Another Bush internet.tv mod thread

[IanJeffray]

This implies there’s an empty space for a VGA crystal on the IBX pcb?

There is a space for the XTAL. From the RPC schematic it looks like it should be 24Mhz.

The "VGA crystal" on all other machines (pre RiscPC/A7000) is 25.175MHz (Though some A5000s have just a 25.000MHz fitted). The reason this was added is that VIDC can't make a true VGA signal from 24MHz, so on earlier machines without a VIDC Enhancer, the VGA output actually scans slowly (56Hz IIRC). Perhaps VIDC20 can derive from 24MHz, but it'd seem an odd choice to me.

[baz4096]

Perhaps VIDC20 can derive from 24MHz, but it'd seem an odd choice to me.

From the TRM:

In addition to clk64 there is a 24MHz clock source which
provides a reference clock to both VIDC20 and the
FDC37C665 device.

and

By programming the r, v and n registers in
VIDC the Pixclk output can be varied anywhere from 8Mz
up to in excess of 110 MHz. The comparison frequency
is determined by the ratio: 2412r and the resulting pixel
clock is equal to 24 x (vir) x (11n) MHz.

[thecellartroll]

The "VGA crystal" on all other machines (pre RiscPC/A7000) is 25.175MHz (Though some A5000s have just a 25.000MHz fitted). The reason this was added is that VIDC can't make a true VGA signal from 24MHz, so on earlier machines without a VIDC Enhancer, the VGA output actually scans slowly (56Hz IIRC). Perhaps VIDC20 can derive from 24MHz, but it'd seem an odd choice to me.

25.175Mhz makes sense as that is what I found most commonly when googling the XTAL in VGA circuits. I can't find a picture of the A7000 board which shows the XTAL value clearly (I've actually been tempted to buy CJEs knackered A7000 board for reference purposes!) used on that, which I assume would be the same.

It is of course likely that I'm not reading the RPC schematic properly but it looks like VIDC20 is getting the clock signal from 24Mhz X2 (Refclk). It could be that RPC is totally different though. As far as I can see it doesn't use separate XTALs for VGA and sound.

[IanJeffray]

it looks like VIDC20 is getting the clock signal from 24Mhz X2 (Refclk)

More's the interesting question then perhaps how it's generating any video signal without any crystal? 7500 must be different to VIDC20 in that respect.

[thecellartroll]

More's the interesting question then perhaps how it's generating any video signal without any crystal? 7500 must be different to VIDC20 in that respect.

For reference I should probably say that unplugging my HSYNC bodge turns off the VGA monitor so it isn't just using the RGB SCART signals to produce a picture.

[IanJeffray]

More's the interesting question then perhaps how it's generating any video signal without any crystal? 7500 must be different to VIDC20 in that respect.

For reference I should probably say that unplugging my HSYNC bodge turns off the VGA monitor so it isn't just using the RGB SCART signals to produce a picture.

No, what I mean is that it IS generating a video output WITHOUT the 24MHz xtal fitted. How?
Looking at the VIDC20 manual it becomes clear - VIDC20 has 3 clock sources.

[thecellartroll]

Actually the A7000 board doesn't help much. Stolen from CJE's ebay page and annotated by scouring all the pictures of rotten A7000s/A7000+s:

[thecellartroll]

What may be more revealing is that 25.175MHz seems to be readily available with the same footprint that is on the board, but 24Mhz does not.

[IanJeffray]

Actually the A7000 board doesn't help much. Stolen from CJE's ebay page and annotated by scouring all the pictures of rotten A7000s/A7000+s:
A7000 board.jpg

Well, it does help insomuch as the location of the 24MHz is by the Superio. Now Acorn are notorious for dragging clocks all over the board willynilly, but that'd be a poor location for something that really was shared with VIDC20 in the 7500.

Note also that it's a 24MHz crystal not an oscillator, so the drive must be from the Superio, further suggesting it's only used for the Superio.

[IanJeffray]

Additional note - the 7500FE datasheet goes to some length to eulogise how "an enormous number of frequencies are possible" from the 32MHz internal reference clock alone, further suggesting that no other would be required. (If you have 32 and VCO, there's not much need for 24)

[baz4096]

Well, it does help insomuch as the location of the 24MHz is by the Superio. Now Acorn are notorious for dragging clocks all over the board willynilly, but that'd be a poor location for something that really was shared with VIDC20 in the 7500.

I think perhaps the 7500 generates this reference clock internally:

From the 7500 datasheet:

The internal i_oclk32 signal, derived from I_OCLK input provides a reference clock which is recommended to be 24MHz.

So that probably rules out the missing crystal as to why VGA isn't working.

[thecellartroll]

Additional note - the 7500FE datasheet goes to some length to eulogise how "an enormous number of frequencies are possible" from the 32MHz internal reference clock alone, further suggesting that no other would be required. (If you have 32 and VCO, there's not much need for 24)

It does make one wonder why the extra XTAL position on the board. Is there a socket for those? This board really doesn't like repeated soldering at the same point so I'd rather fit a socket so that I can just plug the XTAL in and see if it works. I think I'll try 25.175MHz.

[IanJeffray]

It does make one wonder why the extra XTAL position on the board. Is there a socket for those? This board really doesn't like repeated soldering at the same point so I'd rather fit a socket so that I can just plug the XTAL in and see if it works. I think I'll try 25.175MHz.

25.175 would be a weird choice when that's not fitted to any A7000+

You can make a crystal socket out of a machined-pin DIP socket - just push out the pins that aren't required.

[thecellartroll]

I suppose it makes sense to mention that the missing XTAL delivers a clock to HCLK (pin 75) on the SoC via a resistor which is also not fitted.

[IanJeffray]

I suppose it makes sense to mention that the missing XTAL delivers a clock to HCLK (pin 75) on the SoC via a resistor which is also not fitted.

Makes sense. Looks like that's where the 'missing' XO on the A7000 board would also feed?

[thecellartroll]

Makes sense. Looks like that's where the 'missing' XO on the A7000 board would also feed?

I have no way to confirm that, but the IBX board already has a 48mhz XTAL present.

[thecellartroll]

Limit the VCO range and use the prescaler within the video and sound macrocell to obtain a lower pixel
rate than the VCO frequency. The VCO range may require constrainment so much that it cannot provide
the highest frequencies for operation of the video and sound macrocell. For this scenario, a single highfrequency clock can be fed into CL-PS7500FE on the HCLK pin and selected for the pixel clock

This could explain it? Though the table below suggests a frequency of 120mhz might be required?

[IanJeffray]

Though the table below suggests a frequency of 120mhz might be required?

No. That's a table of the frequencies that can be VCO generated using different r and v

[IanJeffray]

Makes sense. Looks like that's where the 'missing' XO on the A7000 board would also feed?

I have no way to confirm that, but the IBX board already has a 48mhz XTAL present.

That's the main system clock isn't it?

[thecellartroll]

That's the main system clock isn't it?

I believe the main system clock is the 64Mhz XTAL above it (which gives a system clock speed of 56mhz unless you add a resistor).

[IanJeffray]

That's the main system clock isn't it?

I believe the main system clock is the 64Mhz XTAL above it (which gives a system clock speed of 56mhz unless you add a resistor).

I believe it's running at 48MHz. The 64M XO is the memory clock. The resistor mod uses that clock as the system clock.
The IBX250 part is 56MHz rated, but AFAIK it's only run at 48MHz - you'd need a funky divider to get 56MHz from 64MHz - basic dividers are always integer.

[thecellartroll]

So perhaps the issue is that the system clock from the SoC is used for the FDC and therefore can't be used for VGA and that's why the extra XTAL position is there?

So then we're back to trying to guess what value it might be. Shame no-one sells a big bag of different values for a tenner so I could just try swapping in different units.

[IanJeffray]

So perhaps the issue is that the system clock from the SoC is used for the FDC and therefore can't be used for VGA and that's why the extra XTAL position is there?

Is it? Have you actually confirmed that by tracing the clock in to the SMSC? Hmm.

So then we're back to trying to guess what value it might be. Shame no-one sells a big bag of different values for a tenner so I could just try swapping in different units.

I'd start by just having it set up as per the A7000+. Just sticking anything in, in vaguely the right sort of range, would let you see if it magically caused the output to work in non-TV modes.

It's a clock input, not a crystal, so you could just feed it in from a sig gen if you want to "play with values". If you have a sig gen with even a few MHz bandwidth, but a decent sample rate ARB, you can stick in a squarewave as the user waveform and usually get something acceptable out in the tens of MHz range and use the base rate as the multiplier (I did exactly this just recently when @AndyMc1280 asked me to try overclocking his Alpha A5000 -- sadly it wasn't stable above the stock 33MHz).

[thecellartroll]

Is it? Have you actually confirmed that by tracing the clock in to the SMSC? Hmm.

CLK14 on the FDC appears to get its clock signal from the CPU clock through an ACT74 (IC19)

It's a clock input, not a crystal, so you could just feed it in from a sig gen if you want to "play with values". If you have a sig gen with even a few MHz bandwidth, but a decent sample rate ARB, you can stick in a squarewave as the user waveform and usually get something acceptable out in the tens of MHz range and use the base rate as the multiplier (I did exactly this just recently when @AndyMc1280 asked me to try overclocking his Alpha A5000 -- sadly it wasn't stable above the stock 33MHz).

Interesting. I don't have anything that will make a clock signal close, but some Googling tells me I might be able to persuade a Raspberry Pi to act as a signal generator (output at 3.3v) to try different values.

[thecellartroll]

Also, leaving it alone for two weeks seems to have killed the keyboard all that happens now with the keyboard plugged in is the lights all flash three times and the IBX boots without recognising any keypresses at all. I wonder if some bit on the CMOS RAM can get stuck on a setting that kills the keyboard.

[thecellartroll]

The last point is confirmed. Changing the CMOS RAM chip brought the keyboard back to life.

I changed it thus:

Thanks to @Myelin for his Acorn Electron CMOS RAM board design

[thecellartroll]

Interesting. I don't have anything that will make a clock signal close, but some Googling tells me I might be able to persuade a Raspberry Pi to act as a signal generator (output at 3.3v) to try different values.

Bah, it doesn't work on Pi 3 or Pi400. Maybe Pi Zero, but the Raspberry Pi Foundation should have been called the "Different video connector on every model foundation!"

[thecellartroll]

Update: now that (some of) my mail has finally arrived I can confirm that a 24Mhz XTAL in the X3 position does not cause VGA to work properly. Can't test 25.175 yet as Hermes has it. I did fit the missing resistors and can confirm that the clock signal does reach the SOC.

Oh well.

It is a bit less annoying to play with with a proper CMOS RAM circuit in place. I am also now convinced that delete power on does absolutely nothing.

[thecellartroll]

Mr Evri came bashing at my door today so...

Alas still no change in behaviour. Configuring VGA or SVGA monitor type still turns off video output entirely. Bah.

Nice clock to the SoC...

Bizarrely the screechy Acorn CRT now works with Monitor Type 0 with or without X3. Bizarre.

[thecellartroll]

Thought people might be interested to see some up-close pictures of the IBX200 board since there aren't many online.

Quite a lot in common with the IBX100 board of course but there are some differences:

No SIMM slot: the RAM is soldered to the board.

No 4 pin XTALs.

The PS2 connector is on the front of the board, which would make it possible to do a nice case mod. I think it might not work though due to the unpopulated resistors around the header.

There is a 7 pin cable header next to it. Interesting that the actual header is soldered in. Perhaps this board was designed for something with a keyboard socket on the back of the case?

There are two more mysterious unpopulated headers either side of the FDC. One 14 pin and one 5 pin. Perhaps this board has easier access to the additional features of the FDC?

I can't see RPC style audio anywhere, even unpopulated. Maybe this board can't have sound added.

IBX200 is a single board design so the SCART sockets, the LEDs and the IR sensor are hard wired to it. There is no video header so I don't think that putting RISC OS ROMs in the same way as the IBX100 board would give video out as there would then be nothing to switch the SCART socket on.