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

[Henry Kiefer]

| Henry Kiefer wrote:
|
| > | "Here, this has 65 parts and costs $2.50 including PCB and final test.
| > | Can you beat that?"
| >
| > parts and test:
| > That would be impossible to make in Germany or even the EU. So here it would be cheaper to use a costly PSoC. OK, in China they
do
| > it.
| >
|
| Yes, it's Chinese production. But who would not produce it there? Thing
| is, if you come up with a more integrated solution at $4 and the
| competitior does it all in discrete for $2.50 they'll have you for lunch.

Yes :-(

I don't like my job ;-(

- Henry

 

[Joerg]

| Henry Kiefer wrote:
|
| > | "Here, this has 65 parts and costs $2.50 including PCB and final test.
| > | Can you beat that?"
| >
| > parts and test:
| > That would be impossible to make in Germany or even the EU. So here it would be cheaper to use a costly PSoC. OK, in China they
do
| > it.
| >
|
| Yes, it's Chinese production. But who would not produce it there? Thing
| is, if you come up with a more integrated solution at $4 and the
| competitior does it all in discrete for $2.50 they'll have you for lunch.

Yes :-(

I don't like my job ;-(

Actually, I like mine. I've always enjoyed to squeeze the last penny out
of any given design. It might sound silly to ponder the usage of a BJT
plus resistor versus a FET. Heck, we've even gone custom on an inductor
that was available off-the-shelf because we couldn't get the price down
far enough. When there is a $0.02 cost differential and they make
thousands of units per week even a whole hour spent on matters like that
pays back quickly for the client. It also keeps the skills honed.

However, sometimes it seems that larger semi mfgs don't quite understand
that level of cost pressure yet. IOW there really has to be a $0.25 PSoC
and a $0.15 16-bit micro.

 

[Joel Kolstad]

However, sometimes it seems that larger semi mfgs don't quite understand
that level of cost pressure yet. IOW there really has to be a $0.25 PSoC and
a $0.15 16-bit micro.

I suspect some of them are consciously *choosing* not to join the "race to the
bottom" -- for many semiconductor companies it's a lot easier to compete on
all tose "fluffly" things like customer service and support than simply price.
Many companies don't have someone like you around to squeeze out the last
penny, after all.

Of course there are some manufacturers out there who don't seem to be good on
price or relationships, e.g., those with really poor web sites!

 

[Joerg]

I suspect some of them are consciously *choosing* not to join the "race to the
bottom" -- for many semiconductor companies it's a lot easier to compete on
all tose "fluffly" things like customer service and support than simply price.

That's a dangerous path to be on these days. You've got to keep a keen
eye on where that bottom is. Because most likely your competitor does.

Many companies don't have someone like you around to squeeze out the last
penny, after all.

Whenever I open a consumer device it seems that companies in Asia all do
))

I have become pretty hardened as to how far I'll go to squeeze costs but
sometimes they still blow me away. Like when I saw a design where they
used a substrate diode as a rectifier and (most probably) a nifty dummy
load scheme in firmware to keep Icc somewhat even so they could also get
away without a regulator. The way I found out was by accidentally
stopping the clock: KABOOM.

Of course there are some manufacturers out there who don't seem to be good on
price or relationships, e.g., those with really poor web sites!

Yeah, and Cypress isn't all that stellar with their web site. WRT
support we shall see. That was one of the issues that has caused the
loss of several design-ins for TI in my case. They wouldn't answer a few
crucial questions that went beyond data sheet info. So I did it like
usual, all analog.

 

[werty]

I am EE , i do signal processing . Proportional-Integral-Derivative
is for college teachers , not real engineers like us who
"do" / invent / create / fix

We engineers dont use text to create machines , we use
the picture side of our brain . So after trying unsucessfully
to figure what the Luddite Proff is saying , we toss it
and put our hands on the hardware and learn
Electronics ( Me , 70's , ) .

You are just using buzz words , hoping for a job
laying out mo buzz words .

You dont work ....

Now on to rpoductive ...

ARM 9 is so low cost , it will be the only mcu .
I am now redefinig ASCII as a temp way for computer
to talk to us .
Ojective is to keep each Ascii char general , use more
to make it specific . This allows the mcu to use the chars
on the left as a classing/typing method .

Dont use < to mean GT , only , it limits its use for more powerful
contexts .
Add char's to the Right, to adds classes to that object .
Now a user, at keyboard will be able to use these new
meanings to quickly put the mcu right .
imagine a PC , misspells a name , and you
----- < R N ------------------
< expanded the hi-lited letters displayed on LCD .
R moved cursor right one alpha ,
N generally means no, undo , reverse ,backup , BS , last MRU

PC swaps these chars , for chars from MRU edit strings place ....

This is old stuff , not copyrighted , u can use it ..

I am the fastest systems programmer on Earth ....

 

[Joerg]

I am EE , i do signal processing . Proportional-Integral-Derivative
is for college teachers , not real engineers like us who
"do" / invent / create / fix

Really? Then how do you propose to control a slow plant? And here I mean
a plant where you do not know its inherent (and usually thermal) delays.

We engineers dont use text to create machines , we use
the picture side of our brain . So after trying unsucessfully
to figure what the Luddite Proff is saying , we toss it
and put our hands on the hardware and learn
Electronics ( Me , 70's , ) .

You are just using buzz words , hoping for a job
laying out mo buzz words .

You dont work ....

As it stands I've got mo work than I'd like.

Now on to rpoductive ...

ARM 9 is so low cost , it will be the only mcu .

Sorry to disagree on that one. Might change my mind some day but not
right now

Ever seen an ARM that can live on a few microamps while running on a
watch crystal? It's been a while but most of those that I checked out
needed multiple supplies that had to come up in a very particular
fashion and they were quite the power hogs compared to modern
16-bitters. While 32 bits are nice they aren't always needed.

I am now redefinig ASCII as a temp way for computer
to talk to us .
Ojective is to keep each Ascii char general , use more
to make it specific . This allows the mcu to use the chars
on the left as a classing/typing method .

Dont use < to mean GT , only , it limits its use for more powerful
contexts .
Add char's to the Right, to adds classes to that object .
Now a user, at keyboard will be able to use these new
meanings to quickly put the mcu right .
imagine a PC , misspells a name , and you
----- < R N ------------------
< expanded the hi-lited letters displayed on LCD .
R moved cursor right one alpha ,
N generally means no, undo , reverse ,backup , BS , last MRU

PC swaps these chars , for chars from MRU edit strings place ....

This is old stuff , not copyrighted , u can use it ..

I am the fastest systems programmer on Earth ....

And here I was hoping you'd suggest a slick way to do a PID job in a
"non-luddite" way...

 

[Klaus Kragelund]

Joerg skrev:

Yes, I realized some of that during the seminar. For example, if you
want an ADC that has to run at more than 100Hz you don't get many bit
with PSoC. But you can hook a PGA up front. If this is in a regulator
loop you can then re-program the gain on the fly and effectively put in
a vernier drive.

Actually the opamps are fast for the price (5MHz GBW) and the ADC runs
at >5kHz with 8bit. (2kHz 10bit)

The next gen PSOC will have ADC that is a lot faster

Regards

Klaus

 

[Joerg]

Joerg skrev:



Actually the opamps are fast for the price (5MHz GBW) and the ADC runs
at >5kHz with 8bit. (2kHz 10bit)

The GBW is ok. The offset isn't much to write home about but I can
usually live with that. WRT the ADC capabilities the engineer from
Cypress was honest with us (which I very much appreciated) and told us
not to expect more than a few ten "real" Hertz out of the slope
converter variant.

Most of the time I can make do with lots of PWM-DACs plus comparators.
That should work nicely with PSoC.

The next gen PSOC will have ADC that is a lot faster

Yes, but I am also hoping that they work on extending the lower tier
segment. That's where I'd have potential applications but not if they
cost $1 or more.

 

[Henry Kiefer]

| > Actually the opamps are fast for the price (5MHz GBW) and the ADC runs
| > at >5kHz with 8bit. (2kHz 10bit)
| >
|
| The GBW is ok. The offset isn't much to write home about but I can
| usually live with that. WRT the ADC capabilities the engineer from
| Cypress was honest with us (which I very much appreciated) and told us
| not to expect more than a few ten "real" Hertz out of the slope
| converter variant.
|
| Most of the time I can make do with lots of PWM-DACs plus comparators.
| That should work nicely with PSoC.

I like the possibility to select the right ADC on the same silicone. Even it is late and you need to change it entirely in software.
Or left it out if I don't need it and use the spare module for other purposes.

The flexibility have it's price of course.

- Henry

 

[Joerg]

| > Actually the opamps are fast for the price (5MHz GBW) and the ADC runs
| > at >5kHz with 8bit. (2kHz 10bit)
| >
|
| The GBW is ok. The offset isn't much to write home about but I can
| usually live with that. WRT the ADC capabilities the engineer from
| Cypress was honest with us (which I very much appreciated) and told us
| not to expect more than a few ten "real" Hertz out of the slope
| converter variant.
|
| Most of the time I can make do with lots of PWM-DACs plus comparators.
| That should work nicely with PSoC.

I like the possibility to select the right ADC on the same silicone. Even it is late and you need to change it entirely in software.
Or left it out if I don't need it and use the spare module for other purposes.

The flexibility have it's price of course.

Just an hour ago I happened to get a request for a new design that might
be a near perfect fit for PSoC. One of those where the specs are not yet
cast in stone (the usual...).

So I am poring over the CY8C21x23 data sheet, trying to figure out what
this charge pump is all about or how I can "divert" its usage, how to
get these things out of sleep mode and so on. It'll be interesting.

 

[Robert Latest]

["Followup-To:" header set to sci.electronics.design.]

Actually, I like mine. I've always enjoyed to squeeze the last penny out
of any given design. It might sound silly to ponder the usage of a BJT
plus resistor versus a FET. Heck, we've even gone custom on an inductor
that was available off-the-shelf because we couldn't get the price down
far enough. When there is a $0.02 cost differential and they make
thousands of units per week even a whole hour spent on matters like that
pays back quickly for the client. It also keeps the skills honed.

I've always noted that priority in your posts. However, whenever
you mentioned something definite about your (possibly former)
field of work it was about "medical ultrasound", hardly a 1000/wk
item.

So what's all this cheap analog consumer crap you seem to be
working on?

Just curious:

robert

 

[Klaus Kragelund]

Joerg skrev:

The GBW is ok. The offset isn't much to write home about but I can
usually live with that. WRT the ADC capabilities the engineer from
Cypress was honest with us (which I very much appreciated) and told us
not to expect more than a few ten "real" Hertz out of the slope
converter variant.

The offset can be nullified easily. Just use a configuration with a DAC
to trim the offset away. Then you have zero drift over temperature,
time and other parameters

Most of the time I can make do with lots of PWM-DACs plus comparators.
That should work nicely with PSoC.

Yes, they are fast and low current (at least the comparators)

Regards

Klaus

 

[Klaus Kragelund]

Joerg skrev:

Just an hour ago I happened to get a request for a new design that might
be a near perfect fit for PSoC. One of those where the specs are not yet
cast in stone (the usual...).

So I am poring over the CY8C21x23 data sheet, trying to figure out what
this charge pump is all about or how I can "divert" its usage, how to
get these things out of sleep mode and so on. It'll be interesting.

There is no PGA, DAC or voltage reference in these parts. You can make
a simple DAC using a PWM output though


Regards

Klaus

 

[Henry Kiefer]

| Just an hour ago I happened to get a request for a new design that might
| be a near perfect fit for PSoC. One of those where the specs are not yet
| cast in stone (the usual...).
|
| So I am poring over the CY8C21x23 data sheet, trying to figure out what
| this charge pump is all about or how I can "divert" its usage, how to
| get these things out of sleep mode and so on. It'll be interesting.

Out of memory I would say you looked up the wrong type. It is not always best to just sort the variants in dollars.
OK, that is not much help.

Start with the 29466 or such to prototype. It is easy to replace if burned (socket!) and have rich resources to play. So much
resources that is possible to make target and test generator on the same part for development.

If you know later on what you really want and how to implement, then downgrade. The 24123 or such is a nice little 8-pin device and
fits the same socket as the 29466 if you can switch the VCC pin. There is an app note with all types listed!

One trick with PSoC is to choose a smaller Flash(ed) device just for speedy loading of code and then later replace the part with a
bigger Flash(ed) one if your code sizes...

The charge pump on the SMP pin is just a boost converter to run PSoC out of a small battery. As always: There is an app note ;-)

Sleep mode -> app note!

Read the PSoC Technical Manual.


- Henry

 

[Joerg]

Joerg skrev:



There is no PGA, DAC or voltage reference in these parts. You can make
a simple DAC using a PWM output though

I was just using it as an example, plus because they are in stock
locally. On this one I'll have to run a rather fast switcher, basically
like creating a 120V inverter with a true sine wave. It has to run above
100kHz because there isn't much space for magnetics. Have to talk to the
client tomorrow. If it's like usual where they need it "yesterday" I
might do it analog just one more time (and ask for forgiveness...).

 

[Joerg]

Joerg skrev:



The offset can be nullified easily. Just use a configuration with a DAC
to trim the offset away. Then you have zero drift over temperature,
time and other parameters

Either that or use a mux up front and ping-pong, then average in SW. In
the analog world I usually just clamp it away. That's why the sales guys
have such a hard time talking me into opamps that cost more than 10
cents ;-)

Yes, they are fast and low current (at least the comparators)

Also in standby. I was pleasantly surprised when I saw that PSoC can
live off a few uA. Not as good as an MSP430 but good enough even for a
coin cell.

 

[Joerg]

| Just an hour ago I happened to get a request for a new design that might
| be a near perfect fit for PSoC. One of those where the specs are not yet
| cast in stone (the usual...).
|
| So I am poring over the CY8C21x23 data sheet, trying to figure out what
| this charge pump is all about or how I can "divert" its usage, how to
| get these things out of sleep mode and so on. It'll be interesting.

Out of memory I would say you looked up the wrong type. It is not always best to just sort the variants in dollars.
OK, that is not much help.

Yes, it was just an example. Express won't let you select a type that's
too cramped.

Start with the 29466 or such to prototype. It is easy to replace if burned (socket!) and have rich resources to play. So much
resources that is possible to make target and test generator on the same part for development.

If you know later on what you really want and how to implement, then downgrade. The 24123 or such is a nice little 8-pin device and
fits the same socket as the 29466 if you can switch the VCC pin. There is an app note with all types listed!

I also would have to make sure that whatever I use is either small or
available in MLF or similar. Can't get any big fat chip into this design.

One trick with PSoC is to choose a smaller Flash(ed) device just for speedy loading of code and then later replace the part with a
bigger Flash(ed) one if your code sizes...

The charge pump on the SMP pin is just a boost converter to run PSoC out of a small battery. As always: There is an app note ;-)

Sleep mode -> app note!

Read the PSoC Technical Manual.

Yep, I will. Looks like quite a learning curve.

 

[Henry Kiefer]

| I was just using it as an example, plus because they are in stock
| locally. On this one I'll have to run a rather fast switcher, basically
| like creating a 120V inverter with a true sine wave. It has to run above
| 100kHz because there isn't much space for magnetics. Have to talk to the
| client tomorrow. If it's like usual where they need it "yesterday" I
| might do it analog just one more time (and ask for forgiveness...).

EL converter?
Hm.

- Henry

 

[Henry Kiefer]

| > Out of memory I would say you looked up the wrong type. It is not always best to just sort the variants in dollars.
| > OK, that is not much help.
| >
|
| Yes, it was just an example. Express won't let you select a type that's
| too cramped.

Oh sorry, I forgot you use Express. It don't come to mind how to select a device there. Sorry, I use Designer.


|
|
| > Start with the 29466 or such to prototype. It is easy to replace if burned (socket!) and have rich resources to play. So much
| > resources that is possible to make target and test generator on the same part for development.
| >
| > If you know later on what you really want and how to implement, then downgrade. The 24123 or such is a nice little 8-pin device
and
| > fits the same socket as the 29466 if you can switch the VCC pin. There is an app note with all types listed!
| >
|
| I also would have to make sure that whatever I use is either small or
| available in MLF or similar. Can't get any big fat chip into this design.

The app note lists all devices in nice table with package type.


- Henry

 

[Joerg]

| I was just using it as an example, plus because they are in stock
| locally. On this one I'll have to run a rather fast switcher, basically
| like creating a 120V inverter with a true sine wave. It has to run above
| 100kHz because there isn't much space for magnetics. Have to talk to the
| client tomorrow. If it's like usual where they need it "yesterday" I
| might do it analog just one more time (and ask for forgiveness...).

EL converter?
Hm.

No. But my lips must remain sealed for now. The whole thing is slow but
it has to run off a battery voltage that's only 1/8th of what the load
needs and in about the space of a slim USB stick.