CH341A Serial Memory Programmer Power Supply Fix

[magic]

I suppose the 3.3V reg is used to power USB so the USB interface operates at reduced voltage without this connection because of losses in the internal regulator. Not a big deal if it works, but I would fix it.

[Mcast]

I suppose the 3.3V reg is used to power USB so the USB interface operates at reduced voltage without this connection because of losses in the internal regulator. Not a big deal if it works, but I would fix it.

I tried to do what you said, I disconnected pin 28 and left only pin 9 powered, the device didn't turn on. I believe that there are other components within the integrated circuit that need power from pin 28.

[magic]

No, that's not what I meant. It looks like everything on this chip runs from VCC except for USB communications which require 3.3V.
When VCC=5V the chip generates V3 internally. When VCC=3.3V the internal regulator will output less than 3.3V on V3 so USB may work if you are lucky or it may sometimes fail to work. That's why you short V3 toVCC.

[fearconte]

Hi,

I have a CH341A Pro black "edition" and I tried to read some chips using clamp (chips are on board soldered), first the Detect doesn't work - ok china - i have selected the correct chip from list and read the chip.
on second read some data shifted and I was like this is not ok the chip is a Winbond W25X10
I have tried on different chip a PMC PM25LV512A which was read ok multiple times - erase - write - all good, strange but works with PMC not with Winbond which I need.

So I have searched the web and saw this post about the voltage, checked mine - 5V everywhere and I have started the mod using the OP
Lifted the pin 28 connect to C3 then to 1117 tested - success 3.3V - tried to Detect the chip still not working, tried to read, erase, program all error
Then I have tried next mod pin 28 to 1117 and pin 9 to 1117 - 3.3V but the same errors

Any ideas?

Thank you!

[Mcast]

No, that's not what I meant. It looks like everything on this chip runs from VCC except for USB communications which require 3.3V.
When VCC=5V the chip generates V3 internally. When VCC=3.3V the internal regulator will output less than 3.3V on V3 so USB may work if you are lucky or it may sometimes fail to work. That's why you short V3 toVCC.

Now I get it. I measured it and the difference between Vcc and V3 was about 130 millivolt, there is a loss of voltage too.
It's like 78xx, the regulator starts to become unstable (with noise, variation and the whole thing) when the input voltage stays the same with the output voltage.
I made the change, thank you all!

[Ian.M]

@Mcast, that's a nice neat job you should be proud of.

If you want to mod it to be a switchable 3.3V/5V programmer, in addition to the Vcc pin lift already discussed, you need to add a tiny DPDT switch with one pole switching Vcc between 3.3V from the LM1117-3.3V, and 5V (via a polyfuse), and the other shorting V3 to Vcc in the 3.3V position to bypass the internal regulator.  Where to mount it is the problem - if its small enough it may be possible to solder the switch body to the ground track along the board edge next to the three capacitors by the regulator, otherwise epoxying it on top of the CH341A chip would be a sensible option.

There's a good page about this programmer here: https://www.onetransistor.eu/2017/08/ch341a-mini-programmer-schematic.html

Finally, for those of you less experienced with SMD rework and board mods *PLEASE* for your and our sanity, and to reduce the risk of snapping the Vcc pin, get some 28 or 30 AWG Kynar wirewrap wire or magnet wire to do the mod with! If you *MUST* use stranded wire it needs to be something really thin and flexible, but with a good quality core,e.g. a single wire stripped from an old 80 wire UDMA PATA cable, ends carefully stripped, pre-tinned and trimmed to length *BEFORE* making the joints, and you need an order of magnitude more skill to get a decent result with stranded wire.

[Mcast]

@Mcast, that's a nice neat job you should be proud of.

If you want to mod it to be a switchable 3.3V/5V programmer, in addition to the Vcc pin lift already discussed, you need to add a tiny DPDT switch with one pole switching Vcc between 3.3V from the LM1117-3.3V, and 5V (via a polyfuse), and the other shorting V3 to Vcc in the 3.3V position to bypass the internal regulator.  Where to mount it is the problem - if its small enough it may be possible to solder the switch body to the ground track along the board edge next to the three capacitors by the regulator, otherwise epoxying it on top of the CH341A chip would be a sensible option.

There's a good page about this programmer here: https://www.onetransistor.eu/2017/08/ch341a-mini-programmer-schematic.html

Finally, for those of you less experienced with SMD rework and board mods *PLEASE* for your and our sanity, and to reduce the risk of snapping the Vcc pin, get some 28 or 30 AWG Kynar wirewrap wire or magnet wire to do the mod with! If you *MUST* use stranded wire it needs to be something really thin and flexible, but with a good quality core,e.g. a single wire stripped from an old 80 wire UDMA PATA cable, ends carefully stripped, pre-tinned and trimmed to length *BEFORE* making the joints, and you need an order of magnitude more skill to get a decent result with stranded wire.

I was really thinking about this upgrade, but there are 2 obstacles for me: I really care about the aesthetics of the adapter and the most of the memories I use, work at 3.3 volts. Maybe I will make a separate box to hold the new sockets and switches. Thanks for the tip!

[protocold]

It was mentioned in this link (in Russian): https://savagemessiahzine.com/forum/index.php?showtopic=884713&st=1880

that there is yet another fault in the design to reset / set the QE bit in the second status register. (please scroll down till you reach the supported chip model section)

It included some pictures but they were not very clear. (I have pulled the pictures and attached here)

Is this a known issue or could someone better at this shed some light into this please?

[Ian.M]

The common CH341A programmer boards cant use quad SPI mode for faster programming as the extra data I/O lines  from the CH341A chip aren't connected to anything.  It would be theoretically possible, *if* the right connections were present, but would need software support and would probably be incompatible with the use of the usual software and ordinary non-quad SPI FLASH chips in the same socket.

There is however a limitation of how the ZIF socket is wired, that's incompatible with certain SPI FLASH chips *IF* they have been configured for quad SPI mode, which uses those pins as I/O so naturally cant tolerate them being shorted to rails.   Unfortunately the bit in question in the FLASH chip's status register is non-volatile so quad SPI mode is 'sticky', as opposed to the sane alternative of making that bit volatile and having the host processor configure the chip appropriately for the application circuit on startup.  See Winbond's W25Q64FV datasheet for details.

The proposed mod is simply to make three track cuts disconnecting those pins from the Vcc trace on the bottom side running under the ZIF socket, to add a wire link to patch the interrupted Vcc track back together, leaving out the /HOLD pin, then to fit pullup resistors from the Vcc track to the /HOLD and /WP pins.   *IF* you need the capability to program DIP-8 package SPI FLASH chips with non-volatile quad SPI mode, and to be able to enable that mode with the chip in the ZIF socket  you need to do the mod.  Some magnet  wire and a couple of 0402 10K resistors should let you do it very neatly.  Any resistors bigger than 0603 will be problematic due to the limited space between the Vcc track and the /Wp pin.   Otherwise, if you are programming in circuit, simply make sure your test clip lead includes 10K resistors in series with the /HOLD and /WP pins.

[protocold]

I see. Many thanks!

This maybe off topic for this thread. But do you know much about read/writing spi nand with the CH341A? There is a software called SNANDer but I have not been able to get consistent read out of this. Reading a 128MB spi nand chip will always have a few bytes different.

https://github.com/McMCCRU/SNANDer

Any idea? possible bad chip?