[SOLVED] [moved] programmable negative Bias voltage

[zhangz64]

I am trying to generate a programmable negative Bias voltage from 0 to -2V with <0.5mV error.

I was able to generate a programmable positive bias voltage of similar range using DAC. based on this resource .

https://www.ianjohnston.com/index.php/onlineshop/handheld-precision-digital-voltage-source-detail

My idea is to have a negative 5 Voltage supply on the board and connect 0V to 5V, -5V to the ground, So everything is inverted, it will just give me the negative bias voltage with the same accuracy.

However, My problems is that The DAC requires SPI communication from uC which is 3.3 V. I could not talk to the negative section of the board since They are at completely different voltage level.

Is there anyway for me to get around this? Do i have to get another uC on the board just to talk to that DAC?

Thanks a lot for the help.
Eddy

 

[dick_freebird]

Re: programmable negative Bias voltage

Add the negative supply (-5V) and find yourself a rail-rail
op amp, hook it up in inverting configuration with suitable
gain (A=-1 probably) and hope you get one with a decent
Vio.

 

[zhangz64]

Re: programmable negative Bias voltage

Add the negative supply (-5V) and find yourself a rail-rail
op amp, hook it up in inverting configuration with suitable
gain (A=-1 probably) and hope you get one with a decent
Vio.

Thanks for the reply.
I think you are suggesting I use op-amp to invert the positive bias voltage to get negative bias voltage?

The block diagram will be Positive Bias Ouput 0-3.3V -----> Op-Amp(-5V to 5v) -----> Negtive Bias output ?

or Should i invert the SPI to get -3.3V SPI voltage?

Sorry if the question is too obvious.

Eddy

 

[KlausST]

Hi,

Instead of the confusing textual descriptions you should use a simple drawing.

--> It makes no sense to invert SPI lines. Juse invert the DAC output.

Klaus

 

[zhangz64]

Hi,

Instead of the confusing textual descriptions you should use a simple drawing.

--> It makes no sense to invert SPI lines. Juse invert the DAC output.

Klaus

Hi Klasus,

Thanks for the reply.

Here is the circuit i simulated, as you suggested .


I considered resistance mismatch, and offset voltage of the op-amp.

I do have some Questions for my complete understanding:
1) What are the consideration for Resistance value selection?
My thinking is to have 2 resistance value larger than 10K, (With <0.02% Tolerance), So That the trace resistance will not contribute much to the over error.

Also Since My input are DC, I do not care about the speed. Does this make sense?

2) For the op amp,
I know that V_offset will contribute some errors, What other errors from Op-amp could impact to my circuit? et offset current

3)For the op amp, Can i just tie the VCC+ to the ground, and Vcc- to -5V? What is the problem with that?

Also, I was suggested to use rail to rail op-amp, Why does rail to rail matter?

4)For the op amp, I have the option to tie the VCC+/- to 5v or 3.3V. Does it matter in terms of output accuracy?

Thank you so much for your helps,
Eddy

 

[KlausST]

Hi,

0.5mV accuracy at 500mV means 0.1% error.
Therefore you should choose 0.1% tolerated resistors.
But since you have two equal resistors you don't need absolute accuracy, but relative accuracy.
Use two resistors from the same production batch
( I assume 1% tolerated resistors are sufficient, but I can't give a guarantee for that)

Expect trace resistance of only mOhms. Not critical. You can improve performance with a good PCB layout.

Speed doesn't matter. You need high open loop gain, low offset voltage, low offset current (0.0005V/10000Ohm), use a 5k resistor at the noninverting input. Select a unity gain stable precision opamp. Use a capacitor across the feedback resistor to reduce nois (frequency).

If you use GND at the V input, then you need an Opamp capable to input = positive rail. RR input. Output is 0.5V from rail.
--> I recommend to use a true positve supply if available.(No RR opamp necessary)

Klaus

 

[zhangz64]

Hi KlausST,

Thanks for the detailed reply.

I picked a Chopper-Stabilized Operational Amplifier from TI. It has ultra low input offset voltage and offset current.
https://www.ti.com/product/TLC2652A

I also picked a resistor matching network from LT spice. It has 0.01% matching Accuracy and 0.2ppm/°C Matching Temperature Drift. It is less than 10 dollars.
https://www.linear.com/product/LT5400

Sorry for the additional questions

1)Why is 5k resistor recommended? What are the consideration? Why not 100K?

2) For the low offset current, you wrote " (0.0005V/10000Ohm)", How does the offset current impact the accuracy? Where is this 10000Ohm number coming from? Is this the reason i should not be using a >5K resistor?

2)For the capacitor filter cross R-feedback, Should i just pick a large cap (10uF) since We do not care about the speed at all?


Thanks a lot for your help,
Eddy

 

[asdf44]

1/2) Offset current multiplied by the R at the input pin creates an offset just like offset voltage. So smaller R's, like 5k instead of 100k, limit the impact of offset current (and current noise). The tradeoff is the current wasted in the R. Very low power applications would use 100k or more, but if a mA doesn't matter 1-5k is good.

3) 10uF starts to become large physically. 0.1uf would be fine. If you want, run an AC analysis in LTspice to see the impact.

Try to keep in mind an overall balance to the design. 0.2ppm drift is a pretty fancy resistor. If that's what you need make sure the rest of the design is up for it, for example your DAC (linearity? how many bits?) and its reference are just as critical.

 

[KlausST]

Hi,

Chopper stabilized TLC2652. Will do the job. Higher quality than you need.
******
Resistor network: this really is overkill. A 0.1dollar will do.
Why? I see you used worst case conditions to get 0.2mV (but you want less than 0.5mV). This is not realistic. Your result is +/-0.02% tolerance.
The most imortant error sources of a resistor are: initial production accuracy, temperature drift, drift with time.
Initial accuracy:
It may be a bit off the desired value, far less than the specified value, but it has very low error distribution within one batch.
If both resistors are off the same value it does not introduce voltage error in your circuit. (Both 10k give the same output as both 11k.)
Temperature drift:
Since you use the same value resistor from the same production batch, they both use the same resistance material with same temperature drift. They both drift the same value with the same temperature. In your circuit this causes no considerable error in your circuit. The only thing you have to care of is that the resistors should have about the same temperature. Place both resistors close together, keep them away from heat sources.
Drift with time:
The same as with temperture drift: Both will drift about the same value, causing no considerable error.
..remaing mainly the initial production distribution error. Any 0.1% specified resistor will be well within your recommendations.
(But this is only true with two resistors from one batch. If you ever have to replace one of them, you need to replace the other, too).
It will be more challenging to clean the PCB after soldering to maintain precision.

Why 5k: 10k || 10k give 5k. This is to match the input impedance of both Opamp inputs. Then only offset_current counts and not bias_current(without resistor).

Offset current:
input_offset_current x feedback_resistor R3 = output_offset_voltage (independent of R2)

Capacitor:
No electrolytic, no tantalum because of leakage current.
Use a high quality ceramics with relatively high voltage rating...maybe 100V
Best are foil capacitors, like MKT. Material not critical. Use a molded one.

******
True: Reference accuracy, DAC accuracy. Keep thermal influence in mind (thermocouple effects on each solder joint).
Any flux residuals may cause errors. GND guard rings around IN+ and IN- may prevent from creepage currents causing errors.

Klaus

 

[zhangz64]

Hey Klaus,

We got the parts this Monday,

The TI Op amp did not work out, It has a lot of chopper noise.
I ended up used the op amp from ST (TSZ121)
(https://www.st.com/content/st_com/e...lt50-mhz/low-power-op-amps-lt1-ma/tsz121.html)

The circuit worked very well, I see virtually no error from the op amp inverting stage.

Thanks a lot for the help,
Cheers,
Eddy