confimed at 1Hz same result.

This is junction of diode and 10K and Inv input.

Top trace is current into R1, you can't see it because its blue.

I can't explain this.

Regards, Dana.
 

Attachments

Last edited:

ebeowulf17

Joined Aug 12, 2014
3,307
Getting to realise rail to rail output ain't quite what it says. Staggered to find TLE2021 even worse
What makes you think this should be a rail to rail device? It's datasheet indicates the low output voltage and high output voltage limitations. Offset voltage is a different spec.
Thanks for that.
A tad above my op amp knowledge/capabilities.
Getting to realise rail to rail output ain't quite what it says. Staggered to find TLE2021 even worse - pushing 700-mV voltage follower with input grounded.
Only one op amp that I have ever used comes really close (within approximately 1 or 2 mV to true grounded input at output) and that's MAX407. Expensive, but dependable.
I'm going to try an LM7705 charge pump with a -ve 0.23-V output to see if I can get the elusive zero
The 700mV figure shouldn't be staggering. The datasheet predicts such performance. That's what datasheets are for!

I don't remember the exact figures that I've measured, but I've been recently impressed with LMC6484 performance, both in terms of rail to rail headroom and offset voltage.

Rail to rail headroom depends a lot on the input impedance of whatever you're driving with this op amp. What's the output connected to?
 

Phil-S

Joined Dec 4, 2015
238
What makes you think this should be a rail to rail device? It's datasheet indicates the low output voltage and high output voltage limitations. Offset voltage is a different spec.

The 700mV figure shouldn't be staggering. The datasheet predicts such performance. That's what datasheets are for!

I don't remember the exact figures that I've measured, but I've been recently impressed with LMC6484 performance, both in terms of rail to rail headroom and offset voltage.

Rail to rail headroom depends a lot on the input impedance of whatever you're driving with this op amp. What's the output connected to?
I'm easily staggered.
Thanks for the LMC6484 info.
I'm monitoring a TSL260 light sensor which is about 10-metres away from the scope. The signal, which is in the ten's to hundreds mV range runs from the source to the scope through CAT5e cable (twisted unshielded pairs). I was getting quite a discrepancy between the scope reading and a DMM at the source. I chose the LM2141 and also tried the LM2021 as the datasheet for the LM2141 was applicable to driving long cable runs. I was looking for an alternative to the MAX407 which works well, but is a touch expensive at over £8. The addition of an op amp was to buffer the signal.
At risk of getting bogged down in op amp niceties, I was assuming that @crutschow's diodes gave a virtual zero measuring point which was above the negative supply point, a bit like getting a temperature sensor like LM35 to operate below the supply negative (ground). I assumed that the forward voltage of the 1N4148 would give sufficient "headroom" at the bottom end which is what I'm interested in to offset the 68-mV I was getting from the LM2141. Reducing that to 11-mV is an improvement, but naively I assumed it would be 0-mV (offset nulling has been tried and makes very little difference).
Previous use of the LM2141 for a similar purpose was only looking for big changes in signal, not the absolute value, so it worked fine.
 

MisterBill2

Joined Jan 23, 2018
18,179
One cheap trick that has not been mentioned is to power the opamp with a split voltage supply. Then the very low voltage would be in the center of the operating range where the opamp should be able to work very well. That is why in the early times we powered the amps with both +15 and -15 volts, but only expected linear operation between +10 and -10 volts out. The only downside is needing a split supply.
 

ebeowulf17

Joined Aug 12, 2014
3,307
I'm easily staggered.
Thanks for the LMC6484 info.
I'm monitoring a TSL260 light sensor which is about 10-metres away from the scope. The signal, which is in the ten's to hundreds mV range runs from the source to the scope through CAT5e cable (twisted unshielded pairs). I was getting quite a discrepancy between the scope reading and a DMM at the source. I chose the LM2141 and also tried the LM2021 as the datasheet for the LM2141 was applicable to driving long cable runs. I was looking for an alternative to the MAX407 which works well, but is a touch expensive at over £8. The addition of an op amp was to buffer the signal.
At risk of getting bogged down in op amp niceties, I was assuming that @crutschow's diodes gave a virtual zero measuring point which was above the negative supply point, a bit like getting a temperature sensor like LM35 to operate below the supply negative (ground). I assumed that the forward voltage of the 1N4148 would give sufficient "headroom" at the bottom end which is what I'm interested in to offset the 68-mV I was getting from the LM2141. Reducing that to 11-mV is an improvement, but naively I assumed it would be 0-mV (offset nulling has been tried and makes very little difference).
Previous use of the LM2141 for a similar purpose was only looking for big changes in signal, not the absolute value, so it worked fine.
That's all understandable. Sorry if I was snarky or judgmental sounding yesterday. Didn't get enough sleep the night before.

Although I'm a fan of the LMC6484, it probably still won't get you zero, so maybe not worth another switch in parts (although certainly a useful chip to have around based on my limited experience.) I also have no idea if it would be better or worse in terms of working with long cable runs.

I agree that a negative supply, whether from a charge pump or some other source, will make it much easier to get any op amp closer to zero.

I'm intrigued by the effects of the long cable run, but I didn't quite follow your description - which components are where? Is just the sensor on the far end of the cable? Is the op amp there too? Where is the power supply? Is the whole circuit on the far end and only the oscilloscope monitoring is being done remotely? Depending on what your present setup is, and how flexible it is, creating a current loop might be a good way to get better signal transmission. Then the voltage drop across a long cable run becomes a non-issue, plus your signal is much less vulnerable to noise pickup.
 

Phil-S

Joined Dec 4, 2015
238
One cheap trick that has not been mentioned is to power the opamp with a split voltage supply. Then the very low voltage would be in the center of the operating range where the opamp should be able to work very well. That is why in the early times we powered the amps with both +15 and -15 volts, but only expected linear operation between +10 and -10 volts out. The only downside is needing a split supply.
Only 0 to 5-V available I'm afraid.
Splitting that and taking off the headroom at both ends won't leave a lot to work with - a minus 1-v and plus 4-V might work, but then the problem gets shifted to the 0 to 5-V ADC. The already good MAX407 only draws 1.2uA so it might be easy to provide a lopsided supply
 

OBW0549

Joined Mar 2, 2015
3,566
Reducing that to 11-mV is an improvement, but naively I assumed it would be 0-mV (offset nulling has been tried and makes very little difference).
Offset nulling acts to correct any slight imbalance in the op amp's differential input stage, thus reducing the input offset voltage to zero (or near zero). It has nothing to do with how close to either supply rail the op amp can drive its output; that is specified by the output voltage range (or output voltage swing).

Input offset voltage and output voltage swing are completely different-- and unrelated-- things.
 

Phil-S

Joined Dec 4, 2015
238
That's all understandable. Sorry if I was snarky or judgmental sounding yesterday. Didn't get enough sleep the night before.

Although I'm a fan of the LMC6484, it probably still won't get you zero, so maybe not worth another switch in parts (although certainly a useful chip to have around based on my limited experience.) I also have no idea if it would be better or worse in terms of working with long cable runs.

I agree that a negative supply, whether from a charge pump or some other source, will make it much easier to get any op amp closer to zero.

I'm intrigued by the effects of the long cable run, but I didn't quite follow your description - which components are where? Is just the sensor on the far end of the cable? Is the op amp there too? Where is the power supply? Is the whole circuit on the far end and only the oscilloscope monitoring is being done remotely? Depending on what your present setup is, and how flexible it is, creating a current loop might be a good way to get better signal transmission. Then the voltage drop across a long cable run becomes a non-issue, plus your signal is much less vulnerable to noise pickup.
All the action, sensor, op amp and power is done at the remote end, though with the CAT5e, I can power it at either end. The cable is just there for me to see what is going on at the remote end (kitchen) on the scope upstairs. The final connection is wireless using XBee's, but I've used up all it's analogue inputs. The light sensor output will go into an Arduino analogue, but to get the set points right, I need to see what it is doing while I'm not there.
To give a bit of background, I monitor all the heating boiler functions and transmit them with the XBee. The Arduino does a bit of signal processing pre-XBee.The analogues are room temperature and boiler flow and return temperature. The light sensor I'm trying to commission monitors the siphon (condensing boiler) for proper fill and empty, and another channel monitors the build-up of debris in the siphon. Several IR LED's shine across the siphon and TSL260R's give me the full, empty and debris levels.
 

MisterBill2

Joined Jan 23, 2018
18,179
There are some quite cheap tiny modules that produce an isolated 5 volts out from 5 volts in. You would need to have that at the power supply end of the cable to avoid getting a whole lot of ground line noise. But then you could have your opamp running on + and - 5 volts and anything near zero would be in the center of the range. That would totally solve the problem of not working next to the negative supply rail. And that supply could probably power a few other amps as well.
 
Top