chegewara wrote: ↑Sat Sep 11, 2021 6:50 pm
Rivers wrote:
Thanks for your idea. Yes that's probably a workaround for hobbyist. But you will end up with 4 cores MCU @ 160MHz along with 2 SPI Flash memories draining lot of power,
As you maybe know there is esp32 SOLO with just 1 core (and maybe embedded flash). Bluetooth is also able to work with MCU @ 80MHz and with 32kHz crystal there is power safe mode.
Sure, maintaining 2 firmware is drawback, but espressif is providing AT command firmware to control bluetooth.
I can see reasons why BT classic may be not existing on S3:
- hardware complexity and production cost,
- firmware and RAM consumption,
- probably more i cant image from top of my head
Im not saying your questions or needs are not important, just saying what you can do to stick to esp32 and keep product cost low.
off topic:
As small off topic i would like to say that espressif could make in future dev kits even cheaper if they use esp32 S2 instead of CP210X thanks to native USB, maybe even with preinstalled binary with option to flash main chip from web browser without implementing it in user app
Hi chegewara,
I will try to elaborate and educate you as much as I can from the RF engineer and the company's point of view in order to prevent others from trying to experiment with your risky proposal.
We are a company and in order to retain competitiveness, the company usually makes products based on existing approved reference designs. They buy IP cores, Source code licenses, etc.. and top up each other with their in-house development to make the products. They do not go into a time-consuming "reinvent the wheel" route that poses many risks. And I'm talking about "Real World Scenarios" as well as in "Commercial off the shelf Products". Have you ever seen company stitches 2 WiFi MCUs together in close proximity?
Your proposal is a "reinvent the wheel" route that has no approved reference implementation (If you have a reference design of your own, you give me the link, FCC approved preferred). If you would like to know more about an RF-approved reference design it'd be better to check out this link (
https://developer.amazon.com/en-US/alex ... s/dev-kits).
No matter there are workarounds using AT commands or migrate to a single core MCU in low power mode. Your proposal of stitching 2 WiFi MCUs together in close proximity will exhibit a risk, and I believe that an Espressif's engineer, as well as all RF engineers, will never recommend doing that at all, due to the RF regulation issues. Because chances are that software/hardware malfunction or even a hacker could lead to accidentally turn on both radios simultaneously, thus violates the RF regulations. This is the answer to your off-topic question regarding why Espressif's engineer will never replace CP210X with S2 for sure.
If you have ever designed a Linux-based WiFi router like us or at least look inside it, you will notice that there are only a single Linux-based WiFi "CPU" (multi-cores if need) couples with multi-band RF AFE ICs and there are a mechanism called packet traffic arbitration (PTA) implemented in the hardware of each RF AFE ICs in order to chop transmitting time frames share among each RF ICs this is required for the coexistence RF like multi-transmitters WiFi and WiFi + Bluetooth which operates on the same RF band.
The ESP32 itself (literally most of WiFi "MCU") doesn't have a mechanism to synchronize through IEEE 802.15.2 packet traffic arbitration (PTA) between each WiFi and Bluetooth transmitter, see this post (
viewtopic.php?f=12&t=8527&sid=ecbdd8f25 ... f22103a393). That's why your approach doesn't exist in real life.
Moreover, it will eventually have to get approved by part 15 of FCC rules which is very costly if you put 2 "Intentional Radiators" on the same frequency band in very close proximity (less than 20 cm) which costs you at least $5000 per radiator. Yes, you can utilize a modular approval of ESP32 RF modules in your design then turn on only ESP32 radios and shut down the S3 radios, but the FCC doesn't care as long as both chipsets are capable of transmitting RF power. They will treat your design falls into "FCC Multi-transmitters/Antennas Co-location" categories then double or triple the certification fees and hence the pre-certified house doesn't help much due to the complexity of the RF radiation pattern. These are the dead-load costs that make products more expensive. Your approach will never get cheap at all.
Espressif is also a company and they make products based on the market demand, so if the demand is high enough I see no reason why they should not make a dual-mode BLE ESP32-Sx series or even a dual-band WiFi 2.4/5GHz variant, check this out $6 module from Realtek (
https://www.cnx-software.com/2021/01/10 ... eless-mcu/). Since they have already obtained the licenses of RISC-V and LX7 along with its own in-house WiFi/Bluetooth RF subsystems. To answer your question regarding how hard to produce such a chip. If you have ever learned VLSI design before, the production cost is not as high as you guess, because they utilize a TSMC high yield maturity 40nm ULP process, the same process used to manufacture low-cost Cortex M0-M7 variants, and the production cost of the chip at this 40nm scale always depends on the "die size" and "royalty fees of IP cores" rather than the complexity of the chip itself.
So it is only the question of when they will announce to the public.
Finally, I'm not talking about cheap design. Instead, I'm talking about a lower price than the existing design. If I were you I would rather recommend Kendryte K210 + ESP32 for this kind of application instead, since its CNN @ 0.5 TOPS is more powerful than the S3 at the same price point and it also has a Tensorflow workspace for K210. But as I said earlier managing 2 OS through OTA update is a tradeoff.
So I would like to get information directly from Espressif's technician for sure when available rather than personal opinions.
Thanks for your corporation.
Rivers
