Single ended to differential circuit assistance

[peter_claybars]

I have a dac that outpus a single ended signal (0-3.3v) which I need to convert into a differential signal. The differential signal needs to range from -6.6v to +6.6v (i.e. channel A ranges from -3.3v to +3.3v and channel B ranges from -3.3v to +3.3v). Channel A and Channel B are centered around "GND" which in this case is 0v of the DAC

When the dac outputs 1.65v (the middle of the range) I wanted the differential output to be zero. When the dac outputs 0 or 3.3v, I wanted the range to be -6.6v or +6.6v.

I built a circuit to accomplish this using two inverting op-amps where the first amplifies the signal with a gain of 2 and the second inverting op-amp is of unity gain.

Because I wanted 1.65v to equate to 0v differential output, I connected the "+" of the first op-amp to +9v via a pot

After building this circuit, I found it didn't work. I couldn't get 1.65v to result in 0v output and the min/max range values were not symmetrical around 0v.


After some research, I found this was due to the "voltage offset" characteristic of op-amps which for the op-amp I am using (tl074) is 3000uV!

I know I can get 5uV or less op-amps, but I was wondering if there is a better solution. The lower voltage offset op-amps will probably be ok for me, but I would really prefer a solution that I can "tune" once, then for about, even if the device's temperature fluctuates +/-10 degrees C.

Any help would be appreciated!

 

[godfreyl]

Depends on the accuracy you need. The TL074's drift is typically only 10uV / degree C so once tuned it should be fairly stable, but you will need separate adjustments for both opamps.

 

[peter_claybars]

Well, it's a 12 bit dac and I would like to have the best resolution I can.. but if I got away with 10 bits, I wouldn't be too upset. I'm updating this at 115khz.

How much will one op-amp vary from another? I.e. if I got a theoretical 100 tl074 op-amps, will they all require the same offset, or will they vary significantly?

Thanks!

edit: I suppose I shouldn't have commented so much on the temperature variation. It's something I'm concerned about.. but right now I'm more interested in the offset issue.

 

[godfreyl]

How much will one op-amp vary from another? I.e. if I got a theoretical 100 tl074 op-amps, will they all require the same offset, or will they vary significantly?

Probably quite a bit of variation. See the difference between "typical" and "maximum" on the datasheet. I have heard that for dual and quad opamps, the opamps on the same chip have similar offset. If that's the case, you might be able to use the offset of one to cancel the offset of another.

If this is for audio, why do you care about a bit of DC offset anyway?

 

[crutschow]

...............................

When the dac outputs 1.65v (the middle of the range) I wanted the differential output to be zero. When the dac outputs 0 or 3.3v, I wanted the range to be -6.6v or +6.6v.

I built a circuit to accomplish this using two inverting op-amps where the first amplifies the signal with a gain of 2 and the second inverting op-amp is of unity gain.

Because I wanted 1.65v to equate to 0v differential output, I connected the "+" of the first op-amp to +9v via a pot

After building this circuit, I found it didn't work. I couldn't get 1.65v to result in 0v output and the min/max range values were not symmetrical around 0v.


After some research, I found this was due to the "voltage offset" characteristic of op-amps which for the op-amp I am using (tl074) is 3000uV!

........................

To get -6.6V to +6.6V from an input of 0 to 3.3V you need an inverting gain of 4 not 2. If you then put 1.32V into the op amp positive input, the output will be 0V with 1.65V input and swing from -6.6V to +6.6V for an input of 0 to 3.3V. The 3mV op amp input offset will be amplified by the non-inverting gain of 5 to give an output offset of 15mV. This can be adjusted to zero if you use a pot for the 1.32V at the plus input.

This offset is fairly random from unit to unit since it is determined by small variations in the fabrication process across the chip. If you want a no adjust circuit, then you could go with a lower offset op amp. But this implies you can generate an accurate 1.32000V voltage for the plus input to your required accuracy without requiring adjustment.

 

[peter_claybars]

If this is for audio, why do you care about a bit of DC offset anyway?

This is not for an audio application

To get -6.6V to +6.6V from an input of 0 to 3.3V you need an inverting gain of 4 not 2. If you then put 1.32V into the op amp positive input, the output will be 0V with 1.65V input and swing from -6.6V to +6.6V for an input of 0 to 3.3V. The 3mV op amp input offset will be amplified by the non-inverting gain of 5 to give an output offset of 15mV. This can be adjusted to zero if you use a pot for the 1.32V at the plus input.

This offset is fairly random from unit to unit since it is determined by small variations in the fabrication process across the chip. If you want a no adjust circuit, then you could go with a lower offset op amp. But this implies you can generate an accurate 1.32000V voltage for the plus input to your required accuracy without requiring adjustment.

The circuit does give me about -6.6v and +6.6v currently (although slightly shifted due to a voltage gain offset so one extreme is a tad high and one extreme is a tad low) so I'm quite certain my gain calculation is correct. The input is centered at 1.65v... so a gain of 2 on the first inverting amp gives me an output of:
with dac output of 0v: 3.3v
with dac output of 1.65v: 0v
with dac output of 3.3v: -3.3v


This is the "Channel A" output which is fed into the 2nd inverting unity gain op amp to allow channel b to have:
with first inverting op-amp output at 3.3v: -3.3v
with first inverting op-amp output at 0v: 0v (but not really because I set the offset of the "+" pin to 0v)
with first inverting op-amp output at -3.3v: 3.3v


Channel A + Channel B outputs combined are:
with dac output of 0v: Channel A to Channel B = 3.3v to -3.3v = 6.6v
with dac output of 1.65v: Channel A to Channel B = 0v (well.. it was supposed to be!)
with dac output of 3.3v: Channel A to Channel B = -3.3v to 3.3v = -6.6v

I have to run to work now, when I get back I'll post a schematic to clarify this

 

[crutschow]

Yes, you need a gain of two. I misread your requirements and thought you wanted plus and minus 6.6V output.

 

[peter_claybars]

Here is the schematic for the version that does not work.


Even if I do add the offset pot circuit to the second op-amp (but using the -9v vs +9v), isn't it just going to flip the issue to the other side?

 

[peter_claybars]

Ok, here's a simpler circuit which still does not do what I would like. If I set the offset value such that 0v from the dac = 3.3v and 3.3v from the dac = -3.3v, then 3.3v/2 from the dac results in 0.15v output! Doh! In this situation, is my only solution an op-amp with lower offset? Thanks!

 

[crutschow]

The offset of the amp should be fairly constant with signal level, so I don't see that you have an offset problem. Any offset should have been removed by the original adjustment. But you could have a tolerance problem with the gain resistors which would cause what you observed.

A better approach might be to adjust the offset to give 0V output with 1.65V input. Then the gain error will result in a just a slight change in the maximum plus and minus output values, not the zero point offset (assuming zero offset is more important than gain accuracy at the limits).

If you need to adjust both offset and gain then you will need to add a pot in series with the feedback resistor to allow tweaking the gain. With that setup, the procedure would be to first adjust the gain to give exactly a 6.6V swing for the 0-3.3V input, and then adjust the offset with a 1.65V input to give 0V output.

 

[Moses_M]

Hello peter_claybars,

I didn't quite understand the reason you have used +9V divided by a Resistor network to feed the reference voltage to your OP-Amp. It would be vunerable at times, I might prefer to use a voltage reference IC instead!

/Moses_M

 

[peter_claybars]

The offset of the amp should be fairly constant with signal level, so I don't see that you have an offset problem. Any offset should have been removed by the original adjustment. But you could have a tolerance problem with the gain resistors which would cause what you observed.

A better approach might be to adjust the offset to give 0V output with 1.65V input. Then the gain error will result in a just a slight change in the maximum plus and minus output values, not the zero point offset (assuming zero offset is more important than gain accuracy at the limits).

If you need to adjust both offset and gain then you will need to add a pot in series with the feedback resistor to allow tweaking the gain. With that setup, the procedure would be to first adjust the gain to give exactly a 6.6V swing for the 0-3.3V input, and then adjust the offset with a 1.65V input to give 0V output.

I have tried using a pot for both the offset and gain and it does not work with that single op-amp circuit. I'm stuck with one of two situations:
1. The ouptut is -3.3v to 3.3v with 0 or 3.3v input respectively, but the 1.65v input results in a non-zero output (by 0.15volts or so).
2. With 1.65v input, the output is 0v, but at the extreme ranges, the output is something like 3.6v to -3.3v (I may have this flipped.. i.e. the negative value is -3.6v and the positive only 3.3v)

Hello peter_claybars,

I didn't quite understand the reason you have used +9V divided by a Resistor network to feed the reference voltage to your OP-Amp. It would be vunerable at times, I might prefer to use a voltage reference IC instead!

/Moses_M

Would this cause the offset/gain issue I am seeing?

 

[Moses_M]

The reference needs to be stable irrespective of power supply ripple as well as temperature effects. So a reference voltage IC is what I would recommend. May be a LM336 2.5V for example can be used. A 10K Bourns pot may be used to trim down the 2.5V constant reference voltage output to about 1.650V & fed to the non-inverting input. By the way what DAC are you using? & what all care u have taken while using only 10bits instead of using 12 bits, reference voltage of DAC etc.?

Also I hope the resistors used here are precision resistors...