RGB to HDMI using a Pi Zero and a small CPLD

[hoglet]

i just burned my third xl9572.. i ordered new one but i am afraid my soldering iron isn't small enough for this little smd ics.. is some RGBtoHDMI board for selling somewhere?

I'm not sure I understand what you mean by "i just burned my third xl9572".

Do you mean you damaged it trying to solder it in? Or there were shorts that destroyed it? Maybe posting a photo would help us understand and offer better advice.

As far as I'm aware, no one is making fully assembled RGBtoHDMI boards yet.

Dave

[hoglet]

Do you think rgbtohdmi will ever support the extra features offered by Rob’s VideoNula?

That's an interesting question... Ian has come a long way down this path with the palette work he's been doing, with some fairly impressive results if you look earlier in this thread:
viewtopic.php?p=229621#p229621

But the programming model is different to Video NuLA (the Palette obviously can't be directly memory mapped at &FE2x, like it is with VideoNuLA). What this means is the all of the great work Rob has been doing (Spectrum emulator, Amstrad CPC emulator, Doom, etc) is not going to "just work".

One way to address this might be to design a new Palette API (that games could use) with separate Video NuLA and RGBtoHDMI implementations. I think this is feasible, but as far as I know, no one is working on it currently.

I should also say, there is much development work (both myself and Ian) happening on RGBtoHDMI (in the dev branch), but we are some way off this being ready for release.

This includes:
- Palette control with in-band data
- Increased capture depth (6 bits/pixel, rather than 3 bits/pixel)
- Increased capture rate
- Sync polarity detection
- UI now usable with lower resolution frame buffers
- A new configuration file format that supports multiple machine profiles in separate files (these replace the horrible cmdline.txt file format)
- Persisting the current calibration to the SD Card
- Screen capture (in PNG format) directly to the SD Card

Some of these features (like the increases capture depth) will require the CPLD to be updated, but we are trying hard to ensure as much as possible still works with all previous CPLD releases.

Dave

[aotta]

Do you mean you damaged it trying to solder it in? Or there were shorts that destroyed it? Maybe posting a photo would help us understand and offer better advice.

I probably made some short in soldering, and burned IC (literaly... with red lamps and smoke!) when power on the BBC and flashing with xilink...

Waiting for new IC, and next time i'll start another PCB from scratch (i have components for one or two).
Thanks Hoglet for support, and my compliments for your great work!

[hoglet]

I probably made some short in soldering, and burned IC (literaly... with red lamps and smoke!) when power on the BBC and flashing with xilink...

I had not done any surface mount soldering before this project. The technique to research is called drag soldering.

Here are a couple of video of it being done perfectly:
https://www.youtube.com/watch?v=nyele3CIs-U
https://www.youtube.com/watch?v=wUyetZ5RtPs

The correct soldering iron bit looks like this:

__1.jpg (5.78 KiB) Viewed 2246 times

It's huge compared to the pins, but that doesn't matter. Don't try to solder each pin individually. And don't use a fine pointed bit.

There are three secrets to getting this right:
- first, make sure the device is perfectly placed before tacking down (I use blu-tac initially)
- second, use lots of extra flux. It works like magic
- third, carefully inspect for shorts (especially at the ends) in good light with some sort of magnification

The CPLD has 0.8mm lead spacing, which is actually not too bad. You can get devices with 0.5mm spacing, but even these are doable with a bit of patience.

Thanks Hoglet for support, and my compliments for your great work!

Much appreciated (and also IanB has also been working on this as well).

Dave

[aotta]

The correct soldering iron bit looks like this:
__1.jpg

It's huge compared to the pins, but that doesn't matter. Don't try to solder each pin individually. And don't use a fine pointed bit.

Dave

Ordered a soldering spade tip.. Thank you!

[aotta]

Waiting new soldering tip, made new try and burned another xilinx... but some progress this time! i flashed the cpld and seen an acorn bbc on screen and the menu on screen for a few second, before smoke started!

But I have a question about link p6, i didn't find the explanation on the wiki: it must be open or closed?

[hoglet]

Waiting new soldering tip, made new try and burned another xilinx... but some progress this time! i flashed the cpld and seen an acorn bbc on screen and the menu on screen for a few second, before smoke started!

Can you possibly post a picture of the soldering? Usually solder bridges are pretty obvious. I wonder if something else is going on here, like somehow you are powering the CPLD from 5V rather than 3.3V.

But I have a question about link p6, i didn't find the explanation on the wiki: it must be open or closed?

It's unused in the current stable release of the software, so just leave it open.

Dave

[aotta]

Thank you Dave for the support.. that's a pic of board (with burned xc9572xl): do you notice some mistakes?

[danielj]

I see a bridge between the pins under c6...

[JonC]

And possibly on the two pins next to U1, it's a bit blurred so hard to tell.

[Ramtop]

I'm not sure how many of those little 44-pin Xilinx CPLDs I've soldered, but it's certainly more than 100 of them at this point.

Dave's advice is good. Position the device carefully, really carefully. Use magnification if your eyes are (like mine) less than perfect.

Flux is critical. Good flux makes a huge difference; I use MG Chemicals liquid flux, but the AMTECH tacky stuff is good as well. Drag or tap soldering a fine pitch QFP chip is so much easier when the solder flows well. Don't stint on the flux, use loads and loads if necessary.

Some kind of magnification is a godsend when checking for bridges. A microscope isn't necessary, a cheap magnifying lamp from Ebay - the kind that nail technicians use - is fine for 0.8mm chips.

Oh, and make sure the chip is oriented correctly. Sounds silly, but it's really easy to do. I killed an XC9536XL stone dead last weekend because I was tired and soldered it in the wrong way round

Not really relevant here, but there's also one thing I've found that PCB designers can do to help with surface mount hand soldering; longer pads. The chip footprints in PCB layout software are designed for machine assembly, so the pads are only just big enough and no more. Double or triple the length of the pad and it makes soldering without bridges so, so much easier because any excess solder has somewhere harmless to go. I adopted this practice after seeing a training PCB that was designed for use in a soldering class, it had extra long pads for all the SMD chips. The image below shows how I do my PCBs now (the CPLD isn't soldered, it's the one I killed included for scale).

[aotta]

the photo was taken after some soldering pass on the xl9572 and there was a bridge, but it's not that the failure.

I found that the pins that go short are the two connected to C5, but only after a few seconds the boad is powered!
Then, after cooled and disconnected, the shorts disappear...
May be the xilinx is damaged? i've no more IC to test...

[myelin]

Not really relevant to the case where you’re just making one board, but if you’re doing more than one (or it’s a board with a ton of pins, or any BGA or QFN parts) using a stencil and a hot plate for soldering is much much easier than doing it by hand, and far less failure prone — you’ll always get the right amount of solder. JLC will do stencils with a PCB order for $8, so they’re quite affordable too

[IanB]

I just figured out a way to test how much lag there is with this converter:
I connected a HDMI to VGA converter lead to the HDMI output and then wrote a program to flash the screen briefly.
The VGA converter lead is just a simple D to A converter and doesn't introduce much delay itself.
I then looked at one of the RGB inputs (from the computer) and one of the RGB outputs from the HDMI to VGA lead on a scope and it shows the total delay is 2.4 milliseconds:

(Yellow is the computer RGB source, Cyan is the output of the converter)
If I switched on multiple buffering, the delay went up by 20ms for each buffer as expected.

[hoglet]

Wow, that's spectacularly good. I'm rather amazed that it's so low.

If you change the vsync line, does that directly alter the latency by the same amount?

[IanB]

If you change the vsync line, does that directly alter the latency by the same amount?

Yes the delay goes up as the vsync line goes down the screen.

[IanB]

I started a thread on the vogons board for PC related applications of the converter if you want to follow that:

https://www.vogons.org/viewtopic.php?f=63&t=66194

[tricky]

That's great.
It feels like you could display a counter on the beeb and connect something via component and a display via HDMI and record the two with a 50hz+ camera to see how many frames/fields delay the TV adds and possibly adjust the vsync line to calculate the sub frame delay.

[Elminster]

Thread has got huge now. Quick question did anyone every design any sort of 3d printed case for the Beeb version? The atom one fits in a project case and got me thinking to putting this in one.

[aotta]

Thread has got huge now. Quick question did anyone every design any sort of 3d printed case for the Beeb version? The atom one fits in a project case and got me thinking to putting this in one.

I think the standard PiZero case used IanB is perfect, look his post viewtopic.php?f=3&t=14430&p=225297

I copied his idea of course:

[Elminster]

Ah yes, missed that post as I spent a month or 2 in the wilderness (figeratively).

Looks the same as this ebay item, except it is cheaper: Pi Case

[aotta]

Ah yes, missed that post as I spent a month or 2 in the wilderness (figeratively).

Looks the same as this ebay item, except it is cheaper: Pi Case

yes, it's the one! youl'll have only to drill some holes for buttons and leds

[trixster]

My two just run naked! maybe I should investigate casing them up

[Elminster]

My two just run naked! maybe I should investigate casing them up

I have soo many things running naked I am afraid they will touch, get over excited, and explode.

[IanB]

Ah yes, missed that post as I spent a month or 2 in the wilderness (figeratively).
Looks the same as this ebay item, except it is cheaper: Pi Case

Yes, they are lots of listings for them at varying prices and quantities:
This is slightly cheaper again:
https://www.ebay.co.uk/itm/Protective-C ... 3487349305
2x works out at £3.60 compared to £3.89 above and 5x is £8.75 compared to £9.68 in my original link.

There is even a transparent option now but it's a bit more expensive (£2.80):
https://www.ebay.co.uk/itm/For-Raspberr ... 3402771283

yes, it's the one! youl'll have only to drill some holes for buttons and leds

Here is a drilling template pdf from kicad:

Drilling template.pdf

(12.11 KiB) Downloaded 25 times

Print and cut it out, push pinholes through the centre points and place face down on the inside of the case. It should just fit in the narrow direction with little or no slack but you will have to centre it in the long direction. The holes should be closest to the side with the hdmi and usb connectors and you should push the template towards that side if there is any slack as the board is mounted slightly off centre in the narrow direction. (You have to make pinholes because the printing is face down on the inside of the case.) As the cases are so cheap, it's worth getting an extra one in case of drilling mistakes.

Also these threaded metal standoffs are exactly the right size to support the board They screw directly into the plastic case and the screws supplied with the plastic case fit into the top of the standoffs to hold hoglet's board in place:

I only used two in the above as the 40 way header provides support on the other side but you could fit four of them.

The standoffs are Ettinger part no. 05.12.113 and they can be found here:
https://uk.farnell.com/ettinger/05-12-1 ... dp/2494583
and here:
https://cpc.farnell.com/ettinger/05-12- ... dp/PC01982

[Elminster]

Thanks, I have a big box of stand offs, probably all the wrong size. Plan to expand from 1 to 3 RGB hdmi boards. And want to stick them out the way without damage them, these all should be ideal.

[Elminster]

Ordered 4 transparent cases, only need 3 but at that price and the fact it takes a while to come, I ordered a spare.

Also ordered 10 v4 boards from seeed.

My current RGBHDMI is from first batch Dave sent out so not sure which hardware revision that was.

I have a spreadsheet with all the bits for beeb hdmi, atom hdmi, atom noise killer and beeb FPGA, just waiting on me checking on what I already have and then ordering the missing bits. (FPGA itself is already in transit)

[sparkyhall]

Also ordered 10 v4 boards from seeed.

If you have any spare bare boards I will buy one from you

Chris

[hoglet]

If you have any spare bare boards I will buy one from you

DanielJ has a bunch of these boards as well I think.

Dave

[sparkyhall]

Thanks Dave, I'll send DanielJ a PM.

Chris