OPAMP with better Steps?

[sgt_bear]

Hello,

I want to have a voltage supply that gives me an output of 25 to 33V with very low noise on the output (2mV max). For this my plan was to use a low noise step-up regulator (35V) and then use an opamp (the LT1001) and regulate the voltage with a DAC

I simulated the schematic in lt spice and it seems to work. My plan was to use a DAC with a 4.096V reference. My only problem here is, that the steps are to small.

For example when i simulate with 1V output of the DAC, the ouput of the opamp becomes only 11V, if i use 4.096V i get my 34V output.
How can i modify the circuit that i only can regulate from 25V to 33V, but have more smaller voltage steps?


Find attached my simulation

[TheUnnamedNewbie]

It would be usefull to add a image of the schematic, so people don't have to download yours and open it up in LTSpice to see what you are trying to do.

From the sounds of it you would want to get a DC offset in your system. This is possible with various opamp schematics, such as referencing your system to a different voltage instead of referencing to ground.

[sgt_bear]

Find attached the image

[Ian.M]

I assume you have a 4.096V Vref externally available.   Put a potential divider between your DAC output and your reference so you have a lesser voltage swing closer to your reference voltage.   N.B. if your DAC output is Hi-Z or its output impedance isn't well controlled, you'll need a unity gain OPAMP buffer between its output and the divider to Vref.    I've also tweaked the LT1001 feedback to reduce the gain and make better use of the DAC's output range.   

On running the sim, choose the .dc sweep to see its effect over the full DAC range.  I've also parametrized the PSU noise so it can easily be switched off (e.g to see the ideal output or to reduce simulation time) or modified.  .param nf controls on/off and .func noise switches it using an if() and contains your noise expression: white(2e7*time)/100

LTspice note: If multiple simulation commands are present, after first selecting a sim, its chosen by default and all the others are commented out. To chose a different sim uncomment its simulation command by changing the leading ';' to '.' and run again. 

Edit: the designator of the resistor above R4, value 3K9 from Vref to LT1001 +in should be 'R3'  not '3K3'  - I clicked the wrong field when experimenting with preferred resistor values to get the desired range.   Also R2 should be 7K23 (E192 series), or 7K32 (E96 series)

[ArdWar]

SLOA097: Designing Gain and Offset in Thirty Seconds

PS: It's more than thirty seconds. 

Are you sure using LT1001? What's the switching frequency of your boost converter?
A lot of precision opamp have poor CMRR at higher frequencies, though 2mVpp ripple output is not that demanding if your supply is not that noisy to start with.

[mikerj]

Your step size for a given voltage range is determined by the DAC resolution, not the op-amp.

[sgt_bear]

Are you sure using LT1001? What's the switching frequency of your boost converter?
A lot of precision opamp have poor CMRR at higher frequencies, though 2mVpp ripple output is not that demanding if your supply is not that noisy to start with.

I want to use LT1001 because its very stable, low ofstet,low ripple. But thats just a "personal choice" any opamp with similar performance will work.

My switching converter is the MAX502 X, with 500kHz
https://datasheets.maximintegrated.com/en/ds/MAX5025-MAX5028.pdf
Page  5 claims ripple less than a mV.

The 2mV should be the absolute maximum output ripple at the opamp output!

[BrianHG]

Use a second DAC, with a resistor divider between the 2, or, a higher resolution DAC to begin with.

Dac #1 does the bulk voltage steps.
Dac #2 does the fine voltage steps.

You can always RC filter the output of 1 pre-buffer opamp which has your voltage gain to get a really ultra clean reference source.

[Ian.M]

Unfortunately the two DAC solution is almost invariably non-monotonic, so you need a higher resolution ADC (+ extra software complexity) to close the loop.

To get better than 2mv accuracy, including ripple, the resolution over your required 8V range (25V-33V) must be significantly better.  A 14 bit DAC will give you approximately 0.5mV resolution, depending on the exact limits of the output voltage range you design for.   Its also possible to use a lower resolution parallel DAC, dither its LSB via PWM and aggressively low pass filter its output to reduce the PWM frequency breakthrough to a low enough level.

[BrianHG]

You are correct and I was contemplating why the OP wouldn't just use a bottom end instrumentation 16bit dac with 20v out to begin with & just multiply that output voltage by a minimal 2.

Example: AD5761ARUZ
Internal reference, programmable for up to -10v to 10v, or, 0-20v output.

However, I assume the OP wanted to use a low resolution DAC because he was using one built into his MCU and wanted to avoid using an external precision DAC.

Datasheet:
http://www.analog.com/media/en/technical-documentation/data-sheets/AD5761_5721.pdf

[sgt_bear]

To get better than 2mv accuracy, including ripple, the resolution over your required 8V range

Just to be clear, ripple/noise is my major concern with the circuit. The "true ouptut" voltage is not as critically as the ripple. For example, it doesn't matter if the output is somwhere between +-50mV of the desired voltage, but the output voltage must be stable

My calculation was as following: 8V range, 50mV/steps = 20step/V = 160 Steps complete.

8-bit DAC would give 4096 steps, but 8bit are quite rare these days, more common are 12/16bit
12-bit DAC would give  4096 values which would be 2mV/step which would be MORE than enough for me!

I have LTC1451 in stock, maybe this would do??
http://www.linear.com/product/LTC1451

[BrianHG]

Ok, based on your comments, if you take the output of your LTC1451, feed it through a 10 to 100 ohm resistor, then place a 100uf cap to GND.  Your output switching bandwidth will be in the 1hz range, but after that 1 second when the 100uf cap settles to your new DAC setting, it will be as rock solid as the GND where the other side of the cap is connected to.  You can do a similar thing to the output of your op-amp and that will kill most noise above 1hz.  Now the only noise you need to worry about is the sub 1hz stuff.  This will be mostly down to the VLF noise of the DAC's internal vref.

I honestly believe with this rc filter, you will need a differential probe and audio band spectrum analyzer to find the noise here.  Your scope's probe in 1x mode with a short GND spring (not GND clip on wire which makes a loop antenna) right on the cap - and tip on + will reveal close to nothing.  Most likely nothing at all if your project is battery powered and your MCU sets the DAC then goes into sleep mode.  Any other noise seen will be coming from other radiated sources.

[MasterT]

It seems to me, that arduino project from this link is what you need.
http://henrysbench.capnfatz.com/henrys-bench/arduino-projects-tips-and-more/a-powerful-little-arduino-programmable-10v-supply/
It does voltage setting properly by simple voltage translation.  So I 'd not recommend of using TIP122, for low current 2N2222 is better choice, since it has higher transition frequency. Any BJT with Ft > 100 MHz is o'k

[sgt_bear]

Ok, based on your comments, if you take the output of your LTC1451, feed it through a 10 to 100 ohm resistor, then place a 100uf cap to GND.

I've changed my circuit, did you mean it like i did?

R5+C5 and R6+C6

[BrianHG]

You only need vdac.  Vref will no longer be used.

I'm not sure about the noise feature for you simulation, but if you can add noise to the vdac signal with a similar level and bandwidth to the data sheet, your simulated output should show you what the noise looks like.  You can open and close the 100uf cap to see the difference on the simulated scope.

[Ian.M]

There's a couple of very odd things with your sim:  Why are you injecting noise with respect to ground in the negative supply to the OPAMP, and why is there a 0.5V offset from Ground there?

Personally I'd use a capacitance multiplier to reduce the 500KHz ripple on the OPAMP's Vcc rail by a couple of orders of magnitude.   Enough HF PSRR for OPAMPS can be hard to come by, and putting a capacitance multiplier in the feed using a transistor with a high enough gain and  f_T can easily be worth an extra 40dB at HF while keeping the total smoothing capacitance on the boost converter output small enough to get a reasonable startup time.

Brian's idea of a passive low pass filter on the output is good - if you can stand that extra 100 ohms of output resistance, otherwise add a capacitor to ground at the OPAMP +in to form a low pass filter with the Thévenin equivalent of the divider resistors.  It wont help with OPAMP PSRR but will help with DAC noise.

N.B. avoid High K ceramic caps - they are excessively microphonic, which will impact stability.