PCB power planes?
Martin Griffith
http://www.gennum.com/video/pdf/32469DOC.pdf
and I havent done any multi layer boards, is fig 5.3 on P24 correct?
martin
John Larkin
2.1.1 is insane. They can't make up their mind whether to use 0r
resistors or beads, so they are still guessing.
Splitting grounds is usually a bad idea. Except when it's a terrible
idea.
Fig 5.3 looks so weird to me that I assume PDF rendering errors. Does
anyone else see huge black triangles?
2.1.2 suggests sequencing or possible latchup problems. Be careful
here; get more info maybe.
John
Martin Griffith
I get the same triangles, silly idiots for letting something out like
that in a ref. design.
I don't think I need all the extra bits, just 1 vid in and an AVR or
8051 to control it, so I was just going to have a vdd and gnd copper
flood on the inner layers, get some cheap boards done to test.
martin
Martin Riddle
"Martin Griffith" <mart_in...@yah00.es> wrote in message news:n9fv64p7i3l1d284q...@4ax.com...
Looks like a pdf problem.
Cheers
John Larkin
<mart_in...@yah00.es> wrote:
>On Sat, 05 Jul 2008 11:53:46 -0700, in sci.electronics.design John
>Larkin <jjla...@highNOTlandTHIStechnologyPART.com> wrote:
>
>>On Sat, 05 Jul 2008 20:32:36 +0200, Martin Griffith
>><mart_in...@yah00.es> wrote:
>>
>>>Just been looking for ideas
>>>http://www.gennum.com/video/pdf/32469DOC.pdf
>>>and I havent done any multi layer boards, is fig 5.3 on P24 correct?
>>>
>>>
>>>martin
>>
>>2.1.1 is insane. They can't make up their mind whether to use 0r
>>resistors or beads, so they are still guessing.
>>
>>Splitting grounds is usually a bad idea. Except when it's a terrible
>>idea.
>>
>>Fig 5.3 looks so weird to me that I assume PDF rendering errors. Does
>>anyone else see huge black triangles?
>>
>>2.1.2 suggests sequencing or possible latchup problems. Be careful
>>here; get more info maybe.
>>
>>
>>John
>I get the same triangles, silly idiots for letting something out like
>that in a ref. design.
>I don't think I need all the extra bits, just 1 vid in and an AVR or
>8051 to control it, so I was just going to have a vdd and gnd copper
>flood on the inner layers, get some cheap boards done to test.
>
>
>martin
If you're going to do a test board, start with one solid ground plane,
with power planes below/above as needed (likely only one power plane
is necessary here) preferably with thin dielectric between the planes.
Handle any signal integrity/noise problems locally.
It's amazing how many really silly reference design/eval boards there
are.
John
Eeyore
Martin Riddle wrote:
> "Martin Griffith" <mart_in...@yah00.es> wrote in message
> | Just been looking for ideas
> | http://www.gennum.com/video/pdf/32469DOC.pdf
> | and I havent done any multi layer boards, is fig 5.3 on P24 correct?
>
>
> Looks like a pdf problem.
Yup, looks odd in V7 reader here.
Graham
Eeyore
John Larkin wrote:
> Splitting grounds is usually a bad idea. Except when it's a terrible
> idea.
Except when it's audio of course and you don'y want digital return
currents you can hear in the audio ground. In which case pkysically
separated but carefully linked planes are the way to go in order to
direct the currents where you want them to go rather than randomly
anywhere.
Graham
Rich Webb
<jjla...@highNOTlandTHIStechnologyPART.com> wrote:
>Fig 5.3 looks so weird to me that I assume PDF rendering errors. Does
>anyone else see huge black triangles?
Yes, with both the Foxit and PDF-XChange viewers so it's probably in the
original.
--
Rich Webb Norfolk, VA
Joerg
Weird. But I'd be a lot more concerned about figure 5.2. Splitting
grounds is rarely a good idea. Might also blow the EMC cert.
--
Regards, Joerg
http://www.analogconsultants.com/
"gmail" domain blocked because of excessive spam.
Use another domain or send PM.
Martin Griffith
Ha.. I remember a student who said he could make a small audio desk
with the equalisers using MF10's.
Never saw him again....I wonder why?
martin
Eeyore
Martin Griffith wrote:
> Eeyore wrote:
> >John Larkin wrote:
> >
> >> Splitting grounds is usually a bad idea. Except when it's a terrible
> >> idea.
> >
> >Except when it's audio of course and you don'y want digital return
> >currents you can hear in the audio ground. In which case pkysically
> >separated but carefully linked planes are the way to go in order to
> >direct the currents where you want them to go rather than randomly
> >anywhere.
>
>
> Ha.. I remember a student who said he could make a small audio desk
> with the equalisers using MF10's.
> Never saw him again....I wonder why?
Was that the NatSemi ? switched capacitor filter ?
No surprise there !
Graham
Eeyore
Joerg wrote:
> Martin Griffith wrote:
> > Just been looking for ideas
> > http://www.gennum.com/video/pdf/32469DOC.pdf
> > and I havent done any multi layer boards, is fig 5.3 on P24 correct?
>
> Weird. But I'd be a lot more concerned about figure 5.2. Splitting
> grounds is rarely a good idea. Might also blow the EMC cert.
Depends how you treat them ! Respect pays dividends.
Graham
Joerg
Even a nice application of Olde English furniture polish won't help :-)
> ... Respect pays dividends.
>
Can you tell that to the current generation of kids?
Joerg
You can design rather excellent audio gear with S/C filters. BTDT. Ok,
not audio but audio range. It was a Doppler receiver for medical
ultrasound. Zero through about 25kHz, dynamic range from here to the
Klondike. Noise from clocks and stuff: None. Sold like hotcakes, still
in use.
The only reason why S/C filters fell from grace with guys like us was
cost. They probably tried to maintain fat profit margins and that backfired.
Joel Koltner
news:x26ck.31503$ZE5....@nlpi061.nbdc.sbc.com...
> The only reason why S/C filters fell from grace with guys like us was cost.
> They probably tried to maintain fat profit margins and that backfired.
So you went back to continuous time filtering? Since it doesn't appear to me
that the margins on DSPs are in any way lower than those on switched cap
devices...
Joerg
Actually they are. You can get a TMS320 for less than four bucks. But
mostly I did go back to continuous time filtering. When it had to be
flexible sometimes heterodyne schemes. Mix to wherever filtering is easy
(meaning cheap ...) and then back.
Then there is the lost art of wave digital filters. Only few people know
it and you can get away with very cheap uC sans HW-multiplier.
Scott Seidman
$ZE5....@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.
--
Scott
Reverse name to reply
Joerg
Yes, you must filter that out. Sometimes. There are apps where spectral
components above 100kHz just don't make a difference. Plus the supplied
clock shall be squeaky clean. Marvelous concept but IMHO the marketing
guys killed it. Tried to make a killing with margins and then the market
kind of imploded around them.
Joel Koltner
> Marvelous concept but IMHO the marketing guys killed it. Tried to make a
> killing with margins and then the market kind of imploded around them.
I don't really know, but the prices also might have remained high due to the
market simply shrinking due to direct competition from DSPs -- for many
applications even at zero margin they probably couldn't have competed.
So what do you think of the price of the venerable LTC1569-7
(http://www.linear.com/pc/productDetail.jsp?navId=H0,C1,C1154,C1008,P1764) ?
---Joel
Joerg
It's around $5. Yikes!
You can get a nice DSP for that kind of money.
John Larkin
<namdie...@mindspring.com> wrote:
Also famous for aliasing inputs, aliasing power supply noise, and
making a heap of noise of their own. But the programmability is nice.
The world needs a general-purpose programmable continuous-time lowpass
filter IC. Really.
John
Scott Seidman
news:fol474hs4vfht56qt...@4ax.com:
They get terribly expensive, even for non tunable filters. I've been
using Frequency Devices single freq analog Bessels, and they run about
$100 a channel.
Krohn-Hite used to make an 8-pole tunable analog filter but it was bench
gear, not an RC.
I think the closest you're going to get today would be biquad stages that
use resistors to tune them, and programmable resistors. At least that's
an all IC approach, and you wouldn't need a billion-stage stacked switch.
Aliasing inputs, by the way, can come from surprising sources. People
tend not to think about bleed through on chopper amp clocks, for example,
but it's there.
Joerg
> On 7 Jul 2008 17:04:56 GMT, Scott Seidman
> <namdie...@mindspring.com> wrote:
>
>> Joerg <notthis...@removethispacbell.net> wrote in news:x26ck.31503
>> $ZE5....@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.
>
If you use the R2R ladders of DACs you could make one around chips like
these:
http://datasheets.maxim-ic.com/en/ds/MAX274-MAX275.pdf
Of course, then you'd be on the phone with Rebecca all the time.
Joel Koltner
> You can get a nice DSP for that kind of money.
Yeah, although can you get even within even an order of magnitude in group
delay (12us w/100kHz cutout) with a $5 DSP? :-)
(While I haven't personally used that IC, other people here have, and I'm told
that's why -- they're trying to minimize some big audio system's group delay,
making it something like <1ms total.)
Joel Koltner
> If you use the R2R ladders of DACs you could make one around chips like
> these
Hey Joerg -- know of anyone who makes R2R-type DACs (so called "multiplying
DACs") with better than 12MHz frequency response? I could certianly use such
items on occasion... (the fastest I know of, at 12MHz, is the AD5546).
Joerg
That's the crux with DSP, they often can't do that. This is where wave
digital filters come in handy but youngsters don't have the foggiest
what that is or think it's some tuning device for a surfboard.
Joel Koltner
> That's the crux with DSP, they often can't do that. This is where wave
> digital filters come in handy but youngsters don't have the foggiest what
> that is or think it's some tuning device for a surfboard.
They are few -- if any? -- undergraduate EE curriculums that even mention
switched cap design... it's pretty much all DSP and lumped-element filters
these days.
I've read the Fettweis paper on WDFs -- I think I posted a copy to ABSE at
some point? -- but I've never had a good application for them. I should
definitely revisit the topic as I didn't realize the group delay could be so
small.
---Joel
Joerg
> "Joerg" <notthis...@removethispacbell.net> wrote in message
> news:Xbtck.1063$zv7...@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.
Scott Seidman
$co7....@nlpi066.nbdc.sbc.com:
>
> If you use the R2R ladders of DACs you could make one around chips like
> these:
> http://datasheets.maxim-ic.com/en/ds/MAX274-MAX275.pdf
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.
Joerg
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" :-)
John Larkin
<notthis...@removethispacbell.net> wrote:
>John Larkin wrote:
>> On 7 Jul 2008 17:04:56 GMT, Scott Seidman
>> <namdie...@mindspring.com> wrote:
>>
>>> Joerg <notthis...@removethispacbell.net> wrote in news:x26ck.31503
>>> $ZE5....@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.
>>
>
>If you use the R2R ladders of DACs you could make one around chips like
>these:
>http://datasheets.maxim-ic.com/en/ds/MAX274-MAX275.pdf
>
>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?
Joerg
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.
Scott Seidman
$Ri....@flpi146.ffdc.sbc.com:
> If you know a good digital-pot bargain let us know. But not from a
> company that starts with "M" :-)
>
http://www.analog.com/en/digital-to-analog-converters/digital-
potentiometers/products/index.html
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.
Joerg
> Joerg <notthis...@removethispacbell.net> wrote in news:Qguck.17293
> $Ri....@flpi146.ffdc.sbc.com:
>
>> If you know a good digital-pot bargain let us know. But not from a
>> company that starts with "M" :-)
>>
>
> http://www.analog.com/en/digital-to-analog-converters/digital-
> potentiometers/products/index.html
>
> 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.
Scott Seidman
$N87....@nlpi068.nbdc.sbc.com:
> That would be way out of league for most of my designs.
It depends on how much you need a tunable filter, and how good you need it
to be.
I used to use a bank of these:
http://www.tucker.com/java/jsp/product_partno3202R_invid9608_condR.htm
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.
John Larkin
<namdie...@mindspring.com> wrote:
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.
John
Scott Seidman
> 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.
>
In my world, you need to filter before you acquire, and the rate you
acquire at isn't necessarily known at design stage, and can change. So the
rules are prefilter with a tunable filter, or use a fixed filter and sample
3-5 times faster than the maximum rate you really need. Past a couple of
MHz it is difficult, but not everyone works up there.
Joerg
I find things to be easier past a couple of MHz because there you can
use inductors and PIN diodes. Ok, depends on what your home turf is.
Active filters aren't mine but MHz stuff is.
John Larkin
<notthis...@removethispacbell.net> wrote:
>John Larkin wrote:
>> On 8 Jul 2008 13:27:47 GMT, Scott Seidman
>> <namdie...@mindspring.com> wrote:
>>
>>> Joerg <notthis...@removethispacbell.net> wrote in news:iUvck.14888
>>> $N87....@nlpi068.nbdc.sbc.com:
>>>
>>>> That would be way out of league for most of my designs.
>>>
>>> It depends on how much you need a tunable filter, and how good you need it
>>> to be.
>>>
>>> I used to use a bank of these:
>>> http://www.tucker.com/java/jsp/product_partno3202R_invid9608_condR.htm
>>>
>>> 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.
>>
>
>I find things to be easier past a couple of MHz because there you can
>use inductors and PIN diodes. Ok, depends on what your home turf is.
>Active filters aren't mine but MHz stuff is.
We generally use LC filters in DDS type applications. Our latest arb
uses a 7-pole transitional Gaussian, -3 dB at about 40 MHz with a
bonus notch around 120. But it's not tunable, and it still hogs about
10x or so as much real estate as the half-a-dual-dac that drives it
[1]. The three Ls are 1206's to keep the Q tolerable.
Why hasn't Analog Devices done such a chip or, even better, added the
filters inside their DDS chips? Many of their datasheets and appnotes
seem to hide the fact that you need a filter at all!
Mini-circuits has some nice looking ceramic-blob LC filters, but they
start at 80 MHz.
John
[1] Of course, the traces from the FPGA to the dac take more room than
the dac, too.
Scott Seidman
> Why hasn't Analog Devices done such a chip or, even better, added the
> filters inside their DDS chips? Many of their datasheets and appnotes
> seem to hide the fact that you need a filter at all!
There's always going to be a gap between having access to the technology,
and knowing how to use the technology.
Joel Koltner
news:89d774tbudghra5v0...@4ax.com...
> We generally use LC filters in DDS type applications...
Differential or do you first convert to single-ended and then filter?
(I asked a question about this for my own needs some months ago and I believe
you responded that you didn't think differential filter was worth it, but I'm
curious what you do on your own boards there...)
John Larkin
We first convert the dac differential current output to single-ended,
bipolar, +-1 volt or so, then use a single-ended filter and output
amp. The diff-current to voltage stage is non-trivial in its own
right, if you need speed and low THD.
We did kick this around and decided that an immediate s-e conversion
made sense, but I can't recall the details.
John
Joel Koltner
"John Larkin" <jjla...@highNOTlandTHIStechnologyPART.com> wrote in message
news:c6j774psg48e34b7q...@4ax.com...
> The diff-current to voltage stage is non-trivial in its own
> right, if you need speed and low THD.
You ended up with an op-amp approach (as opposed to transformer), didn't you?
Joerg
Same for many class-D chip. The sobering awakening and the occasional
"oh s..t!" comes at the EMC lab.
[...]
John Larkin
Certainly; we go down to DC.
John
chris...@googlemail.com
<jjlar...@highNOTlandTHIStechnologyPART.com> 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
>
> >>> Joerg <notthisjoerg...@removethispacbell.net> wrote in news:x26ck.31503
> >>> $ZE5.27...@nlpi061.nbdc.sbc.com:
You could design one. You can get some samples made by MOSIS fairly
cheaply if you use a fairly old process (cheaper than some PCBs these
days). They took their price list off their website but it's still on
www.archive.org. If there is a friendly university near you that
teaches chip design, they might be able to let you use Cadence or
similar on one of their PCs, or there is some open-source software
that would probably be more work to set up. You could even put some
nice mosfets on the side for Joerg and a very low current square wave
oscillator for people into that sort of thing. If you used a half-way
modern process you could add a nice sampler for a fast sampling scope
and maybe the front end of Joerg's cheap spectrum analyser too. If
you have time left over, you could put a little ADC on the side, with
a FIFO and a USB2 interface (you can probably find one on opencores)
to make a handy little 50MSPS (480Mbps) USB2 oscilloscope-on-a-chip-
dongle-thing.
Chris
John Larkin
wrote:
Now that you mention it, my garage door opener is getting a little
flakey, too.
John
Joel Koltner
> Now that you mention it, my garage door opener is getting a little
> flakey, too.
Joerg might have an opening on his "honey do" list one of these upcoming
weekends... he seems pretty good at fixing such things. :-)
JosephKK
<mart_in...@yah00.es> wrote:
>Just been looking for ideas
>http://www.gennum.com/video/pdf/32469DOC.pdf
>and I havent done any multi layer boards, is fig 5.3 on P24 correct?
>
>
>martin
Not sure yet. But the split planes appear to be an isolation issue.
Is this an isolation device as well?
Joel Koltner
> Noooo!
:-)
> I've got to redo the whole north section of the sprinkler hookup. Lotsa
> crawling, coughing up dust and so on.
Do us a favor and wear a dusk mask? We need you to stick around as long as
possible!
Joerg
It won't suppress dusk :-)
I can't breathe well with such a mask when doing heavy work. Last time i
saw a puddle and thought "Oh no, a water leak!" It was my sweat ...
Joerg
Noooo!
I've got to redo the whole north section of the sprinkler hookup. Lotsa
crawling, coughing up dust and so on.
--
JosephKK
<marti...@verizon.net> wrote:
>
>"Martin Griffith" <mart_in...@yah00.es> wrote in message news:n9fv64p7i3l1d284q...@4ax.com...
>| Just been looking for ideas
>| http://www.gennum.com/video/pdf/32469DOC.pdf
>| and I havent done any multi layer boards, is fig 5.3 on P24 correct?
>|
>|
>| martin
>
>Looks like a pdf problem.
>
>Cheers
>
I tried using Acroread 8 for MSWin running in WINE on it and it is
just fine.
It looks fine in Linux Firefox with the AR8 plugin.
Have fun.