snufkin wrote on 2022-01-12, 18:08:

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: htt […]
Show full quote

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.

I submitted a request but we’ll see.

I hate it when you can’t just click order 🤣

Any chance of finding these on old scrap motherboards? 🤔

Transplanting the IC is no biggie I have the stuff for it.

Dot side

Pin 4 from dot is gnd
Pin 4 from other end and pin 3 from other end is gnd

Opposite side from dot:

Pin three from opposite end is gnd

pin 6 from dot end is gnd

I can’t find any direct 5v connections. I assume this means 5v goes through a resistor on it’s way to the chip

Sphere478 wrote on 2022-01-12, 20:20:

Dot side […]
Show full quote

Dot side

Pin 4 from dot is gnd
Pin 4 from other end and pin 3 from other end is gnd

Opposite side from dot:

Pin three from opposite end is gnd

pin 6 from dot end is gnd

I can’t find any direct 5v connections. I assume this means 5v goes through a resistor on it’s way to the chip

Ok, the grounds sound ok, the ones I listed, plus FSEL1 pulled to Ground through a jumper (selecting 66.6?). Bit worried about the input voltage as the input voltages aren't supposed to go above the supply voltage. You're sure that R48 goes to +5? Do R49 and R50 also go to +5? My guess is the Vdd pins are at 3.3V then, and the inputs will clamp to 3.3V. Current draw won't be very large through either 4.7k (R48) or 10k (R49 and probably R50) resistors. It works ok with the current ICS9159C-02 so it shouldn't be a problem. Could you check if the pins I listed for Vdd are at least connected to each other (pin 1 is by the dot, then the pins numbers count up along that side, then back along the other side to pin 28 directly opposite pin 1)? They may connect to the tab of the big regulator in the corner near the CPU sockets, that might be the 3.3V board supply.

Also, it'd be a bit reassuring about the KBC if you can find a pin on the AMIKEY-2 that connects to pin 25 on the clock gen, probably with 22ohms resistance.

Fingers crossed you can get a part.

I now have 4 of these boards and noticed something…

Do any of these support greater than 66mhz?

I think (based on the CGS605 datasheet) that they're all the same.
pdf.datasheetcatalog.com/datasheet/nationalsemiconductor/DS012491.PDF
From page 11, the National CGS605, IC Systems ICS9159-02, IC Works W48C60-402 and Cypress CY2254 are all compatible. The CW28 bit on the ICS part is the package type (CW) and pin count (28). So I think they'll all behave the same. Certainly the CGS605 and IC9159-02 datasheets look the same and they only go up to 66.
Sorry.

snufkin wrote on 2022-02-03, 00:13:

I think (based on the CGS605 datasheet) that they're all the same. pdf.datasheetcatalog.com/datasheet/nationalsemiconductor/DS01 […]
Show full quote

I think (based on the CGS605 datasheet) that they're all the same.
pdf.datasheetcatalog.com/datasheet/nationalsemiconductor/DS012491.PDF
From page 11, the National CGS605, IC Systems ICS9159-02, IC Works W48C60-402 and Cypress CY2254 are all compatible. The CW28 bit on the ICS part is the package type (CW) and pin count (28). So I think they'll all behave the same. Certainly the CGS605 and IC9159-02 datasheets look the same and they only go up to 66.
Sorry.

Darn,

No dirrect cross reference routes from any of those numbers to chips with 75 or 83.3?

Not that I can find. Problem seems to be that chip makers had the base part numbers (like 9159) then modded them to whatever the board manufacturer wanted (frequency outputs, pin counts, everything), and tacked that on as a variant. So the suffix doesn't follow any pattern. So you can't even just look up all 9159, find one that does 83MHz and put that in. You have to find a possible part number, hope the datasheet is out in public, then check off all the inputs and outputs. Best I've found was those ones from before where (if it's even possible to find one) you just need to muck around with the 48MHz output.

Bummer. Idk if I’m ready to do that kind of hackery to these boards yet. But brilliant solution. Thanks 😀

Everything involved in converting single to dual is ok, but swapping one chip, moving one resistor and adding one jumper is too much hackery? Go on, it might even work. Pick the board you like the least (I dunno, most melted ISA slot?) and turn it in to the board you like the most. If it still works.

snufkin wrote on 2022-02-03, 11:12:

Everything involved in converting single to dual is ok, but swapping one chip, moving one resistor and adding one jumper is too much hackery? Go on, it might even work. Pick the board you like the least (I dunno, most melted ISA slot?) and turn it in to the board you like the most. If it still works.

If I recall, clicking the links earlier for chips I wasn’t able to get them. I would have to find a donor board.

Converting to dual isn’t hackery, it’s fixing a factory error! 😇

Sphere478 wrote on 2022-02-03, 20:58:

If I recall, clicking the links earlier for chips I wasn’t able to get them. I would have to find a donor board.

Yeah, that was always going to be the problem. Best I've been able to find would be taking a punt on used parts from utsource (PLL52C59-14ASC) or aliexpress (ICS9159CM-14), where (if the parts are real) I guess someone else has already done the salvaging.

Converting to dual isn’t hackery, it’s fixing a factory error! 😇

They probably meant to fit the faster clock gen as well. That's why they made it easy to change the keyboard clock source...

snufkin wrote on 2022-02-03, 22:08:

Yeah, that was always going to be the problem. Best I've been able to find would be taking a punt on used parts from utsource ( […]
Show full quote

Sphere478 wrote on 2022-02-03, 20:58:

If I recall, clicking the links earlier for chips I wasn’t able to get them. I would have to find a donor board.

Yeah, that was always going to be the problem. Best I've been able to find would be taking a punt on used parts from utsource (PLL52C59-14ASC) or aliexpress (ICS9159CM-14), where (if the parts are real) I guess someone else has already done the salvaging.

Converting to dual isn’t hackery, it’s fixing a factory error! 😇

They probably meant to fit the faster clock gen as well. That's why they made it easy to change the keyboard clock source...

Can you think of any practical benefit to installing the jumper for the keyboard clock source in fact there are a lot of jumpers on the board that I’m not really sure what they are for it would be very interesting to find out

Sphere478 wrote on 2022-02-03, 23:04:

Can you think of any practical benefit to installing the jumper for the keyboard clock source in fact there are a lot of jumpers on the board that I’m not really sure what they are for it would be very interesting to find out

Not that I can think of, but I don't really know enough of the details about how the KBC fits in the whole system. Can't imagine that running it at 14MHz will make any difference to running it a 12MHz. Unless you type really fast.

I'll be watching this thread since I have one of these boards too, but I don't think I'll start swapping clock gens or crystals unless someone else (with more boards than me) does it first. 😀

stamasd wrote on 2022-02-07, 02:22:

I'll be watching this thread since I have one of these boards too, but I don't think I'll start swapping clock gens or crystals unless someone else (with more boards than me) does it first. 😀

For one, I haven’t been able to get a chip to try this with, another reason is that I’m hesitant to try this also. Lol

I’m kinda thinking that maybe the problem is not finding the right chip, it’s adapting to the correct one. I think a interposer needs to be created. Possibly a bga style one to just adapt to a different clock gen all together and add the jumper.

Anyone able to help with a pcb design?

Snufkin, thoughts?

Not immediately taken with the idea to be honest. It feels like it'd quickly turn in to something way more messy then the 'change the crystal and add 2 oscillators' option. Any clock gen chip with roughly the right number and frequency of outputs is going to be about the same size, so an interposer board will end up being bigger than the pads it needs to solder on to, and there are components surrounding that area, so it'd have to stand up on wire legs. Soldering it in place would then become interesting. Maybe a small socket could be glued within the SOIC footprint, with surface wires running to the SOIC pads, which an interposer could then attach on to (overhanging other components), but running fast clock signals through connectors isn't going to help the clock waveform.

The drive for the main crystal would have to go through a couple of vias and a connector, when there are usually strict rules about not doing that to guarantee that the crystal will start to oscillate. So the interposer would probably need to have its own crystal. Clock generators need a fair bit (200mA?) from a stable supply, so it'd also need decoupling capacitors and supply filters (not a major issue, depending on the current capabilities of any connector).

There's also the problem of finding suitable generator chips. I had a quick look on Mouser and can't immediately find anything that'd work (5V tolerant is a problem, which can be fixed with buffers, but that's adding more complications). So that still means having to find a supply of old chips with the right frequency outputs. Taking one-off chips from broken boards won't really work because any interposer will have been designed around a particular chip, and then we're back to the problem of finding broken boards with compatible chips.

It just feels like the amount of work to end up with something that has worse output quality, so less likely to work, than swapping the existing crystal isn't worth it. Swap the main crystal, lift a couple of pins, move a resistor and add a couple of self-contained 5x7mm oscillator chips for the 14.318MHz reference and the 24MHz chipset clocks. Not as neat as being able to just change the chip, but it does mean you can then easily change the CPU/BUS frequency just by changing the crystal, rather than being limited by jumpers. It has been done before, so it's not a complete unknown (from a link majestyk found, in a slightly odd way involving sawing out sections from a dead donor motherboard): http://web.archive.org/web/20070808045424/htt … n/busclock.html

That, plus clocks worry me. Probably the most analog bit of the whole system. Except maybe the sound card output.

Here's a left-field idea.

How about, instead of looking for elusive sources of NOS chips with compatible pinouts and voltages, a more modern approach is taken. I'm talking about using a modern DDS chip like an AD9851 which can be programmed externally (for instance using an arduino) to generate waveforms up to 180MHz (realistically though up to 80-100MHz). It would need of course its own board made and its own crystal. It is compatible with both 3.3V and 5V supplies (range 2.7-5.5V). And the new oscillator board could be plugged into either a SO socket, or SO edge connector soldered instead of the previous clockgen chip.

stamasd wrote on 2022-02-07, 13:02:

Here's a left-field idea.

How about, instead of looking for elusive sources of NOS chips with compatible pinouts and voltages, a more modern approach is taken. I'm talking about using a modern DDS chip like an AD9851 which can be programmed externally (for instance using an arduino) to generate waveforms up to 180MHz (realistically though up to 80-100MHz). It would need of course its own board made and its own crystal. It is compatible with both 3.3V and 5V supplies (range 2.7-5.5V). And the new oscillator board could be plugged into either a SO socket, or SO edge connector soldered instead of the previous clockgen chip.

How big would it be?
Sounds quite large?

snufkin wrote on 2022-02-07, 12:12:

Not immediately taken with the idea to be honest. It feels like it'd quickly turn in to something way more messy then the 'chan […]
Show full quote

Not immediately taken with the idea to be honest. It feels like it'd quickly turn in to something way more messy then the 'change the crystal and add 2 oscillators' option. Any clock gen chip with roughly the right number and frequency of outputs is going to be about the same size, so an interposer board will end up being bigger than the pads it needs to solder on to, and there are components surrounding that area, so it'd have to stand up on wire legs. Soldering it in place would then become interesting. Maybe a small socket could be glued within the SOIC footprint, with surface wires running to the SOIC pads, which an interposer could then attach on to (overhanging other components), but running fast clock signals through connectors isn't going to help the clock waveform.

The drive for the main crystal would have to go through a couple of vias and a connector, when there are usually strict rules about not doing that to guarantee that the crystal will start to oscillate. So the interposer would probably need to have its own crystal. Clock generators need a fair bit (200mA?) from a stable supply, so it'd also need decoupling capacitors and supply filters (not a major issue, depending on the current capabilities of any connector).

There's also the problem of finding suitable generator chips. I had a quick look on Mouser and can't immediately find anything that'd work (5V tolerant is a problem, which can be fixed with buffers, but that's adding more complications). So that still means having to find a supply of old chips with the right frequency outputs. Taking one-off chips from broken boards won't really work because any interposer will have been designed around a particular chip, and then we're back to the problem of finding broken boards with compatible chips.

It just feels like the amount of work to end up with something that has worse output quality, so less likely to work, than swapping the existing crystal isn't worth it. Swap the main crystal, lift a couple of pins, move a resistor and add a couple of self-contained 5x7mm oscillator chips for the 14.318MHz reference and the 24MHz chipset clocks. Not as neat as being able to just change the chip, but it does mean you can then easily change the CPU/BUS frequency just by changing the crystal, rather than being limited by jumpers. It has been done before, so it's not a complete unknown (from a link majestyk found, in a slightly odd way involving sawing out sections from a dead donor motherboard): http://web.archive.org/web/20070808045424/htt … n/busclock.html

That, plus clocks worry me. Probably the most analog bit of the whole system. Except maybe the sound card output.

Okay what about this, raise the chip that it has onto a reiser board that holds these oscilators

I just want it to look professional, ya know, not stuff all over the place.

I’m thinking one small board that sits on the fotprint board with half holes that I solder to the footprint

Then I solder the second board to that one through some kind of the same method. A little stack to raise the area and give room for some stuff. We could even bring the board over areas that we need to run tap wires to which I could tap via a soldered wire beforehand that would be covered by the board and soldered to once the board was installed. Or another small board or something couod be used in that area. It would even give enough room for dip switches or whatever we thought it needed and would look proper

Sphere478 wrote on 2022-02-07, 19:32:

How big would it be?
Sounds quite large?

The chip itself is quite small, about the size of the other clock generators. The development boards which I've worked with (and which have all the components necessary for the chip to work, including crystal and a few caps+resistors) are about 3x5cm. Here's a picture to give you an idea:

stamasd wrote on 2022-02-07, 19:46:

The chip itself is quite small, about the size of the other clock generators. The development boards which I've worked with (and […]
Show full quote

Sphere478 wrote on 2022-02-07, 19:32:

How big would it be?
Sounds quite large?

The chip itself is quite small, about the size of the other clock generators. The development boards which I've worked with (and which have all the components necessary for the chip to work, including crystal and a few caps+resistors) are about 3x5cm. Here's a picture to give you an idea:

Looks doable, especially with a custom layout.

How do you propose to mount/tap it?