fiddled filter design - Page 9

Do you have a question? Post it now! No Registration Necessary

Translate This Thread From English to

Threaded View
Re: fiddled filter design
On Thursday, November 8, 2018 at 3:54:21 PM UTC-5, John Larkin wrote:
Quoted text here. Click to load it

You bring up an interesting point about just needing to work.  The effort o
f engineering is getting something to work with minimal resources, however,
 most of the time it is better to make things work with lots and lots of ma
rgin.  In my business wasting resources (ie extra parts cost)is not a big d
eal because we sell high margin products.  

Control loops are my favorite.  I like control loops to be as slow as possi
ble Super damped response, if the application allows for it.  In our text b
ooks we were taught to have the "ideal" loop of 45 degrees of phase margin  
with a couple  (3) dB of overshoot. Now I realize that if I need a loop lik
e that I am probably going to have problems because that is going to be pro
blematic in production or over temperature etc.  

I agree that a filter that works is adequate.  However, filters are one of  
the easiest things for non experts to get close to right due to all the sof
tware tools available (that actually get it right)

Re: fiddled filter design
Quoted text here. Click to load it


Exactly, much better.
And equipment should be 'repairable',
so selected components that just meet specs are out.


Quoted text here. Click to load it

Agreed.
Never save on parts!
Make the design with a safety margin.


Re: fiddled filter design
wrote:

Quoted text here. Click to load it

I recently found a nice ADC that's $1100. And some FPGAs around
$70,000 each. We might pass on those.


--  

John Larkin         Highland Technology, Inc
picosecond timing   precision measurement  

We've slightly trimmed the long signature. Click to see the full one.
Re: fiddled filter design
Quoted text here. Click to load it

Yes I did not mean that sort of parts,
unless it cannot be avoided.

More like protection ciruits and control loops.
Just added an other fuse and diode to my lnb_reference,
just in case somebody plugs in external power and uses POE at the same time.



Re: fiddled filter design
On 11/10/2018 07:02 AM, snipped-for-privacy@columbus.rr.com wrote:

Quoted text here. Click to load it

There's a certain logic to "Sell to people who can afford to pay high  
prices, as they're the ones with the money" product development.

Conversely the history books are littered with the names of  
out-of-business companies who thought their particular fashion of  
overengineered, high-cost, high-margin product would always have willing  
buyers. And were wrong.

Apple Computer was almost one of the but they pulled it out. How many  
people have heard of "Apricot Computer", though? They used to make some  
very innovative computers also

Over-engineering or "conservative-engineering" whatever you want to call  
it _can_ make for a high quality, reliable product but it isn't a  
guarantee of anything intrinsically. it can also make ya soft and  
inflexible and flapping in the breeze when a competitor figures out a  
way to make the same product at 2/3rds the price.

Re: fiddled filter design
On 11/10/2018 10:41 AM, bitrex wrote:
Quoted text here. Click to load it

Quoted text here. Click to load it

Quoted text here. Click to load it


Quoted text here. Click to load it

Quoted text here. Click to load it

Quoted text here. Click to load it

e.g. Apple Computer didn't get off the ground initially because Woz  
over-engineered the Apple II's floppy drive, it was stripped-down.


Re: fiddled filter design
On Sat, 10 Nov 2018 04:02:29 -0800 (PST), snipped-for-privacy@columbus.rr.com
wrote:

Quoted text here. Click to load it

Parts are cheap, but we try to minimize the number of different parts
on a BOM to make pick-and-place setup easier, minimize the number of
reels that have to be mounted on the machine. That becomes a game, and
it's a little dangerous to use r-packs whose value can't be changed.

Quoted text here. Click to load it

Yes, I trust the Spice simulation of a 10 KHz lowpass filter, so the
risk there is basically zero. Other things can't be trusted, so we
need to leave hooks for tweaking, or breadboard, or something.

I started with a 3-pole, 2-opamp filter, using values from TI's old
Filterpro program. (The new version is terrible.) Then I eliminated
the first opamp by scaling the resistors, and split the first resistor
to add one pole, and tweaked all that in LT Spice.


--  

John Larkin         Highland Technology, Inc

lunatic fringe electronics  


Re: fiddled filter design
All filter designs need acceptance criteria.

If the incoming SNR = is near 0 dB with the noise 1 freq. decade above the signal  ( 50kHz DDR for 5kHz signal )  then 1st you define a spec for:


- output SNR and this SNR improvement ratio.
- bandwidth (-3dB) & gain ripple
- stopband (Hz) & attenuation ( dB) at that f
- max group delay error in passband ( incremental phase shift)

Let's say to want 50 dB SNR output.
A 3rd order filter has 60dB/decade.

If you are only putting out a single sinewave then the group delay may not matter.  

I find the best design tool is TI's Active Filter http://www.ti.com/design-tools/signal-chain-design/webench-filters.html  I can remember what took days in the 70's now takes only a few minutes.

Re: fiddled filter design
On Mon, 12 Nov 2018 06:23:15 -0800 (PST), Anthony Stewart

Quoted text here. Click to load it

The DDS process is complex, but the first important image is -20 dBc
at 45 KHz, before filtering.

The easiest acceptance criteria is to simulate the DDS and the filter,
and tweak filters until you see something that you like.


Quoted text here. Click to load it

I have never got Webench to load and run.


--  

John Larkin         Highland Technology, Inc

lunatic fringe electronics  


Site Timeline