Maker Pro
Maker Pro

Notches in ground planes for multi-power multi-channel board

F

Frank Raffaeli

Jan 1, 1970
0
Thanks, that is very helpful, I will stick with one plane and
hopefully the ground 'slice' between the power planes will help as
well. Also, I am wondering if putting decoupling caps between two
power plane pours will help further

It seems intuitive that a gap across copper will reduce current flow
across the gap. This is true only at low frequencies. If you have any
digital logic on this board, and there are any signal wires across the
"gap", you've made an efficient transmitting / receive antenna. I
think Joerg said if people didn't put gaps in their ground planes, he
would have less work ;-) Sometimes the max current flows across the
center of the gap. It's frequency selective.

Frank
 
J

John Larkin

Jan 1, 1970
0
It seems intuitive that a gap across copper will reduce current flow
across the gap. This is true only at low frequencies. If you have any
digital logic on this board, and there are any signal wires across the
"gap", you've made an efficient transmitting / receive antenna. I
think Joerg said if people didn't put gaps in their ground planes, he
would have less work ;-) Sometimes the max current flows across the
center of the gap. It's frequency selective.

Frank

Slot antennas and slotline transmission lines both suggest modes
better left alone.

John
 
D

Didi

Jan 1, 1970
0
Splitting a ground plane is usually a recipe for disaster. I have yet to
see a case where that really worked and it's been decades now.

I have pointed you to that before so it is not that you have yet to
see one, it is that you have yet to look at one :).
http://tgi-sci.com/tgi/hstb.htm

I gree in the context of the OP splitting the GND plane is a bad idea,
probably very bad. My guess is the split power plane can still be
made having wide enough sections so this will not be an issue, but
I did not get into all the details.

Spliiting the GND plane gets useful way beyond commonplace resolution
and sensitivity, this is why most people have no problem without
splitting it, and being happy with their results think this is a
usiversal rule.
It is not.

Dimiter
 
J

Joerg

Jan 1, 1970
0
John said:
Slot antennas and slotline transmission lines both suggest modes
better left alone.

Mostly people end up with a huge loop antenna that picks up just about
anything. That is because stuff will have to be connected to either side
of a split, usually. Now throw something like Sutro tower into the mix
and the circuit noise can reach the signal levels of a Led Zeppelin concert.
 
J

Joel Koltner

Jan 1, 1970
0
Jim Thompson said:
My people don't decide the architecture. You should know that well.
MARKETING decides ;-)

I'm sure it has something to do with the desire from many companies'
management to be able to hire relatively unskilled engineers (...they're
cheap!) and still have a product.

Arguably that's not a particularly awful incentive, but few companies seem to
realize that the whole approach fails as soon as you want to do something just
a little bit different than what the whiz-bang "all-in-one" IC does.

That's when people give Joerg a call.

Or the company I work at, for that matter. We spend plenty of time these days
building radios from all the discrete mixers/amps/etc. that are now readily
available, although unfortuntaely it's often difficult to compete on power
relative to what some of those old Philips (NXP) parts provided. (Remeber the
UAA2080? Beautiful part... very low power, good quality filters using
gyrators...)

At least battery technology has progressed a bit.

---Joel
 
R

Richard Henry

Jan 1, 1970
0
It seems intuitive that a gap across copper will reduce current flow
across the gap. This is true only at low frequencies. If you have any
digital logic on this board, and there are any signal wires across the
"gap", you've made an efficient transmitting / receive antenna. I
think Joerg said if people didn't put gaps in their ground planes, he
would have less work ;-) Sometimes the max current flows across the
center of the gap. It's frequency selective.

Frank- Hide quoted text -

- Show quoted text -

On the other hand, if your circuitry is completely isolated into
physical/electrical/logical domains (such as in red/black segregation)
then splitting or slotting the ground plane makes sense and may be
required by the physical and/or regulatory realities.
 
D

Didi

Jan 1, 1970
0
Interesting spec... Zero Offset: None

Pretty unique, yes. And guaranteed...
This is the only MCA on the market which allows you to calibrate
for energy using a single point.

Dimiter
 
J

Joerg

Jan 1, 1970
0
Joel said:
I'm sure it has something to do with the desire from many companies'
management to be able to hire relatively unskilled engineers (...they're
cheap!) and still have a product.

Arguably that's not a particularly awful incentive, but few companies seem to
realize that the whole approach fails as soon as you want to do something just
a little bit different than what the whiz-bang "all-in-one" IC does.

That's when people give Joerg a call.

Sometimes they give me a call a bit late, after the purchasing guy
became unable to source some super-duper panacea chip. Then it's often
"tabula rasa", scrap the whole design and start from scratch.

Or the company I work at, for that matter. We spend plenty of time these days
building radios from all the discrete mixers/amps/etc. that are now readily
available, although unfortuntaely it's often difficult to compete on power
relative to what some of those old Philips (NXP) parts provided. (Remeber the
UAA2080? Beautiful part... very low power, good quality filters using
gyrators...)

Personally I am not a fan of such "panacea chips". The millisecond they
lose a key account the manufacturers might stop production and soon
after purge the datasheet.

At least battery technology has progressed a bit.

However, electronics and especially the SW guys keep bloating stuff
faster than battery technology can keep up with. Case in point: My early
90's laptop ran up to six hours on one NiCd charge and so I opted not to
buy the extra capacity battery. Nowadays it's down to two hours with an
expensive LiIon battery that has many times the capacity. And I do the
same type of work on them :)

The concept of a mobile device needing a cooling fan is IMHO just plain
sick.

Or take our Dimango remote doorbell. 20 bucks, 10 (!) years on the first
set of batteries and no slowdown in sight. Of course, the receiver is
completely discrete. This year I am going to change the batteries anyhow
because they might leak just from old age.
 
J

Joerg

Jan 1, 1970
0
John said:
Interesting spec... Zero Offset: None

John

C'mon, Didi lives and works in Bulgaria and probably speaks a language
of Macedonian origin all day long. Quite different from our Anglo-Saxon
language roots. It's much tougher for those guys to master English than
it is for Western Europeans. Guess he meant offset from the zero-line.
 
D

Didi

Jan 1, 1970
0
C'mon, Didi lives and works in Bulgaria and probably speaks a language
of Macedonian origin all day long. Quite different from our Anglo-Saxon
language roots. It's much tougher for those guys to master English than
it is for Western Europeans. Guess he meant offset from the zero-line.

Thanks Joerg,

it is not down to my English, I suppose it is good enough.
I have just switched words, the spec should have been called
offset-0, this is how it is known. I suppose people who know about
MCAs would not pay much attention to my mistake, this is a common
spec on that stuff (exactly what you guessed). I'll fix it, though.

Dimiter
 
J

John Larkin

Jan 1, 1970
0
Pretty unique, yes. And guaranteed...
This is the only MCA on the market which allows you to calibrate
for energy using a single point.

Dimiter

What's the percentage tolerance spec on your zero?

John
 
D

Didi

Jan 1, 1970
0
What's the percentage tolerance spec on your zero?

It can be seen (and specified) as 0%. The INL and DNL are non-0, of
course, so it
is safe to specify that (which, in essence, is what counts - you can
calibrate by 1 point without giving it much thought).
The 0 is mathematically correct, and while the input data will always
carry some error, it is randomized and spread over many samples
(hundreds or thousands), so that specifying a non-0 number (say, 0.1%
LSB...)
would make no sense (and would convey perhaps the wrong message
to users, they would have to look for error where there is none).
The INL is within 1 LSB or so, it safely covers for the offset-0 being
specified as
100% 0. So does even the DNL, BTW, I have specified a conservative
number,
yet the fact is it is so low I could not measure it when I did that
(and I could
measure 0.5% LSB DNL on my older, Wilkinson based devices).

Dimiter
 
M

Michael A. Terrell

Jan 1, 1970
0
Joerg said:
In the army our drill sergeant told two guys to stop singing. One of
them is yours truly.


That's because he as afraid everyone else would go AWOL, or give you
two a GI shower and blanket party. ;-)


--
Service to my country? Been there, Done that, and I've got my DD214 to
prove it.
Member of DAV #85.

Michael A. Terrell
Central Florida
 
J

Joel Koltner

Jan 1, 1970
0
Say Joerg,

I just thought I'd mention...

Some purchasing guy was talking with me about various parts for some project
yesterday and I noticed that we're using Linear Tech's LTC1844 in numerous
different products and at least so far no one's reporting seeing one
misbehave. This is a 150mA LDO (90mV headroom required at 150mA, 30mV at
50mA). The data sheet has a chart that claims it's guaranteed stable so long
as the output capacitance is >=1uF and the cap's ESR is <=300milliohms.

I suspect this is too spendy for most of your applications -- $1.45 in reel
quantities from DigiKey -- but if you ever feel like trying our an LDO again,
there's one candidate...

---Joel
 
J

Joerg

Jan 1, 1970
0
Joel said:
Say Joerg,

I just thought I'd mention...

Some purchasing guy was talking with me about various parts for some project
yesterday and I noticed that we're using Linear Tech's LTC1844 in numerous
different products and at least so far no one's reporting seeing one
misbehave. This is a 150mA LDO (90mV headroom required at 150mA, 30mV at
50mA). The data sheet has a chart that claims it's guaranteed stable so long
as the output capacitance is >=1uF and the cap's ESR is <=300milliohms.

Thanks for the hint. However, quote from datasheet: "If the input supply
voltage drops too low for the LTC1844 to maintain regulation, the
internal feedback loop goes into dropout and the internal pass
transistor turns fully on. If the input supply then suddenly rises, the
output may briefly overshoot the intended output voltage ..."

Can be muffled with a big cap on the output, probably driving the
p-channel into its "ouch" region. I don't really like such behavior.
Nah, not my cup of tea.

I've also had my share of experiences with fast input voltage rise
times. A TPS regulator reacted to that with phssst ... BANG. TI couldn't
explain it, refused to give me the innards for SPICE, refused to run my
circuit on their SPICE. So I just chucked it.

I suspect this is too spendy for most of your applications -- $1.45 in reel
quantities from DigiKey -- but if you ever feel like trying our an LDO again,
there's one candidate...

For a 6.5V/150mA max regulator that is quite expensive.
 
J

Joel Koltner

Jan 1, 1970
0
Hi Joerg,

Joerg said:
...If the input supply then suddenly rises, the output may briefly overshoot
the intended output voltage ..."

Thanks for pointing that out. Wouldn't a lot of regulators (LDO and non-LDO
alike) do this, though? Once you saturate your pass device, if it's slower to
respond than the feedback loop, you're sunk, aren't you? (Seems like to
op-amps where, once you hit the rails, pretty much all bets are off on how it
behaves... although things like the outputs reversing polarity like some old
op-amps did is a bit much...)
For a 6.5V/150mA max regulator that is quite expensive.

You're paying for the "very" LDO in VLDO there, I think.

---Joel
 
J

Joerg

Jan 1, 1970
0
Joel said:
Hi Joerg,



Thanks for pointing that out. Wouldn't a lot of regulators (LDO and non-LDO
alike) do this, though? Once you saturate your pass device, if it's slower to
respond than the feedback loop, you're sunk, aren't you? (Seems like to
op-amps where, once you hit the rails, pretty much all bets are off on how it
behaves... although things like the outputs reversing polarity like some old
op-amps did is a bit much...)

I have never seen a LM317 do that. Usually when they approach dropout
LDOs steer their pass device into full conduction and naturally it'll
take a while to swing back when Vin jumps back up. OTOH non-LDOs are
different, they typically run out of base drive and that happens kind of
gracefully. Their pass device doesn't really leave the linear range, at
least not by much.

It's a bit like jet engines versus regular gas engines. When you pull
the throttle on a jet engine it takes a while until it reduces thrust
while a (non-turbocharged) gas engine reacts immediately.

Take a look at the right-most part of the schematic for the LM317, about
in the middle of this datasheet:
http://www.national.com/ds/LM/LM117.pdf

People have done rather strange things with the LM317. One example is in
the above datasheet on page 18 where they built a switcher with it. I've
used them as AM modulators and such, usually producing some forehead
wrinkles and coughing in design reviews.

You're paying for the "very" LDO in VLDO there, I think.

When I am that tight on voltage overhead I usually build a switcher.
 
Top