As is shown in my oscilloscope, the power has noises of -20mV-+20mV broadband noises.

I suppose most power supplies contain noises, am I correct?

If the power is not clean, how could high precision circuit be made?
For example a 10V,24 bit DAC will have 0.6uV precision.

How could that be possible???
Are by-pass capacitors enough???

Thank you

 

For such precision you need to carefully separate the analog and digital parts of the A/D chain, and use filtration on the power supply for the analgo parts and the A/D converter. Also grounding is critical.

 

For such precision you need to carefully separate the analog and digital parts of the A/D chain, and use filtration on the power supply for the analgo parts and the A/D converter. Also grounding is critical.

Could you please introduce more about the filteration of the power ?

Thank you.

 

Depending on the frequency and amplitude of the noise, some simple RC or LC filtering may be enough.

 

Hello,

Read this thread:
Decoupling or Bypass Capacitors, Why?

I have posted there a PDF on RF prototyping.
In there isolation of power sections is given.

Bertus

 

Go to the TI web and type into the search field "2005 Practical analog design seminar". The Signal Integrity section of this seminar contains the practical info you are looking for. Notice that power and ground are considered to be noise free when they come onto the PCB, this means a lot of prefiltering and decoupling. 24 bit goes beyond this reference because more advanced techniques can be required, but they are available if you will look for them. Don't get enamored by the op amp circuit design program because I don't think it has been maintained.

 

I use toroid cores + magnet wires with good results, as well for AD, I use a 78L05 regulator for the reference voltage.

For 24bits precision you may need quite complicate circuits, as outlined above.

 

If you read the data sheets on high resolution converters, they make it very clear that the least significant bits will be very noisy, and that you need to take extraordinary care when building the circuit to even approach 24 usable bits of resolution. It isn't unusual to have noise even at the 12 to 16 bit level.

You are right that power supply noise will be a factor, but so will the noise in the voltage reference, whatever used for that, and noise in any amplifiers in the signal path, and thermal noise in the resistors, and noise received by local power line radiation, and noise due to capacitor dielectric retaining some of its charge-induced polarity, and noise due to temperature varying the output of the voltage reference; the list goes on and on of what can contribute noise to an ADC output.

 

Hello,

Read this thread:
Decoupling or Bypass Capacitors, Why?

I have posted there a PDF on RF prototyping.
In there isolation of power sections is given.

Bertus

Thanks for your reply. I found strange behavior on power supply today.

When the oscilloscope is not connected with anything, there is no noise. It means that the noise is not received by the probes.

However, when probes connected with the ground of power, EVEN WHEN THE POWER IS OFF. The noises showed up and they cannot be filtered by capacitors

 

I use toroid cores + magnet wires with good results, as well for AD, I use a 78L05 regulator for the reference voltage.

For 24bits precision you may need quite complicate circuits, as outlined above.

Could you please explain more? What is the resolution you need for your application? 1mV?
Did you use toroid cores and magnet for inductance?

 

Could you please explain more? What is the resolution you need for your application? 1mV?
Did you use toroid cores and magnet for inductance?

yes I use small toroid cores but in the end it is just an inductor.

I use 12 bit ADC only. Together with dc/dc converter, this is totally unstable, so I added filtering, and a 78L05 regulator as well. That's pretty much about it.

I also need toroid inductor for using I2C circuit having a long supply wire from an electronic transformer.

 

I use a 78L05 regulator for the reference voltage.

According to your experience, that reference is good up to how many bits?

 

In a NEC datasheet I found just temeprature drift of 0.5mV/K. Thus for a 5V reference each degree celsius shifts the reference at the 14th bit. So I´d say it may be good enough for a 10-bit ADC.
Note that no one specifies drift caused by aging etc.

 

In a NEC datasheet I found just temeprature drift of 0.5mV/K. Thus for a 5V reference each degree celsius shifts the reference at the 14th bit. So I´d say it may be good enough for a 10-bit ADC.
Note that no one specifies drift caused by aging etc.

What about the noises generated by 78L05?

By the way, what is the relation between required precision and the precision of voltage reference?

 

Hello,

Do you have the two capacitor s close to the regulator mounted?

See also the notes in the picture.

Bertus