VOGONS


New clock gen for tyan s1564d (Research)

Topic actions

First post, by Sphere478

User metadata
Rank l33t++
Rank
l33t++

https://datasheetspdf.com/mobile/556939/Integ … /ICS9169C-272/1

Think this will fit?

https://www.ultimateretro.net/fr/motherboards/5131

The chip that it has only goes to 66mhz.

I want 75 and 83 options 🙁 🤣

Attachments

Last edited by Sphere478 on 2022-06-17, 02:58. Edited 1 time in total.

Sphere's PCB projects.
-
Sphere’s socket 5/7 cpu collection.
-
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
-
Tyan S1564S to S1564D single to dual processor conversion (also s1563 and s1562)

Reply 1 of 105, by snufkin

User metadata
Rank Oldbie
Rank
Oldbie

My counting isn't always the best, but it looks like the W48C60 has 28 pins, and that ICS9169 has 32 pins. So it'd need some fairly major modding/adapter to make it fit, which is probably a bad idea on clock lines.

[I find clock gen part numbers are pretty unhelpful. So that ICS9169C-272 is 32 pin, but the ICS9169C-231 is 28 pin. Some ICS9169 are 20 pin. So it looks like it'll be either a case of finding out what chips are available, then trying to find a datasheet to find it does 83MHz, or looking through piles of datasheets trying to find a part number and then seeing if you can actually find the part anywhere. That -231 looks like it does both 75 and 83.

But, can't just find a datasheet for the W48C60-402G, so can't be certain if it'll be pin compatible, given that ICS went as far as using the same base part number for different numbers of pins]

[thought I remembered looking about this before, National Semiconductor CGS605 is apparently directly compatible with the W48C60-402 (it doesn't mention the 'G'), so the datasheet for that can be used to check what might be compatible: http://pdf.datasheetcatalog.com/datasheet/nat … or/DS012491.PDF

So it has 4 Processor clock outputs on pin 6,7,9,10 and 6 Bus clock outputs on 15,16,18,19,21,22, along with a floppy disk controller clock on 24 and a keyboard controller clock on 25

A part like the ICS9169C-231 has 8 Processor clock outputs, 6 Bus clocks, and is close to but not actually pin compatible (e.g. pin 15 is a Bus clock on the -402 and a Processor clock on the -231 ]

Reply 2 of 105, by Sphere478

User metadata
Rank l33t++
Rank
l33t++

🤔 darn

Someone else I thought found a data sheet for it at some point but I can’t figure out where that went. There was a old thread where I was asking about this but can’t remember where it went.

What about some kind of adapter?

Sphere's PCB projects.
-
Sphere’s socket 5/7 cpu collection.
-
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
-
Tyan S1564S to S1564D single to dual processor conversion (also s1563 and s1562)

Reply 4 of 105, by Repo Man11

User metadata
Rank Oldbie
Rank
Oldbie

If you want to search your old posts click on the upper right icon, then your user name, then on the magnifying glass, and they will be shown in chronological order.

Attachments

"I'd rather be rich than stupid" - Jack Handey

Reply 5 of 105, by Sphere478

User metadata
Rank l33t++
Rank
l33t++
Repo Man11 wrote on 2021-12-15, 01:34:

If you want to search your old posts click on the upper right icon, then your user name, then on the magnifying glass, and they will be shown in chronological order.

Yeah, that’s where I looked, can’t seem to remember the correct keywords, I’m not even entirely sure it was on vogons but someone managed to find it.

snufkin wrote on 2021-11-25, 17:34:

Looks like there were lots of variants for frequency generators. Couldn't find a datasheet for the IC Works W48C60-402G on your board, but did find the National Semiconductor CGS605 that claims to be compatible (also the IC Systems ICS9159-02 and the Cypress CY2254). As you say, it only supports 50, 60 and 66MHz for the Processor clock (with the Bus clock fixed at half that). I haven't found a good list of all the possible variants, but did find the ICS9159C-14: https://pdf1.alldatasheet.com/datasheet-pdf/v … CS9159C-14.html

That does have a 75MHz PClk (with a 37.5MHz BClk) and looks to be pin compatible, though I haven't looked closely through both datasheets. The switch from an Enable pin to a Turbo pin might cause problems. That turbo pin is weird, normal PClk outputs are 50, 55, 60, 66.8 MHz, but with the turbo they go to 83.3, 75, 61,6 and 68.4 MHz.

No idea how you'd find one though.

Okay, I was remembering wrong, I think this is the post.

ICS9159C-14 Sounds like it probably won’t work? Whatcha think?

Sphere's PCB projects.
-
Sphere’s socket 5/7 cpu collection.
-
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
-
Tyan S1564S to S1564D single to dual processor conversion (also s1563 and s1562)

Reply 6 of 105, by snufkin

User metadata
Rank Oldbie
Rank
Oldbie

That was it, thought I'd had a quick look at this before. A slightly longer look at the ICS9159C-14, comparing with the CGS605 (which should be the same as the W48C60-402), there are two main differences I can see. Firstly pin 5 is Output Enable (OE) on the -402, and that needs to be pulled high to enable the outputs. I'm assuming that the motherboard doesn't actually use the OE function and just pulls it to 3.3V, maybe through R48. Pin 5 on the -14 is a Turbo input, and when that input is high then the Processor Clock (PClk) output options are 50,55,60,66.8MHz. You'll need to somehow connect it to Ground for the higher frequency options 61.6,68.4,75,83.3MHz. How that's best done will depend on how the pin is currently connected.

The other difference might be more of a problem, depending on which clocks the motherboard actually uses. Pin 25 of the -402 is a 12MHz output for a Keyboard controller (KClk). The same pin on the -14 is a 48MHz output for a USB controller. So if the motherboard uses the 12MHz output for the keyboard, then the keyboard might stop working. The -402 shows that clock to be on a different PLL from the PClk and BClk, so it doesn't need to be in sync with those. So just lifting that pin up and connecting a separate 12MHz oscillator to the pad on the motherboard would probably work. I don't know how accurate that keyboard clock needs to be, so maybe it would work by wiring across to use one of the reference frequency outputs (14.318MHz). That way you get to overclock the keyboard as well as the processor.

I don't know if it'll matter, but on both the chips the BClk output is fixed at 1/2 PClk. That means the PCI bus clock will go up as you increase the processor clock, to about 42MHz with PClk@83.3MHz. PClk and BClk are both run from the same PLL, so I guess that it's important for them to be aligned with each other, so it probably won't work to just connect another clock for the bus. It's also possible that the chipset needs the bus clock to be exactly half the processor clock.

Reply 7 of 105, by Sphere478

User metadata
Rank l33t++
Rank
l33t++
snufkin wrote on 2021-12-15, 11:20:

That was it, thought I'd had a quick look at this before. A slightly longer look at the ICS9159C-14, comparing with the CGS605 (which should be the same as the W48C60-402), there are two main differences I can see. Firstly pin 5 is Output Enable (OE) on the -402, and that needs to be pulled high to enable the outputs. I'm assuming that the motherboard doesn't actually use the OE function and just pulls it to 3.3V, maybe through R48. Pin 5 on the -14 is a Turbo input, and when that input is high then the Processor Clock (PClk) output options are 50,55,60,66.8MHz. You'll need to somehow connect it to Ground for the higher frequency options 61.6,68.4,75,83.3MHz. How that's best done will depend on how the pin is currently connected.

The other difference might be more of a problem, depending on which clocks the motherboard actually uses. Pin 25 of the -402 is a 12MHz output for a Keyboard controller (KClk). The same pin on the -14 is a 48MHz output for a USB controller. So if the motherboard uses the 12MHz output for the keyboard, then the keyboard might stop working. The -402 shows that clock to be on a different PLL from the PClk and BClk, so it doesn't need to be in sync with those. So just lifting that pin up and connecting a separate 12MHz oscillator to the pad on the motherboard would probably work. I don't know how accurate that keyboard clock needs to be, so maybe it would work by wiring across to use one of the reference frequency outputs (14.318MHz). That way you get to overclock the keyboard as well as the processor.

I don't know if it'll matter, but on both the chips the BClk output is fixed at 1/2 PClk. That means the PCI bus clock will go up as you increase the processor clock, to about 42MHz with PClk@83.3MHz. PClk and BClk are both run from the same PLL, so I guess that it's important for them to be aligned with each other, so it probably won't work to just connect another clock for the bus. It's also possible that the chipset needs the bus clock to be exactly half the processor clock.

Almost sounds like a riser board to adapt it all with new jumper options etc, might be in order. Shame that there isn’t a perfect drop in replacement 🙁

Sphere's PCB projects.
-
Sphere’s socket 5/7 cpu collection.
-
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
-
Tyan S1564S to S1564D single to dual processor conversion (also s1563 and s1562)

Reply 8 of 105, by snufkin

User metadata
Rank Oldbie
Rank
Oldbie
Sphere478 wrote on 2021-12-22, 03:28:

Almost sounds like a riser board to adapt it all with new jumper options etc, might be in order. Shame that there isn’t a perfect drop in replacement 🙁

That might work, but clocks are tricky. Every time the trace goes through a via or a connector the signal degrades a bit (if I remember rightly, something to do with impedance mismatches causing signal reflections), and you don't know how close the motherboard layout is to not working at 66MHz. So putting in a couple more connections might make it so 75MHz doesn't work, when it might if the pin goes direct to the pad.

It's still not a direct fit, but just found the PLL52C59-14a: http://www.ryston.cz/petr/mirror/www.phaselin … s/PDF/5914a.pdf . Very similar to the ICS59159C-14, but it has options for the bus to be asynchronous from the cpu, so can generate 75MHz and 83.3MHz CPU with a 32MHz bus. Don't know if that would cause problems with the chipset. It still has the 48MHz USB clock rather than the 12MHz Kbd clock you need. Stick a dual flip-flop between pin 25 and the pad, wired as a 1/4 generator (I'm assuming a 12MHz clock will cope better with the impedance mismatches). Lift pin 5 and fit a resistor+jumper to select between Vdd and Gnd. Can't find any prices for it, but some places are at least claiming they have stock, so it might be available.

[edit: to correct a part number on the 9159C-14]

Reply 9 of 105, by Sphere478

User metadata
Rank l33t++
Rank
l33t++

Snufkin wrote:

Reading that GA-586 mod page it sounds like it has a similar chip, except it uses the 48MHz output when your board needs a 12MHz output for the keyboard.

Go on, see if you can find one of those PLL52C59-14A chips (that's the one I found last, and is mentioned on that GA-586 page, and it looks like it does need to be the 14A version to get 83MHz), then lift the 48MHz output pin and sort out generating the 12MHz keyboard clock. Pretty sure you'd just need either a dual flip-flip and a bit of wiring to divide the 48MHz clock by 4 (e.g. https://uk.farnell.com/texas-instruments/sn74ahc74dr/logic-74ahc-d-flip-flop-soic14/dp/1621017?st=7474 ) or if the keyboard doesn't need to be in phase with the main clock (seems likely) then just a 12MHz oscillator (e.g. https://uk.farnell.com/ael-crystals/o12m000000l008/crystal-oscillator-12mhz/dp/9510117 ).

I ran that modding page through Google translate for the later mod where they swap out the 14.318MHz crystal for a 20MHz version, because I was pretty sure I was reading it wrong. Swapping the crystal overclocks the bus, but also stops the floppy and serial ports working, so they wanted to reintroduce the 14MHz signal. So, if Google translate has this right, they take a broken motherboard and use a saw to cut out the bit of the PCB holding the clock generator and crystal, then use that PCB fragment and wire it up to the modded board. Exciting. Wish there were photos.

^Quote from another thread about this one^

Can you draw me a schematic of what you are proposing?

Drawing it how you think it would physically look would be handy also.

I just did a search, I can’t find them anywhere.

Sphere's PCB projects.
-
Sphere’s socket 5/7 cpu collection.
-
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
-
Tyan S1564S to S1564D single to dual processor conversion (also s1563 and s1562)

Reply 10 of 105, by snufkin

User metadata
Rank Oldbie
Rank
Oldbie

[edit: I just realised that I was assuming this is running at 5V, but it's probably 3.3V. Definitely worth checking what the Vdd value is on the W48C60 if you're thinking of doing any of this.]

So I think it's just two pins that will need changing, pin 5 and pin 25:

W48C60_PLL52C59_Diff.jpg
Filename
W48C60_PLL52C59_Diff.jpg
File size
134.17 KiB
Views
2176 views
File license
Fair use/fair dealing exception

Difficult to say exactly what the connections should be as it'll depend on the current layout, which I can't quite make out enough detail. So before I throw up a wall-'o-text, the short version is that if you're lucky you can remove the W48C60 and just solder the PLL52C59 in its place. Then solder a jumper header between pin 5 and Ground to do frequency selection. Then connect power to a 12MHz oscillator, lift a resistor (to disconnect the 48MHz from the keyboard controller) and connect the clock output to one side of a resistor (to connect 12MHz to the controller).

Now the promised wall-'o-text:

Easy one first: pad and pin 5. On the W48C60 pin 5 is an Output Enable and needs to be pulled high for the output clocks to be on. I'm hoping the PCB pad 5 is pulled high through that R48. Could you check if pin 5 connects to one side of R48 and that the other side connects to +5V. On the PLL52C59 then pin 5 is FSEL2 and needs to be pulled to Ground to enable the higher frequencies. If there's already a resistor pulling that pad high then all you need do is solder pin5 to pad 5 as normal, then connect a wire from that pin to a jumper, with the other side of the jumper connected to a Ground. There are plenty around, so pick one that's convenient; this isn't a noisy switching signal so it's not going to interfere with anything. Then remove the jumper to let R48 pull pin 5 high to select the low speeds, or connect the jumper to pull pin 5 low to select high speeds.

If pad 5 isn't connected to +5V through a resistor and is just connected directly to +5V then there's a bit more work. When you remove the W48C60 then cover over pad 5 and don't solder pin 5 of the PLL52C59 to it. Then you'll need a 3-way jumper, with middle pin to pin 5 of the PLL52C59, one pin to +5V and the other pin to Ground. Then you can use a jumper to tie pin 5 either high or low. Connecting the jumper pins to +5 and Ground should probably be done through a couple of 10k resistors to make sure nothing breaks if there's ever a short somehow.

So that's the frequency select stuff. FSEL0 and 1 on the PLL52C59 match up with SEL0 and 1 on the W48C60 (pins 12&13), with probably pull up resistors R59 and one of the other resistors to the right of the W48C60.

Slightly trickier is where to connect a keyboard clock. W48C60 outputs 12MHz on pin 25, apparently as a keyboard clock. That probably goes to a pin on the Amikey 9738. The PLL52C59 puts 48MHz out on pin 25 as a USB clock source. Obviously that will have to be disconnected, but where's easiest?

Now, I'm hoping there's an easy way to disconnect the 48MHz and attach a 12MHz source instead. There are a bunch of SMD components just to the left of the W48C60. I'm hoping those are source termination resistors and that one of them connects to pad 25. If it does then that means you can just remove the W48C60 and solder the PLL52C59 in directly. Then lift the termination resistor (disconnecting the 48MHz signal), then use it to connect a 12MHz source.

If that's not possible then when you remove the W48C60 you'll have to solder a fine wire to pad 25, cover it over, then lift pin 25 of the PLL52C59 up a little before soldering the rest of the pins down. Then solder a 12MHz source to that wire.

Lastly, where to get a 12MHz source. Either get a 12MHz oscillator, connect to +5 and Ground (maybe put a 1uF ceramic capacitor between the power pins), then connect the output to either the pad 25 or the source termination resistor depending on the previous bit of work. That's definitely easiest, but does mean the keyboard and system clocks won't be in sync with each other. That almost certainly doesn't matter.

If you do want to have them kept in sync then you can connect pin 25 of the PLL52C59 to the first clock input on a dual 7474 flip flop. Connect the !Q output to the D input and to the clock input of the second flip-flop. Connect the second !Q output to the second D input. That output is then running at 1/4 of the input, so 12MHz, and you can also connect it to wherever you want to connect the 12MHz to. It also needs +5V and Ground, and also needs the other inputs pulled high. It's a bit more fiddly, but does mean the clocks are at least in sync.

I've no experience sourcing out of production stuff, but searching for PLL52C59 ASC turned up:
https://www.utsource.net/pro/PLL52C59-14ASC.html (claims to have stock and a price)
https://www.exshinetech.com/products/PLL52C59-14-ASC.html (claims to have stock, enquire for price)
No idea what shipping costs or minimum order quantities would be though.

Could be more certain of what connections will be easiest with a better photo showing the clock generator area, top and bottom of the PCB, and ideally with the chip removed to see where the surface traces go. Obviously there are some risks involved whenever doing work to a board, so keep that in mind. But it could be fun.

Reply 11 of 105, by Sphere478

User metadata
Rank l33t++
Rank
l33t++
snufkin wrote on 2022-01-10, 16:55:
[edit: I just realised that I was assuming this is running at 5V, but it's probably 3.3V. Definitely worth checking what the Vd […]
Show full quote

[edit: I just realised that I was assuming this is running at 5V, but it's probably 3.3V. Definitely worth checking what the Vdd value is on the W48C60 if you're thinking of doing any of this.]

So I think it's just two pins that will need changing, pin 5 and pin 25:
W48C60_PLL52C59_Diff.jpg

Difficult to say exactly what the connections should be as it'll depend on the current layout, which I can't quite make out enough detail. So before I throw up a wall-'o-text, the short version is that if you're lucky you can remove the W48C60 and just solder the PLL52C59 in its place. Then solder a jumper header between pin 5 and Ground to do frequency selection. Then connect power to a 12MHz oscillator, lift a resistor (to disconnect the 48MHz from the keyboard controller) and connect the clock output to one side of a resistor (to connect 12MHz to the controller).

Now the promised wall-'o-text:

Easy one first: pad and pin 5. On the W48C60 pin 5 is an Output Enable and needs to be pulled high for the output clocks to be on. I'm hoping the PCB pad 5 is pulled high through that R48. Could you check if pin 5 connects to one side of R48 and that the other side connects to +5V. On the PLL52C59 then pin 5 is FSEL2 and needs to be pulled to Ground to enable the higher frequencies. If there's already a resistor pulling that pad high then all you need do is solder pin5 to pad 5 as normal, then connect a wire from that pin to a jumper, with the other side of the jumper connected to a Ground. There are plenty around, so pick one that's convenient; this isn't a noisy switching signal so it's not going to interfere with anything. Then remove the jumper to let R48 pull pin 5 high to select the low speeds, or connect the jumper to pull pin 5 low to select high speeds.

If pad 5 isn't connected to +5V through a resistor and is just connected directly to +5V then there's a bit more work. When you remove the W48C60 then cover over pad 5 and don't solder pin 5 of the PLL52C59 to it. Then you'll need a 3-way jumper, with middle pin to pin 5 of the PLL52C59, one pin to +5V and the other pin to Ground. Then you can use a jumper to tie pin 5 either high or low. Connecting the jumper pins to +5 and Ground should probably be done through a couple of 10k resistors to make sure nothing breaks if there's ever a short somehow.

So that's the frequency select stuff. FSEL0 and 1 on the PLL52C59 match up with SEL0 and 1 on the W48C60 (pins 12&13), with probably pull up resistors R59 and one of the other resistors to the right of the W48C60.

Slightly trickier is where to connect a keyboard clock. W48C60 outputs 12MHz on pin 25, apparently as a keyboard clock. That probably goes to a pin on the Amikey 9738. The PLL52C59 puts 48MHz out on pin 25 as a USB clock source. Obviously that will have to be disconnected, but where's easiest?

Now, I'm hoping there's an easy way to disconnect the 48MHz and attach a 12MHz source instead. There are a bunch of SMD components just to the left of the W48C60. I'm hoping those are source termination resistors and that one of them connects to pad 25. If it does then that means you can just remove the W48C60 and solder the PLL52C59 in directly. Then lift the termination resistor (disconnecting the 48MHz signal), then use it to connect a 12MHz source.

If that's not possible then when you remove the W48C60 you'll have to solder a fine wire to pad 25, cover it over, then lift pin 25 of the PLL52C59 up a little before soldering the rest of the pins down. Then solder a 12MHz source to that wire.

Lastly, where to get a 12MHz source. Either get a 12MHz oscillator, connect to +5 and Ground (maybe put a 1uF ceramic capacitor between the power pins), then connect the output to either the pad 25 or the source termination resistor depending on the previous bit of work. That's definitely easiest, but does mean the keyboard and system clocks won't be in sync with each other. That almost certainly doesn't matter.

If you do want to have them kept in sync then you can connect pin 25 of the PLL52C59 to the first clock input on a dual 7474 flip flop. Connect the !Q output to the D input and to the clock input of the second flip-flop. Connect the second !Q output to the second D input. That output is then running at 1/4 of the input, so 12MHz, and you can also connect it to wherever you want to connect the 12MHz to. It also needs +5V and Ground, and also needs the other inputs pulled high. It's a bit more fiddly, but does mean the clocks are at least in sync.

I've no experience sourcing out of production stuff, but searching for PLL52C59 ASC turned up:
https://www.utsource.net/pro/PLL52C59-14ASC.html (claims to have stock and a price)
https://www.exshinetech.com/products/PLL52C59-14-ASC.html (claims to have stock, enquire for price)
No idea what shipping costs or minimum order quantities would be though.

Could be more certain of what connections will be easiest with a better photo showing the clock generator area, top and bottom of the PCB, and ideally with the chip removed to see where the surface traces go. Obviously there are some risks involved whenever doing work to a board, so keep that in mind. But it could be fun.

Oh man, that’s involved…

Pretty sure I can handle the soldering,

I’m so worried about messing up the board. Also was hoping for a more elegant solution. But it kinda seems like it may be time for a reiser board.

Are there chip sockets that fit this footprint or something?🤔

I can get some better pics. One sec

Attachments

  • image.jpg
    Filename
    image.jpg
    File size
    1.58 MiB
    Views
    2150 views
    File license
    Public domain
  • image.jpg
    Filename
    image.jpg
    File size
    1.82 MiB
    Views
    2150 views
    File license
    Public domain

Sphere's PCB projects.
-
Sphere’s socket 5/7 cpu collection.
-
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
-
Tyan S1564S to S1564D single to dual processor conversion (also s1563 and s1562)

Reply 13 of 105, by snufkin

User metadata
Rank Oldbie
Rank
Oldbie

Ok, if the keyboard controller can run at 14.318MHz rather than 12MHz then I think this works out nicely.

Tyan_S1564D_Clock.jpg
Filename
Tyan_S1564D_Clock.jpg
File size
170.02 KiB
Views
2133 views
File license
Fair use/fair dealing exception

It looks like R48 is a 4.7k resistor that connects to the pin 5 Output Enable. The other side of the resistor must go to Vdd for the chip (possibly 3.3V). If you can find a PLL52C59-14A then for the higher speeds that needs to be pulled low. So fit a jumper between that resistor and a ground. It might be nice to fit a jumper as part of J13, I think pins 2 and 4 are probably ground.

More fun is that 12MHz/48MHz problem between the current clock gen and the -14A. I can't see where the trace goes from the pad, it's probably under the chip. But there's a likely looking trace that emerges by the U2 silk screen. So I've joined them with a dotted line. From there things get better. Assuming that is the 12MHz clock then it goes through a 22 ohm source termination resistor, through a 0 ohm link, then off to the rest of the board. Before it heads off though, it goes through the empty pad for a SMD component. And where does the other side of that pad go? It heads off to R31 (100 ohm resistor), the other side of which is driven by one of the 2 Reference Clock (14.318MHz) outputs.

So there's a good chance that if you just move the 0 ohm link down one that it will disconnect the 48MHz signal from the -14A and instead connect a 14.318MHz source to the keyboard controller.

Assuming the KBC can handle that, then it's job done. If it can't then it'll need a separate 12MHz clock, but even that's not too bad with all the options of where to connect it to, which makes it easy to swap out the actual generator.

Things to confirm:
1) R48 goes to some +ve voltage, either +5V or +3.3V probably.
2) pin 25 goes to that 22 ohm resistor, and on to that empty pad I've marked to move the 0 ohm link to.
3) pin 27 goes to the other side of that empty pad, with 100 ohms resistance.

Incidentally, it looks like J7 switches something else on the board from using either 24MHz or 14.318MHz. Currently it's hard wired for 24MHz. So if you even decide to switch out the main 14.318 crystal, then that's a handy spot to connect in a 24MHz oscillator to keep whatever that is running at the right speed.

Reply 14 of 105, by Sphere478

User metadata
Rank l33t++
Rank
l33t++
snufkin wrote on 2022-01-11, 22:20:
Ok, if the keyboard controller can run at 14.318MHz rather than 12MHz then I think this works out nicely. Tyan_S1564D_Clock.jpg […]
Show full quote

Ok, if the keyboard controller can run at 14.318MHz rather than 12MHz then I think this works out nicely.
Tyan_S1564D_Clock.jpg

It looks like R48 is a 4.7k resistor that connects to the pin 5 Output Enable. The other side of the resistor must go to Vdd for the chip (possibly 3.3V). If you can find a PLL52C59-14A then for the higher speeds that needs to be pulled low. So fit a jumper between that resistor and a ground. It might be nice to fit a jumper as part of J13, I think pins 2 and 4 are probably ground.

More fun is that 12MHz/48MHz problem between the current clock gen and the -14A. I can't see where the trace goes from the pad, it's probably under the chip. But there's a likely looking trace that emerges by the U2 silk screen. So I've joined them with a dotted line. From there things get better. Assuming that is the 12MHz clock then it goes through a 22 ohm source termination resistor, through a 0 ohm link, then off to the rest of the board. Before it heads off though, it goes through the empty pad for a SMD component. And where does the other side of that pad go? It heads off to R31 (100 ohm resistor), the other side of which is driven by one of the 2 Reference Clock (14.318MHz) outputs.

So there's a good chance that if you just move the 0 ohm link down one that it will disconnect the 48MHz signal from the -14A and instead connect a 14.318MHz source to the keyboard controller.

Assuming the KBC can handle that, then it's job done. If it can't then it'll need a separate 12MHz clock, but even that's not too bad with all the options of where to connect it to, which makes it easy to swap out the actual generator.

Things to confirm:
1) R48 goes to some +ve voltage, either +5V or +3.3V probably.
2) pin 25 goes to that 22 ohm resistor, and on to that empty pad I've marked to move the 0 ohm link to.
3) pin 27 goes to the other side of that empty pad, with 100 ohms resistance.

Incidentally, it looks like J7 switches something else on the board from using either 24MHz or 14.318MHz. Currently it's hard wired for 24MHz. So if you even decide to switch out the main 14.318 crystal, then that's a handy spot to connect in a 24MHz oscillator to keep whatever that is running at the right speed.

What if instead of replacing the clockgen I added some sort of clock multiplying device to the clockgen output? Lift the pin, glue the device to the clock gen and reintroduce the signal to the pad?

The bottom of r48 is connected to 5v

Your assumption about the 22 ohm and the pin are correct (yellow line)

I am getting 99.9ohms between those two points as described.

On a previous test I actually moved that jumper and it seemed to work fine.

It’s possible I wasn’t using whatever it goes to though. I moved it back.

Sphere's PCB projects.
-
Sphere’s socket 5/7 cpu collection.
-
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
-
Tyan S1564S to S1564D single to dual processor conversion (also s1563 and s1562)

Reply 15 of 105, by snufkin

User metadata
Rank Oldbie
Rank
Oldbie

Good to hear I guessed right about where the yellow line went, and that it sounds like you've already tried the KBC at 14.318MHz and if you used the keyboard at all then it worked. You should find that the pin 25 goes to a pin somewhere on the AMIKEY-2 chip.

Sphere478 wrote on 2022-01-12, 01:24:

What if instead of replacing the clockgen I added some sort of clock multiplying device to the clockgen output? Lift the pin, glue the device to the clock gen and reintroduce the signal to the pad?

That kind of works, except it'll have to be a change on the input to the clock gen. The problem with the outputs is that there are 4 CPU clock outputs and 6 bus clock outputs, and those all need to be exactly in phase with each other, with the bus clock exactly 1/2 of the CPU clock. So inserting anything between the clock gen output pins and those pads is going to be a problem.

Instead, plan B (to up the FSB without changing the clock gen chip) will have to be to change the clock gen input, similar to that GA-586HX mod. The current clock gen outputs a CPU clock of 50, 60 and 66MHz, with a 14.318MHz input. That means it's multiplying the input clock by roughly 3.5, 4.2 and 4.6. If you swap the 14.318MHz crystal for a 20MHz crystal, then the output CPU clocks will be roughly 70, 84 and 92MHz with the bus clock at .35, 42 and 46MHz.

Which if the board is currently stable at 66MHz might be ok. Except for the other outputs from the clock gen. There are 2 reference clocks that should be at 14.318MHz, 1 keyboard clock@12MHz and 1 system clock@24MHz. With a 20MHz crystal they'll run at 20, 17 and 34MHz.
Maybe the KBC would be ok at 17MHz, but the others will be a problem.

The phase for these other clocks compared to the CPU and Bus clocks don't matter (internally they're on a different PLL), so adding in a separate clock source for them shouldn't matter. It sounds like the KBC is ok running at 14.318MHz, so we can probably ignore the 12MHz clock. So that leaves just a 14.318 and 24MHz clock source to add.

24MHz clock
It looks like pin 24 (the 24MHz output) goes to R30. So you can get a 24MHz oscillator and connect up Vdd and Ground (if you can find a ~1uF ceramic capacitor to go between the pins that would be good) and lift R30. Then cover the pad that goes to pin 24 and solder R30 back on to the pad that goes to J7. Then connect the output pin of the 24MHz oscillator to the other side of R30. So now R30 is driven directly by a 24MHz clock and not pin 24 of the clock gen.

14.318MHz clock
The reference output is on pins 27 and 28. They look to go to R31 and R32. So do a similar thing as with the 24MHz clock. Lift both of the resistors and cover the pads that connect to the clock gen (probably easiest done with all 3 resistors off). Then solder them back on to just the one pad. Then connect the output of a 14.318MHz oscillator to both R31 and R32, on the side that used to connect to the clock gen.

Then move that 0 ohm link to connect the KBC to the new 14.318MHz source, rather than the overclocked 17MHz. So now the reference and 24MHz clocks are independent of the main crystal and the KBC is run off of the reference clock. If you fit a socket for the crystal then you can try various frequency options just by swapping them out.

A thing to check is the Vdd for the current clock gen. Given R48 goes to +5 then I expect it is running at 5V, but could you check pins 1,8,14,20&26 connect to 5V?

Possible parts (smaller oscillators exist, but these are probably easier to solder):
20MHz crystal: https://uk.farnell.com/raltron/as-20-000-18-f … 9-4h/dp/1611762
crystal socket: https://uk.farnell.com/fischer-elektronik/pq- … sion/dp/4695422
14.318MHz oscillator: https://www.mouser.co.uk/ProductDetail/ECS/EC … w9IYElxng%3D%3D
24MHz oscillator: https://www.mouser.co.uk/ProductDetail/CTS-El … fDb6F1rNg%3D%3D

Summary
Option 1) Find a PLL52C59-14A and swap it in. Move the 0 ohm link to connect KBC to 14.318MHz clock. Jumper pin 5 to ground.
[edit: oh, I forgot about the ICS9159C-14 / ICS9159CM-14 which are very similar to the PLL52C59-14A, except they don't have the asynchronous bus clock options, which you don't need anyway]

Option 2) Put 20MHz crystal in place of 14.318 crystal. Disconnect inputs to R30,R31,R32 and connect 24MHz and 14.318MHz oscillators to them instead. Move the 0 ohm link.

Reply 16 of 105, by Sphere478

User metadata
Rank l33t++
Rank
l33t++

I’m liking option one. I gather that you agree?

So get new glock gen, add a jumper, and move a 0 ohm link?

Which one should I buy?

Sphere's PCB projects.
-
Sphere’s socket 5/7 cpu collection.
-
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
-
Tyan S1564S to S1564D single to dual processor conversion (also s1563 and s1562)

Reply 17 of 105, by snufkin

User metadata
Rank Oldbie
Rank
Oldbie
Sphere478 wrote on 2022-01-12, 16:51:

I’m liking option one. I gather that you agree?

Option 1 is less messy in the end, but does mean replacing a 28 pin smd. I think it's probably less likely to give problems.

OTOH, Option 2 ends up with more bits glued to the board, but by separating the CPU and bus clocks from the other clocks it does mean you've got greater freedom for playing around with the CPU frequency by changing the main crystal. Maybe it turns out your board can't do 83MHz but could do 80MHz (or 79.5MHz with a 17.28MHz crystal), but you can't select that with Option 1.

So get new glock gen, add a jumper, and move a 0 ohm link?

I think so. Pins all line up, voltage supply for both are 3-5V, can't see anything that stops it working, except for the KBC clock we know about. Jumper gives access to higher speed outputs, 0 ohm link runs the KBC off of the 14.318. So running a bit faster than the 12 it's on at the moment.

Which one should I buy?

As to which to get, I think either the PLL52C59-14A or the ICS9159C-14 will do:

PLL52C59_14A_FreqTable.jpg
Filename
PLL52C59_14A_FreqTable.jpg
File size
24.31 KiB
Views
2047 views
File license
Fair use/fair dealing exception
ICS9159C_14_FreqTable.jpg
Filename
ICS9159C_14_FreqTable.jpg
File size
24.2 KiB
Views
2047 views
File license
Fair use/fair dealing exception

Mind that the PLL one does need to be the -14A, not the -14L or T. And the ICS needs to be the C version. Full part numbers from the datasheets look to be:
PLL52C59-14ASC and ICS9159CM-14 which might help when searching.

Datasheets here:
current gen: https://pdf1.alldatasheet.com/datasheet-pdf/v … 59C-02CW28.html
alt1: http://www.ryston.cz/petr/mirror/www.phaselin … s/PDF/5914a.pdf
alt2: https://pdf1.alldatasheet.com/datasheet-pdf/v … CS9159C-14.html

Things I'd check before ordering anything are to double check the supply voltages on those Vdd pins (1,8,14,20,26) and that the grounds (4,11,17,23) are all shorted to ground. Also try moving that 0 ohm link down and check that everything still works.

Reply 19 of 105, by snufkin

User metadata
Rank Oldbie
Rank
Oldbie

I've no experience with any of these, but there are some places that claim to have them, some as new/unused, others as used:
https://www.kynix.com/Detail/1166633/ICS9159CM-14.html
https://www.worldwayelec.com/pro/ics/ics9159cm-14/4340053
https://www.exshinetech.com/products/PLL52C59-14-ASC.html
https://www.utsource.net/pro/PLL52C59-14ASC.html

Although if the board doesn't work it'll be difficult to tell if that because I'm wrong and this mod won't work, or if the chip's broken/fake.