Video Switch Circuit
Quack
I have made a few 'video switching' circuits that are failing,
hopefully someone can give me some tips .. perhaps my board layout is
not good for this purpose ? (not enough ground plains perhaps ?).
Firstly the board layout can be viewed at
http://www.quack.cc/video.jpg
p15v332-Q PDF can be viewed at http://www.quack.cc/P15V.PDF
The 18 pin IC in the middle is a PIC 16f84 (DIP Through-Board) running
with a 4mhz ceramic resonator, which receives serial signals from a
host and configures its outputs (portA and portB, mixed) to control
the video switch chip, which is a P15v332Q (QSOP surface mount package
chip).
Perhaps the resonator is causing some interference in the circuit ? -
i have never dealt with any RF or video circuits before, maybe im
missing some really simple rules that must be followed for a good,
clean signal ... ?
The problem:
After soldering them up, they work nicely for a brief period of time.
After about an hour or so, the video signal gets visible iterference,
it seems to accumulate. After a day or so its completely unusable.
Some of them i did a re-soldering job on the P15V (surface mount
component), basically i just melted the solder again, and this makes
it 'work' cleanly once again. some of them continue to work forever,
some of them stuff up again.
But all of them have a slightly 'dirty' signal, even the 'working'
ones.
I have not used SM components before, so perhaps my technique is not
good.
I have tried quite a few techniques, but i find the following to be
the easiest to do and achieve what i think is a good result. I dont
have special tools or anything for surface mount.
The technique i use it to make my PCB using BLUE PNP paper printed
from a laser printer (part# PNPB100 - www.www.techniks.com), i iron
that on, peel it off and etch my boards in Ferric Chloride. This gives
a fairly good result, and no shorts and good, solid tracks.
I then coat the board in "flux 140" which is a dispenser pen which has
both flux and a protective coating for the board to stop oxidization.
This has worked wonders for all other work i have done. (part#
363-6539 @ £7.29 from RS components, www.rswww.com if anyone is
interested)
Then to install the Surface Mount video switch, i wet each pad with
just enough solder to 'paint it silver", (using 60/40, flux loaded
solder).
(i read somewhere that silver loaded solder is better for SM
components because it conducts the heat faster and therefore minimises
component damage because its soldered on quicker, and something todo
with metal bonding ?? could this be my problem perhaps ?)
I then position the chip under magnification and hold it down while i
lightly and quickly touch the top of each pin with a fine solder tip
at about 400' degrees. this melted the pads just enough for the legs
to stick nicely, and leaves a good looking clean job. I then proceed
to test its connection with a continuity tester, to ensure that the
top of each leg does infact have connectivity to its track, and that
there are no shorts. these tests work out fine.
(even after the 'signal degradation', the continuity test still shows
no problems, even when they completely dont work !! what is going on
?!?)
Im at a loss why these are failing ? i cant see any problem.
Any hints or suggestions would be great! :)
Alex.
PS : i cant change 'video switch' chip brands, due to the large stock
i now have of them, the P15V332 must be used, and in SM form. I wish i
could have got them in through-hole form, but unfortunantly they just
didnt come that way. I chose these largely because of their
availablitity and price. They cost only $2.10 from www.digikey.com.
Robert Baer
First, a few comments concerning "silver loaded" solder: There are at
least 3 variants.
1) 2% eutectic which is recommended and used in old tube Tektronics
scopes for repair on the ceramic standoffs as well as work on any other
components that are silver plated. The eutectic alloy prevents leaching
of silver from those components. There is no significant change in the
melting point compared to the 60/40 or 63/37 alloys. There is *no* other
reason to use this 2% eutectic alloy.
2) the 2 other silver alloys have zero lead and have a significantly
higher melting temperature. Their main useage is for environmentially
"picky" assembly as well as for making PCB assemblys that must perform
to at least 175C. There is *no* other reason to use these alloys.
3) Even if one could use pure silver *and* have the same exact melting
point as 60/40 solder (clearly impossible - but just imagine it), the
greater thermal conductivity would make *zero* difference in soldering
time fiven the exact same soldering techniques.
4) There are no significant or known differences in metal bonding
between any of these alloys.
****
One thing that may help and certainly will give better impedance
matching, is to use a copper ground plane on the back; make sure you
scatter vias from topside ground traces to the ground plane liberally -
seperation not more than 1/6 wavelength of highest frequency (as a rough
guide).
Any place where there is need of a bypass capacitor (eg: supply
voltage on all ICs), make *damn* sure there is one as close to the
device pin as possible on the bypass end, and a via to the ground plane
on the other. One can use size 1026 chip capacitors at the physical
largest; they are large enough to handle easily; if hand soldering, one
can hold down the SMD with a toothpick to prevent movement.
It sounds like you may have one or more open pins that "float" and
maybe the close RF signal biases the pin until its "bias" increases to a
non-operational state.
If you can put the PCB in the oven and bake it at 125C for a few hours
or so, and if that "fixes" the problem which then repeats in a similar
timeframe, that is a virtual guarantee of "biasing" that latches CMOS
inputs.
If you have reasonable assurance the connections are OK, check very
carefully that DC bias + RF peak-to-peak levels on inputs do *NOT*
exceed the supply voltage, or go below ground.
Also, it does not hurt to double check and even triple check the
circuit - start with only scratch paper and the PCB, and without looking
anywhere else, trace out what you have.
Robert Baer
Oops! Typo-- should have been size 1206 chip capacitor!
Quack
Wow, great info, thats for the help :)
> First, a few comments concerning "silver loaded" solder: There are at
> least 3 variants.
Okay, so silver loaded solder wont help me much ... thats good to know
:).
I cant remember where i heard about it ...
> One thing that may help and certainly will give better impedance
> matching, is to use a copper ground plane on the back; make sure you
> scatter vias from topside ground traces to the ground plane liberally -
> seperation not more than 1/6 wavelength of highest frequency (as a rough
> guide).
Sorry, i *think* i understood what your saying, but that would imply
im using a double-sided PCB ? its only single side, the PIC and other
components are 'through-hole' and the video switch is mounted on the
bottom of the board.
By 'copper ground planes' you mean leaving a lot of board area
un-etched and lots of attachments from that to any other ground in the
circuit ?
Is my circuit visibly lacking in ground planes and vias ?
> Any place where there is need of a bypass capacitor (eg: supply
> voltage on all ICs), make *damn* sure there is one as close to the
> device pin as possible on the bypass end, and a via to the ground plane
> on the other. One can use size 1026 chip capacitors at the physical
oooh.. i dont have any capacitors in this circuit at all :(. should i
have ?
i noticed in my PIC book it recommends a .1uF cap going to ground from
the pic's ground pin. is this what you would call a 'bypass capacitor'
?
> largest; they are large enough to handle easily; if hand soldering, one
> can hold down the SMD with a toothpick to prevent movement.
Yup, i have soldered plenty of SM resistors and caps ... theyr easy,
not much chance of shorting their pins :).
> It sounds like you may have one or more open pins that "float" and
> maybe the close RF signal biases the pin until its "bias" increases to a
> non-operational state.
> If you can put the PCB in the oven and bake it at 125C for a few hours
> or so, and if that "fixes" the problem which then repeats in a similar
> timeframe, that is a virtual guarantee of "biasing" that latches CMOS
> inputs.
ahh, this is sounding likely .. although i dont understand how
exactly.
all the pins, PORTA and PORTB are set as outputs in the PIC program.
Although not all pins have any load or connection at all - are these
the ones that could be 'floating' ? and if they are, how would they
(being unused) affect the rest of the circuit ?
Or are you saying that the i/o pins i am using to control the video
switch (p15v) may be floating around ? - they are connected to the
inputs of the other chip and set as outputs only (100% of the time),
surely this would not allow them to 'float' ?
Or maybe, i dont know much about this ... if this was the problem,
what can i do to fix it ? use a pull-up resistor or something ?
> If you have reasonable assurance the connections are OK, check very
> carefully that DC bias + RF peak-to-peak levels on inputs do *NOT*
> exceed the supply voltage, or go below ground.
right over my head on that one!
> Also, it does not hurt to double check and even triple check the
> circuit - start with only scratch paper and the PCB, and without looking
> anywhere else, trace out what you have.
Yup, 100 checks ... not *All* of them have this problem. some seem to
be working nicely. so the circuit must be functional ... Which leads
me to believe it must be a physical problem. Someone suggested i use
'flux remover' after soldering (not included in the process i
described above) as i am using flux-loaded solder and flux-protectant
on the board.
I didnt have any cleaner, so i used some 'isopropyl alcohol', i think
its the main ingredient anyway - and i got a bad one to work!. perhaps
the flux was conducting ?!?
I havent had this problem on my other boards because i dont use any SM
components so things are farther away and i guess the flux residue
doesnt affect them at that distance ...
Im about to make a new batch today - but this time i am scrapping the
P15V chip and going with some NPN transistors to switch the inputs
on/off. I made a small test on a breadboard, seems to work well -
we'll see how well it works as a completed circuit.
Thanks :)
Alex.
Robert Baer
I *did* mean a 2-sided PCB, as with a certain trace width, the result
is a transmission line near 50 ohms (it has been over 30 years so i do
not remember the correct width, but i think it should be 0.1").
For traces longer than 1/8 wavelength, that becomes significant, and
long transmission lines should be terminated in their characteristic
impedance - else one gets standing waves at constant frequencies.
Pulses get reflected at unterminated ends, which could give voltages
greater than "nominal".
Having a ground plane also reduces coupling of signal between parallel
traces.
However, if the traces are all short relative to 1/8 wavelength of the
highest frequency (look at risetime and fall times for that if pulses
are there) then a ground plane per-se might not be needed.
Having ground traces everywhere, especially for bypassing can help in
that case.
There is a "dirty" trick to make PCBs with very little etching, that
can produce a ground trace around everything else: using paper and a
drafting pen with black ink, draw "islands" that are the traces you
need; put dots where you want holes. This gives you a paper negative
that can be archived.
You might be able to shave off some length of the PCB and shorten many
of the traces to the SOIC by moving it closer to the 18 pin device.
At least add in bypassing on that SOIC and the 18 pin device with
ceramic chip capacitors....
**
Regarding RF levels and DC bias on each and every input pin of the
SOIC (i get the impression that is where the problem is):
1) Say the supply is +5V and that is the only supply.
2) Say it is CMOS.
3) Say logic inputs are DC biased at logic levels near 0V and 5V because
the inputs are being driven by CMOS logic devices.
Now all one needs is to have long and unterminated traces. Rapid rise
times will produce reflections, leading to inputs being driven well
above the supply voltage.
4) Say signal inputs have a nominal DC bias of 2.5V, and signal levels
are RF, peak-to-peak of 4.9V.
Now all one needs is that bias level to increase more than 0.1V *or*
the RF amplitude to increase more than 0.1V --- the signal input then is
driven more positive than the supply.
Either condition #3 or #4 can lead to the symptoms you mention.
If you take that SOIC unit that is rather bad and heat it for a decent
time, the damage will anneal out and it will work OK (for a while).
***
A capacitor from a supply pin to a good ground trace is a bypass
capacitor.
Ground pins are already connected to ground...
***
*ALL* unused input pins should be connected to ground, if that biasing
is OK for the circuit. If a supply bias is recommended, use a 1K to 10K
resistor to the supply and add a bypass capacitor to ground. A floating
input pin can "pick up" RF and become driven above the supply voltage,
causing the same problem i previously mentioned.
***
Input pins to the SMD that are driven by a programmable pin that
(after programming) acts like an output may or may not cause problems.
Some programmable I/O devices are "open drain" which require an
external pullup; look carefully for this possibility.
When in doubt, add a 10K resistor to the supply; if an open drain, the
rise time may be a bit slow, but it will work for all cases. The
resistor certainly will protect both inputs (ESD, etc) when there is no
power, and before programming to "out".
Quack
Wow thanks for all your help :)
Although i have found these other chips, MAX4141's, and they do check
the sync signal on all 4 inputs, have buffers or something and switch
cleanly.
I have ordered their 'eval' board so i can see how its implemented and
basically copy that into my circuit. Although it seems simple enough
from reading their PDF .. (although thats what i thaught about the
p15v chip!).
Any comments about that ?
Also on www.digikey.com it lists their package as SOIC, i found a
description on google somewhere, saying "small outline plastic
packages" for SOIC.. is that right ? - doesnt seem too descriptive..
is it surface mount ? :)
Thanks again!,
Alex.
Robert Baer <rober...@earthlink.net> wrote in message news:<3F5AD5C5...@earthlink.net>...
Glenn Gundlach
> Hi Robert,
>
> Wow thanks for all your help :)
>
> Although i have found these other chips, MAX4141's, and they do check
> the sync signal on all 4 inputs, have buffers or something and switch
> cleanly.
>
> I have ordered their 'eval' board so i can see how its implemented and
> basically copy that into my circuit. Although it seems simple enough
> from reading their PDF .. (although thats what i thaught about the
> p15v chip!).
> Any comments about that ?
>
> Also on www.digikey.com it lists their package as SOIC, i found a
> description on google somewhere, saying "small outline plastic
> packages" for SOIC.. is that right ? - doesnt seem too descriptive..
> is it surface mount ? :)
>
> Thanks again!,
>
> Alex.
I looked at that Maxim chip. It did not seem to be anything more than
a good switch. It didn't appear to do anything more with sync than
simply pass it through. You may need to DC restore the signals before
performing the switch and they will need to be synchronized if you
want no disturbance in the output. Also, you may want to switch during
the vertical interval.
GG