Author Topic: bandwidth of d-element active filter opamps(gyrator/general impedance converter)  (Read 6220 times)

0 Members and 1 Guest are viewing this topic.

Offline CopperConeTopic starter

  • Super Contributor
  • ***
  • Posts: 1415
  • Country: us
  • *knock knock*
I am looking at using a 2nd order active filter thats built using synthetic inductors.

This is the base used for the gyrator

http://www.ee.nmt.edu/~wedeward/EE212L/SP15/ImpedanceConvertersFig2.jpg

In general with this topology, can you estimate the gain of the op-amp? I have no clue where to begin with the analysis.

For instance,



This guy has a bandwidth of 50KHz at 1MHz.

Now, I am curious because currently I use one of these at a much lower (audio) bandwidth, but if the Op-amp GBW is good (Mine is 55MHz), perhaps I could do parts replacements to shift, since the topology will be the same, during prototyping.

Any guess to how much GBW you would need there? The analysis seems hairy.

I never thought about it, since with audio your GBW is like 3 orders of magnitude above what your working with.

Otherwise, I need to make a jumpered magnetic based filter option

If someone can show me how to calculate it, that would be useful to see whats the practical range of this filter, e.g. some opamps have a gainbandwidth of 1.2GHz, but these are CFB, idk if they would work here... I don't really like magnetics
« Last Edit: January 26, 2018, 12:06:26 am by CopperCone »
 

Offline The Electrician

  • Frequent Contributor
  • **
  • Posts: 747
  • Country: us
A first step might be to get an expression for Zin of the circuit you linked to: http://www.ee.nmt.edu/~wedeward/EE212L/SP15/ImpedanceConvertersFig2.jpg

in terms of the Z elements and let the opamp gain be A.  Then you can allow A to have a single pole roll off as a first approximation.

Have a look at these:

https://electronics.stackexchange.com/questions/130632/general-impedance-converter

 

Offline Wimberleytech

  • Super Contributor
  • ***
  • Posts: 1135
  • Country: us
I would avoid using opamps in a common-mode configuration (having a common-mode input) as is the case for these gyrators.
Filter topologies where the performance is distributed among all components (such as RLC ladder filters) are the most insensitive to component tolerance, but the most difficult to debug IMHO.

Cascading biquads is far easier to understand, build, and debug (and tune)...and there are millions of examples of them in production.

However, if you are determined to use this architecture, I am pretty sure there is a paper that analyzes effects of GBW.  Go search IEEEXPLORE.  Active filters were a hot topic in research back in the mid to late 70's into 80's.  I just did a search gyrator&gain&bandwidth and got a lot of interesting hits.

Alternatively, you could simulate this in SPICE very easily.  Build a very simple single-pole opamp macromodel.  Dial in various GBWs and see how that works.

---
Looking back at your post, I want to comment about your audio 3 orders of magnitude comment.  Remember, GBW is a figure indicating the frequency at which the open-loop gain is unity.  I don't think a gyrator is very happy at an A(OL)=1.  Lets just say that a gyrator needs at least 60dB at 20kHz to work correctly (probably too low).  It will need an opamp with a GBW of 20MHz.
« Last Edit: January 28, 2018, 02:40:44 pm by Wimberleytech »
 

Offline CopperConeTopic starter

  • Super Contributor
  • ***
  • Posts: 1415
  • Country: us
  • *knock knock*
Thats pretty high, 60db.

Will this circuit work with very high speed current feedback op-amps ?

What kind of modifications can be expected to make a circuit work if it does not, other then component choice? (e.g. potential places to put feedback capacitors, etc)

I wonder if the AD8009 1GHz op-amp is up to the task.

I unfortunately do not have money for IEEE membership or journal downloads.

Anyone wanna get this for us?
https://www.researchgate.net/publication/3439479_Gyrator_Bandwidth_Limitations
« Last Edit: January 29, 2018, 01:57:20 am by CopperCone »
 

Offline danadak

  • Super Contributor
  • ***
  • Posts: 1875
  • Country: us
  • Reactor Operator SSN-583, Retired EE
Some ref material -


https://inst.eecs.berkeley.edu/~ee247/fa07/files07/lectures/L4_f07.pdf


http://www.sophphx.caltech.edu/Physics_5/Experiment_5.pdf


https://core.ac.uk/download/pdf/36706499.pdf


http://downloads.hindawi.com/journals/apec/1989/039762.pdf


http://iowahills.com/9RFFiltersPage.html


http://dergipark.gov.tr/download/article-file/99252


Consider using a simple LC filter design software, to get ideal values. Then do a sim
of one section and OpAmp of choice. To get the BW you need most OpAmps capable
are two pole, and stability will be a concern. Possibly simple TD analysis with a scope,
impulse source, and looking at deviation from ideal LC will give some indication


Regards, Dana.

« Last Edit: January 29, 2018, 02:39:14 am by danadak »
Love Cypress PSOC, ATTiny, Bit Slice, OpAmps, Oscilloscopes, and Analog Gurus like Pease, Miller, Widlar, Dobkin, obsessed with being an engineer
 

Offline The Electrician

  • Frequent Contributor
  • **
  • Posts: 747
  • Country: us
A standard nodal analysis with ideal opamps shows that your circuit has a center frequency of about 847 kHz and a bandwidth of about 42.5 kHz:

 

Offline danadak

  • Super Contributor
  • ***
  • Posts: 1875
  • Country: us
  • Reactor Operator SSN-583, Retired EE
The Electrician, what tool did you use that produces the
matrix printout you posted ?


Regards, Dana.
Love Cypress PSOC, ATTiny, Bit Slice, OpAmps, Oscilloscopes, and Analog Gurus like Pease, Miller, Widlar, Dobkin, obsessed with being an engineer
 

Offline Nitrousoxide

  • Regular Contributor
  • *
  • Posts: 157
  • Country: au
The Electrician, what tool did you use that produces the
matrix printout you posted ?


Regards, Dana.

Looks like Wolfram's Mathematica.

Very powerful and beautiful program, just not as useful as Matlab.  ::)
 

Offline CopperConeTopic starter

  • Super Contributor
  • ***
  • Posts: 1415
  • Country: us
  • *knock knock*
Yes the filter was made to use 20% tollerances on the capacitor so it would default to nice capacitor values with foik capacitors.

Im kind of curious to see what will happen with a 55MHz opamp. Not quite 60db of gain margin but i soldered up half of it already

I was just gonna make it, test for oscillation with input grounded then plug it in into.a e5000 series 300MHz network analyzer to get a bode plot.

Can you clarify other test metrics? What kind of impulse testing should i do? I have a. Variable rise time pulse generator (big ass wavetek) and a 70picosecond pulse generator (old tektronix type 264).

My goal is stablity i found in simulations that this type of filter has the highest stability. Pp Capacitors have like 160ppm drift and in second order configuration this filter just about doubles the drift to around 320ppm according to montecarlo analysis of around 20000 iterations. Salen key etc were much worse. If ceramic caps are used (and possibly polyphelenine sulfide) this might only mean -~40ppm drift which is very respectable for a 2nd order filter.  I dont think you can get within an order of magnitude of this stability with magnetics. I found during my own tests that some random ferrite bourns magnetics had approximately 600ppm drift which is then combined with the drift of the capacitors; possibly made worse by a second order.

I did read something however that said an inductors tcl corresponds to about 70 percent of its mu. I dont know if i measured an inductor with a extremely high mu or if my test setup was wrong or if i tested to too high a temperature, i.e. i tested to a 40c delta above room temp and the mu*0.7 rule only holds true close to room temp.

I see there are materials that have mu´s of 14 or so, if this is true then it is highly possible a gyrator could be easily beaten by the right choice of magnetics. It appears i need to do some research on what i tested. I used a de5000 esr meter for my test a long while back.

I see a similar filter is possible with inductors at the range of microhenries and icapacitors around nanofarads, but the required values are not easily obtained. I guess i will play around in a simulator after taking some measurements.
« Last Edit: January 29, 2018, 06:03:59 pm by CopperCone »
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22435
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
A standard nodal analysis with ideal opamps shows that your circuit has a center frequency of about 847 kHz and a bandwidth of about 42.5 kHz:

https://www.eevblog.com/forum/projects/bandwidth-of-d-element-active-filters-(gyratorgeneral-impedance-converter)/?action=dlattach;attach=390679

Nice; it looks like this does not model op-amp GBW though, can you squeeze that in too?  (Hmm, as simple as substituting A --> 1 / (1/A + s/GBW) I think..?!)

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22435
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
I did read something however that said an inductors tcl corresponds to about 70 percent of its mu. I dont know if i measured an inductor with a extremely high mu or if my test setup was wrong or if i tested to too high a temperature, i.e. i tested to a 40c delta above room temp and the mu*0.7 rule only holds true close to room temp.

Material TC is all over the place.  It's usually a lumpy response with respect to temperature, as different domains become activated.  (Materials not optimized for flatness typically have mu ~proportional to temp, on the whole.)

Mu in most ferrites will be large (2000ish), but this doesn't matter much because average mu (what matters to the inductor in terms of turns and electrical properties) is reduced by an air gap.  Air gap is usually dominant (i.e., mu_eff << mu_r), so that you'd only expect to see the tempco of the airgap dominate, which should be a modest positive tempco corresponding to material expansion rate.  (Because, the air gap distance increases, reducing inductance, but the cross sectional area increases twice over, giving a net increase.)

Powdered iron depends on the mix used.  I don't remember offhand how it usually varies.

Air core inductors are typically positive tempco, though you can use differential TCEs to make them behave differently.  IIRC, some PCB mount "spring" inductors have near-zero TCE when mounted on FR-4 (which has a much higher expansion rate, stretching the inductor faster than it expands).

Resistors and the better capacitor types can have quite low TC, which is nice.  Beware the GBW contribution, which itself has a TC -- and good luck getting a guarantee on that parameter from the manufacturer!

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline The Electrician

  • Frequent Contributor
  • **
  • Posts: 747
  • Country: us
The Electrician, what tool did you use that produces the
matrix printout you posted ?


Regards, Dana.

Mathematica
 

Offline The Electrician

  • Frequent Contributor
  • **
  • Posts: 747
  • Country: us
Here are some results with various finite opamp parameters.  Av is low frequency gain, GBW and BW are the usual.

First, with GBW of 50 MHz and Av of 10000.  The center frequency is reduced and so is the gain:




Now with GBW reduced to 40 MHz; the changes are obvious:



Next with GBW back to 50 MHz but Av reduced to 1000:


« Last Edit: January 29, 2018, 09:39:06 pm by The Electrician »
 

Offline The Electrician

  • Frequent Contributor
  • **
  • Posts: 747
  • Country: us
The Electrician, what tool did you use that produces the
matrix printout you posted ?


Regards, Dana.

Looks like Wolfram's Mathematica.

Very powerful and beautiful program, just not as useful as Matlab.  ::)

Mathematica can do everything Matlab can do (and more), but you might have to work harder for some things.  It does have a steep learning curve.
 

Offline CopperConeTopic starter

  • Super Contributor
  • ***
  • Posts: 1415
  • Country: us
  • *knock knock*
Stupid question, but what corresponds on the circuit to Av being changed in your simulations?
 

Offline CopperConeTopic starter

  • Super Contributor
  • ***
  • Posts: 1415
  • Country: us
  • *knock knock*
I made a thread pertaining to inductor tempco here as a branch, I think the topics are too different
https://www.eevblog.com/forum/projects/temperature-coefficent-of-inductance-and-mu-factor/
 

Offline The Electrician

  • Frequent Contributor
  • **
  • Posts: 747
  • Country: us
Stupid question, but what corresponds on the circuit to Av being changed in your simulations?

The open loop response of the opamps is as shown here; Av is the gain at low frequencies:



 

Offline CopperConeTopic starter

  • Super Contributor
  • ***
  • Posts: 1415
  • Country: us
  • *knock knock*
Well I built one using a quad op-amp that has a 55MHz gain bandwidth, it appears to oscillate at 550KHz with a 1vpp square wave on the output.

I need to check my soldering and try a different IC, but it does not look good  :-[

I used carbon film resistors and ceramic capacitors (NPO probobly, don't know). Not ideal but I thought i'd get some kind of result.

Any idea of what the problem might be? Montecarlo did not seem to show problems with oscillation and component tolerance.
« Last Edit: January 30, 2018, 08:07:42 pm by CopperCone »
 

Offline The Electrician

  • Frequent Contributor
  • **
  • Posts: 747
  • Country: us
You have two GICs in this circuit: http://www.ee.nmt.edu/~wedeward/EE212L/SP15/ImpedanceConvertersFig2.jpg

What happens if you have the two GICs energized stand-alone with the top end (the Zin node) either short circuited to ground or left open, or with 10k to ground.  Do they oscillate in any of the 3 cases?
 

Offline Wimberleytech

  • Super Contributor
  • ***
  • Posts: 1135
  • Country: us
Well I built one using a quad op-amp that has a 55MHz gain bandwidth, it appears to oscillate at 550KHz with a 1vpp square wave on the output.

I need to check my soldering and try a different IC, but it does not look good  :-[

I used carbon film resistors and ceramic capacitors (NPO probobly, don't know). Not ideal but I thought i'd get some kind of result.

Any idea of what the problem might be? Montecarlo did not seem to show problems with oscillation and component tolerance.

What opamp did you use?
 

Offline CopperConeTopic starter

  • Super Contributor
  • ***
  • Posts: 1415
  • Country: us
  • *knock knock*
 

Offline CopperConeTopic starter

  • Super Contributor
  • ***
  • Posts: 1415
  • Country: us
  • *knock knock*
Also, do you think the design of it could be why its oscillating?



I think my breadboard is fucked up so I soldered it together into a DIP machine pin socket.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf

 

- - -
-->