fujinet #FujiNet Testing and Bug Reporting Thread

[+mytek]

Ok got some more info from Mr Robot, specifically the part number for the diode that is being used in the FujiNet board which is a DSS12UTR.

 

That particular part appears to have a forward voltage drop (Vf) of 500mv, which is not really that much better than a generic silicone diode which has a Vf= 750mv. Although its only going to set you back $0.15.

 

However Infineon has a Schottky Diode with a Vf of only 300mv which is perhaps enough to overcome the problem. It's P/N is BAT60AE6327HTSA1, and costs $0.46 at Mouser Electronics (can also be purchased from Digi-Key for the same price). It appears to be a drop-in for the DSS12UTR in the present FujiNet board layout (EDIT: just a wee bit smaller, so it might take some creative soldering), having what looks to be the same footprint. Might be worth a try to see if that fixes things.

 

Edit: I have something I'll be shipping to Mr Robot in about a week, and I have some parts I need from Mouser that I'll be ordering today, so I was thinking of adding a couple of these diodes to my order and tossing them into Mr Robot's shipment if he's game to change out the diode in his FujiNet to see if that works with the XEL. Of course he'll have to first temporarily bypass the diode on the FTDI board that he already installed (jumper it). However I don't want Mr Robot to feel obligated to do this, since he's already done more than enough. So if there is someone else with both a FujiNet and an unmodified XEL available for this test, I would be very willing to drop those diodes in the mail to them.

 

Edit2: Here's one with the exact same package type having a 340mv Vf MBR120VLSFT3G.

 

[mozzwald]

4 hours ago, mytek said:

There is an SMD dual gate open drain buffer that would probably work quite well, while also not taking much space (SN74LVC2G07) and is available from Digi-Key as P/N 296-13494-1-ND costing only $0.39 in single quantities.

 

This would be ideal, since the 1088XEL has been in production and offered as DIY for 3 years now, with quite a few out in the field. And it wouldn't be practical to expect all of those systems to be modified.

 

Edit: although I could see adding the SMT diode to the FTDI board from this point forward in all 1088XEL builds just to be on the safe side, and to better accommodate future SIO devices.

 

Yes Mr Robot's solution is good where someone wishes to use the current rev of FujiNet hardware, but obviously not ideal for the unskilled to implement on their own.

 

---------------

 

Also I was thinking that for the current crop of FujiNet devices that maybe there is a better diode to be used. Something with a much lower voltage drop. Without scanning through this the original very long FujiNet thread, I don't know what you are presently using and whether that is perhaps already the best device you can find.

 

After I finish this first batch of boards I will look into adding the buffer. That 2 line buffer chip looks promising; tiny and cheap. Thanks for the part tip.

 

In the mean time, I hope those wishing to use FujiNet on their 1088XEL systems are ok with adding the diode. If this is a no go for anyone, they can send it back for a refund and wait for the next revision.

2 hours ago, mytek said:

Ok got some more info from Mr Robot, specifically the part number for the diode that is being used in the FujiNet board which is a DSS12UTR.

 

That particular part appears to have a forward voltage drop (Vf) of 500mv, which is not really that much better than a generic silicone diode which has a Vf= 750mv. Although its only going to set you back $0.15.

 

However Infineon has a Schottky Diode with a Vf of only 300mv which is perhaps enough to overcome the problem. It's P/N is BAT60AE6327HTSA1, and costs $0.46 at Mouser Electronics (can also be purchased from Digi-Key for the same price). It appears to be a drop-in for the DSS12UTR in the present FujiNet board layout (EDIT: just a wee bit smaller, so it might take some creative soldering), having what looks to be the same footprint. Might be worth a try to see if that fixes things.

 

Edit: I have something I'll be shipping to Mr Robot in about a week, and I have some parts I need from Mouser that I'll be ordering today, so I was thinking of adding a couple of these diodes to my order and tossing them into Mr Robot's shipment if he's game to change out the diode in his FujiNet to see if that works with the XEL. Of course he'll have to first temporarily bypass the diode on the FTDI board that he already installed (jumper it). However I don't want Mr Robot to feel obligated to do this, since he's already done more than enough. So if there is someone else with both a FujiNet and an unmodified XEL available for this test, I would be very willing to drop those diodes in the mail to them.

 

Edit2: Here's one with the exact same package type having a 340mv Vf MBR120VLSFT3G.

 

The v1.0 FujiNet's use a PMEG2010ER for the TX line diode and has a 340mV forward voltage so it's on par with the ones you've suggested. I'm pretty sure the vertical style prototypes I shipped out also use this diode (the 2 I have here both have PMEG2010ER). The DSS12UTR was probably only used on the earlier prototypes (ice cream sandwich). @Mr Robot could confirm this by looking at the diode marking on D7 which should be "B5".

 

I will need to work with some folks who already have a FujiNet and 1088XEL to do some testing when I get the buffer chip design ready.

 

Edit: I think it best to move any more discussion on this over to the hardware thread so we don't drown out other issues

 

Edited August 4, 2020 by mozzwald
Add Hardware thread

[Mr Robot]

3 hours ago, jamm said:

At any rate, try connecting again after the first failure. It worked in my case.

 

I must have tried this dozens of times now in as many ways I can think of and nothing doing.

 

If someone can explain to me how to send a debug version of the fujinet firmware to the device without using the fujinet-flasher I'd be willing to try a version that never times out, just to see if it ever completes. I can always flash the latest normal one again afterwards.

 

Annoyingly when I got this Fujinet a while ago now, the firmware as it stood then, connected to my network OK. Before that I was on a Marsbar beta hardware and that connected OK toward the end but struggled at the beginning of development on that device. So I've had a small window of time where both Marsbar and the Latest vertical fujinets worked ok.

 

I'm going to try the marsbar again, it still has that firmware, it'll be an interesting test if nothing else!

 

 

[Mr Robot]

12 minutes ago, mozzwald said:

I will need to work with some folks who already have a FujiNet and 1088XEL to do some testing when I get the buffer chip design ready.

@DrVenkman has both devices, as do I.

I have another FTDI board on order and mine is socketed so I'll be able to go back and forth between diode/no diode.

 

I'll take my Fujinet apart and verify the diode later

 

[+DrVenkman]

49 minutes ago, Mr Robot said:

@DrVenkman has both devices, as do I.

 

Indeed I do, though my FTDI board is soldered in place. To access it for any minor surgery, I'd have to physically remove my XEL from its case, which, with all the front panel connections for the LEDs and XEL-CF3 interface, is kind of a PITA. I could do it (For Science!™) but I'd prefer not to have to.  

[+DrVenkman]

In other news, I flashed my FujiNet device this afternoon and tried out some XEX files. Very nice to have this functionality added back in. Interesting thing about Yoomp!, at least when tested with my 1200XL: when using HSIO, it loads in fits and starts, with an definite perceptible pause between each block, and stuttering to a complete stop after anywhere from a few blocks to just one or two. This behavior is the same whether loading from the FujiNet.Online TNFS host over the internet, or from my local server here on an RPiZero. If I disable HSIO, it loads just like you would expect a file to load - stead, even, perfectly-spaced SIO blocks with no hesitations or pauses. Just slower, as you would expect. 

 

[tschak909]

Just now, DrVenkman said:

In other news, I flashed my FujiNet device this afternoon and tried out some XEX files. Very nice to have this functionality added back in. Interesting thing about Yoomp!, at least when tested with my 1200XL: when using HSIO, it loads in fits and starts, with an definite perceptible pause between each block, and stuttering to a complete stop after anywhere from a few blocks to just one or two. This behavior is the same whether loading from the FujiNet.Online TNFS host over the internet, or from my local server here on an RPiZero. If I disable HSIO, it loads just like you would expect a file to load - stead, even, perfectly-spaced SIO blocks with no hesitations or pauses. Just slower, as you would expect. 

 

It's because Yoomp! has its own SIO loader which doesn't do well with speed divisors over 0x0A or so.

-Thom

[Mr Robot]

OK...

 

D7 is indeed a B5

 

I tried both of the Fujinet Marsbar boards I have here, both needed to have the Wi-Fi set up again due to my reconfiguring the Wi-Fi for the latest Fujinets. Both work fine to connect to the network... I timed the initial connection and for one it was 17 seconds and the other it was 21 seconds. It was much faster on subsequent reboots so it looks like just upping the connect timeout to ~30 secs should be ample. I wonder why Apple routers are so slow to connect to!?

 

I couldn't load any files off the tnfs servers I had configured (mine and fujinet.online) I got a file listing but couldn't select any files, I guess that has something to do with the new version of tnfsd and the old firmware?

 

 

[+DrVenkman]

1 minute ago, tschak909 said:

It's because Yoomp! has its own SIO loader which doesn't do well with speed divisors over 0x0A or so.

-Thom

Yet it has no problems loading from the SIDE Loader, like a lot of other programs do with their own custom SIO stuff going on (some of the Lucasfilm games, for instance). It also loads fine at Divisor 0 over an SIO2PC device, at least on my machines. 

[tschak909]

5 minutes ago, Mr Robot said:

OK...

 

D7 is indeed a B5

 

I tried both of the Fujinet Marsbar boards I have here, both needed to have the Wi-Fi set up again due to my reconfiguring the Wi-Fi for the latest Fujinets. Both work fine to connect to the network... I timed the initial connection and for one it was 17 seconds and the other it was 21 seconds. It was much faster on subsequent reboots so it looks like just upping the connect timeout to ~30 secs should be ample. I wonder why Apple routers are so slow to connect to!?

 

I couldn't load any files off the tnfs servers I had configured (mine and fujinet.online) I got a file listing but couldn't select any files, I guess that has something to do with the new version of tnfsd and the old firmware?

 

 

yup, TNFS protocol has changed slightly to vastly improve performance (the read directory call has changed)

-Thom

[jamm]

1 hour ago, DrVenkman said:

Yet it has no problems loading from the SIDE Loader, like a lot of other programs do with their own custom SIO stuff going on (some of the Lucasfilm games, for instance). It also loads fine at Divisor 0 over an SIO2PC device, at least on my machines. 

It's your wireless network.  Or at least the ESP32 is having a hard time with your network.  I have around 5 different ATR/XEX versions of Yoomp here and they all loaded at POKEY/0 without a hitch.  There was a blip once, but it was not repeatable (dropped packet at some point).

 

Try tweaking settings on your WiFi access point or router to see if you can get the behavior to be more consistent.  One user with an ASUS router was having a terrible time until they changed their channel width settings from 20MHz to 20/40 automatic.  (Which shouldn't have made a difference, but did...)

 

Edited August 4, 2020 by jamm

[+DrVenkman]

Just now, jamm said:

It's your wireless network.  Or at least the ESP32 is having a hard time with your network.  I have around 5 different ATR/XEX versions of Yoomp here and they all loaded at POKEY/0.  There was a blip once, but it was not repeatable (dropped packet at some point).

 

Try tweaking settings on your WiFi access point or router to see if you can get the behavior to be more consistent.  One user with an ASUS router was having a terrible time until they changed their channel width settings from 20MHz to 20/40 automatic.  (Which shouldn't have made a difference, but did...)

 

Well, okay. Let's call it what it is then, really: it's the ESP32 itself. My wifi is as close to perfect as wireless can get - no trouble with multiple simultaneous 4K streams all over the house on 802.11ac; multi-hundred-MBPS downloads on devices ranging from Apple TV's, an LG TV with integrated Roku; iPads; iPhones; PC laptops ... None of them have file transfer hiccups and stutters, let alone when communicating at the blistering speed of 125 kilobits per second.   

 

 

[jamm]

1 hour ago, Mr Robot said:

 I wonder why Apple routers are so slow to connect to!?

I wouldn't blame it on any particular router. The ESP32's WiFi has proven a little finicky for some people, unfortunately.  (See post to @DrVenkman above.)  The good news is that they're frequently providing new updates to the developer libraries, so it's entirely possible it might improve over time.  In the mean time, our only choice is to play around with WiFi settings... so far we haven't run into a system it wouldn't connect to at all, at least.

[jamm]

4 minutes ago, DrVenkman said:

Well, okay. Let's call it what it is then, really: it's the ESP32 itself. My wifi is as close to perfect as wireless can get - no trouble with multiple simultaneous 4K streams all over the house on 802.11ac; multi-hundred-MBPS downloads on devices ranging from Apple TV's, an LG TV with integrated Roku; iPads; iPhones; PC laptops ... None of them have file transfer hiccups and stutters, let alone when communicating at the blistering speed of 125 kilobits per second.   

 

 

I never said it was anything other than the ESP32 being touchy.  That's life on the radio waves, man.

 

[+DrVenkman]

10 minutes ago, jamm said:

I never said it was anything other than the ESP32 being touchy.  That's life on the radio waves, man.

 

Except that it’s not even that. It’s Yoomp! and how it loads on FujiNet’s loader. I made a comparison. Both loaded from the same little RPiZeroW.  
 

“Let’s go to the video tape ... “

 

 

[jamm]

9 minutes ago, DrVenkman said:

Except that it’s not even that. It’s Yoomp! and how it loads on FujiNet’s loader. I made a comparison. Both loaded from the same little RPiZeroW.  
 

“Let’s go to the video tape ... “

 

 

 

So the differences between what you're doing and what I'm doing are:

1. We may not be loading the same XEX file. I tried several versions, so I doubt that's it. Here are 3 SHA1 hashes for XEX's I tried:

322C6346CEACABC3C311F5F5D8C0645AF906D549
960F0626BD7AFC1DEADF6D14708DF2DD9C946C80
172D135044A35C460A6AF754117DEE6A7D3402C3

2. We're not talking to the same TNFSD server. Unlikely to be an issue, otherwise you'd have issues loading other things, too.

3. We're not on the same Atari hardware.  I'm running on my Incognito 800.

4. We're not on the same wireless network. Your video suggests at least some things load fine, so it might not be this, but I've seen enough WiFi weirdness not to write this off.

 

Unless we can reproduce your problem, the only way to make any headway is to see the debug output from FujiNet. Are you able to collect that during your load of Yoomp and post it?

 

[+DrVenkman]

7 minutes ago, jamm said:

Unless we can reproduce your problem, the only way to make any headway is to see the debug output from FujiNet. Are you able to collect that during your load of Yoomp and post it?

FWIW, I get the very same kind of stuttering, herky-jerky load from FujiNet.Online as I do from my local server, so it's not my little RPi chugging, I don't think. Other files load much, much more smoothly and faster - I tried booting up The Goonies from my local server; Spelunker from FujiNet.Online; then Hitchhiker's Guide to the Galaxy from my local server ... all load up fast and smoothly as expected.

 

I'm powering the FujiNet via a separate microUSB power supply. I can try later or tomorrow sometime to power it from my laptop and try to catch the debug output at the same time. Right now I'm watching a livestream of SpaceX's attempted "hop" of their Starship test vehicle, and later tonight there's a near-perfect ISS pass coming, so I might not get to it.

[jamm]

1 minute ago, DrVenkman said:

FWIW, I get the very same kind of stuttering, herky-jerky load from FujiNet.Online as I do from my local server, so it's not my little RPi chugging, I don't think. Other files load much, much more smoothly and faster - I tried booting up The Goonies from my local server; Spelunker from FujiNet.Online; then Hitchhiker's Guide to the Galaxy from my local server ... all load up fast and smoothly as expected.

 

I'm powering the FujiNet via a separate microUSB power supply. I can try later or tomorrow sometime to power it from my laptop and try to catch the debug output at the same time. 

You could also try copying to file to SD and loading it from there.  The same bootloader is used regardless of whether it comes from SD or TNFS.

[jamm]

Posting this mostly to play around a little with my cheapie USB HDMI capture stick, but also shows how fast Yoomp can load even over WiFi.

 

 

[+DrVenkman]

24 minutes ago, jamm said:

You could also try copying to file to SD and loading it from there.  The same bootloader is used regardless of whether it comes from SD or TNFS.

Okay, I tried your suggestion. The very same XEX exhibits the very same behavior when loaded from the SD card.  I then swapped in my Incognito 800 and see the exact same behavior.

 

So ... maybe something weird about this particular XEX file.  As I said before, it loads perfectly via an SIO2PC at POKEY divisor 0, and also loads perfectly from the SIDE Loader on my U1MB-equipped machines.

 

If you want to try to figure it out, here you go.

 

 

Yoomp! _ V1.1 _ NTSC.xex

[StickJock]

38 minutes ago, DrVenkman said:

Right now I'm watching a livestream of SpaceX's attempted "hop" of their Starship test vehicle

I got goosebumps watching this!

[jamm]

13 minutes ago, DrVenkman said:

Okay, I tried your suggestion. The very same XEX exhibits the very same behavior when loaded from the SD card.  I then swapped in my Incognito 800 and see the exact same behavior.

 

Well, that definitely rules out WiFi, then! 

 

I compared your Yoomp upload with the files I have and it matches one of the ones I tested against, so it's not the XEX, either.

 

And you've tried with your Incognito 800, so that doesn't leave us a lot of options...

 

I went so far as to load up a different FujiNet box to test with and... it's also fine.

 

All I can think of at this point is that some weird edge case got unintentionally fixed in firmware between the version you're running and the version I have - but there have been very few changes recently.  That's the last thing left to try, though, so please update to the firmware @mozzwald just posted this afternoon and see if anything changes.

 

[+DrVenkman]

3 minutes ago, jamm said:

All I can think of at this point is that some weird edge case got unintentionally fixed in firmware between the version you're running and the version I have - but there have been very few changes recently.  That's the last thing left to try, though, so please update to the firmware @mozzwald just posted this afternoon and see if anything changes.

I did. No change.  Doesn't matter if it's off Wifi or the SD card, and on two different machines (Incognito 800 and 1200XL+U1MB). 

Can you post the file from your test video? I can try that one on the SD card and my LAN server. 

[+DrVenkman]

9 minutes ago, StickJock said:

I got goosebumps watching this!

That was badass. Who knew grain silos could fly? FWIW, SpaceX has FAA approval already for a 20km test flight and landing no earlier mid-August. That one will have 3 engines rather than just one, nose cone and fins. 

[jamm]

5 minutes ago, DrVenkman said:

I did. No change.  Doesn't matter if it's off Wifi or the SD card, and on two different machines (Incognito 800 and 1200XL+U1MB). 

Can you post the file from your test video? I can try that one on the SD card and my LAN server. 

That's wild.

I'm pretty sure the one I posted the video of is binary-identical to the one you posted, but here it is anyway.  Yoomp_NT is the one in the video and the same as yours.  Yoomp1 looks like an earlier version and results in a blank screen after loading on my Atari, but you can see the loading process and hear the music when it loads.  The third version I have is PAL, so it stops loading after it detects NTSC.

 

 

yoomp_nt.xex yoomp1.xex