Hi guys,

I'm running into some problems designing a 10MHz 8th order sallen-key butterworth lowpass filter.
My requirements are:

• cornerfrequency f_c = 10 MHz
• max passband gain of 0.5dB
• filterorder of 8
• available components: resistors, capacitors, op-amps

What I have tried so far:

I found this TI app-note http://dea.unsj.edu.ar/sredes/sloa088.pdf and tried using the approach from p.16/17/21 of cascading 4 second-order SK stages.







The values calculated are not so practical to begin with but I guess using a different given value for C_1 - C_5 might solve the problem. However simulating the circuit does not give me the expected result:


I also tried using the analog devices filterwizard to aid with the design, but it used an inductance which I can not use:

If somebody could point me in the right direction I would be very grateful as I do not know what I am doing wrong.

Cheers

 

You've got several challenges to overcome:

1. An ADA4807-2 op amp is likely not fast enough for your application; its gain-bandwidth product is only 18X the cutoff frequency of your filter. This will cause the filter behavior to depart significantly from that calculated. You need a MUCH faster op amp.

2. An 8-pole Butterworth S-K filter is extremely difficult to implement at ANY frequency, let alone at 10 MHz, while obtaining the passband flatness you desire. With that many poles, the required component tolerances become extremely tight.

Implementing a 2- or 3-pole S-K filter is a trivial task. 4 to 6 poles takes great care. And 7 or 8 poles becomes the stuff of nightmares.

Good luck.

 

Are you required to use a cascade of four 2nd-order stages?
For an 8th-order lowpass it is much better to use "direct" realization - which means: Active realization of a passive RLC reference structure (ladder topology).
Of course, I did not check the parts values you have calculated - however, something must be totally wrong. At least for very low frequencies (down to DC) the output must be flat (DC gain). However, your Bode diagram starts at -96 dB.
By the way: 3 of the handmade circuit diagrams are wrong.

What about using a filter design tool?

 

Hi there if I may be so bold I will not tell you what you're doing wrong I believe this is going to homework and you have done most of the work may I suggest.
MAX7480
8th-Order, Lowpass Butterworth Filter
Low Noise and Distortion: -73dB THD + Noise
Clock-Tunable Corner Frequency (1Hz to 2kHz)
100:1 Clock-to-Corner Ratio
+5V Single-Supply Operation
Low Power
2.9mA (Operating Mode)
0.2µA (Shutdown Mode)
Available in 8-Pin SO/DIP Package
Low Output Offset: ±5mV

 

Hi there if I may be so bold I will not tell you what you're doing wrong I believe this is going to homework and you have done most of the work may I suggest.
MAX7480
8th-Order, Lowpass Butterworth Filter
Low Noise and Distortion: -73dB THD + Noise
Clock-Tunable Corner Frequency (1Hz to 2kHz)
100:1 Clock-to-Corner Ratio
...

And just how does that work at 10MHz???

 

And just how does that work at 10MHz???

Maybe about as well as the original with the selected opamp?

 

I also tried using the analog devices filterwizard to aid with the design, but it used an inductance which I can not use:

And the reason for not using an inductor?

As has been said, 8-pole at 10MHz is going to need much faster amps, it will be hard to get working and very hard to keep stable, all sorts of parasitic effects will come in to play. Even if the simulator says it'll work, PCB layout will be critical, even minor voltage fluctuations on the supply will have odd effects...

 

Maybe about as well as the original with the selected opamp?

Probably a lot worse...

 

The filter wizard obviously was not cognizant of your stated requirements. That's a problem with wizards, they do what they want, which is not necessarily what you want.

 

Probably a lot worse...

Yaeah..OK. When a design does not work, a miss by an inch is not much different than a miss by a mile.

 

I have what may be an excessively picky point, but don't you determine the required filter order from the requirements instead of making an a priori determination. In other words, how do you know you need an 8th order filter?

 

Quote: I also tried using the analog devices filterwizard to aid with the design, but it used an inductance which I can not use:

They are using inductances for active filters? I cannot believe.Perhaps you have used a tool for passive structures.
There are many design tools available for active RC filters.
Again: Are you required to use the cascade approach?

 

The filter wizard obviously was not cognizant of your stated requirements. That's a problem with wizards, they do what they want, which is not necessarily what you want.

I cannot confirm this general statement. I know some programs which do what I want....filter topology, parts selection, resistor niveau,...

 

Are you required to use a cascade of four 2nd-order stages?
For an 8th-order lowpass it is much better to use "direct" realization - which means: Active realization of a passive RLC reference structure (ladder topology).
Of course, I did not check the parts values you have calculated - however, something must be totally wrong. At least for very low frequencies (down to DC) the output must be flat (DC gain). However, your Bode diagram starts at -96 dB.
By the way: 3 of the handmade circuit diagrams are wrong.

What about using a filter design tool?

The hand drawn diagram was just for quick visualization, I went ahead and corrected it:

Yes I am also curious to find out why the simulation shows more of a bandpass characteristic.
Sadly I can not use inductances and thus the second order stages are more or less required(I think).
If you had a recommendation for a filter tool I would be glad to hear it as I have not yet found one that does what I am looking for.

 

I cannot confirm this general statement. I know some programs which do what I want....filter topology, parts selection, resistor niveau,...

It was not supposed to be a general statement. In this particular instance the output contained an inductor.
The specifications I put in the tool are the following:

 

Filter free (Nuhertz Technologies)
Filterlab (Microchip Technology)
FilterPro (Texas Instruments)
Filter Wiz Pro (Schematica Software, www.schematica.com)

For "direct" realzation you are using the passive RLC topology for reference purposes only!
No inductors for realization. They are replaced by "active inductors" (opamp blocks).
___________________________________________________
Quote: It was not supposed to be a general statement. In this particular instance the output contained an inductor.
The specifications I put in the tool are the following

Where is the circuit? Did you choose "passive" realization?
An active Filter has no inductors - because THIS ist the main advantage of acrive RC blocks!

 

Given the stated requirements of -85 dB over 2 octaves, I think you might want to pick a different topology to reduce the required order. I'd try a Chebyshev Type I with 0.1 dB ripple in the passband. It will be hard to see the ripple on a standard Bode plot and the rolloff will be a good bit steeper. Do you know how to compute the required order given the specification?

 

Filter free (Nuhertz Technologies)
Filterlab (Microchip Technology)
FilterPro (Texas Instruments)
Filter Wiz Pro (Schematica Software, www.schematica.com)

For "direct" realzation you are using the passive RLC topology for reference purposes only!
No inductors for realization. They are replaced by "active inductors" (opamp blocks).
___________________________________________________
Quote: It was not supposed to be a general statement. In this particular instance the output contained an inductor.
The specifications I put in the tool are the following

Where is the circuit? Did you choose "passive" realization?
An active Filter has no inductors - because THIS ist the main advantage of acrive RC blocks!

Sadly Filter free does not seem to be free, but does cost a lot of money, in Filterlab you can only design up to 1Mhz and the TI FilterPro has only very few opamps to choose from and even less have spice models available. I will try if Filter Wiz Pro works tomorrow.
I do not know about the "direct" realization and will try to find out more about it but if you have literature suggestions I will be very happy to read up on them.

 

Given the stated requirements of -85 dB over 2 octaves, I think you might want to pick a different topology to reduce the required order. I'd try a Chebyshev Type I with 0.1 dB ripple in the passband. It will be hard to see the ripple on a standard Bode plot and the rolloff will be a good bit steeper. Do you know how to compute the required order given the specification?

I did a quick hand calculation with the Butterworth transfer function to determine the filter order so I assume that it is basically the same but with the Chebyshev transfer function, is that correct?

 

Where is the circuit? Did you choose "passive" realization?
An active Filter has no inductors - because THIS ist the main advantage of acrive RC blocks!

The circuit is the last one in my original post, there is no option to choose passive realization.