🚀 go-pugleaf

Vitaliy wrote:

> Hello,
> 
> My appplication is light detection -> using photodiodes and
> transimpedeance amplifier (OPA657 from Texas Instruments). At this
> point I have 2-layer working design. I need to increase the bandwidth
> of the detector (from 10MHz to 40MHz) and try to reduce noise (also
> reducing noise is always welcome, of course). I'm thinking about using
> two inner layers for +5V and -5V power layers (I'm using free space on
> top and bottom layers for ground layer). +-5V is currently connected to
> the board from external power supply by soldering two wires (one for
> +5V and one for -5V). I am using +-5V to bias photodiodes and to power
> the opamp. The software I'm using is OrCAD. Even if going to 4 layer
> from two wouldn't give me a significant improvement in terms of
> reducing the noise level, I would like to know the answers for the
> following questions.
> 
> 1) Will traferring +-5V to separate layers (rather than just have the
> power traces with filters) work? Since the inner layer is most likely
> to be .5oz copper as opposed to 1oz copper of outer layers. There is no
> way to specify that in gerber file, right? Will creating a power layer
> give me more stable voltage (and high capacitance)? That's why people
> are using multilayers as opposed to two layers, right? I guess I will
> still need some decoupling capacitors at the vias.
> 
> 
> 2) The price differential between 2 and 4 layer boards is quite steep,
> and I can easily fit a few prototype boards on one board and then cut
> them (or have it cut by manufacturer). What is the easiest way to place
> array of differemt designs on one board? I only have access to OrCAD,
> but it can generate gerber files, so maybe some other software can be
> used. Are the manufacturers ok with this? I would prefer them to cut
> the boards, if possible, since I probably wouldn't be able to do it as
> well without special equipment. (Something tells me they woudn't be
> willing to make x by x board, and then cut it in y pieces, but just
> checking). I do realize I might be able to place three designs side by
> side and not require any cutting, but that might not always work for
> me.
> 
> Merry Christmas,
> Vitaliy
> 
single sided Ceramic daughter board for the OP-AMP and INPUT?
Also a hole in the board so that the chip is suspended via the
rails and output legs only also may help. I did something like that
years ago to remove the cap around the structure.


-- 
"I'm never wrong, once i thought i was, but was mistaken"
Real Programmers Do things like this.
http://webpages.charter.net/jamie_5

All components are smt (except for photodiodes). I also have the
components (resistors and capacitors) placed on both bottom and top
layers (2 layer board), with opamp being on the top layer. Top and
bottom are just relative terms in my case.


Vitaliy

Jamie wrote:
> single sided Ceramic daughter board for the OP-AMP and INPUT?
> Also a hole in the board so that the chip is suspended via the
> rails and output legs only also may help. I did something like that
> years ago to remove the cap around the structure.
>

On 24 Dec 2006 22:55:41 -0800, "Vitaliy" <m.vitaliy@gmail.com> wrote:

>Hello,
>
>My appplication is light detection -> using photodiodes and
>transimpedeance amplifier (OPA657 from Texas Instruments). At this
>point I have 2-layer working design. I need to increase the bandwidth
>of the detector (from 10MHz to 40MHz) and try to reduce noise (also
>reducing noise is always welcome, of course). I'm thinking about using
>two inner layers for +5V and -5V power layers (I'm using free space on
>top and bottom layers for ground layer). +-5V is currently connected to
>the board from external power supply by soldering two wires (one for
>+5V and one for -5V). I am using +-5V to bias photodiodes and to power
>the opamp. The software I'm using is OrCAD. Even if going to 4 layer
>from two wouldn't give me a significant improvement in terms of
>reducing the noise level, I would like to know the answers for the
>following questions.
>
>1) Will traferring +-5V to separate layers (rather than just have the
>power traces with filters) work? Since the inner layer is most likely
>to be .5oz copper as opposed to 1oz copper of outer layers. There is no
>way to specify that in gerber file, right? Will creating a power layer
>give me more stable voltage (and high capacitance)? That's why people
>are using multilayers as opposed to two layers, right? I guess I will
>still need some decoupling capacitors at the vias.

Gerber files only specify the copper patterns that are to appear on
the board.  Any other construction details, including copper weight,
and the order of layers on a multi-layer board, must be specified
separately.  I have a standard format for a text file that includes
such information.

When laying out a multi-layer board, you should put a wide track at
the edges of the board on the power or ground plane layers to keep the
copper on those layers away from the edge of the board.  (Anything you
place on a plane layer will end up as "no copper" on the finished
board.)
>
>
>2) The price differential between 2 and 4 layer boards is quite steep,
>and I can easily fit a few prototype boards on one board and then cut
>them (or have it cut by manufacturer). What is the easiest way to place
>array of differemt designs on one board? I only have access to OrCAD,
>but it can generate gerber files, so maybe some other software can be
>used. Are the manufacturers ok with this? I would prefer them to cut
>the boards, if possible, since I probably wouldn't be able to do it as
>well without special equipment. (Something tells me they woudn't be
>willing to make x by x board, and then cut it in y pieces, but just
>checking). I do realize I might be able to place three designs side by
>side and not require any cutting, but that might not always work for
>me.

Discuss the matter with your board shop.  If I have two or more boards
that I want built on a common panel, I'll usually produce separate
gerbers for each board, then ask the board maker to combine them on a
single panel.  However, a co-worker places multiple boards on a panel
himself, and sends gerbers for the full panel to the board shop.
Either way, the board shop will separate the individual boards if
asked.  If you panelize the boards yourself, you should first ask the
board shop what separation they want between boards - they need some
unused space to leave room for the router bit to go between boards.

>
>Merry Christmas,
>Vitaliy

-- 
Peter Bennett, VE7CEI  
peterbb4 (at) interchange.ubc.ca  
new newsgroup users info : http://vancouver-webpages.com/nnq
GPS and NMEA info: http://vancouver-webpages.com/peter
Vancouver Power Squadron: http://vancouver.powersquadron.ca

Vitaliy wrote:
> Hello,
>
> My appplication is light detection -> using photodiodes and
> transimpedeance amplifier (OPA657 from Texas Instruments). At this
> point I have 2-layer working design. I need to increase the bandwidth
> of the detector (from 10MHz to 40MHz) and try to reduce noise (also
> reducing noise is always welcome, of course). I'm thinking about using
> two inner layers for +5V and -5V power layers (I'm using free space on
> top and bottom layers for ground layer). +-5V is currently connected to
> the board from external power supply by soldering two wires (one for
> +5V and one for -5V). I am using +-5V to bias photodiodes and to power
> the opamp. The software I'm using is OrCAD. Even if going to 4 layer
> from two wouldn't give me a significant improvement in terms of
> reducing the noise level, I would like to know the answers for the
> following questions.
>
> 1) Will traferring +-5V to separate layers (rather than just have the
> power traces with filters) work? Since the inner layer is most likely
> to be .5oz copper as opposed to 1oz copper of outer layers. There is no
> way to specify that in gerber file, right? Will creating a power layer
> give me more stable voltage (and high capacitance)? That's why people
> are using multilayers as opposed to two layers, right? I guess I will
> still need some decoupling capacitors at the vias.

The power planes on 4 layers boards are more useful for high switching
current and high speed digital stuff. There is a good chance it may
make little or no difference for application depending on your layout.
4 layers will give you greater ability to achieve a lower noise layout,
but if you don't know exactly how to achieve this then it's not going
to help you much, it could even make things worse, I have seen that
happen before.

> 2) The price differential between 2 and 4 layer boards is quite steep,
> and I can easily fit a few prototype boards on one board and then cut
> them (or have it cut by manufacturer). What is the easiest way to place
> array of differemt designs on one board?

Simply ask the manufacturer to do it, do waste time and effort
yourself. Just supply them the layout for one board and tell them you
want X number of boards. For the prototype services they will usually
fit as many as they can on the one panel for you, you pay per panel.

For high production designs it's usually different, you have to worry
about panel handling and tooling hole/marks etc.

> I only have access to OrCAD,
> but it can generate gerber files, so maybe some other software can be
> used. Are the manufacturers ok with this? I would prefer them to cut
> the boards, if possible, since I probably wouldn't be able to do it as
> well without special equipment. (Something tells me they woudn't be
> willing to make x by x board, and then cut it in y pieces, but just
> checking). I do realize I might be able to place three designs side by
> side and not require any cutting, but that might not always work for
> me.

Some manufacturers will do up to say 3 different designs on the one
prototype panel for you, if that's what you want.

Dave :)

"Vitaliy" <m.vitaliy@gmail.com> writes:

> Hello,
>
> My appplication is light detection -> using photodiodes and
> transimpedeance amplifier (OPA657 from Texas Instruments). At this
> point I have 2-layer working design. I need to increase the bandwidth
> of the detector (from 10MHz to 40MHz) and try to reduce noise (also
> reducing noise is always welcome, of course).

These requirements tend to be mutually exclusive... i.e., more
bandwidth lets in more noise.

What is the source of the "noise"? E.g. source noise, photocurrent
shot noise, transimpedance amplifier noise, external interference?

Until you do this it seems pointless to speculate about numbers of
layers in boards.

<SNIP>

> I'm thinking about using two inner layers for +5V and -5V power
> layers (I'm using free space on top and bottom layers for ground
> layer). +-5V is currently connected to the board from external power
> supply by soldering two wires (one for
> +5V and one for -5V). I am using +-5V to bias photodiodes and to power
> the opamp. The software I'm using is OrCAD. Even if going to 4 layer
> from two wouldn't give me a significant improvement in terms of
> reducing the noise level, I would like to know the answers for the
> following questions.

>
> 1) Will traferring +-5V to separate layers (rather than just have the
> power traces with filters) work? Since the inner layer is most likely
> to be .5oz copper as opposed to 1oz copper of outer layers. There is no
> way to specify that in gerber file, right? Will creating a power layer
> give me more stable voltage (and high capacitance)? That's why people
> are using multilayers as opposed to two layers, right? I guess I will
> still need some decoupling capacitors at the vias.
>
>
> 2) The price differential between 2 and 4 layer boards is quite steep,
> and I can easily fit a few prototype boards on one board and then cut
> them (or have it cut by manufacturer). What is the easiest way to place
> array of differemt designs on one board? I only have access to OrCAD,
> but it can generate gerber files, so maybe some other software can be
> used. Are the manufacturers ok with this? I would prefer them to cut
> the boards, if possible, since I probably wouldn't be able to do it as
> well without special equipment. (Something tells me they woudn't be
> willing to make x by x board, and then cut it in y pieces, but just
> checking). I do realize I might be able to place three designs side by
> side and not require any cutting, but that might not always work for
> me.
>
> Merry Christmas,
> Vitaliy
>

-- 

John Devereux

On 24 Dec 2006 22:55:41 -0800, "Vitaliy" <m.vitaliy@gmail.com> wrote:

>Hello,
>
>My appplication is light detection -> using photodiodes and
>transimpedeance amplifier (OPA657 from Texas Instruments). At this
>point I have 2-layer working design. I need to increase the bandwidth
>of the detector (from 10MHz to 40MHz) and try to reduce noise (also
>reducing noise is always welcome, of course). I'm thinking about using
>two inner layers for +5V and -5V power layers (I'm using free space on
>top and bottom layers for ground layer). +-5V is currently connected to
>the board from external power supply by soldering two wires (one for
>+5V and one for -5V). I am using +-5V to bias photodiodes and to power
>the opamp. The software I'm using is OrCAD. Even if going to 4 layer
>from two wouldn't give me a significant improvement in terms of
>reducing the noise level, I would like to know the answers for the
>following questions.
>
>1) Will traferring +-5V to separate layers (rather than just have the
>power traces with filters) work? Since the inner layer is most likely
>to be .5oz copper as opposed to 1oz copper of outer layers. There is no
>way to specify that in gerber file, right? Will creating a power layer
>give me more stable voltage (and high capacitance)? That's why people
>are using multilayers as opposed to two layers, right? I guess I will
>still need some decoupling capacitors at the vias.
>
>
>2) The price differential between 2 and 4 layer boards is quite steep,
>and I can easily fit a few prototype boards on one board and then cut
>them (or have it cut by manufacturer). What is the easiest way to place
>array of differemt designs on one board? I only have access to OrCAD,
>but it can generate gerber files, so maybe some other software can be
>used. Are the manufacturers ok with this? I would prefer them to cut
>the boards, if possible, since I probably wouldn't be able to do it as
>well without special equipment. (Something tells me they woudn't be
>willing to make x by x board, and then cut it in y pieces, but just
>checking). I do realize I might be able to place three designs side by
>side and not require any cutting, but that might not always work for
>me.
>
>Merry Christmas,
>Vitaliy

More layers won't necessarily improve speed or reduce noise. If you
increase trace capacitance (from thinner dielectrics) it might make
things worse.

10 or even 40 MHz isn't very fast for a photodetector; I've got 180
MHz from cheap opamps. What detector device are you using? Increasing
the photodiode back-bias should help speed a bunch, maybe 2:1 if you
go from 5 volts to something like 20.

What's the application? How much noise do you see, in equivalent
optical power?

John

> Gerber files only specify the copper patterns that are to appear on
> the board.  Any other construction details, including copper weight,
> and the order of layers on a multi-layer board, must be specified
> separately.  I have a standard format for a text file that includes
> such information.
>
> When laying out a multi-layer board, you should put a wide track at
> the edges of the board on the power or ground plane layers to keep the
> copper on those layers away from the edge of the board.  (Anything you
> place on a plane layer will end up as "no copper" on the finished
> board.)
> >
I think I'm missing something: should I leave some space between the
edge of the board and power plane (which will be internal plane)?


> Discuss the matter with your board shop.  If I have two or more boards
> that I want built on a common panel, I'll usually produce separate
> gerbers for each board, then ask the board maker to combine them on a
> single panel.  However, a co-worker places multiple boards on a panel
> himself, and sends gerbers for the full panel to the board shop.
> Either way, the board shop will separate the individual boards if
> asked.  If you panelize the boards yourself, you should first ask the
> board shop what separation they want between boards - they need some
> unused space to leave room for the router bit to go between boards.
>
OK, I will look into panelizing the boards myself (also I would like to
panelize different designs (same # of layers though) on the same
board). If you can suggest a program that can take two gerber files and
create one board will be great (I'll check if this can be done in
OrCAD).


PS. I will respond to other suggestions a bit later on today.

Thanks a lot,
Vitaliy

On 26 Dec 2006 23:50:13 -0800, "Vitaliy" <m.vitaliy@gmail.com> wrote:

>> When laying out a multi-layer board, you should put a wide track at
>> the edges of the board on the power or ground plane layers to keep the
>> copper on those layers away from the edge of the board.  (Anything you
>> place on a plane layer will end up as "no copper" on the finished
>> board.)
>> >
>I think I'm missing something: should I leave some space between the
>edge of the board and power plane (which will be internal plane)?
>
Yes.  If you let the copper on the planes extend to (and beyond) the
edge of the board, you may get short circuits between planes or
between a plane and a metal card guide.
>
>> Discuss the matter with your board shop.  If I have two or more boards
>> that I want built on a common panel, I'll usually produce separate
>> gerbers for each board, then ask the board maker to combine them on a
>> single panel.  However, a co-worker places multiple boards on a panel
>> himself, and sends gerbers for the full panel to the board shop.
>> Either way, the board shop will separate the individual boards if
>> asked.  If you panelize the boards yourself, you should first ask the
>> board shop what separation they want between boards - they need some
>> unused space to leave room for the router bit to go between boards.
>>
>OK, I will look into panelizing the boards myself (also I would like to
>panelize different designs (same # of layers though) on the same
>board). If you can suggest a program that can take two gerber files and
>create one board will be great (I'll check if this can be done in
>OrCAD).

I use Protel, which does allow cutting-and-pasting from one board
design to another, or within a board design.  When I make a panel,
I'll finish all the board designs in individual files, then copy and
paste them into a new blank design.  In Protel, when pasting designs,
you have to use a "Paste Special" command, and tell the program to
permit duplicate reference designators - otherwise Protel will
re-number components on the second and later boards.

-- 
Peter Bennett, VE7CEI  
peterbb4 (at) interchange.ubc.ca  
new newsgroup users info : http://vancouver-webpages.com/nnq
GPS and NMEA info: http://vancouver-webpages.com/peter
Vancouver Power Squadron: http://vancouver.powersquadron.ca

> The power planes on 4 layers boards are more useful for high switching
> current and high speed digital stuff. There is a good chance it may
> make little or no difference for application depending on your layout.
> 4 layers will give you greater ability to achieve a lower noise layout,
> but if you don't know exactly how to achieve this then it's not going
> to help you much, it could even make things worse, I have seen that
> happen before.
>
To start off, I will read you tutorial from pcb123 and hopefully
improve a few things in my design from that manual. I want to add the
power layers to reduce the length of power traces (I might need to add
the 2nd stage to the detector, and that will definitely require
more/longer power traces without power layers).
>
> Simply ask the manufacturer to do it, do waste time and effort
> yourself. Just supply them the layout for one board and tell them you
> want X number of boards. For the prototype services they will usually
> fit as many as they can on the one panel for you, you pay per panel.
>
> For high production designs it's usually different, you have to worry
> about panel handling and tooling hole/marks etc.
>
Any recommendations, I'm using pcbexpress, will try to get a hold of
them after New Year's day, local manufacturers are charging per panel
(but I will see if they can put a few designs on the same panel).

John Devereux wrote:
>
> These requirements tend to be mutually exclusive... i.e., more
> bandwidth lets in more noise.
>
> What is the source of the "noise"? E.g. source noise, photocurrent
> shot noise, transimpedance amplifier noise, external interference?
>
> Until you do this it seems pointless to speculate about numbers of
> layers in boards.
>
> John Devereux
What I meant was: I must increase the BW but see if there is a way to
reduce the noise (it does not have to be less than what I have right
now).
I was thinking differently: if I can provide more stable voltage to
both opamp and photodiodes, I can get less fluctuations during current
generation and amplifications and power plane should give more stable
voltage, right? (I can be totally wrong about this)

John Larkin wrote:
> More layers won't necessarily improve speed or reduce noise. If you
> increase trace capacitance (from thinner dielectrics) it might make
> things worse.
>
> 10 or even 40 MHz isn't very fast for a photodetector; I've got 180
> MHz from cheap opamps. What detector device are you using? Increasing
> the photodiode back-bias should help speed a bunch, maybe 2:1 if you
> go from 5 volts to something like 20.
>
> What's the application? How much noise do you see, in equivalent
> optical power?
>
> John
I'm using InGaAs PIN (Small Area/Fiber-Optic) photodiodes, which
according to the datasheet should be biased at 5V, wavelenth of the
signals varies between 800nm to 1700nm. NEP of the photodiode is <0.02
pW/sqrt(Hz) @1550nm according to the datasheet.


Vitaliy

PCB stack is an invention to minimise the EMI radiation and improve the
speed of signals talking in terms of microwave or high frequency
propagation.
Jumping from 2 to 4  layers will not gave you many improvements in EMI
as long the major signals are still routed on top and bottom. The first
stack with some performance is a 4 layer stack (counted from top to
bottom)as PWR-signal-signal-GND and 6 layer stack
(top-signal-PWR-GND-signal-GND-bottom or other similar versions) where
top and bottom are mixed signal/ground planes
Will be no noise improvement if +5 and -5V routes becomes planes, could
be worse.
In analogic and RF design the supply lines are usually routed radially
from the source (supply) to destination. This not means a contiguous
supply plane could be worst as long is in the neigborhood af quiet
ground planes.

best regards,
Vasile Surducan
senior application engineer
National Institute R&D for Isotopic and Molecular Technologies
Cluj-Napoca
Romania



Vitaliy wrote:
> Hello,
>
> My appplication is light detection -> using photodiodes and
> transimpedeance amplifier (OPA657 from Texas Instruments). At this
> point I have 2-layer working design. I need to increase the bandwidth
> of the detector (from 10MHz to 40MHz) and try to reduce noise (also
> reducing noise is always welcome, of course). I'm thinking about using
> two inner layers for +5V and -5V power layers (I'm using free space on
> top and bottom layers for ground layer). +-5V is currently connected to
> the board from external power supply by soldering two wires (one for
> +5V and one for -5V). I am using +-5V to bias photodiodes and to power
> the opamp. The software I'm using is OrCAD. Even if going to 4 layer
> from two wouldn't give me a significant improvement in terms of
> reducing the noise level, I would like to know the answers for the
> following questions.
>
> 1) Will traferring +-5V to separate layers (rather than just have the
> power traces with filters) work? Since the inner layer is most likely
> to be .5oz copper as opposed to 1oz copper of outer layers. There is no
> way to specify that in gerber file, right? Will creating a power layer
> give me more stable voltage (and high capacitance)? That's why people
> are using multilayers as opposed to two layers, right? I guess I will
> still need some decoupling capacitors at the vias.
>
>
> 2) The price differential between 2 and 4 layer boards is quite steep,
> and I can easily fit a few prototype boards on one board and then cut
> them (or have it cut by manufacturer). What is the easiest way to place
> array of differemt designs on one board? I only have access to OrCAD,
> but it can generate gerber files, so maybe some other software can be
> used. Are the manufacturers ok with this? I would prefer them to cut
> the boards, if possible, since I probably wouldn't be able to do it as
> well without special equipment. (Something tells me they woudn't be
> willing to make x by x board, and then cut it in y pieces, but just
> checking). I do realize I might be able to place three designs side by
> side and not require any cutting, but that might not always work for
> me.
> 
> Merry Christmas,
> Vitaliy