I have randomly assigned myself the project of designing an analog to digital converter out of fairly basic parts. My envisioned design will involve a digital to analog converter as a subcomponent first, so I am current working on that. I am current in the early stages and am attempting to use a unity gain 741 as a buffer to my voltage dividers. However, when I attempt to simulate it in multisim, the 741 is outputting around 800mV when it should not (Figure 1). When I did further analysis (figure 2), I found the very behavior of the op amp to be very odd. Can anyone find a flaw in my design or some non ideal 741 behavior that I was not aware of that would explain this?

Figure 1

Figure 2

 

A 741 opamp needs more than +5V to operate.
Typically, it requires a dual power supply, +15V and -15V.

 

Thank you. I gave its own -15V and 15V power supply and it now only outputs a more reasonable 500uV.

 

A 741 is an ancient op amp design with poor AC and DC characteristics. You should use something more modern.

If you want to use just a 5V single supply you can use a rail-to-rail type op amp that is designed to operate from 5V.

 

is LM358 OK as standard OpAmp these days? I never tried it at 5 volts...and yes it can work from single supply.

 

No such thing as standard, but yes, the LM358 is pretty good (can sense the negative rail with single supply) as long as you don't need high frequency, over 2kHz or so. I don't recall the exact limit, but if it matters, you have to look it up anyway. I use the LM358 as my general purpose op-amp because I bought a small pile of them a while back. Cheap and easy to find.

 

The LM358 is ok as a general purpose op amp as long as you don't need frequency response above a few kHz (GBW <1MHz), or low DC offset, or have the output go completely to the positive rail (the output will go to ground).

 

The LM358 is ok as a general purpose op amp as long as you don't need frequency response above a few kHz (GBW <1MHz), or low DC offset, or have the output go completely to the positive rail (the output will go to ground).

Couldn't you actually run it closer to a megahertz at unity gain? My design is based on using the digital to analog converter to match the input every sample, so to be accurate this must all happen in a small fraction of the time beftween samples. (So it might run at 10MHz, but only do pulses of say 25 cycles every 20kHz).

In a rough form of psuedocode it would work something like this:

Vout=0
For i from 2 to 9
{if Vin>(Vout+(2^-i))
{Vout=Vout+(2^-i)}}

 

Couldn't you actually run it closer to a megahertz at unity gain? ...........................

Depending upon the supply voltage (which somewhat affects the GBW) the response would be about 3dB down near 1 MHz. In your case the settling time is likely of interest which can be found on page 7 of the data sheet. The follower large signal response is 5-7μs to settle and the follower small signal respose is about 2.5μs to settle.