Designing A Differential Amp

[BrownOut]

I'm making a tri-output signal generator, the last stage of which is a triangle to sine wave convertor. I have the stage working and generating a nice sine wave, but I want to drive it with a differential signal. The attached circuit is intended to accept the single-ended tirangle wave, and produce a differential triangle wave to drive the convertor. The voltage driving the amp is identical to the signal I'm presently generating in the 'real' version ( linear trialgle wave from +4Vmin to +8Vmax )

Clearly, I have a problem in that the differential driver is easily driven into the non-linear region, Q1 being driven into cutoff. There are several solutions available to prevent this.

1) Increase the value of R5

2) Adjust R9/R8 voltage divider

3) Implement feedback (R4 to Q2 base)

Either of these solutions would work well for simulation. Does anyone have an opinion what would work best for real devices?

Waveform Guidance

Red: Source Voltage
Blue: Q1 Collector Voltage
Green: Q2 Collector Voltage

 

[ericgibbs]

hi,

I would use a constant current source in place of R1 and R2, connect the two emitters together.
Reduce the value of the two collector resistors.


Post your *.asc.

 

[vne147]

I know this doesn't answer your question and maybe you already know this and you're just making your signal generator from scratch for fun but they make waveform generator ICs that you could probably use for your project. Check out this datasheet:

http://www.datasheetcatalog.org/datasheets2/15/153108_1.pdf

 

[BrownOut]

Hi Eric, Here is my asm file (never posted one before, hope I get it right). I'm working on using negative feedback to control the gain. After thinking about it, I think that would be the method least sensitive to component variations. I'll post later about my first attempt at using feedback.

 

[BrownOut]

Hi vne... Yeah, I'm exploring a different way to generate these signals. But I do intend to build and use a permanent version of my circuit. Thanks for your reply.

 

[BrownOut]

Here is my attempt to use feedback. The gain is set as R10/R9. I'm pleased with the results of the sumulation. However, I used a voltage (V4) to cancel the DC component of the input ( pwl min=4v, max=8v) I need to replace it with a component. Once I started using feedback, I could no longer null the output using base bias resistors.

 

[BrownOut]

Final design. Worked all dang day on this. Good thing I've been in disability today LOL!

The sim attempted to treat R12 as a variable resistor, to null the output.

 

[BrownOut]

Me again. I have the whole circuit now, including the sine convertor. I'll prototype it this weekend, and post the results, if anyone is interested.

In the image, Green is the linear ramp from V3. Red is the resulting sine wave taken from the collector of Q7.

 

[unclejed613]

you might try this. basically it's a simplified op amp with a differential output. the nonlinearity problem you experienced is because the output of the diff pair itself is a current, and it must be converted and amplified to be linear. attached is a simple op amp, with outputs taken from both legs of the input diff pair. the transistor types aren't really important. i use 2N3904's and 2N3906's, or 2N2222's and 2N2907's for the NPN/PNP types you might want to play with the values for the collector resistors on the outputs. anything between 4.7k and 20k should work. the other resistor values aren't critical as long as R7 and R9 are equal. the gain of the amp is (R7/R8)+1. if you want a gain of 1, replace R7 with a short. R3 and R4 should also be equal. the linearity of this amp is better than a diff amp with resistors feeding the input emitters, because of the current source Q3. the diff amp current is determined by R1. R1 will always have about 0.7 volts across it, so the current source current is 0.7/R1. i set it at about 1mA. higher current=higher OPEN LOOP gain (which lowers the distortion in a closed loop amp). diff amp currents are usually anywhere from 20uA up to about 2 or 3mA. setting it too high will cause thermal drift. setting it too low will increase distortion.

 

[ericgibbs]

hi brownout,
If you have a couple of mid range OPA's on the shelf, this is a simple option.
Outputs are referenced to 0V.

Plots for BW and Signal

 

[BrownOut]

unclejed, thanks for your reply. The reason for the non-linearity in my original circuit was because the differential amp has very high gain, which quickly drives the amp into the nonlinear region. Your circuit solves the problem the same way mine does, by providing feedback to reduce the gain. Your Q4 cross-couples a portion of the output back to the input, establishing a closed loop. I lineraized my circuit in a very similar way, but my method uses two transistors, while yours uses only one. I was trying to remember how to do that without cracking my old textbooks. I might try to use your cross couple method in my circuit. At this point, the simulation is working. Time to fire up the soldiering iron.

 

[BrownOut]

Eric, thanks I can always count on you to come up with something For now, I'm going to stick with the discrete solution, as I'm trying to achieve some performance gains without using any extravgant IC's. I'll file your suggestions away for future references.

Right now, I have the circuit running about 300Khz. I'm hoping to push that up to about 1Mhz as I transfer it from and plug-in breadboard to a more permanent fixture.

I know there is probably some wiz-bang op-amp that will run at those kinds of frequencies. However, I like the durability and robustness of my all-transistor projects. I have a habit of operating my projects in hostile conditions. I do alot of ignition-system testing and high-voltage/high frequency experimentation. Every time I've done a post-mortem on a blown-up sig-gen, I've found fried IC's, but the discreets have always survived. Even though I use transient-absorbing components in my projects, I routinely get very fast 400V+ spikes that somehow find their way through the protection devices and do damage. Modern, IC based power supplies fry instantly, while old, transistor and SCR based power supplies take the hit and keep working. Further, I can get the performance I want without worring too much about layout issues.

Someday, when my process of building high performance projects is refined, I'll take another look at those hot-rod op-amps.

~Danny

 

[BrownOut]

I decided to try a different approach. Following unclejed613's example, I thought it would be good to use a cross-coupled feedback scheme, because it allows me to adjust the gain using only one resistor. It took me awhile to get this correct. It also uses a complementary driver for feedback, which exploits the VBE multiplier. We had a thread about this some time ago, so some might be interested in this as an example. Transistor Q9 balances the output voltages. You'll notice a slight difference in in the differential outputs, which is due to the circuits marginal common mode rejection. However, I'm only interested in getting a decent differential signal, so I'm calling it quits at this point, working to improve the CM rejection isn't really worth it for this project.

I did't want to re-invent the opamp for this project, but every 'simpler' circuit didn't give acceptable results. I still think I need to use discreets to get the performance and ruggedness I want, however.

 

[unclejed613]

it's best to have the Vbe multiplier transistor thermally coupled to the "output" transistors, if you want their operating current stabilized.... with small to-92 devices it's not as important as it would be for output stages of 10 watts or more.