VOGONS


Short check: oscillators on 386sx board

Topic actions

Reply 60 of 90, by Marco

User metadata
Rank Member
Rank
Member

Hi again. Thanks for following up.

No I have nothing grounded or anything else done on hardware layer.

The behavior is sometimes even stranger. Sometimes the pc won’t do start at all. Only after a reset the beeps do come then.
Another time the beep sound was just like fading away: beep beep beep beeeaaaaaaaaau

Anything else that I could try?

1) VLSI SCAMP 311 / 386SX@27,5 (non-W95: stable at 30MHz) / 8MB / CL-GD5428 / LAPC-I / Fast-SCSI AHA 1542CP (...which is disappointing)
2) SIS486 / 486DX/2 66 / 32MB / TGUI9440 / SCC-1&MT32 / CT2830

Reply 61 of 90, by mkarcher

User metadata
Rank Oldbie
Rank
Oldbie
Marco wrote on 2022-10-13, 20:36:
Hi again. Thanks for following up. […]
Show full quote

Hi again. Thanks for following up.

No I have nothing grounded or anything else done on hardware layer.

The behavior is sometimes even stranger. Sometimes the pc won’t do start at all. Only after a reset the beeps do come then.
Another time the beep sound was just like fading away: beep beep beep beeeaaaaaaaaau

Anything else that I could try?

As it seems, the BIOS chip is accessed faster than allowed. This causes the processor to get garbled instructions. Which instructions are garbled in what way is kind of random, so getting slightly different symptoms after each reset attempt is not that surprising. Can you verify whether busosc at the chipset is directly connected to the output pin of the 16MHz oscillator? If yes, can you check wheter there is a pull-up or pull-down resistor installed on that line (just measure the oscillator output pin while no oscillator is installed. If it is low (permanently running on sclkdiv), please check for pullup or pulldown resistors configuring sclkdiv.

Reply 62 of 90, by Marco

User metadata
Rank Member
Rank
Member

Ok thanks will do so. I assume while power off via multimeter is it?

1) VLSI SCAMP 311 / 386SX@27,5 (non-W95: stable at 30MHz) / 8MB / CL-GD5428 / LAPC-I / Fast-SCSI AHA 1542CP (...which is disappointing)
2) SIS486 / 486DX/2 66 / 32MB / TGUI9440 / SCC-1&MT32 / CT2830

Reply 63 of 90, by mkarcher

User metadata
Rank Oldbie
Rank
Oldbie
Marco wrote on 2022-10-13, 21:21:

Ok thanks will do so. I assume while power off via multimeter is it?

The check of the voltage at the oscillator output pin has to be performed while the power is on. You can put the black ground probe of the multimeter at a safe grounded space (like a screw hole for a slot cover, so you can focus on placing the red "live" probe and the meter display. When power is off, voltage checks always yields 0V (except in the small part powered by the CMOS battery).

Reply 64 of 90, by Marco

User metadata
Rank Member
Rank
Member

Guys. Sorry I am on holidays so I will come back after my return.

1) VLSI SCAMP 311 / 386SX@27,5 (non-W95: stable at 30MHz) / 8MB / CL-GD5428 / LAPC-I / Fast-SCSI AHA 1542CP (...which is disappointing)
2) SIS486 / 486DX/2 66 / 32MB / TGUI9440 / SCC-1&MT32 / CT2830

Reply 65 of 90, by Marco

User metadata
Rank Member
Rank
Member

Thanks to Rasz_pl and zyga I reached follow update:

Code checked. This is the result:
I assume it means divider = /2?
It would be true since I run ext Osc Busosc 24mhz/2

Attachments

1) VLSI SCAMP 311 / 386SX@27,5 (non-W95: stable at 30MHz) / 8MB / CL-GD5428 / LAPC-I / Fast-SCSI AHA 1542CP (...which is disappointing)
2) SIS486 / 486DX/2 66 / 32MB / TGUI9440 / SCC-1&MT32 / CT2830

Reply 66 of 90, by Marco

User metadata
Rank Member
Rank
Member

Deleted as wrong measured

Last edited by Marco on 2023-01-20, 12:09. Edited 1 time in total.

1) VLSI SCAMP 311 / 386SX@27,5 (non-W95: stable at 30MHz) / 8MB / CL-GD5428 / LAPC-I / Fast-SCSI AHA 1542CP (...which is disappointing)
2) SIS486 / 486DX/2 66 / 32MB / TGUI9440 / SCC-1&MT32 / CT2830

Reply 67 of 90, by Marco

User metadata
Rank Member
Rank
Member

Deleted as wrong measured

Last edited by Marco on 2023-01-20, 12:09. Edited 1 time in total.

1) VLSI SCAMP 311 / 386SX@27,5 (non-W95: stable at 30MHz) / 8MB / CL-GD5428 / LAPC-I / Fast-SCSI AHA 1542CP (...which is disappointing)
2) SIS486 / 486DX/2 66 / 32MB / TGUI9440 / SCC-1&MT32 / CT2830

Reply 68 of 90, by Marco

User metadata
Rank Member
Rank
Member

Deleted as wrong measured

Last edited by Marco on 2023-01-20, 12:09. Edited 1 time in total.

1) VLSI SCAMP 311 / 386SX@27,5 (non-W95: stable at 30MHz) / 8MB / CL-GD5428 / LAPC-I / Fast-SCSI AHA 1542CP (...which is disappointing)
2) SIS486 / 486DX/2 66 / 32MB / TGUI9440 / SCC-1&MT32 / CT2830

Reply 69 of 90, by rasz_pl

User metadata
Rank l33t
Rank
l33t
Marco wrote on 2023-01-20, 12:01:
Quote from rasz_pl: "find pin 115 and measure its resistance to ground and 5V supply, on powered off system obviously 😀" […]
Show full quote

Quote from rasz_pl:
"find pin 115 and measure its resistance to ground and 5V supply, on powered off system obviously 😀"

Result:
I am really struggling with this one - sorry 😀 The CPU pins are that thin and I cant even see where the right corner of the cpu is in accordance to the manual (the special edge "\" I cannot find as from top all CPU edges are exactly the same. Anyway assuming the top view is in accordance with the way the CPU specs where printed on I anyway struggle. I only get sth like 3mOhm while measuring against ground (PS screwhole). Not a big input from me isnt it 🙁

chipset pin, I think I linked chipset documention in same post

Reply 70 of 90, by Marco

User metadata
Rank Member
Rank
Member

Damn it. Sorry all back to zero.

PLS IGNORE ALL MY POSTS SO FAR (except for debug)

1) VLSI SCAMP 311 / 386SX@27,5 (non-W95: stable at 30MHz) / 8MB / CL-GD5428 / LAPC-I / Fast-SCSI AHA 1542CP (...which is disappointing)
2) SIS486 / 486DX/2 66 / 32MB / TGUI9440 / SCC-1&MT32 / CT2830

Reply 71 of 90, by Marco

User metadata
Rank Member
Rank
Member

1st Quote mkarcher:
Can you verify whether busosc at the chipset is directly connected to the output pin of the 16MHz oscillator?

Result:
No direct connect. Possible clear since Chrystal signal has to be transformed to ttl compatible

2md quote mkarcher:
Pls measure voltage on Chrystal with ground (screw hole).

Result:
1,4V for Pin1. 1,0V for Pin2.
I don’t know what the correct pin is so I measured both.

3rd quote rasz_pl:
find pin 115 and measure its resistance to ground and 5V supply, on powered off system obviously 😀

Result:
To be tested yet. Still my equipment is to thick for that.

Any idea for next steps?

1) VLSI SCAMP 311 / 386SX@27,5 (non-W95: stable at 30MHz) / 8MB / CL-GD5428 / LAPC-I / Fast-SCSI AHA 1542CP (...which is disappointing)
2) SIS486 / 486DX/2 66 / 32MB / TGUI9440 / SCC-1&MT32 / CT2830

Reply 72 of 90, by zyga64

User metadata
Rank Oldbie
Rank
Oldbie

I also wanted to try what you are doing, on my computer with a 286 on a VL82C311L 😀
I have PCCHIPs M218 - looks like this one: Re: PCChips M218 286 motherboard info needed

On this motherboard BUSOSC (pin121) input is connected to 14.7 MHz signal, and ISA CLK signal (in PIN B20) was half of that (measured by osciloscope).
So it looks like divider was set to 2 (reading from CLKCTL register gives me 4 - which means divider is 2 when BUSOSC is not connected).

Then I lifted up BUSOSC pin and ISA CLK signal on PIN B20 becomes 20MHz 😀 Computer still booted normally but very fast !
Now Wolf3d demo is 13.9 FPS... (CL-GD5422)

But I assume that this is not very safe setting, so I lowered it to 10MHz this way:

c:\>Debug
-o fb ff
-o ec 7
-o ed c
-q

Dummy write to FB port (in my case 0ffh, but can be anything) enables write to Configuration registers. (dummy write to F9 port disables write to those registers).
Then I wrote 0ch (00001100) to CLKCTL register - setting up bit 3 - sets clock divider to 4 (40/4 = 10 MHz).

1) VLSI SCAMP /286@20 /4MB /CL-GD5422 /CMI8330
2) i420EX /486DX33 /16MB /TGUI9440 /YMF718+GUS
3) i430FX /200MMX /64MB /Rage XL PCI /ES1370+ALS100
4) i440BX /P!!!750 /256MB /MX440 /SBLive!+Vibra16s
5) iG31 /E8400 /4GB /X1950GT /HDA

Reply 73 of 90, by mkarcher

User metadata
Rank Oldbie
Rank
Oldbie
Marco wrote on 2023-01-21, 15:50:
1st Quote mkarcher: Can you verify whether busosc at the chipset is directly connected to the output pin of the 16MHz oscillator […]
Show full quote

1st Quote mkarcher:
Can you verify whether busosc at the chipset is directly connected to the output pin of the 16MHz oscillator?

Result:
No direct connect. Possible clear since Chrystal signal has to be transformed to ttl compatible

2md quote mkarcher:
Pls measure voltage on Chrystal with ground (screw hole).

Result:
1,4V for Pin1. 1,0V for Pin2.
I don’t know what the correct pin is so I measured both.

Ouch! I was in a wrong mindset the whole time. I assumed (instead of looking at the picture in the OP) that the source of the 16MHz signal would be a 4-pin boxed crystal oscillator circuit like the 50MHz oscillator for the FSB, and you removed that oscillator circuit. But instead, you were talking about removing just a two-pin crystal, with the oscillator circuit being on board electronics. Looking at the board layout, the oscillation is likely performed by the 74F04 next to both crystals. It is likely that a circuit similar to this circuit employing half a 74F04 is used. Your board has two crystals, so likely three inverters are used for either crystal. Try to beep/trace out which inverters are used for the 16MHz crystal. If the circuit is as expected, try to pull up the pin of the crystal that is connected to an inverter input to +5V using a resistor around 100 Ohms. The inverter then inverts it to 0V at its output, which is again inverted by another inverter to yield a good TTL signal, so the output will be high again. A high signal at BUSOSC tells the 82C311 to run on FCLKDIV, which should divide the clock by a sufficiently high divider to permit booting.

Reply 74 of 90, by Marco

User metadata
Rank Member
Rank
Member

Thanks also all.
Will now try to understand and then take actions.

@zyga64;
Ah your MB also uses an external Chrystal for isa bus speed. In your case it’s the 14,3MHz.
1. what do you mean with „lifting up“ Busosc?
2. do you see any possibility to just change from Busosc Chrystal input towards cpu oscillator (sync mode) just by changing registers?
3. are you really sure that your 286 ran at 20MHz Bus!speed while finishing wolf3d bench even?

@mkarcher:
Thanks. So I will measure, buy stuff and then implement.

Thanksssss

1) VLSI SCAMP 311 / 386SX@27,5 (non-W95: stable at 30MHz) / 8MB / CL-GD5428 / LAPC-I / Fast-SCSI AHA 1542CP (...which is disappointing)
2) SIS486 / 486DX/2 66 / 32MB / TGUI9440 / SCC-1&MT32 / CT2830

Reply 75 of 90, by zyga64

User metadata
Rank Oldbie
Rank
Oldbie

@Marco
1. Desoldering (using flux) pin from board, and moving it slightly up (by tweezers) at the same time. Pin121 is located on the corner of the IC, so it was relatively easy.
2. You may try (carefully) to do as I did. Floating BUSOSC is in high state 😀 If the system gets up after this operation, you will be able to change divisors in debug 😀
3. Yes. Checked by oscilloscope on ISA PIN B20. System (DOS 5.0) seems perfectly stable, but right now only VGA and Multi I/O cards are put in. More to test 😀

Attachments

  • 20mhz_ISA_B20.jpg
    Filename
    20mhz_ISA_B20.jpg
    File size
    662.61 KiB
    Views
    147 views
    File license
    Public domain

1) VLSI SCAMP /286@20 /4MB /CL-GD5422 /CMI8330
2) i420EX /486DX33 /16MB /TGUI9440 /YMF718+GUS
3) i430FX /200MMX /64MB /Rage XL PCI /ES1370+ALS100
4) i440BX /P!!!750 /256MB /MX440 /SBLive!+Vibra16s
5) iG31 /E8400 /4GB /X1950GT /HDA

Reply 76 of 90, by Marco

User metadata
Rank Member
Rank
Member

@zyga64:

I might try first mkarchers proposal as I feel more comfortable than soldering that fine in first place.

Regarding point 2: this might not work for me as isaclock seems to be SYSCLK (=40MHz/2 for you). On my end it would be 27,5MHz (as I use a 55MHz osci). I see veryyyyyyyyy little change for that 😀

SHORT OFFTOPIC:
Btw if you could run another benchmark at 20MHz bus I would be interested to compare with mine (just short off topic).

3d bench (slow pcs): 9,9
Pcpbench vga: 2,1 forget that one. Requires dos extender

1) VLSI SCAMP 311 / 386SX@27,5 (non-W95: stable at 30MHz) / 8MB / CL-GD5428 / LAPC-I / Fast-SCSI AHA 1542CP (...which is disappointing)
2) SIS486 / 486DX/2 66 / 32MB / TGUI9440 / SCC-1&MT32 / CT2830

Reply 77 of 90, by Marco

User metadata
Rank Member
Rank
Member

Short off topic 2: why ext. Osci?

As zyga is the 2nd example of external (asynchronous bus clock provider - at least to a certain extend async) I wonder why they built it like that back in the days.
Possibilities:
- to be fully independent from the cpu they built in
- the be able to run a fpu at a different speed? I red about that here in vogons

1) VLSI SCAMP 311 / 386SX@27,5 (non-W95: stable at 30MHz) / 8MB / CL-GD5428 / LAPC-I / Fast-SCSI AHA 1542CP (...which is disappointing)
2) SIS486 / 486DX/2 66 / 32MB / TGUI9440 / SCC-1&MT32 / CT2830

Reply 78 of 90, by zyga64

User metadata
Rank Oldbie
Rank
Oldbie
Marco wrote on 2023-01-21, 18:16:
SHORT OFFTOPIC: Btw if you could run another benchmark at 20MHz bus I would be interested to compare with mine (just short off t […]
Show full quote

SHORT OFFTOPIC:
Btw if you could run another benchmark at 20MHz bus I would be interested to compare with mine (just short off topic).

3d bench (slow pcs): 9,9
Pcpbench vga: 2,1 forget that one. Requires dos extender

With 20MHz ISA clock (div/2): 7,5
With 10 MHz ISA clock (div/4): 7,3
With 6,66 MHz ISA clock (div/6): 7,0

Almost no change. CPU bottleneck 😀

I would expect bigger differences with HDD transfers. Which tool will you propose for HDD transfer test ?

1) VLSI SCAMP /286@20 /4MB /CL-GD5422 /CMI8330
2) i420EX /486DX33 /16MB /TGUI9440 /YMF718+GUS
3) i430FX /200MMX /64MB /Rage XL PCI /ES1370+ALS100
4) i440BX /P!!!750 /256MB /MX440 /SBLive!+Vibra16s
5) iG31 /E8400 /4GB /X1950GT /HDA

Reply 79 of 90, by Marco

User metadata
Rank Member
Rank
Member

1st: thanks
2nd: hope we don’t break the original topic here
3rd: for quick first impression I simply use Norton utilities system information (sysinfo). Part of Phil’s bench lab. Ä
1. with 6MHz DMA / 12 MHz ISA speed I reach: 2mb/s
2. with 12MHz DMA / 12 MHz ISA speed I reach: 2,8 mb/s

4th: let’s just break down your results per Megahertz on 3dbench: 9,9/27,5*20=7,2. so your results are just in line with my 386sx. Nope. Even slightly better 🫣😅

1) VLSI SCAMP 311 / 386SX@27,5 (non-W95: stable at 30MHz) / 8MB / CL-GD5428 / LAPC-I / Fast-SCSI AHA 1542CP (...which is disappointing)
2) SIS486 / 486DX/2 66 / 32MB / TGUI9440 / SCC-1&MT32 / CT2830