Replies: 3 comments 2 replies
-
|
Hi, yes it makes sense to use an additional chip for pwm drive and control loop than the ESP32(S3). I think the main advantages are:
Here we just began with the espressif chips, and the Wi-Fi enables some neat features for prototyping. Considering espressif MCUs only, the ESP32 appears to better suited for this job since it has a high-speed PWM driver for glitch-free transition (whatever this means, the PWM clock has the same 80 Mhz as the esp32s3), and its is the only esp chip with an adc that can run at 2 Msps, whereas the other chips only support 83KHz. Why would we need such high bandwidth? We can implement very fast shut-down path in over-load conditions, and we can place the current sensor across the inductor, which adds the ability to detect inductor core saturation and a more accurate diode emulation. The stm32f4 adc can do even 6 Msps, and probably has a much lower linearity error and noise, as compared to the esp32. The dsPIC33CK64MC105 appears to do the job even better, it is designed for DC/DC converter applications and has a PWM resolution of 2ns (400mhz) and fast adc triggering. |
Beta Was this translation helpful? Give feedback.
-
|
The dsPIC33CK64MC105 does look like an interesting chip... im curious though, it has a high speed PWM but the core runs at 100mhz and it only has 64k of flash and 3.5msps? I've not used PIC's all that much myself and i think i've made 2 boards based around pic32's chips and they ended up being somewhat complex to assemble. There is a single dev board I can find built around that particular chip and its startingly expensive for what it is. The circuit diagram here for it has some interesting text on it though I cant quite understand what that would be useful for given the chip says 3.0 to 3.6v input, but presumably if you remove that and everything is running at 3.3v alot of circuit would simplify a fair bit. The chip they use for the debugger is a pretty impressive little chip though, ATSAMD21E18 and looks like it has code to support drap-and-drop flashing so it would probably easy enough to replace that with something lighter that just allowed serial programming over usb. The STM32F407 though is very readily available for pretty small $$ for existing dev boards. They're pretty trivial to make into an on-board chip as well (as in I was thinking of putting a chip straight onto the fugu board rather than designing around an existing dev board using the various pill boards) There is a STM32H7 as well that runs at 400mhz, but i dont think it has a great PWM or ADC but its just as trivial to make a dev board for and there are a number available pretty cheaply. |
Beta Was this translation helpful? Give feedback.
-
|
the core speed of 100mhz is the frequency instructions are executed, which doesn't limit the PWM driver frequency. you can consider the 100mhz as an upper limit the PWM can be updated. the PWM circuit is basically just a counter and a programmable comparator which can run at a higher frequency than core. 64k of flash is enough, this will just be the digital control loop and there will be no network code, which usually takes most of the space. I've seen dev kits starting at $90 (TI, Renesas), the $15 for the EV88G73A appears to be reasonably priced, considering it comes with a plug&play USB debugger on-board. You're right, for programming we would just need a UART line, which would reduce the BOM tremendously. I am pretty sure the ESP32 could be used to program the dsPIC33CK64MC105 . Where do you find the STM32F407 / dev boards cheap? |
Beta Was this translation helpful? Give feedback.
-
|
I get the PWM can run at a high speed, its more that doesn't the cpu speed limit how quickly you can read the ADC and respond to changes (or does it not matter much)?
Yeah, i realized after I posted it'd be coding the chip rather than trying to design a usb type interface for a pc for the PIC which should be easier in a circuit but coding the ESP to do the job would be interesting.
You're right, that EV88G73A is indeed $15 which makes it more then affordable. Strange because that was the same board I was looking at earlier but it was significantly more expensive (like $65 or so). On aliexpress/ebay you can find a tonne of stm32f407 dev boards in the sub $15 range, but in truth you can almost use the chip straight off a usb port with almost nothing beyond a regulator (though you really do want a resonator cause the internal oscillator kinda sucks). I think the smallest board i've made for the stm32's has only 6 actual components. Having said that, it looks like if you were putting a dsPIC33CK64MC105 on a board next to an ESP, it looks like you need even less componentry, the dev board doesn't even have a resonator populated and if im reading the circuit diagram correctly it only really needs a single resistor and single cap. The datasheet for the chip also suggests you can write a boot loader and have it self-program which could mean you can simplify the interface between an ESP and the PIC, though you still need a way to get the bootloader on in the first place. On another note though, buspirate seems to suggest it can code a dspic33 directly. |
Beta Was this translation helpful? Give feedback.
-
|
Concerning your 2nd question: I think the same, it works well enough to connect MPPTs in parallel. |
Beta Was this translation helpful? Give feedback.
-
Howdy,
To be honest, my knowledge of DC analogue electronics is a little poor, so while I understand the circuit to a degree, there's lots of things I don't and so apologies in advance if i ask a stupid question.
It says that more accurate pwm isn't possible because its limited by the 80mhz on the ESP's, what if I were to replace it with another chip (say an stm32f4 @ 168 or 180mhz or some other similar chip) and split the buck/boost operation's to the STM32 and shunt the control/human interface to an ESP8266/32? What (if any) advantages does it provide and is it worth the effort? (i only mention stm32f4's cause I'm intimately familiar with the chip and i've stacked esp's and stm's before).
Second question was about synchronized charging... most of what I read about mppt controllers suggests that even different mppt chargers can be connected in parallel as long as they are setup as similar as possible. So presumably two fugu's would operate happily in parallel? Some mppt controller can talk to each other so they output the same voltage and enter charging modes at the same time. Is there an advantage to looking to add that type of functionality to fugu or is it "good enough" to just have controllers run in parallel with the same parameters?
Beta Was this translation helpful? Give feedback.
All reactions