Predicting filter performance with low-cost measurement

[Junus2012]

Dear friends

I have designed a tunable integrated LPF. Right now I am evaluating the performance with Cadence simulation. However, when it is coming to reality, measurement is needed. I am not trying to avoid the measurement but I am trying to have fast method in which for example by DSP unit can tell me about the filter properties.
I have seen something like that but couldn't find details

It will be very helpful if you could suggest me such a solution, especially that based on digital processing

Thank you
Best Regards

 

[KlausST]

Hi,

Your text includes no single technical specification.
This makes helping difficult, because we have to guess.
Is it audio frequency range? Is it Gigahertz range?
What 'items' do you want to measure..with which resolution, with which precision?

For audio range you may use the soundcard of a PC and free software.

Klaus

 

[Junus2012]

Thank you klaus for your reply

I am designed 4th order Butterworth LPF with tunable fc upto 5 MHz based on sallen key unity gain filter
It is targetted for 10 bit resolution

 

[KlausST]

Hi,

Again: "what" do you want to measure?
Noise, THD, amplitudes, phase shift....

Klaus

 

[danadakk]

Nano VNA would cover most of the range (it measures > 50 Khz).

Or do a bode plotter with an oscilloscope.

DIY Network Analyzer / Bode Plotter with Oscilloscope and Function Gen - Page 1

DIY Network Analyzer / Bode Plotter with Oscilloscope and Function Gen - Page 1

www.eevblog.com

Creating A Bode Analyzer From A Microcontroller

Electrical engineers will recognize the Bode plot as a plot of the frequency response of a system. It displays the frequency on the x-axis and the phase (in degrees) or magnitude (in dB) on the y-a…

hackaday.com

Shop - Red Pitaya

Products Archive - Red Pitaya

www.redpitaya.com

Or just manually with a function generator, scope, and paper and pencil.....


Regards, Dana.

 

[Junus2012]

Hi,

Again: "what" do you want to measure?
Noise, THD, amplitudes, phase shift....

Klaus

Dea Klaus, most important thing for me is THD and cutoff frequency and the filter should always be Butterworth......

--- Updated Nov 23, 2020 ---

Nano VNA would cover most of the range (it measures > 50 Khz).

Or do a bode plotter with an oscilloscope.

DIY Network Analyzer / Bode Plotter with Oscilloscope and Function Gen - Page 1

DIY Network Analyzer / Bode Plotter with Oscilloscope and Function Gen - Page 1

www.eevblog.com

Creating A Bode Analyzer From A Microcontroller

Electrical engineers will recognize the Bode plot as a plot of the frequency response of a system. It displays the frequency on the x-axis and the phase (in degrees) or magnitude (in dB) on the y-a…

hackaday.com

Shop - Red Pitaya

Products Archive - Red Pitaya

www.redpitaya.com

Or just manually with a function generator, scope, and paper and pencil.....


Regards, Dana.

Thank you dear friends,

Indeed I am searching for alternate solution vased on DSP, to predict, tune and calibrate my chip during system operation....just imagine you have a filter in your phone in which be need calibration, so real measurement is not feasible for my application

 

[KlausST]

Dea Klaus, most important thing for me is THD and cutoff frequency and the filter should always be Butterworth......

How do you want to measure whether the characteristic is butterworth or not?

I am searching for alternate solution vased on DSP

You may use a PC you may use any microcontroller...

so real measurement is not feasible for my application

What deos this mean?
a simulated measurement?

Klaus

 

[danadakk]

Possibilities for the DSP + Core -

https://www.ti.com/processors/digital-signal-processors/products.html?pqs=paqs&familyid=44#p2100=50;100

Regards, Dana.

 

[Junus2012]

How do you want to measure whether the characteristic is butterworth or not?
You may use a PC you may use any microcontroller...
What deos this mean?
a simulated measurement?

Klaus

Dear Klaus

My filter has to be tuned in similar way in the figure below

Built-In Self Optimization for Variation Resilience of Analog Filters

At deep sub-micron integrated circuit technologies and beyond, characteristics of analog circuits are increasingly affected by process variations and device aging. Although design-time approaches are able to take these effects into account, it is very difficult for a single static design to...

ieeexplore.ieee.org

The author used bank of variable resistors to tune this BPF, I am not sure if you have an access to the paper or not, but the author was tunning for Q and bandwidth, so he used on chip digital approach for characterising the filter, no external equipment. In my case I am allowed to use DSP or Rassbarypi for signal processing, but I am not aware of technique in DSP that can extract fitler response properties, and that let me say my question

--- Updated Nov 24, 2020 ---

Possibilities for the DSP + Core -

https://www.ti.com/processors/digital-signal-processors/products.html?pqs=paqs&familyid=44#p2100=50;100

Regards, Dana.

Dear Dana

Thank you for sharing the link, it shoes only DSP units by TEXASINSTRUMENTS, I am interested on the idea of how to do that with DSP

 

[danadakk]

One approach is DAC tuned State Variable filter, but I think the 5 Mhz
response might be a challenge.

https://www.ti.com/lit/ug/tidu543/tidu543.pdf

The fig you show in the paper is showing a similar approach using banks of Rs
or digital pots. Again at 5 Mhz thats a bot of a challenge. Reason is you need
very wideband OpAmps, and layout, stabilization is not easy. Sub 1 Mhz these
problems much easier to handle.

Fast filter - https://www.ti.com/lit/an/sloa032/sloa032.pdf?ts=1606179083533

Of course you can do DSP in this range, takes a lot of power, feed the filter
white noise and look at spectral output. Phase measurement harder, best done
feeding it a sine from a DDS source.

There is always an approach, just use DDS, build a fast true RMS converter to
look at output (analog devices lots of info on that true RMS design). Or use
fast peak detector. That way you can get your 5 Mhz upper limit.


Regards, Dana.

 

[KlausST]

Hi,

your IEEE link does not work for me. Maybe one needs to be logged in.

I´m confused.
* DSP or not DSP?
* Measurement or simulation?

....

We have to guess. Do you want an "on chip" solution? Inside your ASIC?
If so, then a couple of questions arise:
* I´ve no experience in chip design. But I guess if there is an experienced member ... he/she needs to know what type of ASIC and what software you use.
* and what exactly do you need? calculations / mathematical background? code? IP?

Please clarify.

Klaus

 

[Junus2012]

Dear friends,

Thank you very much for your replies, that was very useful even thaugh still I need to clarify the problem

The filter which I have constructed is integrated is part of read out circuit. This chip is connected to a processing unit built by Rassberypi (you can think of it as my DSP unit).
The chip has to be in system tunned, means I have no possibility to disconnect the chip for measurement by instruments like network analyzer and so on.

Therefore, I am searching for a method that can tune/calibrate my filter in this chip. As I already mentioned, important factors for me is the cutoff frequency. Suppose it is shifting by time or temperature or in case I need to vary the fc, So how can the processing uint be able to predict the filter performance ? which kind of digital processing I have to apply for this purpose ?

Another possible solution presented by dannak is to have DDS unit that can apply variable test frequency signal to the amplifier input and recording its response, like a traditional approach of measuring filter transfer function. I would say I liked this solution in case there is no easier solution can be provided by the power of DSP

Thank you

 

[KlausST]

Hi,

It's time for a drawing.

Klaus

 

[Junus2012]

Hi,

It's time for a drawing.

Klaus

Do you mean to draw the circuit

 

[KlausST]

Hi,

The description still is unclear from the beginning. We don´t know what you want.
--> Any drawing that helps to clarify what you need will be useful.

I´m on the edge to leave this discussion, because I´m not patient enough to ask for clarification again and again.
In the end it´s you who gains from a good description.

Klaus