Understanding PIC pricing

[luxfx]

I'm following several beginning PIC tutorials and have bought a couple each of the chips they refer to. I have a goal project in mind that I've selected a different chip for. I don't understand how they're pricing these though!

Comparison: PIC16F684, PIC16F1824, PIC16F18324 (PIC16F325 is the chip I've selected myself, but I'm using 18324 for comparison)

Data from MAPS --

684 (midrange): $1.19 for one, $1.03 each for 1000
3.5K flash, 128 bytes RAM, 256 bytes EEPROM, 20MHz, 2x8,1x16 bit timers

1824 (enhanced midrange): $1.14 for one, $0.94 each for 1000
7K flash, 256 RAM, 256 EEPROM, 32MHz, 4x8,1x16 bit timers

18324 (enhanced midrange): $0.89 for one, $0.68 each for 1000
7K flash, 512 RAM, 256 EEPROM, 32MHz, 4x8,3x16 bit timers

For three different chips all still in production, I would expect the more power and options, the higher the price. This is the opposite! I assume it has something to do with which are the new hotness models and which are kept around just for older designs. Perhaps there's also some price jacking in there to get more money out of supplying parts for older designs, leveraging production lines that are stuck with the older chips.

But if this is the case, are there any resources for listing these new hotness models to make it easier for a newb to select a chip - based on price/value rather than something like MAPS which is useful for spec parameters.

[jmsigler]

Without looking deeper into the datasheets, it is most likely based on supply/demand. If big electronics companies are using the chip, its logically going to be much cheaper than a counterpart with similar performance.

[Maxlor]

So many factors go into the pricing of these that it's essentially nondeterministic, particularly with something like 8-bit PICs where there are hundreds of different models. It might have something to do with die size, but it the price might just as well have been set to a certain value because of a competitor, or because a single client ordered 1 million of one chip but not the other, or whatever other reason.

But if you're a hobbyist, what do you care. Does it really matter whether the chip costs 1$, or slightly more, or slightly less?

[dannyf]

"would expect the more power and options, the higher the price. "

Newer chips tens to less (not more) expensive. Older chips require set up costs and inventorying or warehousing costs.

So always use a newer chip and don't carry inventory yourself.

[mikeselectricstuff]

Microchip tend to introduce new chips at lower prices, then increase the pricing on older parts, so they look competitive compared to others, but they make more on older parts designed into products where the customer hasn't updated to newer devices.

[JPortici]

Microchip almost NEVER phase out a product (basically it only happens to failed products or to acquisitions, parts that they didn't have direct control over the manufacturing.. yet still most of them should be available as a special order)
what it does is increase the price of older parts so you eventually upgrade to the plethora of pin to pin compatible chips, unless it is mandatory that you use the original one.

Also, newer chips usually come with newer manufactury process which means a smaller die -> way way way cheaper to produce

[MT]

MCU's based on ARM dont follow any market logic's at all.

[David Hess]

Microchip tend to introduce new chips at lower prices, then increase the pricing on older parts, so they look competitive compared to others, but they make more on older parts designed into products where the customer hasn't updated to newer devices.

If older parts are built on larger processes, then I would expect a price premium but Microchip is limited to processes which support floating gate MOSFETs for their Flash and EEPROM.  That rules out dense logic processes.

Other than that, I would expect parts with greater amounts of memory and large peripherals like USB to cost more.

[MarkF]

I suggest looking at the
pic18f2550
   32K program memory
   2048 SRAM memory
   48 MHz max clock
   USB capable
   28 Pin PDIP
or
pic18f2620
   64K program memory
   3936 SRAM memory
   40 MHz max clock
   28 Pin SPDIP

I find that it doesn't take long to fill up program memory.

[dannyf]

Microchip almost NEVER phase out a product

This horrible used car almost NEVER breaks down;
That terrible monster almost NEVER hurt you;
That sharp blade almost NEVER cut anyone;
....

The art of telling a lie without lying.

[luxfx]

Microchip almost NEVER phase out a product (basically it only happens to failed products or to acquisitions, parts that they didn't have direct control over the manufacturing.. yet still most of them should be available as a special order)
what it does is increase the price of older parts so you eventually upgrade to the plethora of pin to pin compatible chips, unless it is mandatory that you use the original one.

Also, newer chips usually come with newer manufactury process which means a smaller die -> way way way cheaper to produce

Thanks @JPortici and @mikeselectronicstuff, that's the kind of insight I was hoping to get. That kind of pressured obsolescence makes sense. How do I interpret the product numbers, or otherwise look up chips by how recent they were created? I'd love so see some kind of product map, like a family tree or something.

[luxfx]

I suggest looking at the
pic18f2550
or
pic18f2620

I find that it doesn't take long to fill up program memory.

Thanks for the recommendations, but $$YIKES$$   I'm looking at the $1.00 range for my goal project, and so far I'm pretty happy with the bang-for-the-buck I'm getting with the PIC16f18325. However, if I had some kind of product family tree where I could more easily see which are the more recent products with price cuts, that might get me to change my mind.

[Ian.M]

Microchip launched their first EPROM PIC, the PIC16C54 back in 1989, two years after they were spun off from G.I. Its still available in quantity, and not at a gold-plated legacy price.  It looks like its still in production:  http://www.microchipdirect.com/ProductDetails.aspx?Category=PIC16C54

Other microcontroller companies that can still supply their >25 year old product range, to original specs, in quantity are rare like hen's teeth.   If you are a small volume OEM and need design longevity, then  Microchip's track record  speaks for itself.

On the subject of which PICs to use - if you want a high performance  8 bit PIC and don't want USB, CAN or other highly specialised interfaces, look at the PIC18FxxK22 families.  They are loaded with peripherals, and give you more memory and performance for your buck than the PIC18Fxx20  devices in the same pinout.  Don't consider any PIC18 part ending '50 unless you are building a USB device.

OTOH the enhanced midrange PIC12/16F1xxx parts are Microchip's current budget medium performance 8 bit range.  The newest ones tend to have 5 digits after the F/LF. 

You can download Microchip's complete FLASH MCU product range as a spreadsheet here: http://www.microchip.com/ParamChartSearch/chart.aspx?branchID=1005.  Click 'Show ALL Products' then check 'Show All Specs' then download the data in CSV format.  It includes all the specs you can see in the product selector and indicative pricing, so you can sort and search by whatever criteria you desire.  Caution: there are some errors in the database so always check the datasheet and errata before ordering.

[JPortici]

How do I interpret the product numbers, or otherwise look up chips by how recent they were created?

I can't really help you with that.. from what i know
PIC16FXXX: the old boys like the 876A midrange core, i would avoid those (slow, poor on peripherals, "power hungry")
PIC16F1XXX: enchanced midrange. Some more instruction to make your life sligly easier, better peripherals overall. XLP and "L" Versions powered down to 1.8V in some cases.
of these i know this sub-families
PIC16F15XX: lower cost, but lesser features. be careful that some are without debugging on chip
PIC16F18XX: better featured version of the 15XX (more peripherals)
(these two also have equivalents in the PIC12 range, so 8 pin, basic core and absudly low price)
PIC16F17XX: Family aimed at intelligent power design with core independent peripherals (the hardware on board should be enough to make a SMPS controller without touching the registers, once set.. there is an application note on their pages)
PIC16F14XX: PIC16 with USB
then there are the newest of all, the PIC16F18XXX. For example the PIC16F18323 is a 16F1823 with a couple of things more and a lower cost

the other numbers will tell the pin count and the memory size (but there isn't a standard i know of, like in the pic18 and upwards)

for a pic18, unless you really need usb or other special features (for example the guy looking for 7 UARTs) i would always suggest the 26k22.

at work for 8bitters we use a ton of 18f26k22 and 16f1823 for smaller/simpler stuff

[mikeselectricstuff]

Microchip's parametric selector on their website has price, so select the features you want then sort by increasing price

[AndyC_772]

The process I follow to choose a PIC is:

1) Go to the MAPS page at http://www.microchip.com/maps/microcontroller.aspx

2) Select the features that are pre-requisites for the project, ie. number of op-amps, CTMU, PWM, voltage range and so on.

3) Perform search

4) Sort results by price

5) Work my way up the list of results, starting at the cheapest, to find one I'm happy with.

[RogerRowland]

The process I follow to choose a PIC is:

1) Go to the MAPS page at http://www.microchip.com/maps/microcontroller.aspx

2) Select the features that are pre-requisites for the project, ie. number of op-amps, CTMU, PWM, voltage range and so on.

3) Perform search

4) Sort results by price

5) Work my way up the list of results, starting at the cheapest, to find one I'm happy with.

6) Order some free samples.

[Ian.M]

No.
 
6. RTFM !
7. Check availability
8. Order free samples.

The product database MAPS and the other parametric search tools work from has quite a few errors, so you need to confirm the features you need are actually present (from the datasheet) and functional (i.e no grossly objectionable errata).

[dannyf]

The process I follow to choose a PIC is:

My approach is the opposite: find the biggest chip I have that envelope the project. Develop my code for it. Think about what features may be needed to future proof the project and pick the favorite chips / corporate chips for that project, and retarget my code.

[deephaven]

When you find a chip you think you would like to use, do check the errata to make sure there's nothing there to spoil your project.

[igendel]

Microchip tend to introduce new chips at lower prices, then increase the pricing on older parts

Microchip almost NEVER phase out a product (basically it only happens to failed products or to acquisitions

Apparently we're getting the best of both worlds now, with the very significant increases in AVR prices 

[ebclr]

If you chose PIC you chose legacy convenience, for any PIC chip on the market will have another processor on the market from another source with a more modern structure with same or better specs 40% cheaper, Pic is a monopoly reign, and is priced  on that way

[David Hess]

Apparently we're getting the best of both worlds now, with the very significant increases in AVR prices 

I assumed that would be the case when Microchip bought ATMEL so I immediately wrote off using any ATMEL products.

[AndyC_772]

If you chose PIC you chose legacy convenience, for any PIC chip on the market will have another processor on the market from another source with a more modern structure with same or better specs 40% cheaper, Pic is a monopoly reign, and is priced  on that way

Which other families will run from 5V, at 150C, and have a decent range of analogue peripherals built in? Those are the design criteria for most of my work that currently uses PICs.

If I care about performance, I go straight to STM32.

[ebclr]

Can you show me what other complimentary components did you use for 150C like capacitors, crystals and even connectors, and what kind of eccentric application need to be running at 150C and 5V

[AndyC_772]

Harsh environment sensors, which are required to have a 0-5V output, and which are bolted straight to something that gets very hot in use.

Supporting components really aren't the problem; parts like R's and C's, crystals, voltage regulators are all readily available with 150C+ temperature ratings, albeit at higher cost than commodity parts.

[Kilrah]

He's talking of completely diffrerent products being cheaper, not of finding cheaper PICs...

[KL27x]

Makes perfect sense, the pricing. Look at offerings from other manufacturers. They drop older stuff to concentrate on the latest and greatest. Which is the most efficient way to do it.

Microchip taps market demand for older devices. Producing smaller amounts of a vast number of products costs more money. The guys that move to newest and best devices should not pay that tax. The guys using the older devices must support that cost. They are the ones that benefit, for w/e reason they are using the older device. For hobbyist and small production, it may be cheaper to continue using older devices for who knows how long? Even in large scale manufacturing, it  might be cheaper to continue using legacy device, particularly where reliability is paramount, and testing and liability could be very expensive. 

If you get a discount on another manufacturer device, it may be because they are planned to go out of production. If you are buying from Microchip, that thing might be around, forever.

[dannyf]

"Makes perfect sense, the pricing."

It is the razor and blades equivalent  in the chip world. Customers using older chips are essentially captured by high software costs, or their lack of reporting the code to newer chips, or product life cycle, etc.

I wonder if someone will take this strategy to the extreme of giving away new chips and charging progressively for them as the chips age.

[dannyf]

"40% cheaper where?"

Stm8 for example.

Also the answer depends on what you meant by spec. For example, avrs can run on 16M or more mips. Many mid range pics run atva quarter of that.

Without laying out a common agreed up set of criteria, it is hard to assert or refute any claims.

[wraper]

He's talking of completely diffrerent products being cheaper, not of finding cheaper PICs...

He is saying comparable or better performance.

I've gotten PICs at far less then what MC advertises on there site for large volume pricing in prototype quantities so the price you see has flexibility.

For example Universal Bee from silicon labs. MCU with USB and CPU which is much faster faster than any of 8 bit PICs (it's not classic 8051 core, just instruction set compatible), 12 bit @ 200ksps /10 bit @ 800ksps ADC and very precise internal oscillator for EUR 0.68 at quantity of 1 http://lv.farnell.com/silicon-labs/efm8ub10f8g-b-qfn20/mcu-8bit-8051-50mhz-qfn-20/dp/2468099
Also contains internal Vreg which can supply up to 100 mA and can power external circuit too. 16KB version is below EUR 1 too. Also they have more compact code than PIC, so less flash actually needed for the same thing.
64KB for EUR 1.39 at single quantity http://lv.farnell.com/silicon-labs/efm8ub20f64g-a-qfn32/mcu-8bit-8051-48mhz-qfn-32/dp/2468104

[luxfx]

then there are the newest of all, the PIC16F18XXX. For example the PIC16F18323 is a 16F1823 with a couple of things more and a lower cost

Awesome information, thank you. I'm pleased to hear my choice of a PIC16F18325 is of the most recent products!

6) Order some free samples.

I'm a hobbyist right now, and from what I can tell free samples caters mostly towards businesses that might be influenced to purchase large quantities. Is there a good way to get some samples as a hobbyist?

If you chose PIC you chose legacy convenience, for any PIC chip on the market will have another processor on the market from another source with a more modern structure with same or better specs 40% cheaper, Pic is a monopoly reign, and is priced  on that way

I've looked a little at EFM8 but was a little surprised at how few of their options came in PDIP, which I'm interested in for testing/breadboarding purposes. (I have done some MSSOP soldering onto adapter boards but would like DIP for learning)  Do you have any other recommendations? I'm mostly focusing on PIC because of the large amount of educational material online and Mac support for the software, and because the PIC16F18325 that I found has dual I²C peripherals that support PMBus in the $1.00 range.

[ebclr]

For hobbyist Ti & analog devices  is the sample king!  It's easier  to get in with a gun on the white house than get a free sample from microchip

[JPortici]

and i have never been refused samples a single time in 6 years! (Academic email, though)
i don't know if i ever ordered some with my personal email.. but now i just add a couple of parts to the week's/mont's order

Anyway, yes. i thought to replace the 16f18323 as the low-pin-count-8bit where i work (three new projects coming) but it's too new a part, the datasheet is incomplete in the electrical caracteristics and we needed to be sure that it had no issues with low power modes (in at least two of them. one is a coin cell transponder so that was critical) and in the last couple of years it happened too frequently that the new parts across all families had many critical bugs in the first release... so i settled for the 1823: a proven, reliable micro (at revision 9!) all the peripherals needed, enough low power and only some cents more than the 323, for now.

[Ian.M]

A good few years back, Microchip clamped down on samples due to scammers ordering them to resell on EBAY (and being dumb enough to sell them with the sample label still on the packaging).   They blacklisted all the common free email providers.  From time to time they add more free providers to the blacklist.

All you need to qualify for free samples is to register at microchip.com or microchipdirect.com with an email address at a domain that doesn't offer free email. This is a separate registration to your Microchip forum logon - there is no problem with using a disposable  email address for their forum.

YMMV as I believe the policy varies from country to country, but AFAIK I have described the requirements for EU and North America

[RogerRowland]

I'm a hobbyist right now, and from what I can tell free samples caters mostly towards businesses that might be influenced to purchase large quantities. Is there a good way to get some samples as a hobbyist?

I'm a hobbyist too, and I've had no problem getting Microchip samples for the last 4-5 years. I'm so impressed with the ease of it, and grateful for the chance to test/prototype different devices that I always order from Microchipdirect when I need more, even if I can get slightly better deals elsewhere. As some of my hobby projects have required 50 to 100 PICs, it's a reasonable expense for a hobby, but I also bought my PICKit3 direct too. I guess they are being paid back for the samples, which in my mind is how it's supposed to work.

[KL27x]

It is totally ok to get free samples from Microchip as a student.

As long as you use a non POP/anon email. (which I suppose could be abused by someone who could turn a profit on sample... like somewhere in a third world country or something).

I have gotten loads. But honestly, I order stuff from electronics distributors so often, what's the point saving a few dollars? Unless it's a part that is unavailable in prototyping quantities, I'll usually just buy it. Delivered in 2 days. Rather than wait up to a week for a free sample from Microchip Direct.

They want students to get free samples. Students are future customers. You also have to answer some questions which gives them meta data on their customer (or prospective customer) base. Which is worth something to them.

[Kilrah]

I have ordered samples many times but that was +10 years ago while I was a student. Some I've used, some I haven't... But I did end up buying a few hundred parts to build into small volume products I've then designed/sold.

I still like their parts, but haven't used them recently mainly because of their policy about development tools. It's gotten better now they harmonized a decent in-house solution with the XC series compilers, but I still think that charging for development tools is ridiculous nowadays, a manufacturer should do everything they can to facilitate and encourage adoption and good and free tools is the main port of call. I much prefer ARM's attitude of integrating support and maintaining GCC for their processor lines.

[wraper]

It's gotten better now they harmonized a decent in-house solution with the XC series compilers, but I still think that charging for development tools is ridiculous nowadays, a manufacturer should do everything they can to facilitate and encourage adoption and good and free tools is the main port of call.

Yes it's ridiculous and other guys do understand it. For example Silicon Labs you gives a full licence of Keil compiler together with Simplicity studio for free. They pay for you to encourage you using their MCUs. They also give free USB VID/PID pairs to encourage using their MCUs with built in USB.

[JPortici]

They also give free USB VID/PID pairs to encourage using their MCUs with built in USB

i believe microchip does that as well, at least i'm sure it did in the past

i can't speak for XC32 but XC16 optimization is already pretty good with "O.1" which you can use with the free version.. for the 8 bitters there isn't much hope at an efficient c code, i almost always resort to ton of inline assembly

[mikeselectricstuff]

They also give free USB VID/PID pairs to encourage using their MCUs with built in USB

i believe microchip does that as well, at least i'm sure it did in the past

i can't speak for XC32 but XC16 optimization is already pretty good with "O.1" which you can use with the free version.. for the 8 bitters there isn't much hope at an efficient c code, i almost always resort to ton of inline assembly

XC16 and 32 are very useable in free mode, it's only the 8 bit version that's badly crippled.

[RogerRowland]

XC16 and 32 are very useable in free mode, it's only the 8 bit version that's badly crippled.

But don't forget that XC32 now obliges you to use Harmony by default and doesn't include the old plib anymore. I've found Harmony to be such a pain to get to grips with that I'm now investigating switching to STM32 ARM instead. Maybe I'll suffer the same pain with ST's Cube stuff, but we'll see .... I guess the only other option is to do away with frameworks altogether but I'm not sure that's realistic if you want to do some quick prototyping.

[mikeselectricstuff]

XC16 and 32 are very useable in free mode, it's only the 8 bit version that's badly crippled.

But don't forget that XC32 now obliges you to use Harmony by default and doesn't include the old plib anymore. I've found Harmony to be such a pain to get to grips with that I'm now investigating switching to STM32 ARM instead. Maybe I'll suffer the same pain with ST's Cube stuff, but we'll see .... I guess the only other option is to do away with frameworks altogether but I'm not sure that's realistic if you want to do some quick prototyping.

Obliges you how ?
I use the latest XC32 without it.

[RogerRowland]

Obliges you how ?
I use the latest XC32 without it.

Well, what I mean is that the legacy peripheral libraries are deprecated and no longer supplied with the latest XC32 compiler. So, unless you do all of the register settings etc. manually, using the datasheets (which I guess is what you do?), then you must either downgrade the compiler to use plib.h or install Harmony and try to climb that learning curve. So, I'm just saying it's a bit of a PITA if you want to knock up something quickly, a bit like you can with 8-bit PICs and the code configurator.

[dannyf]

then you must either downgrade the compiler to use plib.h or install Harmony and try to climb that learning curve.

You shouldn't use such libraries directly. I have a set of my own "libraries" that perform certain functions. Those libraries are identical from chip to chip and often from one family to another family. Those libraries are built on a variety of approaches, direct register access, or different 3rd party libraries or OEM libraries. My user code will never have direct access to any of the lower level libraries and instead only through this middle layer.

With that, when I migrate from one approach to another, or from one library to another, I only need to rewrite the middle layer and i'm done.

Using OEM libraries directly in user code is suicidal in my view.

[RogerRowland]

You shouldn't use such libraries directly. I have a set of my own "libraries" that perform certain functions. Those libraries are identical from chip to chip and often from one family to another family. Those libraries are built on a variety of approaches, direct register access, or different 3rd party libraries or OEM libraries. My user code will never have direct access to any of the lower level libraries and instead only through this middle layer.

With that, when I migrate from one approach to another, or from one library to another, I only need to rewrite the middle layer and i'm done.

Using OEM libraries directly in user code is suicidal in my view.

Yes, I get what you're saying - and if you've already developed your own solution over time, then that's infinitely better. I was talking about prototyping something quickly with a new device. If you're new to PIC32, it's helpful to have a tool that generates working code easily, so that you can just get something running. As you say, one wouldn't choose to rely on that in a real project - burnt too many times on the non-embedded side to make that mistake.

[mikeselectricstuff]

I generally avoid libraries-been  bitten too many times by bugs & other issues in them. For simple functions it's often just as easy to learn the reg settings than the library functions anyway.
Is there anything stopping you using the plib files from older compilers,apart from new device support?

[RogerRowland]

I generally avoid libraries-been  bitten too many times by bugs & other issues in them. For simple functions it's often just as easy to learn the reg settings than the library functions anyway.
Is there anything stopping you using the plib files from older compilers,apart from new device support?

No, it's really an irritation rather than a problem, and maybe I'm half using it as an excuse to explore ARM. After all, they're only just down the road from where I work now and recently in the news, so I'm probably too easily influenced  This is still only a hobby for me, but one that's drawing me slowly but surely closer to the hardware. After nearly 40 years writing software, I enjoy new challenges.

[dannyf]

The same problem exists in the ARM world as well, and it varies from vendor to vendor but generally more severely if your focus is on software across multiple platforms.

I would point to STM, TI and NXP as some of the worst offenders there.

[ebclr]

mbed joins all arm together on same platform, Of course you must be smart enough to make a custom board or be limited only to dozen of platforms from different vendors

https://www.mbed.com/en/