Tír Na HiFi

Here's the correspondence with him - first option B - €3,000 for hardware & software development project

If we are building our own board, I would suggest:
* Use STM32F767 instead of STM32F407 as the main chip, mainly for its ability to use a newer 4-bit SDIO and I2S_CKIN at the same time, instead of being stuck with a buggy SDIO interface choked down to an 1-bit interface. This breaks the 32-bit 192kHz barrier too since the SDIO now maxes out at 96Mbps. STM32F7 series is also much faster.
* Since STM32F767 supports it, we can use 64MB external SDRAM instead of that 1MB low power SRAM as the buffer memory, if we ever need it. SDRAM is cheaper, faster and more spacious than the low power SRAM.

For a STM32-based stack, here is my idea:

* Chip: STM32F767ZI
* Core: Arm Cortex-M7F @216MHz
* RAM buffer: 64MB SDRAM
* Storage interfaces: microSD card slot in 4-bit mode or eMMC in 8-bit mode
* Connectivity: USB (hopefully high-speed if the pins for an external PHY can be arranged) and Wi-Fi (based on an ESP8266 module)
* Output interface: independently-clocked I2S

The files are read in from SD card or USB stick (if we implemented that - the USB hardware on STM32F767 does supports that,) decoded and placed in the DRAM buffer. The separately clocked I2S peripheral DMA’s the samples from the buffer to external DAC.

Need some more detail, Max
- What is needed to provide a fixed price?

Price is dependent on the exact list of features to implement for each installment. Since there is flexibility being designed in, prices can come in installments that corresponds to feature sets, with 3000 euros for the hardware and base software, and additional installments for additional software features.

- Describe the deliverables, please, both hardware & software?

First installment:
Hardware: one PCB design in KiCad of your preferred license, one or two assembled units, a few blank boards.
Software: Base software includes the drivers for SDIO, I2S and SDRAM, middleware for MBR, GPT, FAT32 filesystems and uncompressed WAV, AIFF, AIFC-sowt formats, as well as an basic audio playback application, all in a plugin-based architecture for future expansion.

Future installments are all software-only:
Drivers: USB (big driver,) Flashair-specific iSDIO, Ethernet (also a big one,) graphics LCD
Middlewares: decoders (FLAC, ALAC, AAC, Vobris, etc.) format containers (MP4, MKV, etc,) GUI,metadata reading, full UNICODE, etc.

- When you say no compressed format - you mean it reads WAV but not FLAC files?

Kind of yes: WAV, AIFF, AIFC-sowt are the three uncompressed formats. To me there is no difference between various container or compression algorithms, being uncompressed means there is a “null” compression algorithm. The software uses a plugin-based architecture, so it wouldn’t be hard to add new algorithms down the road, just not now in the first installment, since external libraries still need tweaking and wrapping to work.

- The software drivers you are referring to are USB drivers, right?

Everything needs drivers - USB, SD, Ethernet, I2S, SDRAM, all of them. For basic playback only SD, I2S and maybe SDRAM are needed.

Option A - €2,000 if we provide the board schematic

On May 22, 2018, at 5:36 PM, John Kenny <john.f.kenny@gmail.com> wrote:

OK, what would be your price to just develop the software for base level functionality on the existing SD player board
Functionality:
- read SD card directory/file structure & display on attached screen

Driver: SDIO
Middleware: MBR, GPT, FAT32

- allow navigation of files/folders using existing buttons on board

Drivers: Graphics, GPIO user interface
Middleware: GUI

- allow file selection playback/pause
- read selected uncompressed audio file off SD card

Application: basic playback

- process into I2S output

Driver: I2S

For this system, you need to confirm that the following risks are okay to bear:
* It is very possible that the board’s maker have locked the chip from being reprogrammed, forcing a PCB redesign.
* The 1-bit fast SD mode will result in compatibility problems, as you have experienced.
* Even with the 1-bit fast SD mode, the maximum bitrate of the audio stream, at the best case scenario, is limited to 24Mbps.
* There is no special support implemented for FlashAir, resulting in potential conflicts.

whats the advantage of option A over just using the installed software ?
am i missing something?

on option B its a little alarming he wants to change it so much, although i can see his point,
id be afraid of tipping the baby out with the bath water?

back to this internal 16mhz sck, that sort of goes against our synchronous theory?
ive two other stm sd card players not nearly as good
one with audio clks and that akm chip something like 4118 interfering for spdif?
and another with 8mhz sck with no audio clks (if i remember right very low noise)

is there no way to use audio clks as system clock?

nige2000 wrote: ↑Wed May 23, 2018 11:29 am whats the advantage of option A over just using the installed software ?
am i missing something?

We would have the source code for a working SD player with excellent sound !!
Far easier to & can enhance/modify it ourselves than write it from scratch
So, we could change the existing functionality of the software - to remove the issues we may find when we begin using it i.e. with WAV tags issues or changes needed in trying to seamlessly connect it to a good user interface as we are planning.

For instance, I thought the best scheme so far was that the SD card player worked as a dumb audio file player - just playing the first file it finds in the playlist folder & deleting it after it has finished & playing the next file in the list (if any). At the moment a lot of these steps would have to be achieved by the control software on PC/phone sending control codes to the SD player to play the file, delete the file, play the next file in the playlist, etc.
All this control logic would be better done in the software on the STM. We could strip the STM code down to go headless - removing all code for screen display, button sensing, etc

Later on, we could enhance/change the software to work on our own hardware design or do things that it currently doesn't do - maybe output other than I2S signals for other DAC inputs?

on option B its a little alarming he wants to change it so much, although i can see his point,
id be afraid of tipping the baby out with the bath water?

Well, what would you want to see in option B?

back to this internal 16mhz sck, that sort of goes against our synchronous theory?
ive two other stm sd card players not nearly as good
one with audio clks and that akm chip something like 4118 interfering for spdif?
and another with 8mhz sck with no audio clks (if i remember right very low noise)

is there no way to use audio clks as system clock?

Exactly - we thought the secret sauce for why this SD card player sounded so good was the audio clock & system clock being the same - turns out that's not how the card works so we don't really know why it sounds so good. Just doing option B or any new hardware design could well leave us with just another SD card player that sounds OK - that was the point I was making - option B is a gamble & it seems the odds are against us in getting the same great sound.

I don't see what the point is of chasing what turned out to be a false premise - the audio clock & system clock being the same? It's something we could investigate down the road but it's not the cause of the great sound of the current SD player

That's why I favour option A - getting working software on the existing hardware. If it doesn't sound as good as the existing SD player then we know the secret sauce is in the software or in the way the software is using the hardware

If we do get a solution that sounds as good as the existing SD player then any changes we make to the hardware/software can be evaluated for impact on sound. We might, for instance, change the on 1MB board memory to SLC memory - we might increase the memory size & buffer more of the SD card reading (if that's the function that the memory is being used for?). We might decide that the purist way to do this is to read all of the file off the SD card into on board SLC memory & playback from that memory instead of what seems to be currently happening - the file is being read off SD card while it is being processed to I2S

Lots to think about

If it helps keep this alive John i am willing to stump up 1k of the cost seeing i cannot contribute much by way of software or hardware input.
If it does not work and the money does not bear fruit then so be it, i have spent much more over recent years on this hobby with less to show for it compared to the potential i see in this project.

Worth a punt from me.
We can sort out bank transfers by PM if we get a few more backers.

According to this mega thread linked below, the only SLC storage that sounded much better than the rest should be 4GB Panasonic micro SDHC card

http://www.usaudiomart.com/forum/viewtopic.php?t=1172

Since not all SLC chips are created equal, we better identify those NAND chips with ChipGenius before we decide what should be used

http://www.softpedia.com/get/System/Sys ... nius.shtml

After that, we could take a look at Digi-Key or Mouser etc. and figure out how much would 50 pieces of NAND chips cost. Of course we better pay attention to the date of EoL and secure them ASAP if necessary. Lots of SLC chips should be phased out sooner rather than later, that's why we've gotta act quickly just in case we might miss the boat.

Of course we'll also need to know what NAND flash controllers are compatible with those chips afterwards, then ask that Chinese STM programmer and see if a specific controller were any good for our project since that particular part should also cost extra.

Could finding the right controller for our designated SLC NAND chips be rather complicated?

Sligolad wrote: ↑Wed May 23, 2018 4:16 pm If it helps keep this alive John i am willing to stump up 1k of the cost seeing i cannot contribute much by way of software or hardware input.
If it does not work and the money does not bear fruit then so be it, i have spent much more over recent years on this hobby with less to show for it compared to the potential i see in this project.

Worth a punt from me.
We can sort out bank transfers by PM if we get a few more backers.

That's a very generous offer, Pearse
If there's interest, maybe we could structure this like a kickstarter campaign but some work in defining the product is probably needed first?
- contributions of €1,000 receive ....
- contributions of €500 receive ....
- contributions of €200 receive ....
- contributions of €100 receive ....
-

seeteeyou wrote: ↑Wed May 23, 2018 6:00 pm According to this mega thread linked below, the only SLC storage that sounded much better than the rest should be 4GB Panasonic micro SDHC card

http://www.usaudiomart.com/forum/viewtopic.php?t=1172

OK, good info - didn't know that - will look into that thread. Is the Panasonic 4GB specifically?) sound noticeably better than other size Panasonic SLC SD cards & other brand SLC cards?

I don't know if it was a Panasonic 4GB that we listened to on Sat in Tony's - maybe Pearse can confirm?

Since not all SLC chips are created equal, we better identify those NAND chips with ChipGenius before we decide what should be used

http://www.softpedia.com/get/System/Sys ... nius.shtml

After that, we could take a look at Digi-Key or Mouser etc. and figure out how much would 50 pieces of NAND chips cost. Of course we better pay attention to the date of EoL and secure them ASAP if necessary. Lots of SLC chips should be phased out sooner rather than later, that's why we've gotta act quickly just in case we might miss the boat.

Right, are SLC NAND memory going EOL?

Of course we'll also need to know what NAND flash controllers are compatible with those chips afterwards, then ask that Chinese STM programmer and see if a specific controller were any good for our project since that particular part should also cost extra.

Could finding the right controller for our designated SLC NAND chips be rather complicated?

Right, the Chinese guy says that " NAND chips, complicates programming. complicated. Even when the ECC is implemented in hardware, it appears that write balancing and bad block management algorithms are needed in software"

seeteeyou wrote: ↑Wed May 23, 2018 6:00 pm According to this mega thread linked below, the only SLC storage that sounded much better than the rest should be 4GB Panasonic micro SDHC card

http://www.usaudiomart.com/forum/viewtopic.php?t=1172

Since not all SLC chips are created equal, we better identify those NAND chips with ChipGenius before we decide what should be used

http://www.softpedia.com/get/System/Sys ... nius.shtml

I'm following that tread, but I don't remember anyone comparing the sound between the two different SLC mSD cards!!!

There were comparation between the SLC USB sticks vs SLC mSD cards, where the mSD card sounds better!
I have 64GB Mach Extreme SLC USB stick and compared to my 2GB Apacer SLC mSD card, the Apacer card sounds better.

I believe that the whole idea about the Panasonic SLC mSD card, and the whole idea about using mSD SLC cards for playback, started from the developer of the "Infinity Blade HQ" player.

Again, I'm not denying that the 4GB Panasonic micro SDHC card might be the best, but I just didn't see any comparation between the different mSD SLC cards!
I'm very curious about that!

It seems that it's mainly Tubelover2 promoting the Panasonic 4GB m-SD SLC cards on that forum & that is plugged into a USB 3.0 Startech USB hub it seems (so there may be power quality considerations heere)

On Jplay forum we see (I don't know what hardware the SD card is plugged into)
http://jplay.eu/forum/index.php?/topic/ ... st&p=36855

"I have finally received my p-SLC 8gb micro sd card from Taiwan.......it took 3 weeks.

https://www.amazon.c...0?ie=UTF8&psc=1

I compared it to the Panasonic 8gb SLC micro sd card.

http://www.digikey.c...dchips_standard

I have to say that the performance of the Taiwanese p-slc card is nearly identical to the Panasonic. I can definitely recommend this card, especially at 30% of the cost of the Panasonic.

The Sandisk regular micro sd card was slow loading and the SQ was significantly less than the p-slc and slc cards. Don't waste your time or money on that for this application.

https://www.amazon.c...k 16gb micro sd

One worrying thing however is this http://jplay.eu/forum/index.php?/topic/ ... st&p=39648

I'm not sure how many of you are using Hiroyuki's recommended Panasonic microSD SLC card, but I am and the it sounds very great.

The problem is, the load times can be very long. In fact, when loading a 24/192 album the load time can be over 10 minutes.