PIC16F1783 - great chip for PSU makers

[ptricks]

I noticed a new pic on the microchip site and sample are available now, this chip has lots of features for power supply control . Cost is only $2 per chip QTY 1 and it comes in SPDIP format too.  I am halfway through building another supply but may stop and re-consider a re-design with this chip, it looks like it could cut my parts count in half.
http://www.microchip.com/wwwproducts/Devices.aspx?dDocName=en552741

Internal 32MHz oscillator
2 x PSMC (Programmable Switch Mode Controller)
Dedicated 16bit PWM
Digital and/or analog feedback control of PWM frequency
Dead-band control / auto-shutdown & restart
3 x Fast Comparators with selectable Voltage Reference
2 x Operation Amplifiers (rail-to-rail in and out)
11 x 12-bit ADC with Voltage Reference
8-bit DAC
MI2C / SPI / EUSART w/auto baud
2 x Compare-Capture-PWM
Four 8-bit Timer (TMR0/TMR2/TMR4/TMR6)
One 16-bit Timer (TMR1)

[MikeK]

Sounds cool.  Unfortunately, Microchip stopped sending samples to the U.S. years ago.  It's now on my Mouser "to-order" list.

[bxs]

Great, have you read the errata?

You if try to make a PSU with the current revisions I'm almost sure it will blow up 

[ivan747]

One thing I don't like about having a MCU in a control loop is that it becomes tedious to make the circuit work. You now have something you have to program in order to get your basic power supply working.

Here's an example: If the MCU only told the supply what voltage to output using a DAC it would be easier to test the power supply. That's because you don't have to figure out how to use the DAC before confirming the rest of circuit works fine.

[jeroen74]

I believe these kinds of ICs, just like the AVR PWM series, are really meant for fairly steady loads like electronic drivers for HID lamps, or fluorescent tubes. Or motor control which is fairly slow too.

I don't see the added value of a MCU inside the control loop of a PS. Not every single little analogue thing needs MCUfication.

[ptricks]

Sounds cool.  Unfortunately, Microchip stopped sending samples to the U.S. years ago.  It's now on my Mouser "to-order" list.

When ? I just got a sample order from Microchip two days ago. They do fedex them from Thailand, really weird since they also have a warehouse in Austin, TX. I guess they got a deal with fedex on the low cost shipping.

[ptricks]

I believe these kinds of ICs, just like the AVR PWM series, are really meant for fairly steady loads like electronic drivers for HID lamps, or fluorescent tubes. Or motor control which is fairly slow too.

I don't see the added value of a MCU inside the control loop of a PS. Not every single little analogue thing needs MCUfication.

Variable output switch mode supplies can use this to easily put together everything but the output stage with the FET. A lot of the supplies currently use one of the URxxxxx chips along with a couple extra chips. There is an app note for using it to charge lithium batteries, LED controls,  as well as 3 phase motor control.     For the price it is a lot of features in a small package.

[ptricks]

One thing I don't like about having a MCU in a control loop is that it becomes tedious to make the circuit work. You now have something you have to program in order to get your basic power supply working.

Here's an example: If the MCU only told the supply what voltage to output using a DAC it would be easier to test the power supply. That's because you don't have to figure out how to use the DAC before confirming the rest of circuit works fine.

If all you want is preset voltages then there are easier ways to do it. If you want variable voltages controlled by a processor I can't think of an easier way than this chip.

[jeroen74]

Those are applications where transient response can be fairly slow. PWM resolution is also a problem. Of course, I have not studied these devices in any detail, so maybe they have some features to help improve transient response greatly (the analogue feedback part maybe?)

I know for fact that in eHID drivers, that a PWM output running at 96MHz does not provide enough time resolution to keep the output stable enough (lamp flicker), so a dithering circuit is needed that basically modulates the duty cycle in a certain pattern so the apparent time-averaged resolution is increased by 8.

I think you can control the output voltage of a conventional switcher without this kind of MCUs.

[ivan747]

I think both of the integrated comparators can be linked with the special PWM peripheral for a fast response.

I was using a completely different PIC the other day to design a cool circuit. This is an example of what can be done with some peripherals. This was an LED backlight driver. It worked from 3V.

I used a capacitive voltage doubler, the low resolution DAC and one of the built in comparators to make a simple LED backlight driver.
Since the DAC can be wired internally to the comparator, I did that. I could also disable the PWM output using the comparator output internally. I only used a couple of caps and diodes and a few resistors to get it working.

The most time consuming task was setting up all the fuses and configuration bits and software. In the end the PIC is pretty much a black box.

[nctnico]

Microchip is wasting their time. I have made a 300kHz power supply controller using an ARM microcontroller 6 years ago.

[metalphreak]

Microchip is wasting their time. I have made a 300kHz power supply controller using an ARM microcontroller 6 years ago.

The Wright brothers made an aircraft fly in 1903, but that didn't stop Boeing and hundreds of others building planes... 

[BravoV]

...<snip>....

Not every single little analogue thing needs MCUfication.

Sorry for ot, just love this one, feel the strong urge that its deserved as my new fav quote. 

[nctnico]

Microchip is wasting their time. I have made a 300kHz power supply controller using an ARM microcontroller 6 years ago.

The Wright brothers made an aircraft fly in 1903, but that didn't stop Boeing and hundreds of others building planes... 

Yes, but the Wright brothers made their aircraft BEFORE Boeing made the 747. Not the other way around.

[jerry507]

I'm at a complete loss to see why Microchip shouldn't do what every other microcontroller vendor does, offer their own version of things other companies do. It's just another addition to their CPU independent stuff.

http://ww1.microchip.com/downloads/en/DeviceDoc/41565d.pdf

[JoannaK]

Microchip is wasting their time. I have made a 300kHz power supply controller using an ARM microcontroller 6 years ago.

The Wright brothers made an aircraft fly in 1903, but that didn't stop Boeing and hundreds of others building planes... 

Yes, but the Wright brothers made their aircraft BEFORE Boeing made the 747. Not the other way around.

I know this may backfire, but at what pricepoint did your power supply to be? Was it produced in masses, like part of some highly competetive package or was it One-off build?

Even though Microchip chips can be used for 1-off protos, their real value is on mass production.

 

[nctnico]

It is a mass produced product.

The volume must be really high (>100 000 units) for the total cost of a PIC to drop below that of a modern microcontroller. Most of the cost of a microcontroller based product goes into developing the software. For most products I've worked on about 95% of the price is the cost of developing software.  Because the other components, PCB, casing, testing also cost money it really doesn't matter if I use a $2 or $5 controller. It will change the price only by like 0.5%. But if I choose a controller which takes me twice as long to write software for then the price of a product suddenly almost doubles.

For example ARM simply is much more convenient to program and existing software (like algoritms you can find on internet) is easier to get running on ARM. PICs (especially the 8 bit ones) can't deal with pointers very well so you have to (re)write software in a clunky way which is more difficult to understand and maintain. This may seem a minor problem but when summed up over a large period you'll see the losses in time and money grow to staggering amounts if a product has to do more than read a few buttons and print some text on a display.

[JoannaK]

I agree, that if the product needs something like Tcp/Ip (aka ethernet) or bluetooth etc then those pics are definitley suboptimal.

But, if it will be stand.alone power supply, the those pics can do a quite well job.

I admit, I have not checked what kind Soc-ARM.s there are available. back in days (not so many years ago) the 32-bit cores would definitely have been too much overkill to any power supply.

For the reference (and related to my own Fujtsu thread), is there any particular family of Arm chips you'd recommend to get a closer look? I know there are n+1 variants from all new companies, but if we try to limit the list to those avaible at hobbyist friendly parts and devtools, open documentation (no NDA or registration)  and honest Erratas.

[nctnico]

The question is whether you'd have to put a microcontroller in a standalone power supply...

Anyway, the ARM devices from NXP have been serving me very well. The errata sheets aren't very long and all you need to program them is a tool called 'flash magic' and a serial port. No hassle with JTAG dongles and so on. But.. like with all microcontrollers it does take time and reading to get up to speed.  Fortunately you can download a lot of software from NXP's website to see how the peripherals need to be setup.

[JoannaK]

As saying goes: Have a hammer and everything looks like a nail .. 

[JoannaK]

I decided to check those Nxp:s and they indeed look nice for what they are. That is if you are looking for a mpu with PIO-pins and some Uart+Spi+I2c. But the on the controller IO and functionality side those are quite lacking. Ok, the Arm core is neat and the basic IO:s modules are well available, but even when compared to this pic16 (in subject) they lack oomph on HW-driving. Not to mention those real beasties Motorola (later Freescale) has been making for over a decade.

I also realized how relative the world 'difficult' is when we talk about micro controller setup. For something I'd agree on difficult (some might say 'extra hard') would be something like MPC555 TPU-s (Data sheet not for the faint hearted)..
http://cache.freescale.com/files/microcontrollers/doc/user_guide/MPC555UM.pdf

Yep.. that's over decade old chip and pricey as hell (was not cheap even back then). But having 32bit PPC(Float) core plus complex IO support, and with 10+year availability it's not too bad.

Addendum: I had to check, Freescale uses ARM only at those Infotainment chips. Reason or other, all their system critical control (motor, transmission, breaks) chips are PPC based.

[nctnico]

I decided to check those Nxp:s and they indeed look nice for what they are. That is if you are looking for a mpu with PIO-pins and some Uart+Spi+I2c. But the on the controller IO and functionality side those are quite lacking.

Then its time for you to get a new pair of glasses. Most peripherals are quite versatile and can do a lot more than it seems at first sight. The SPI interface for instance has a variable word width up to 16 bit and a hardware buffer. The SPI interface also has several modes which also allow it to interface with several kinds of audio codecs. Depending on the device timers can be used as versatile PWM generators or even to trigger the ADC. For the projects I came across the NXP controllers turned out to be the best fit compared to TI's and ST's ARM offerings. Another neat feature is that the flash has multi bit error correction which improves reliability.

[jerry507]

Picking microcontrollers is largely personal preference masked by technical requirements.

[nctnico]

Nonsense. A device fits the project or it doesn't.

[JoannaK]

In a sense I see a point. Since designs are at least partly made with compromises. And question usually is what is one willing to give up to make it fit and work.

So far, of the Arm cores controlers (m0 or m0+) I have seen the Freescale KL2 series seems to be least ackward. Ok, their 'free compiler' that is avialable with the 10-20$ protoboards is apparently more limited than what's available at NXP, but IO modules have bit more oomph and documentation seems more trustworthy.