> Yeah, although can you get even within even an order of magnitude in group > delay (12us w/100kHz cutout) with a $5 DSP? :-)
> (While I haven't personally used that IC, other people here have, and I'm told > that's why -- they're trying to minimize some big audio system's group delay, > making it something like <1ms total.)
That's the crux with DSP, they often can't do that. This is where wave digital filters come in handy but youngsters don't have the foggiest what that is or think it's some tuning device for a surfboard.
> That's the crux with DSP, they often can't do that. This is where wave > digital filters come in handy but youngsters don't have the foggiest what > that is or think it's some tuning device for a surfboard.
They are few -- if any? -- undergraduate EE curriculums that even mention switched cap design... it's pretty much all DSP and lumped-element filters these days.
I've read the Fettweis paper on WDFs -- I think I posted a copy to ABSE at some point? -- but I've never had a good application for them. I should definitely revisit the topic as I didn't realize the group delay could be so small.
Joel Koltner wrote: > "Joerg" <notthisjoerg...@removethispacbell.net> wrote in message > news:Xbtck.1063$zv7.897@flpi143.ffdc.sbc.com... >> That's the crux with DSP, they often can't do that. This is where wave >> digital filters come in handy but youngsters don't have the foggiest what >> that is or think it's some tuning device for a surfboard.
> They are few -- if any? -- undergraduate EE curriculums that even mention > switched cap design... it's pretty much all DSP and lumped-element filters > these days.
Shhht! Don't tell. Keeps us busy :-)
> I've read the Fettweis paper on WDFs -- I think I posted a copy to ABSE at > some point? -- but I've never had a good application for them. I should > definitely revisit the topic as I didn't realize the group delay could be so > small.
You need to select a processor that can do stuff like fetch, shift and shift-add in very few machine cycles, ideally in one. Should be 16-bit, too, but that's no big deal anymore these days.
AFAICT after Professor Fettweis retired the usual happened. This whole research area became unglued and withered. It's sad, there is hardly any continuance in academia. That's what I like about working in industry, you continue the product lines whether you like it or not. Better for the customers.
Plenty of companies package that sort of thing in a nice "programmable resistor" package, and give you a lovely cascadable serial bus to program them as well.
> Plenty of companies package that sort of thing in a nice "programmable > resistor" package, and give you a lovely cascadable serial bus to program > them as well.
Yes, but then it gets expensive because you need a whole lot of them. With DACs you can luck out and find a cheap 8-pack.
If you know a good digital-pot bargain let us know. But not from a company that starts with "M" :-)
<notthisjoerg...@removethispacbell.net> wrote: >John Larkin wrote: >> On 7 Jul 2008 17:04:56 GMT, Scott Seidman >> <namdiestt...@mindspring.com> wrote:
>>> Joerg <notthisjoerg...@removethispacbell.net> wrote in news:x26ck.31503 >>> $ZE5.27...@nlpi061.nbdc.sbc.com:
>>>> Noise from clocks and stuff: None. >>> That's not my experience. Switched cap filts are famous for clock bleed >>> through, and you need to be sure your circuit can tolerate this. The clock >>> frequency is usually an order of magnitude or more higher than the filter >>> freq, so you can usually filter out the bleedthrough with an RC.
>> Also famous for aliasing inputs, aliasing power supply noise, and >> making a heap of noise of their own. But the programmability is nice.
>Look at the bright side: You can make really nice mixers and stuff with >them. If they just weren't so darn expensive.
>> The world needs a general-purpose programmable continuous-time lowpass >> filter IC. Really.
John Larkin wrote: > On Mon, 07 Jul 2008 11:05:28 -0700, Joerg > <notthisjoerg...@removethispacbell.net> wrote:
>> John Larkin wrote: >>> On 7 Jul 2008 17:04:56 GMT, Scott Seidman >>> <namdiestt...@mindspring.com> wrote:
>>>> Joerg <notthisjoerg...@removethispacbell.net> wrote in news:x26ck.31503 >>>> $ZE5.27...@nlpi061.nbdc.sbc.com:
>>>>> Noise from clocks and stuff: None. >>>> That's not my experience. Switched cap filts are famous for clock bleed >>>> through, and you need to be sure your circuit can tolerate this. The clock >>>> frequency is usually an order of magnitude or more higher than the filter >>>> freq, so you can usually filter out the bleedthrough with an RC. >>> Also famous for aliasing inputs, aliasing power supply noise, and >>> making a heap of noise of their own. But the programmability is nice.
>> Look at the bright side: You can make really nice mixers and stuff with >> them. If they just weren't so darn expensive.
>>> The world needs a general-purpose programmable continuous-time lowpass >>> filter IC. Really.
>> Of course, then you'd be on the phone with Rebecca all the time.
> MSOP-8:
> V+ > V- > GND > IN > OUT > CLK > DATA > CS-
> SPI interface sets pole/zero locations and gain.
> Why not?
If even Maxim doesn't make any "exists at least in glossy print" chip like that I guess there aren't enough of us, meaning no market to write home about.
Analog devices, about $0.60 each in quantity, probably come to about 2 bucks for each pole pair on a UAF42. Still not cheap, but not out of the realm of possibility.
I'm not thrilled with some of the "M" stuff either.
> Analog devices, about $0.60 each in quantity, probably come to about 2 > bucks for each pole pair on a UAF42. Still not cheap, but not out of the > realm of possibility.
That would be way out of league for most of my designs. Another option is to servo a dual FET as a variable resistor. If you can get one, that is. Man, that so so easy during the days when the SD5400 was cheap and available. Done it many times.
> I'm not thrilled with some of the "M" stuff either.
It's the only company where I only did design-outs. Mostly because of part unavailability.
At $500 per channel used for a 4-pole tunable filter, 4 bucks a channel for digital pots plus the UAF42 is quite a bargain. Of course, you need to build your serial bus and then write some code to support it, but still the bargain.
>At $500 per channel used for a 4-pole tunable filter, 4 bucks a channel for >digital pots plus the UAF42 is quite a bargain. Of course, you need to >build your serial bus and then write some code to support it, but still the >bargain.
Sure you can make your own, but it will be 20 times the size, and 2x the cost, of the DDS chip or dac you're trying to post-filter. And past a couple of MHz it will be difficult.
