Hacking Hardware Picofly - a HWFLY switch modchip

[Uberfish]

I have new theory on why the cmd resistor died. Try to change the cmd resistor in pico which is 47ohms into around 100ohms.

The cmd resistor (on the board) is a pullup resistor 4k7 to 1.8v. I am not sure, whether picofly send something to cmd line, which i assume it is. And i am not sure on what voltage the rp2040 send the signal, which i assume its 3.3v. So It could be that the current flows in the cmd line is too big for that cmd resistor (4k7) to handle, so pico burns it. The easy one is to increase the resistor on the pico to 100 ohms. Its still on the range of emmc specification, and also limiting the current. Might solve the mysterious cmd burns issue.


Test the D line of mosfet to GND, if its 12 Ohms then the D and S is soldered correctly. If is read 0-3 ohms, means the solder between 2 pads of the cap are short circuit. If its read huge resistance or open circuit then its means the soldered tin (D and/or S) is not touching the capacitor pad.

Check also the resistance D-G and S-G also S-GND. Just to make sure. D-G should be high resistance, S-G high resistance and S-GND 0-3 ohms.

There are 6 pin on the flex, The middle pin 3,4 is the G, the else you need to check yourself, i forgot.

Any idea how to check the G S and D with the cable installed? I tried searching for a schematic of the cable to no success.

 

[jkyoho]

For some reason, I found I have IRFHS8342 keep failing on me since last time I did one switch lite with TOSHIBA eMMC, but I occasionally use a CPU mosfet from XBOX (TPN8R903NL), with nothing change just mosfet replace, the v2 I just got today instantly glitch success instead of ==* code when using IRFHS8342. I guess will stick with TPN8R903NL since they have easier solder even you wanna have the 1k resistor between Gate and Source.

 

[abal1000x]

For some reason, I found I have IRFHS8342 keep failing on me since last time I did one switch lite with TOSHIBA eMMC, but I occasionally use a CPU mosfet from XBOX (TPN8R903NL), with nothing change just mosfet replace, the v2 I just got today instantly glitch success instead of ==* code when using IRFHS8342. I guess will stick with TPN8R903NL since they have easier solder even you wanna have the 1k resistor between Gate and Source.

I've check the price, and its 10x pricier than IRF8342

 

[jkyoho]

I've check the price, and its 10x pricier than IRF8342

I have it 30cents difference per mosfet, but I think I could make them even when ordering from China

 

[abal1000x]

Any idea how to check the G S and D with the cable installed? I tried searching for a schematic of the cable to no success.

If you have spare flex cable, it will be more easier to track the trace for each pin.
I don't have any flex cable right now.
You could check it using multimeter and/or microscope.

Post automatically merged: Jun 17, 2023

For some reason, I found I have IRFHS8342 keep failing on me since last time I did one switch lite with TOSHIBA eMMC, but I occasionally use a CPU mosfet from XBOX (TPN8R903NL), with nothing change just mosfet replace, the v2 I just got today instantly glitch success instead of ==* code when using IRFHS8342. I guess will stick with TPN8R903NL since they have easier solder even you wanna have the 1k resistor between Gate and Source.
View attachment 378451
View attachment 378452

Let me guess, the cable you use for the mosfet D is 1mm cable which hammered until little bit flat, right?

 

[jkyoho]

Let me guess, the cable you use for the mosfet D is 1mm cable which hammered until little bit flat, right?

It's just 0.53mm OD enameled jumper wire, I think it's overkilled and after flatten a bit it has quite narrow space between cpu caps plus my twisting skill sucks

 

[abal1000x]

For some reason, I found I have IRFHS8342 keep failing on me since last time I did one switch lite with TOSHIBA eMMC, but I occasionally use a CPU mosfet from XBOX (TPN8R903NL), with nothing change just mosfet replace, the v2 I just got today instantly glitch success instead of ==* code when using IRFHS8342. I guess will stick with TPN8R903NL since they have easier solder even you wanna have the 1k resistor between Gate and Source.
View attachment 378451
View attachment 378452

Have read both datasheet and compare, seems the major difference is as usual TPN8R903NL have lower Rds(on), just like what we already suspect. And the other part is the switch timing which is faster. I think i need to appologize to someone whose said, the switching time could be a factor of glitching. Sorry pal, it seems switching time did have an effect on the glitching.

 

[jkyoho]

My initial mod switch lite has single NP2016 MOSFET setup and glitching fast as well, I guess having 10x 8342 msofet in the drawer was a bad move/investment

 

[abal1000x]

My initial mod switch lite has single NP2016 MOSFET setup and glitching fast as well, I guess having 10x 8342 msofet in the drawer was a bad move/investment

Unfortunately in the datasheet the voltage Vds and Vgs they used are different. Assume that the switching time is not that variate in 4.5V-10V, then the NP2016 have very bad switching time. So i think is not better than 8342. Seems the Xbox mosfet the good one.

TPN8R903NL
Rds 10.2 ~ 12.7 mΩ
tr 2.4 ns
td(off) 5.2 ns
tf 2.1 ns
td(on) 8.3 ns

IRFHS8342
Rds 20 ~ 25 mΩ
tr 15 ns
td(off) 14 ns
tf 5 ns
td(on) 5.9 ns

NP2016
Rds 9 ~ 12 mΩ
tr 11 ~ 25 ns
td(off) 35 ~ 70 ns
tf 30 ~ 60 ns
td(on) 10 ~ 20 ns

 

[frozenboy]

I just wanted to share a random pic I took during an install today to appreciate the picture quality of my microscope. It's an Andonstar AD207S. I recommend it

Is this stable with 1 cap only. Things just get interested. I've done 20 picofly with most of it using one mosfet and single line though both caps. It could be even better with only single cable to one cap.

 

[jkyoho]

Unfortunately in the datasheet the voltage Vds and Vgs they used are different. Assume that the switching time is not that variate in 4.5V-10V, then the NP2016 have very bad switching time. So i think is not better than 8342. Seems the Xbox mosfet the good one.

TPN8R903NL
Rds 10.2 ~ 12.7 mΩ
tr 2.4 ns
td(off) 5.2 ns
tf 2.1 ns
td(on) 8.3 ns

IRFHS8342
Rds 20 ~ 25 mΩ
tr 15 ns
td(off) 14 ns
tf 5 ns
td(on) 5.9 ns

NP2016
Rds 9 ~ 12 mΩ
tr 11 ~ 25 ns
td(off) 35 ~ 70 ns
tf 30 ~ 60 ns
td(on) 10 ~ 20 ns

Interesting, I was waiting 8342 shipment and found NP2016 on the v1 flex. Didn't really AB test between them but NP2016 has occasional 5s+ glitching which is acceptable for me.

 

[abal1000x]

Interesting, I was waiting 8342 shipment and found NP2016 on the v1 flex. Didn't really AB test between them but NP2016 has occasional 5s+ glitching which is acceptable for me.

Noted.
Back to the same conclusion then, switching time doesn't have correlation to the glitching time.
My last argument is the NP2016 which known to be used on hwfly have a very slow switching time compared to IRFHS8342. That is why i conclude that only Rds is matter, switching time is not. Smaller Rds better glitching time. And the best one is AON6512 which have 2.4 milliohms. I assume the number in datasheet is legitimate.

 

[Adran_Marit]

I don't check in for like a week and the thread catches on fire

I believe I saw rehius mention somewhere some of the issues are caused by the 2.74 firmware with 300mhz speed?

curious

 

[Uberfish]

If you have spare flex cable, it will be more easier to track the trace for each pin.
I don't have any flex cable right now.
You could check it using multimeter and/or microscope.

Post automatically merged: Jun 17, 2023

Let me guess, the cable you use for the mosfet D is 1mm cable which hammered until little bit flat, right?

I have a spare, I'm guessing it's something like this? I'm not totally sure.

edit: Thinking about it, the G pads should be on the inside pads, not the labelled ones...?

 

[abal1000x]

I have a spare, I'm guessing it's something like this? I'm not totally sure.

edit: Thinking about it, the G pads should be on the inside pads, not the labelled ones...?

Now you only need to check using multimeter, each pin goes to where.

The one that goes to the picofly is the Gate(G).

I remember that on the flex pin theres also the D part and the S/GND part.
You might test it using multimeter to know which pin.

From that you could check the D and S part should be like 12Ohms. And the S to GND should be zero ohms. And the D to GND should be 12ohms.

This is why i think that the hwfly design is good. I could add in the firmware to detect whether the connection is correct or not, since D, S, and G is given to the board. But picofly original design didn't so only G comes to the picofly. And we couldn't detect the condition of the connection of D and S. We assume the installer already installed it correctly.

 

[deeps]

Is this stable with 1 cap only. Things just get interested. I've done 20 picofly with most of it using one mosfet and single line though both caps. It could be even better with only single cable to one cap.

it's how I do all my installs. No issues.

 

[Danook28]

Just check if they are shorted to ground , on side is always grounded the other ones shouldn't be

No short... That meaning it is fine ic

 

[Dee87]

No short... That meaning it is fine ic

yes , u can also check m92 and and pi or usb for shorts if there are shorts then ur switch doesnt boot

but whats the issue with the board sinc eu that the bq chip was defektive?
like i said its normal that it gets warm since thats the charging ic

 

[Browbon]

CLear

 

[Switxh]

Sorry for the unregarding question because i have i idea where to ask. Today i open my Keychron K6 to clean. Unfortunaly, i dropped it. The usb c pad ripped off from the pcb, i try to fix it but unsucessful. Anyone know where i can get a solution or the alternative point for the usb c? Mine is 16 pin usb c. I appreciate any help.

No offense but I don't see why you would potentially derail this thread over this instead of just asking on reddit. There's a subreddit for literally everything. This isn't even remotely related to the topic we're discussing :/