Predistortion of LC filter

[Philip Newman]

Hi,
I have desiged an 9th order elliptic LC low pass filter with cutoff at
111.55MHz. The stop-band attenuation is 50dB and should be at
121.55MHz. I have designed this with ideal components, which give an
ideal response. however, when I simulated with standard library
components, not surprisingly, the response was poor, due to the low Q
of the inductors. The filter was designed using parallel resonant
circuits, so the capacitors were tweaked to try to tune the circuit to
move the zeros back in place, but to no avail.

I wish to try to re-design the filter with predistortion, to try and
compensate for this effect. however, I can find no literature on the
internet that will help me. Can anybody point me in the right
direction, send a link or two, or even reccomend a book? I am using
Agilent's ADS for design and simulation.

Cheers,

Phil

 

[Don Pearce]

Hi,
I have desiged an 9th order elliptic LC low pass filter with cutoff at
111.55MHz. The stop-band attenuation is 50dB and should be at
121.55MHz. I have designed this with ideal components, which give an
ideal response. however, when I simulated with standard library
components, not surprisingly, the response was poor, due to the low Q
of the inductors. The filter was designed using parallel resonant
circuits, so the capacitors were tweaked to try to tune the circuit to
move the zeros back in place, but to no avail.

I wish to try to re-design the filter with predistortion, to try and
compensate for this effect. however, I can find no literature on the
internet that will help me. Can anybody point me in the right
direction, send a link or two, or even reccomend a book? I am using
Agilent's ADS for design and simulation.

Cheers,

Phil

Phil, if the unloaded Q of your resonators won't allow it, even
predistortion won't bring this filter back to the necessary
performance. Predistortion only really helps in flattening out a saggy
passband, not in making the edge plummet.

A book I can recommend though is Electronic Filter Design Handbook, by
Arthur Williams and Fred Taylor, ISBN 0-07-070434-1

d
Pearce Consulting
http://www.pearce.uk.com

 

[Chuck Olson]

Phil, if the unloaded Q of your resonators won't allow it, even
predistortion won't bring this filter back to the necessary
performance. Predistortion only really helps in flattening out a saggy
passband, not in making the edge plummet.

A book I can recommend though is Electronic Filter Design Handbook, by
Arthur Williams and Fred Taylor, ISBN 0-07-070434-1

d
Pearce Consulting
http://www.pearce.uk.com

You might also dig into Anatol I. Zverev"s "Handbook of Filter Synthesis",
1967, John Wiley & Sons Inc., ISBN 0 471 98680 1, sometimes known as
"The Big Red book". "Zverev" is surprisingly understandable and useful
almost to the point of being indispensable if you do much filter design. As
Zverev shows on page 121, compared to the lossless design, predistortion of
a Chebyshev filter, for example, restores flatness in the passband (with
some added average attenuation) and pretty nearly restores the sharpness of
the cutoff shoulder, but the stopband is slightly more attenuated with about
the same rate of rolloff until the null, which is rounded off and shallower.
These differences will be more pronounced as the losses in the components
increase.

 

[Norm Dresner]

Phil, if the unloaded Q of your resonators won't allow it, even
predistortion won't bring this filter back to the necessary
performance. Predistortion only really helps in flattening out a saggy
passband, not in making the edge plummet.

A book I can recommend though is Electronic Filter Design Handbook, by
Arthur Williams and Fred Taylor, ISBN 0-07-070434-1

How about recommending a book that's still in print. B&N has no copies and
Amazon has 2 used from $125. They also claim to have some used paperbacks
which originally sold for $25 and are now selling starting at ~$65.

Norm

 

[Don Pearce]

How about recommending a book that's still in print. B&N has no copies and
Amazon has 2 used from $125. They also claim to have some used paperbacks
which originally sold for $25 and are now selling starting at ~$65.

Norm

Oops! Oh well, I have a mint copy - those numbers look interesting...

Although I'm guessing pretty much any filter book is going to cover
predistortion.

d
Pearce Consulting
http://www.pearce.uk.com

 

[Philip Newman]

Oops! Oh well, I have a mint copy - those numbers look interesting...

Although I'm guessing pretty much any filter book is going to cover
predistortion.

d
Pearce Consulting
http://www.pearce.uk.com

Thanks Don, I already have a copy of the book you suggest, and it
doesn't really cover predistortion design. I might try to get a copy
of Zverev though.

Back to the drawing board I guess....

Phil

 

[Don Pearce]

Thanks Don, I already have a copy of the book you suggest, and it
doesn't really cover predistortion design. I might try to get a copy
of Zverev though.

Back to the drawing board I guess....

Phil

My copy has four pages on predistortion (3.12 - 3.15), and appears to
cover the principles pretty well; has it been dropped from later
editions, perhaps?

d
Pearce Consulting
http://www.pearce.uk.com

 

[Philip Newman]

Phil, if the unloaded Q of your resonators won't allow it, even
predistortion won't bring this filter back to the necessary
performance. Predistortion only really helps in flattening out a saggy
passband, not in making the edge plummet.

What do you mean by Unloaded Q, and loaded Q? The filters band-edge
has been atteunated due to the useless crappy inductors, and is not as
sharp as the ideal response. The rejection limit and stop-band floor
is fine, and the steepness of the transition band is fine. I just
want to keep the flat pass-band and a sharp cut-off.

I have tried amplitude equalisation by using a series LC resonator
with a DC attenuator, which helps to a certain point, but what about
pre-distortion?

I presume pre-distortion is a bit too complex to try and explain on
usenet? I think there is a copy of Zverev in the library.

So, if the LC architecture doesn't work, what else can I try? I think
the shape factor is too demanding for an active filter, and the
frequencies are too high for a digital filter? Actually, I think this
particular filter is an Anti Aliasing filter, so perhaps now. But why
do AAFs hav to be before the DSP? Can the filter be part of the DSP
program, and make sure it is first bit of processing it does?

Cheers,

Phil

 

[Philip Newman]

ah yes. sorry. i will work through that section tomorrow.

cheers,

phil

 

[Don Pearce]

What do you mean by Unloaded Q, and loaded Q? The filters band-edge
has been atteunated due to the useless crappy inductors, and is not as
sharp as the ideal response. The rejection limit and stop-band floor
is fine, and the steepness of the transition band is fine. I just
want to keep the flat pass-band and a sharp cut-off.

I have tried amplitude equalisation by using a series LC resonator
with a DC attenuator, which helps to a certain point, but what about
pre-distortion?

I presume pre-distortion is a bit too complex to try and explain on
usenet? I think there is a copy of Zverev in the library.

So, if the LC architecture doesn't work, what else can I try? I think
the shape factor is too demanding for an active filter, and the
frequencies are too high for a digital filter? Actually, I think this
particular filter is an Anti Aliasing filter, so perhaps now. But why
do AAFs hav to be before the DSP? Can the filter be part of the DSP
program, and make sure it is first bit of processing it does?

Cheers,

Phil

You have just described exactly what happens in an audio DAC. The
first bit of filtering is done digitally at a high rate
(oversampling), and then a cheap and simple analogue filter can deal
with what little remains.

d
Pearce Consulting
http://www.pearce.uk.com

 

[Philip Newman]

You have just described exactly what happens in an audio DAC. The
first bit of filtering is done digitally at a high rate
(oversampling), and then a cheap and simple analogue filter can deal
with what little remains.

d
Pearce Consulting
http://www.pearce.uk.com

OK thanks. Any answers to the previous questions about Q and
alternative filter architectures? Is it possible to raise the
absolute stop-band attenuation floor, in order to raise the band-edge?
Is it possible to have a digital filter working at 100's MHz?

I am only working on google-groups as i can't get to outlook express
newsgroups, as i am working on NT on a corporate network and don't
know the news-server address!

Cheers,

Phil

 

[Philip Newman]

My copy has four pages on predistortion (3.12 - 3.15), and appears to

cover the principles pretty well; has it been dropped from later
editions, perhaps?

It doesn't cover predistortion for an elliptic function, which is what I am using.

Phil

 

[Don Pearce]

It doesn't cover predistortion for an elliptic function, which is what I am using.

Phil

The principles still hold - find out how much sag at the top end you
want to compensate for, the introduce that amount of loss at the
bottom end.

d
Pearce Consulting
http://www.pearce.uk.com