Re: PSoC Express: Does it work for semi-analog designs?

On Fri, 29 Dec 2006 10:31:38 +0100, "Henry Kiefer"
<otc_friend@gmx.net> wrote:

>Your' probably total wrong. The first 8051 had 40 pins. Have you ever
>thought that this guy will someday have 8 pins?

I predate the 8051 in embedded work.  So yes, I remember those days
and before and, no, I didn't imagine sot23-6 for PICs or 8051's back
then.

We shall see.  Situation is substantially different than then, in
diversity of options, as well as application spaces.  So I've no idea
why you feel that historical period should predict a new one in an
vastly different context.  On the other hand, given the length of time
since that time, I'd be willing to imagine some nice -6 and -8 parts
for ARM.  But by then, I can also imagine so many other changes.

For now, I'll stay with my suspicions that folks manufacturing ARMs
don't see 8- and 16-bit register application spaces with small numbers
of highly-repurposable I/O pins as their near-term future.  And I'll
be _very_ glad to be wrong about it!

Jon

"Jonathan Kirwan" <jkirwan@easystreet.com> schrieb im Newsbeitrag
news:bsp9p2p2237c5kfibb6msfn83nuncnqeno@4ax.com...
> On Fri, 29 Dec 2006 10:31:38 +0100, "Henry Kiefer"
> <otc_friend@gmx.net> wrote:
>
> >Your' probably total wrong. The first 8051 had 40 pins. Have you ever
> >thought that this guy will someday have 8 pins?
>
> I predate the 8051 in embedded work.  So yes, I remember those days
> and before and, no, I didn't imagine sot23-6 for PICs or 8051's back
> then.
>
> We shall see.  Situation is substantially different than then, in
> diversity of options, as well as application spaces.  So I've no idea
> why you feel that historical period should predict a new one in an
> vastly different context.  On the other hand, given the length of time
> since that time, I'd be willing to imagine some nice -6 and -8 parts
> for ARM.  But by then, I can also imagine so many other changes.
>
> For now, I'll stay with my suspicions that folks manufacturing ARMs
> don't see 8- and 16-bit register application spaces with small numbers
> of highly-repurposable I/O pins as their near-term future.  And I'll
> be _very_ glad to be wrong about it!

Once upon time the functional chip itself was big and the border around was
filled with i/o pads. Process shrunked a lot and now the core is very small
and there is a problem how to effective add the pads around if the inner
area is so small. That can leave chip area unused and that is very costly
and so prohibited.

So the natural way is that newer chips CAN have less pin-count (That means a
smaller case too). Probably newer ones will go more in the direction of
having more complex high-speed serial i/o.
The other point is that casing is costly (production money and space), so a
smaller case is better.

Hope the best ;-)

- Henry

--
www.ehydra.dyndns.info

On 29 Dec 2006 06:41:28 -0800, "rickman" <gnuarm@gmail.com> wrote:

>Jonathan Kirwan wrote:
>> On 28 Dec 2006 17:55:12 -0800, "rickman" <gnuarm@gmail.com> wrote:
>>
>> >Jonathan Kirwan wrote:
>> >> On Thu, 28 Dec 2006 23:29:41 +0100, "Henry Kiefer"
>> >> <otc_friend@gmx.net> wrote:
>> >>
>> >> ><snip>
>> >> >No one can foresee the future. But as out of my experience the ARM
>> >> >architecture will drive at least the next 20 years and will completely
>> >> >replace (almost) all 8051.
>> >> ><snip>
>> >>
>> >> When it comes in SOT-23-6 and 8-MSOP, it may begin to replace PIC and
>> >> 8051.  So far, I think the smallest package ARM I've seen is in the
>> >> 48-pin arena.  They also come with far too much on-chip resource,
>> >> which I pay for in terms of active power and pricing.  Way above what
>> >> I want, most of the time.
>> >>
>> >> I'm not holding my breath for an 8-MSOP ARM.  I hope to see a 28/32
>> >> pin device, someday.  That might actually reach down into the upper
>> >> end of my territory.
>> >
>> >Check out the Luminary Micro parts.  They have a low end ARM Cortex-M3
>> >chip in a 28 pin SOIC, IIRC, that can go for under $1 in huge quantity.
>>
>> No PWM until I get to 48-pin devices; a 32kHz input siphons off one
>> CCP pin and the LM3S101 only has one of them; etc.
>
>So what are your requirements again?  Seems like they are a lot more
>extensive than you had originally stated.

I didn't state them, earlier.  At least, not with any intention.

>I don't see where the CCP and the 32kHz are mutually exclusive.  I
>think you may be mistaken on this.

I was looking at this site:

http://www.luminarymicro.com/products/product_selector_guide.html

If you scan the list, on the left side you will see this description,
"External 32 KHz Clock (uses CCP pin)"  This is where I got it.

>I also don't get your aversion to the 48 pin TQFP package.  They are
>not hard to solder and they are a lot smaller than the 28 pin package
>you seem to think is ok.  The lead pitch is finer, but I have not found
>many applications where 20 mil lead pitch is a problem.

The 8-MSOP mentioned isn't quite .5mm spacing, but close at .65mm. For
larger packages, say 16-pin, they are also .65mm spacing and 4mm on a
side.

I take clear note of the fact that you seem to be arguing (separately
from Henry) that no one should have any problem with 48-pin packages
and that anyone suggesting fewer pins for their application is simply
wrong about it.  If that is your position in some attempt to defend
Henry's point, "the ARM ... will completely replace (almost) all
8051," I can't agree with you or Henry about this.

>If you look in more detail at the Philips line you will find any number
>of devices with a single supply.

I will do that, then.  But since I was already looking at devices with
more flash and more ram than I want, and they were the smallest
devices I could easily find, my suspicion is that finding single
supply devices will push me up the curve a bit into territory even
further away from where I'm usually at.  But I'll look more.

>All of the Atmel parts are single
>supply and they have a new AT91SAM7S16(1) coming out in the 48 pin
>packages (both TQFP and QFN).

I know this, at one time, wasn't true.  Because I actually went
through Atmel's list (as I know you have, too) and found dual supply
ARM devices (from their France operation.)  Perhaps that has changed,
but I doubt it.  Are you certain that _all_ Atmel ARM devices are
_not_ dual supply?

Not that it will matter to me.  For now, I won't spec them.

>It should be around $2 in quantity.

As I said, I won't spec them right now.  But in the 2k-5k area?

>The
>lowest price I have seen from Atmel is $3.35 on one of the two SAM7S32
>flavors.

Oh, I must have misunderstood your context then about the $2 figure.
Were you talking about Atmel, then?

>There was some confusion on pricing between the SAM7S32 and
>SAM7S321.  The SAM7S321 is the same part in the 64 pin package with all
>the same functionality of the other parts in the series while the
>SAM7S32 is in the 48 pin package with reduced pinout and functionality.

I'm confused.

> The pricing I received was within $0.06 of each other so maybe the
>higher pin count for the same price will do something for you.  BTW,
>these quantities are for 4K per year.

Okay.  But really, I don't want higher pin counts.  I'm already fine
with 16-20 pins and with 6 and 8 pins, as well.  Debating over 48 and
64 pins is for someone else to engage.

Jon

Henry Kiefer skrev:
> "Jonathan Kirwan" <jkirwan@easystreet.com> schrieb im Newsbeitrag
> news:bsp9p2p2237c5kfibb6msfn83nuncnqeno@4ax.com...
> > On Fri, 29 Dec 2006 10:31:38 +0100, "Henry Kiefer"
> > <otc_friend@gmx.net> wrote:
> >
> > >Your' probably total wrong. The first 8051 had 40 pins. Have you ever
> > >thought that this guy will someday have 8 pins?
> >
> > I predate the 8051 in embedded work.  So yes, I remember those days
> > and before and, no, I didn't imagine sot23-6 for PICs or 8051's back
> > then.
> >
> > We shall see.  Situation is substantially different than then, in
> > diversity of options, as well as application spaces.  So I've no idea
> > why you feel that historical period should predict a new one in an
> > vastly different context.  On the other hand, given the length of time
> > since that time, I'd be willing to imagine some nice -6 and -8 parts
> > for ARM.  But by then, I can also imagine so many other changes.
> >
> > For now, I'll stay with my suspicions that folks manufacturing ARMs
> > don't see 8- and 16-bit register application spaces with small numbers
> > of highly-repurposable I/O pins as their near-term future.  And I'll
> > be _very_ glad to be wrong about it!
>
> Once upon time the functional chip itself was big and the border around was
> filled with i/o pads. Process shrunked a lot and now the core is very small
> and there is a problem how to effective add the pads around if the inner
> area is so small. That can leave chip area unused and that is very costly
> and so prohibited.
>
> So the natural way is that newer chips CAN have less pin-count (That means a
> smaller case too). Probably newer ones will go more in the direction of
> having more complex high-speed serial i/o.
> The other point is that casing is costly (production money and space), so a
> smaller case is better.
>

A smaller case is not always cheaper. Smaller means the prodcution
equipment has to be of better precision (more expensive), and the
Pick&Place machine also is more expensive

For example we have looked at the resistor arrays with 4x0603 in one
1206 pack, but our production cant solder that reliable. The placement
error rate rises dramatically.

Also using fine pitch devices (TSSOP or smaller), there is an increased
risks of solder bridges.

Regards

Klaus

Jonathan Kirwan wrote:
> On 29 Dec 2006 06:41:28 -0800, "rickman" <gnuarm@gmail.com> wrote:
> >I don't see where the CCP and the 32kHz are mutually exclusive.  I
> >think you may be mistaken on this.
>
> I was looking at this site:
>
> http://www.luminarymicro.com/products/product_selector_guide.html
>
> If you scan the list, on the left side you will see this description,
> "External 32 KHz Clock (uses CCP pin)"  This is where I got it.

You might try reading the data sheet.  I am no expert on this part, but
when I look at the pin lists and the block diagrams, I don't see the
CCP and 32 kHz clock being mutually exclusive.


> >I also don't get your aversion to the 48 pin TQFP package.  They are
> >not hard to solder and they are a lot smaller than the 28 pin package
> >you seem to think is ok.  The lead pitch is finer, but I have not found
> >many applications where 20 mil lead pitch is a problem.
>
> The 8-MSOP mentioned isn't quite .5mm spacing, but close at .65mm. For
> larger packages, say 16-pin, they are also .65mm spacing and 4mm on a
> side.
>
> I take clear note of the fact that you seem to be arguing (separately
> from Henry) that no one should have any problem with 48-pin packages
> and that anyone suggesting fewer pins for their application is simply
> wrong about it.  If that is your position in some attempt to defend
> Henry's point, "the ARM ... will completely replace (almost) all
> 8051," I can't agree with you or Henry about this.

I am not trying to argue anything.  I am suggesting that you have not
explained the issue with using a 48 pin TQFP in place of a 28 pin SOIC.
 I read in another post where someone had trouble assembling 1206
multicaps.  I don't know the resolution required for this particular
part, but these days an assembly house that can't mount a 0.5 mm pitch
part is getting pretty long in the tooth.  I understand the issues of
amortizing equipment, but at some point the incremental cost of staying
current (or at least within 10 years of current) gets pretty low.

But if you are locked into using assembly houses that are this out of
date, then those are your requirements.


> >If you look in more detail at the Philips line you will find any number
> >of devices with a single supply.
>
> I will do that, then.  But since I was already looking at devices with
> more flash and more ram than I want, and they were the smallest
> devices I could easily find, my suspicion is that finding single
> supply devices will push me up the curve a bit into territory even
> further away from where I'm usually at.  But I'll look more.

The Philips line is a bit convoluted.  They have a number of single
supply parts and a number of dual supply parts and never the twain
shall meet.  By that I mean, unlike the Atmel parts that let you go
either way, the single supply parts from Philips do not allow you to
separately power the core if you want.  There is also no coorellation
to the other features on the chips.

There is a site with a comparison chart of various ARM7 devices.  You
can get a quick view of what is available.  Go to www.gnuarm.com and
click to the Resources page.  Then scroll down to the section and click
on the ARM Device Comparison chart.


> >All of the Atmel parts are single
> >supply and they have a new AT91SAM7S16(1) coming out in the 48 pin
> >packages (both TQFP and QFN).
>
> I know this, at one time, wasn't true.  Because I actually went
> through Atmel's list (as I know you have, too) and found dual supply
> ARM devices (from their France operation.)  Perhaps that has changed,
> but I doubt it.  Are you certain that _all_ Atmel ARM devices are
> _not_ dual supply?

Sorry, not all ARM devices, all SAM7 devices.  They can be used either
way.  They bring the LDO output to a pin which you connect to the Vcore
pin or you can power Vcore separately.  The internal LDO has two modes
to save on quiescent current when you have throttled back the CPU.

Atmel makes other, older ARM devices that are pretty universally larger
and/or more power hungry.


> Not that it will matter to me.  For now, I won't spec them.
>
> >It should be around $2 in quantity.
>
> As I said, I won't spec them right now.  But in the 2k-5k area?
>
> >The
> >lowest price I have seen from Atmel is $3.35 on one of the two SAM7S32
> >flavors.
>
> Oh, I must have misunderstood your context then about the $2 figure.
> Were you talking about Atmel, then?

I don't have a quote on the SAM7S16(1) parts, I am speculating that
given the pricing on the other parts, the new 16 KB part should be
around $2.  It is due out in the summer.

Klaus Kragelund wrote:

> Henry Kiefer skrev:
> 
>>"Jonathan Kirwan" <jkirwan@easystreet.com> schrieb im Newsbeitrag
>>news:bsp9p2p2237c5kfibb6msfn83nuncnqeno@4ax.com...
>>
>>>On Fri, 29 Dec 2006 10:31:38 +0100, "Henry Kiefer"
>>><otc_friend@gmx.net> wrote:
>>>
>>>
>>>>Your' probably total wrong. The first 8051 had 40 pins. Have you ever
>>>>thought that this guy will someday have 8 pins?
>>>
>>>I predate the 8051 in embedded work.  So yes, I remember those days
>>>and before and, no, I didn't imagine sot23-6 for PICs or 8051's back
>>>then.
>>>
>>>We shall see.  Situation is substantially different than then, in
>>>diversity of options, as well as application spaces.  So I've no idea
>>>why you feel that historical period should predict a new one in an
>>>vastly different context.  On the other hand, given the length of time
>>>since that time, I'd be willing to imagine some nice -6 and -8 parts
>>>for ARM.  But by then, I can also imagine so many other changes.
>>>
>>>For now, I'll stay with my suspicions that folks manufacturing ARMs
>>>don't see 8- and 16-bit register application spaces with small numbers
>>>of highly-repurposable I/O pins as their near-term future.  And I'll
>>>be _very_ glad to be wrong about it!
>>
>>Once upon time the functional chip itself was big and the border around was
>>filled with i/o pads. Process shrunked a lot and now the core is very small
>>and there is a problem how to effective add the pads around if the inner
>>area is so small. That can leave chip area unused and that is very costly
>>and so prohibited.
>>
>>So the natural way is that newer chips CAN have less pin-count (That means a
>>smaller case too). Probably newer ones will go more in the direction of
>>having more complex high-speed serial i/o.
>>The other point is that casing is costly (production money and space), so a
>>smaller case is better.
>>
> 
> 
> A smaller case is not always cheaper. Smaller means the prodcution
> equipment has to be of better precision (more expensive), and the
> Pick&Place machine also is more expensive
> 
> For example we have looked at the resistor arrays with 4x0603 in one
> 1206 pack, but our production cant solder that reliable. The placement
> error rate rises dramatically.
> 

That looks like a pricey boutique part. Interesting. Anyhow, I'd make 
sure your production can at least handle regular 0402 parts. That is 
pretty much state of the art these days. I just debugged a design I had 
completed today and it had some of those teeny parts. Luckily there 
wasn't much debugging needed. My eyes don't get any younger ;-)


> Also using fine pitch devices (TSSOP or smaller), there is an increased
> risks of solder bridges.
> 

There I'd really sit down for a chat with the production manager. TSSOP 
is absolutely standard. There are many parts that you simply cannot 
obtain in any package larger than TSSOP.

BTW, none of my current clients assembles circuit boards in house 
anymore. Some used to do that but, as you mentioned above, the 
investment in equipment can become prohibitive if you don't have enough 
volume to operate most of the machines around the clock minus regular PM 
intervals. Not all of them outsource to Asia, some have it produced here 
in the US but at companies that are highly specialized in SMT assembly.

Layout plays a big role which is why I am using a layouter with about 
three decades under the belt. It is important to talk about production 
capabilities with the factory before beginning the layout. In his case 
that's often not needed because he already knows. The footprint 
suggested by a part manufacturer is not always the cat's meouw and only 
experienced layouters or SMT production folks can really tell you how to 
optimize.

-- 
Regards, Joerg

http://www.analogconsultants.com

Joerg skrev:
> Klaus Kragelund wrote:
>
> > Henry Kiefer skrev:
> >
> >>"Jonathan Kirwan" <jkirwan@easystreet.com> schrieb im Newsbeitrag
> >>news:bsp9p2p2237c5kfibb6msfn83nuncnqeno@4ax.com...
> >>
> >>>On Fri, 29 Dec 2006 10:31:38 +0100, "Henry Kiefer"
> >>><otc_friend@gmx.net> wrote:
> >>>
> >>>
> >>>>Your' probably total wrong. The first 8051 had 40 pins. Have you ever
> >>>>thought that this guy will someday have 8 pins?
> >>>
> >>>I predate the 8051 in embedded work.  So yes, I remember those days
> >>>and before and, no, I didn't imagine sot23-6 for PICs or 8051's back
> >>>then.
> >>>
> >>>We shall see.  Situation is substantially different than then, in
> >>>diversity of options, as well as application spaces.  So I've no idea
> >>>why you feel that historical period should predict a new one in an
> >>>vastly different context.  On the other hand, given the length of time
> >>>since that time, I'd be willing to imagine some nice -6 and -8 parts
> >>>for ARM.  But by then, I can also imagine so many other changes.
> >>>
> >>>For now, I'll stay with my suspicions that folks manufacturing ARMs
> >>>don't see 8- and 16-bit register application spaces with small numbers
> >>>of highly-repurposable I/O pins as their near-term future.  And I'll
> >>>be _very_ glad to be wrong about it!
> >>
> >>Once upon time the functional chip itself was big and the border around was
> >>filled with i/o pads. Process shrunked a lot and now the core is very small
> >>and there is a problem how to effective add the pads around if the inner
> >>area is so small. That can leave chip area unused and that is very costly
> >>and so prohibited.
> >>
> >>So the natural way is that newer chips CAN have less pin-count (That means a
> >>smaller case too). Probably newer ones will go more in the direction of
> >>having more complex high-speed serial i/o.
> >>The other point is that casing is costly (production money and space), so a
> >>smaller case is better.
> >>
> >
> >
> > A smaller case is not always cheaper. Smaller means the prodcution
> > equipment has to be of better precision (more expensive), and the
> > Pick&Place machine also is more expensive
> >
> > For example we have looked at the resistor arrays with 4x0603 in one
> > 1206 pack, but our production cant solder that reliable. The placement
> > error rate rises dramatically.
> >
>
> That looks like a pricey boutique part. Interesting. Anyhow, I'd make
> sure your production can at least handle regular 0402 parts. That is
> pretty much state of the art these days. I just debugged a design I had
> completed today and it had some of those teeny parts. Luckily there
> wasn't much debugging needed. My eyes don't get any younger ;-)
>

Mine doesn't either (we are prototyping using 0805 resistors since any
smaller is a pain to debug/solder out/in)

Regarding the 4x array, it is fairly cheap (about 15 times a single
0805, in the 5% version about 3 times a 0805 resistor). So a quad
respack compares to a 0805 if you can live with the 5% tolerance (we
often can, since only a portion of a design needs precision resistors)

But here is the point, since our placement costs are high, using a
resistor array means ony one placement instead of 4 placements and thus
the total price is much lower. We cannot use them however since the
production cannot solder them reliable. We have used the standard
layout from the manufactor and enhanced it further but they could not
get the failure rate down :-( We can however handle 0402 parts reliably

Another really nice part is the dual BJT: BC847BPN - you should check
it out (it's cheap also)

>
> > Also using fine pitch devices (TSSOP or smaller), there is an increased
> > risks of solder bridges.
> >
>
> There I'd really sit down for a chat with the production manager. TSSOP
> is absolutely standard. There are many parts that you simply cannot
> obtain in any package larger than TSSOP.
>

Good point. I have tried for the last year, but my suspicion is that
the production guys simply is overbooked with work and also that they
lack the experience and drive to improve the failure rates of special
components. Its quite sad to try to optimize a design and being told
the production cannot use the suggested parts.

I guess another solution would be to find an expert consultant, that
can bring us further

> BTW, none of my current clients assembles circuit boards in house
> anymore. Some used to do that but, as you mentioned above, the
> investment in equipment can become prohibitive if you don't have enough
> volume to operate most of the machines around the clock minus regular PM
> intervals. Not all of them outsource to Asia, some have it produced here
> in the US but at companies that are highly specialized in SMT assembly.

We have a pretty high volume (more than 100mill SMD part mounted per
year), but prefer to mount only SMD inhouse and take the
labor-extensive work abroad

Regards

Klaus

Klaus Kragelund wrote:


[...]
>>>
>>>A smaller case is not always cheaper. Smaller means the prodcution
>>>equipment has to be of better precision (more expensive), and the
>>>Pick&Place machine also is more expensive
>>>
>>>For example we have looked at the resistor arrays with 4x0603 in one
>>>1206 pack, but our production cant solder that reliable. The placement
>>>error rate rises dramatically.
>>>
>>
>>That looks like a pricey boutique part. Interesting. Anyhow, I'd make
>>sure your production can at least handle regular 0402 parts. That is
>>pretty much state of the art these days. I just debugged a design I had
>>completed today and it had some of those teeny parts. Luckily there
>>wasn't much debugging needed. My eyes don't get any younger ;-)
>>
> 
> 
> Mine doesn't either (we are prototyping using 0805 resistors since any
> smaller is a pain to debug/solder out/in)
> 

On one of my latest designs I had a few 0805 and some SOT23. Under the 
microscope they looked like huge boulders compared to all the rest.


> Regarding the 4x array, it is fairly cheap (about 15 times a single
> 0805, in the 5% version about 3 times a 0805 resistor). So a quad
> respack compares to a 0805 if you can live with the 5% tolerance (we
> often can, since only a portion of a design needs precision resistors)
> 
> But here is the point, since our placement costs are high, using a
> resistor array means ony one placement instead of 4 placements and thus
> the total price is much lower. We cannot use them however since the
> production cannot solder them reliable. We have used the standard
> layout from the manufactor and enhanced it further but they could not
> get the failure rate down :-( We can however handle 0402 parts reliably
> 

I guess you really have to work on it to get competitive placement costs.

Check out these guys for stuff you don't want to move outside Europe:
http://www.jabil.com/
Don't know if they are also in Denmark since their locations page 
requires some stupid flash player. Web site designers are .... no, no, 
just bit my lips, I am not going to say that.


> Another really nice part is the dual BJT: BC847BPN - you should check
> it out (it's cheap also)
> 

Yes, that one is nice. Not exactly cheap, between 3-4c in the US.

> 
>>>Also using fine pitch devices (TSSOP or smaller), there is an increased
>>>risks of solder bridges.
>>>
>>
>>There I'd really sit down for a chat with the production manager. TSSOP
>>is absolutely standard. There are many parts that you simply cannot
>>obtain in any package larger than TSSOP.
>>
> 
> 
> Good point. I have tried for the last year, but my suspicion is that
> the production guys simply is overbooked with work and also that they
> lack the experience and drive to improve the failure rates of special
> components. Its quite sad to try to optimize a design and being told
> the production cannot use the suggested parts.
> 
> I guess another solution would be to find an expert consultant, that
> can bring us further
> 

If you can't get anywhere right now then a consultant might really be 
needed. I believe that either your own production needs to buy into the 
future or you need to outsource it.

When it comes to the "convincement meeting" which usually has to include 
your corporate top brass a very powerful method of convincing most 
everyone is this: Take a modern miniaturized appliance, for example a 
cell phone, an iPod, a kids toy. Something that everyone knows. Pry the 
circuit board out and pass that around. Or, what we often do, take a 
photo with a small coin next it and show it on the projector. Then ask 
the magic question "This is the future. If we don't get there the 
competition will have us for lunch. So how do we get there?"

> 
>>BTW, none of my current clients assembles circuit boards in house
>>anymore. Some used to do that but, as you mentioned above, the
>>investment in equipment can become prohibitive if you don't have enough
>>volume to operate most of the machines around the clock minus regular PM
>>intervals. Not all of them outsource to Asia, some have it produced here
>>in the US but at companies that are highly specialized in SMT assembly.
> 
> 
> We have a pretty high volume (more than 100mill SMD part mounted per
> year), but prefer to mount only SMD inhouse and take the
> labor-extensive work abroad
> 

Well, if the inhouse guys don't have the tools for fine-pitch they need 
to get those machines then. If they don't want to, what choice other 
than outsourcing do you really have? Or to say it more bluntly, building 
only tractors because the folks can only build big stuff isn't going to 
fly in the long run. The US auto makers have taken that road to some 
degree. Look where that got them. Every time a new environmental law 
comes out they whine and complain while their Japanese colleagues roll 
up their sleeves and get to work.

-- 
Regards, Joerg

http://www.analogconsultants.com

Joerg skrev:
> Klaus Kragelund wrote:
>
>
> [...]
> >>>
> >>>A smaller case is not always cheaper. Smaller means the prodcution
> >>>equipment has to be of better precision (more expensive), and the
> >>>Pick&Place machine also is more expensive
> >>>
> >>>For example we have looked at the resistor arrays with 4x0603 in one
> >>>1206 pack, but our production cant solder that reliable. The placement
> >>>error rate rises dramatically.
> >>>
> >>
> >>That looks like a pricey boutique part. Interesting. Anyhow, I'd make
> >>sure your production can at least handle regular 0402 parts. That is
> >>pretty much state of the art these days. I just debugged a design I had
> >>completed today and it had some of those teeny parts. Luckily there
> >>wasn't much debugging needed. My eyes don't get any younger ;-)
> >>
> >
> >
> > Mine doesn't either (we are prototyping using 0805 resistors since any
> > smaller is a pain to debug/solder out/in)
> >
>
> On one of my latest designs I had a few 0805 and some SOT23. Under the
> microscope they looked like huge boulders compared to all the rest.
>

I have looked at some of the smaller packages, but I often see higher
prices for the small stuff. For example:

BC847 (SOT-23): 0.17USD
BC846BM3T5G (SOT-723): 0.37USD

Both prices are from mouser just to compare them fairly. So I think in
some cases a larger design is actually cheaper since the parts run in
high volume from the manufactor and it is a well known production
process

>
> > Regarding the 4x array, it is fairly cheap (about 15 times a single
> > 0805, in the 5% version about 3 times a 0805 resistor). So a quad
> > respack compares to a 0805 if you can live with the 5% tolerance (we
> > often can, since only a portion of a design needs precision resistors)
> >
> > But here is the point, since our placement costs are high, using a
> > resistor array means ony one placement instead of 4 placements and thus
> > the total price is much lower. We cannot use them however since the
> > production cannot solder them reliable. We have used the standard
> > layout from the manufactor and enhanced it further but they could not
> > get the failure rate down :-( We can however handle 0402 parts reliably
> >
>
> I guess you really have to work on it to get competitive placement costs.
>
> Check out these guys for stuff you don't want to move outside Europe:
> http://www.jabil.com/
> Don't know if they are also in Denmark since their locations page
> requires some stupid flash player. Web site designers are .... no, no,
> just bit my lips, I am not going to say that.
>

http://www.jabil.com/ looks very interesting - i will try to get a BOM
they can qoute (will be very interesting)

>
> > Another really nice part is the dual BJT: BC847BPN - you should check
> > it out (it's cheap also)
> >
>
> Yes, that one is nice. Not exactly cheap, between 3-4c in the US.
>

The same here - for comparison we give about 1c for a standard BC847B
(high volume)

> >
> >>>Also using fine pitch devices (TSSOP or smaller), there is an increased
> >>>risks of solder bridges.
> >>>
> >>
> >>There I'd really sit down for a chat with the production manager. TSSOP
> >>is absolutely standard. There are many parts that you simply cannot
> >>obtain in any package larger than TSSOP.
> >>

Yes - it seems its time to dig deeper into the production problems

> >
> >
> > Good point. I have tried for the last year, but my suspicion is that
> > the production guys simply is overbooked with work and also that they
> > lack the experience and drive to improve the failure rates of special
> > components. Its quite sad to try to optimize a design and being told
> > the production cannot use the suggested parts.
> >
> > I guess another solution would be to find an expert consultant, that
> > can bring us further
> >
>
> If you can't get anywhere right now then a consultant might really be
> needed. I believe that either your own production needs to buy into the
> future or you need to outsource it.
>
> When it comes to the "convincement meeting" which usually has to include
> your corporate top brass a very powerful method of convincing most
> everyone is this: Take a modern miniaturized appliance, for example a
> cell phone, an iPod, a kids toy. Something that everyone knows. Pry the
> circuit board out and pass that around. Or, what we often do, take a
> photo with a small coin next it and show it on the projector. Then ask
> the magic question "This is the future. If we don't get there the
> competition will have us for lunch. So how do we get there?"
>

Good recommendation. Actually we often buy pumps from Asia competitors
to look how they can do it so cheap. Often they still use leaded
assembly and cheap FR4 boards (so our quality guys just say they can
sell it so cheap because the quality is bad)

An iPod would be a good choice since it has to have long durability,
but ofcourse it is expensive

Thanks

Klaus

Klaus Kragelund wrote:

> Joerg skrev:
> 
>>Klaus Kragelund wrote:
>>
>>
>>[...]
>>
>>>>>A smaller case is not always cheaper. Smaller means the prodcution
>>>>>equipment has to be of better precision (more expensive), and the
>>>>>Pick&Place machine also is more expensive
>>>>>
>>>>>For example we have looked at the resistor arrays with 4x0603 in one
>>>>>1206 pack, but our production cant solder that reliable. The placement
>>>>>error rate rises dramatically.
>>>>>
>>>>
>>>>That looks like a pricey boutique part. Interesting. Anyhow, I'd make
>>>>sure your production can at least handle regular 0402 parts. That is
>>>>pretty much state of the art these days. I just debugged a design I had
>>>>completed today and it had some of those teeny parts. Luckily there
>>>>wasn't much debugging needed. My eyes don't get any younger ;-)
>>>>
>>>
>>>
>>>Mine doesn't either (we are prototyping using 0805 resistors since any
>>>smaller is a pain to debug/solder out/in)
>>>
>>
>>On one of my latest designs I had a few 0805 and some SOT23. Under the
>>microscope they looked like huge boulders compared to all the rest.
>>
> 
> 
> I have looked at some of the smaller packages, but I often see higher
> prices for the small stuff. For example:
> 
> BC847 (SOT-23): 0.17USD
> BC846BM3T5G (SOT-723): 0.37USD
> 
> Both prices are from mouser just to compare them fairly. So I think in
> some cases a larger design is actually cheaper since the parts run in
> high volume from the manufactor and it is a well known production
> process
> 

That's if you buy one or two. Else the SOT23 is under $0.02 at Mouser:
http://www.mouser.com/search/ProductDetail.aspx?R=BC847AMTFvirtualkey51210000virtualkey512-BC847AMTF

Never seen SOT723 though.

> 
>>>Regarding the 4x array, it is fairly cheap (about 15 times a single
>>>0805, in the 5% version about 3 times a 0805 resistor). So a quad
>>>respack compares to a 0805 if you can live with the 5% tolerance (we
>>>often can, since only a portion of a design needs precision resistors)
>>>
>>>But here is the point, since our placement costs are high, using a
>>>resistor array means ony one placement instead of 4 placements and thus
>>>the total price is much lower. We cannot use them however since the
>>>production cannot solder them reliable. We have used the standard
>>>layout from the manufactor and enhanced it further but they could not
>>>get the failure rate down :-( We can however handle 0402 parts reliably
>>>
>>
>>I guess you really have to work on it to get competitive placement costs.
>>
>>Check out these guys for stuff you don't want to move outside Europe:
>>http://www.jabil.com/
>>Don't know if they are also in Denmark since their locations page
>>requires some stupid flash player. Web site designers are .... no, no,
>>just bit my lips, I am not going to say that.
>>
> 
> 
> http://www.jabil.com/ looks very interesting - i will try to get a BOM
> they can qoute (will be very interesting)
> 
> 
>>>Another really nice part is the dual BJT: BC847BPN - you should check
>>>it out (it's cheap also)
>>>
>>
>>Yes, that one is nice. Not exactly cheap, between 3-4c in the US.
>>
> 
> 
> The same here - for comparison we give about 1c for a standard BC847B
> (high volume)
> 
> 
>>>>>Also using fine pitch devices (TSSOP or smaller), there is an increased
>>>>>risks of solder bridges.
>>>>>
>>>>
>>>>There I'd really sit down for a chat with the production manager. TSSOP
>>>>is absolutely standard. There are many parts that you simply cannot
>>>>obtain in any package larger than TSSOP.
>>>>
> 
> 
> Yes - it seems its time to dig deeper into the production problems
> 

This could be a difficult mission. Be prepared for some stone-walling 
and excuses that are brought onto the table. You'll need upper 
management participation and support as well as iron-clad quotes from 
outside assemblers for existing products of your company. Then you'll 
have hard data along the lines of "It costs this much to assemble here 
and XYZ Corporation would charge us only that much to do it outside".
> 
>>>
>>>Good point. I have tried for the last year, but my suspicion is that
>>>the production guys simply is overbooked with work and also that they
>>>lack the experience and drive to improve the failure rates of special
>>>components. Its quite sad to try to optimize a design and being told
>>>the production cannot use the suggested parts.
>>>
>>>I guess another solution would be to find an expert consultant, that
>>>can bring us further
>>>
>>
>>If you can't get anywhere right now then a consultant might really be
>>needed. I believe that either your own production needs to buy into the
>>future or you need to outsource it.
>>
>>When it comes to the "convincement meeting" which usually has to include
>>your corporate top brass a very powerful method of convincing most
>>everyone is this: Take a modern miniaturized appliance, for example a
>>cell phone, an iPod, a kids toy. Something that everyone knows. Pry the
>>circuit board out and pass that around. Or, what we often do, take a
>>photo with a small coin next it and show it on the projector. Then ask
>>the magic question "This is the future. If we don't get there the
>>competition will have us for lunch. So how do we get there?"
>>
> 
> 
> Good recommendation. Actually we often buy pumps from Asia competitors
> to look how they can do it so cheap. Often they still use leaded
> assembly and cheap FR4 boards (so our quality guys just say they can
> sell it so cheap because the quality is bad)
> 

Why is FR-4 so bad? Believe it or not but some of my designs were even 
run on good old phenolic. The stuff still lasts decades. There has to be 
a very compelling reason to go beyond FR-4, even for RF designs. Maybe 
if your pump controllers have to work in outer space or something like 
that :-)

As for RoHS, well, don't get me started. That was one of the more stupid 
decisions by Eurocrats but that's a whole 'nother topic. The main goal 
should probably be to push for an exemption.


> An iPod would be a good choice since it has to have long durability,
> but ofcourse it is expensive
> 

Or maybe they added a humongous profit margin. Kids pay whatever it 
takes to be at least as hip as their class mates and the industry knows 
that well. Since you are in Scandinavia you might pick a more local 
example that is not expensive. I have a Nokia 2115i cell phone which 
should definitely contain lots of really tiny parts. Still it cost me 
only about $10 with a $40 usage commitment. Without a long-term plan so 
besides maybe half of those $40 there cannot be any subsidies or loss 
tendering in it.

-- 
Regards, Joerg

http://www.analogconsultants.com