LC ladder filter questions

[John Woodgate]

I read in sci.electronics.design that maxfoo <maxfooHeadFromButt@punkass
..com> wrote (in <c6dg70pdj7qc3ij3tdc0mtevao724vnde0@4ax.com&gt about 'LC
ladder filter questions', on Sat, 10 Apr 2004:

On Sat, 10 Apr 2004 07:42:40 -0700, John Larkin
jjlarkin@highlandSNIPtechTHISnologyPLEASE.com> wrote:


OK, we have a low-order, third maybe, LC filter, which has nice
reasonable L and C values. We terminate it and run the signal through
an opamp buffer. The buffer has an input capacitance of 1 pF, an
output impedance of a few ohms, and it's flat to over 2 GHz. After the
buffer, we have another low-order LC filter. So we've made a
higher-order LC filter out of simpler sections, just like we do when
we make an active filter out of 2nd-order unilateral sections. Since
interactions are reduced, we have another degree of freedom in the
design.

So why does the presence of the buffer increase component sensitivity?
If you can explain why it does, I'll be suitably embarassed. [1]


John

What's that part number on the buffer op amp that has a flat response from
dc-2400mhz? I want to buy out the stock....

The pat number is '20 AWG'. Note that the spec given above doesn't
mention the input resistance.(;-)
--
Regards, John Woodgate, OOO - Own Opinions Only.
The good news is that nothing is compulsory.
The bad news is that everything is prohibited.
http://www.jmwa.demon.co.uk Also see http://www.isce.org.uk

 

[John Larkin]

On Sat, 10 Apr 2004 20:08:13 +0100, John Woodgate
<jmw@jmwa.demon.contraspam.yuk> wrote:

I read in sci.electronics.design that maxfoo <maxfooHeadFromButt@punkass
.com> wrote (in <c6dg70pdj7qc3ij3tdc0mtevao724vnde0@4ax.com&gt about 'LC
ladder filter questions', on Sat, 10 Apr 2004:
On Sat, 10 Apr 2004 07:42:40 -0700, John Larkin
jjlarkin@highlandSNIPtechTHISnologyPLEASE.com> wrote:


OK, we have a low-order, third maybe, LC filter, which has nice
reasonable L and C values. We terminate it and run the signal through
an opamp buffer. The buffer has an input capacitance of 1 pF, an
output impedance of a few ohms, and it's flat to over 2 GHz. After the
buffer, we have another low-order LC filter. So we've made a
higher-order LC filter out of simpler sections, just like we do when
we make an active filter out of 2nd-order unilateral sections. Since
interactions are reduced, we have another degree of freedom in the
design.

So why does the presence of the buffer increase component sensitivity?
If you can explain why it does, I'll be suitably embarassed. [1]


John

What's that part number on the buffer op amp that has a flat response from
dc-2400mhz? I want to buy out the stock....

The pat number is '20 AWG'. Note that the spec given above doesn't
mention the input resistance.(;-)

The THS4302 typical input impedance is 1.6 megs plus 1 pF, and output
impedance is 0.2 ohms, good for 180 mA drive. Its -3 dB point is
claimed to be 2.4 GHz typ, and they seem to be at least that good in
real life. Transient response is absurdly clean as compared to MMICS
that are rated at much higher frequencies. My only small complaint is
the 6 volt max supply rating... I'd prefer 30 or so.

Semiconductors have been getting faster steadily since 30 KHz
point-contact transistors, but people still are surprised.

John

 

[John Larkin]

On Sat, 10 Apr 2004 20:33:24 GMT, maxfoo
<maxfooHeadFromButt@punkass.com> wrote:

THS4302.

Noise: 2.8 nV/Hz that's too high, wanted to use it for an active loop filter to
replace the ths4130 1.3nV/Hz I'm using now. My PLL phase noise spec is 140dBc/Hz
@100khz offset. Wish it was pin compatible

Can't be... it doesn't have pins!

John

 

[Tony]

On Thu, 08 Apr 2004 20:35:12 -0700, John Larkin
<jjlarkin@highlandSNIPtechTHISnologyPLEASE.com> wrote:

I am suggesting using LCs as the filter elements, but with a few
essentially transparent buffers breaking up sections (and providing
distributed gain, if we want it.) Once sections are unilateralized,
the filter design becomes much simpler, and I'd guess that
sensitivities would be much lower. A multistage LC filter often has
awkward component ratios, but the sectionalized thing I'm suggesting
would have much reduced interaction, so each little clump can be
whatever impedance that's handy, and use a more practical set of
values.

John

Like you, I would once have expected this to be the case, and I was quite
surprised the first time I simulated a 9 pole Butterworth LPF. As an active
filter I needed considerably more precise parts than the junk box could provide,
but as a single L-C network the response was still tolerable even when after
only minor juggling I just dropped in the nearest E6 values (of course in
practice the Q of the chokes I had caused some minor discrepancies, but even
that was easily allowed for). It seems that the interaction between L-C elements
actually works in your favour when variations occur.

Tony (remove the "_" to reply by email)

 

[John Larkin]

On Sun, 11 Apr 2004 08:56:56 +1000, Tony <tony_roe@tpg.com.au> wrote:

On Thu, 08 Apr 2004 20:35:12 -0700, John Larkin
jjlarkin@highlandSNIPtechTHISnologyPLEASE.com> wrote:
[snip]
I am suggesting using LCs as the filter elements, but with a few
essentially transparent buffers breaking up sections (and providing
distributed gain, if we want it.) Once sections are unilateralized,
the filter design becomes much simpler, and I'd guess that
sensitivities would be much lower. A multistage LC filter often has
awkward component ratios, but the sectionalized thing I'm suggesting
would have much reduced interaction, so each little clump can be
whatever impedance that's handy, and use a more practical set of
values.

John

Like you, I would once have expected this to be the case, and I was quite
surprised the first time I simulated a 9 pole Butterworth LPF. As an active
filter I needed considerably more precise parts than the junk box could provide,
but as a single L-C network the response was still tolerable even when after
only minor juggling I just dropped in the nearest E6 values (of course in
practice the Q of the chokes I had caused some minor discrepancies, but even
that was easily allowed for). It seems that the interaction between L-C elements
actually works in your favour when variations occur.

Tony (remove the "_" to reply by email)

OK, is it really the case that many-section LC filters have less
component sensitivity than a cascaded-section, buffered filter of
equivalent transfer function?

John