Discrete Rail to Rail Output Stage

[nickm]

Does anyone have an example of a discrete rail to rail output stage?  The part i'm looking for is what is used to generate the signals which drive the output transistors.  Along those lines how reliable are they compared to a totem poles which are very simple and robust?  And by discrete I don't mean only transistors, I just mean the final output devices are transistors but maybe opamps for the circuitry which generates the drive signals.

[David Hess]

Pages 4 and 5 of Linear Technology application note 18 have examples of discrete rail-to-rail output stages. 

There is another clever configuration which places the common emitter output transistors into the supply lines of a single operational amplifier which has its output grounded through a resistor which also provides local current mode feedback.  This page shows a complex version, second schematic from the bottom, with current mirrors replacing the single transistors.  Q3 and Q4 are not strictly needed but do help:

http://how-to.wikia.com/wiki/How_to_boost_the_output_voltage_swing_of_an_operational_amplifier

The last schematic shows a higher voltage version which adds a pair of cascodes in the supply lines.

The additional difficulty in these designs comes from the common emitter output stages which have voltage gain that varies with load making frequency compensation more challenging.  Adding local feedback or emitter degeneration helps a lot with this issue.

[void_error]

This is something I was looking for.
Do the output voltage swing booster circuits work well with single/asymmetrical supply voltages?

EDIT: Also found this: http://www.ti.com/lit/an/snoa600b/snoa600b.pdf

[David Hess]

I think that is originally an old National Semiconductor application note.  Texas Instruments has been working to erase all credit to National since they bought them so web searches tend to be unreliable in finding them.

Do the output voltage swing booster circuits work well with single/asymmetrical supply voltages?

There is no specific reason that they will not; it just depends on the details.  I have used several different designs for rail-to-rail power output stages with single supplies.

The place where it gets tricky is common mode input voltage ranges.  This is not a problem for inverting amplifiers or non-inverting amplifiers which have enough gain but making a high voltage operational amplifier driven buffer or low gain non-inverting amplifier requires special consideration.

[dannyf]

Basically, use the collector as the output, rather than the emitter. With that approach, you get to within 100mv of the rail easily.

However, it is quite difficult to get such an amp to stablize, particularly with high (open loop) gains.

[David Hess]

I would not say it is difficult so much as both the local and global frequency compensation will need to be more seriously considered.

[Jay_Diddy_B]

Hi,
The original poster was asking about rail-to-rail output stages. Most of the links that have been provided are for booster stages to extend the voltage swing.

Here is the LM10 schematic from National Semiconductor's AN-211:



The LM10 was designed by Bob Widlar, it was the first op-amp to have a rail-to-rail output stage. To obtain a rail-to-rail output it is necessary to saturate the output transistors. Saturating the output transistor can lead to other problems like over-drive recovery.

You can use the circuit to make a non-rail op-amp into a rail-to-rail circuit. (Note: this is not a mistake the outputs are taken from the supply pins on the op-amp):



The circuit has been set up with a gain of -10. This shows the output been driven very close to the rail:



If I increase the input voltage to 1.05V, to cause the output to clip I see some nasty over-drive recovery problems:




Increasing R3, can help a little.

I have attached a zip file with the LTspice model.


Regards,

Jay_Diddy_B

[David Hess]

If I increase the input voltage to 1.05V, to cause the output to clip I see some nasty over-drive recovery problems:

The recovery delay in this case occurs because of the frequency compensation around the operational amplifier so adding local feedback from the discrete output to the node at the output of the operational amplifier will help with overdrive recovery by allowing less feedback capacitance.

[Jay_Diddy_B]

Hi group,

The recovery delay in this case occurs because of the frequency compensation around the operational amplifier so adding local feedback from the discrete output to the node at the output of the operational amplifier will help with overdrive recovery by allowing less feedback capacitance.

I have made some changes to improve the over-drive recovery. I have clamped the voltage on the voltage on the compensation network with diodes D1 and D2. I have also reduce the values of the feedback resistors R4 and R5.

New model:



Over-Drive test results:



I have attached a zip file with the modified model.

I am not sure what performance the original poster is looking for?

Regards,

Jay_Diddy_B

[BravoV]

...<snip>...

(Note: this is not a mistake the outputs are taken from the supply pins on the op-amp):

To be honest as a noob, I fail big time understanding this concept alone. 

Time to print the circuit on the paper, and going to study it for hours until I fall asleep on the bed, hopefully there will be an eureka moment. 

Intriguing, thanks Jay_Diddy_B, bookmarked !

[David Hess]

...<snip>...

(Note: this is not a mistake the outputs are taken from the supply pins on the op-amp):

To be honest as a noob, I fail big time understanding this concept alone. 

It makes it easier when you consider that this configuration turns a voltage feedback amplifier with one inverting input and one non-inverting input into a current feedback amplifier with two inverting inputs and one non-inverting input.  Or maybe it does not make it easier.

[DanielS]

To be honest as a noob, I fail big time understanding this concept alone. 

Using the power pins as outputs is a fairly unconventional way of using an opamp. In that sort of configuration, it is all about how much current the opamp is sourcing or sinking from/to ground.

I would have never thought of using an opamp this way either.

[David Hess]

Using the power pins as outputs is a fairly unconventional way of using an opamp. In that sort of configuration, it is all about how much current the opamp is sourcing or sinking from/to ground.

I would have never thought of using an opamp this way either.

The oldest design which used this configuration that I know of is the Tektronix 7A29 vertical amplifier designed before 1986 using the 741 which was released in 1968.

[GK]

Based on a 741, had been the generic audio power amplifier circuit topology used repeatedly in the ARRL handbook for decades - from my 1984 copy to my 2000 copy. It probably dates to the seventies, but I don't have any editions from that decade.


EDIT: Just flicked through by 2000 edition of the ARRL handbook and couldn't find any examples of it. My memory must be hazy, but I have seen it in relatively modern ARRL publications, possibly QST. Don't know where my 1984 edition is.

[Jay_Diddy_B]

Based on a 741, had been the generic audio power amplifier circuit topology used repeatedly in the ARRL handbook for decades - from my 1984 copy to my 2000 copy. It probably dates to the seventies, but I don't have any editions from that decade.

Hi Group,


Glen is right. My earliest memory of using the power pins as the output is from an audio amplifier, may be 30 years ago. The circuit looked something like this:



I have taken a generic op-amp circuit that I was developing in another thread and wrapped the rail-to-rail output stage around it. I have enclosed the op-amp in a box and marked the five terminals that we are used to seeing.

The other thread is: https://www.eevblog.com/forum/beginners/simple-bjt-amp-stagelow-thd/msg494304/#msg494304

This model gives some insight into how these circuits work:



The performance in over-drive is reasonable:




With +/- 10V rails and 19V pk-pk output signal the THD is 0.003709% at 1kHz.

I have attached a zip file with the LTspice model for those playing along at home.

These circuits are not particularly good. It is difficult to get reliable biasing of the output stage, or you can use them in class B and get crossover distortion.



Regards,

Jay_Diddy_B

[GK]

The topology was popular for portable, battery powered gear because you got the most output voltage swing from the limited rail voltage and current consumption is low as the power output transistors do not have a quiescent current. It's also simple to build as it doesn't require quiescent current adjustment nor temperature compensation with a bias generator physically/thermally coupled to the output devices. Suffers badly from crossover distortion though, particularly over the top octaves of the audio frequency spectrum. Though that doesn't matter much for radio speech <3 kHz.
       

[dannyf]

Using the power pins as outputs is a fairly unconventional way of using an opamp.

It is fairly conventional way to boost the output voltage.