VOGONS


First post, by bracecomputerlab

User metadata
Rank Newbie
Rank
Newbie

I have not posted here since March 2025 (see this long post here) about my retro mainboard development project.
Just to recap, it is a hardware open source project to port VIA Technologies VT5228C Apollo Pro133A Reference Design Schematics (a Slot 1 mATX mainboard with Apollo Pro133A and VT82C686A chipset) and Intel 440BX AGPset Uniprocessor Customer Reference Schematics (a Slot 1 ATX mainboard with 440BX chipset) to KiCad, and actually produce functioning mainboards.
I have since bought most major components (ICs, connectors, etc.) for both mainboards except for one hard to obtain one (Semtech SC1105 voltage selector chip for VT5228C's universal AGP slot's AGP VDDQ voltage selection).
About 3 months ago, I started creating a schematic symbol for VT82C686A "Super South" south bridge chip.
In retrospect, I probably should have started with Apollo Pro133A north bridge chip first since it is easier for a KiCad beginner to work with a simpler symbol first before moving onto a more complicated, feature rich VIA south bridge chip (i.e., chip revision differences and multi-function pin handling relating to APIC).
Anyway, working on and off for the past 2 1/2 months, I finally finished VT82C686A south bridge symbol 3 weeks ago.
I was able to finish Apollo Pro133A symbol 2 weeks ago, and finished its BGA footprint in 1 week (took 3 days off out of 7 days).
Discovered that KiCad has a BGA footprint wizard on 4th of July by accident, used it for generating a footprint for VT82C686A, and finished that one in less than 2 hours.
As just noted, my first manual design of a BGA footprint took 4 days, so huge time savings.
I started to work on Intel Slot 1 connector for the past couple of days.
I finished the Slot 1 symbol except the power / ground portion.
I downloaded a bunch of Intel Slot 1 related documents all the way from 1997 to 1999 using archive.org, and downloaded it on 4th of July.
I started to work on Slot 1 connector footprint yesterday afternoon, and finished it today.
I guess I am getting better at creating my own symbol / footprint in KiCad.
I started to work on SDRAM DIMM connector today.

For the most of March and April, I spent enormous amount of time on eBay and AliExpress hunting for the NOS (New Old Stock) parts for both mainboards.
Most of the components used in this project were manufactured 25+ years ago, so obviously you can imagine that it is not that easy to procure the parts in Year 2025.
Did not make any progress on the design portion due to this.
I really felt like hunting down the relevant parts for this project was, in of itself, a full time job, and I can imagine that parts ordering is done by full time staffs dedicated to this task at high volume mainboard vendors.
The reason I eventually started to order parts from AliExpress is because I was not able to source Harris Semiconductor/ Intersil HIP6004 (part name is HIP6004CB) on eBay.
A similar part named HIP6004B (part name is HIP6004BCB) is available, but is a slightly different part (VID definition is slightly different) than what the VIA reference design specifies, so I did not want to use it.), but some in China had the stock.
It was about $1 / chip + shipping (about $5), and bought 5 of them.
I did get the correct part, but all 5 of the date codes are different, and I wonder if the parts I received are refurbished (i.e., pull off a discarded mainboard).
No, they did not come in a tape carrier.
The devices came in a small ziplock bag.
The part does not exactly look new.
I had no choice.

Anyway, for the VIA mainboard, I am still missing 2 pieces of Linear Technology LT1587 LDO voltage regulator (costs $5 per device NOS or $13.87 over at Digi-Key and very expensive for what it does), a few p and n channel discrete FETs for STR (Suspend to RAM) circuitry, TI TPA122D amplifier for audio, and the aforementioned Semtech SC1105 voltage regulator for AGP VDDQ generation (i.e., 1.5 V or 3.3 V selection).
For the Intel mainboard, I am missing Motorola MMFT3055EL.
This tiny chip is used in conjunction with, yet again, Linear Technology (the former 76% gross margin firm now a part of Analog Devices) LT1575 (a tiny 8 pin SOIC package device) that cost me $4 per chip!
Although I got the device, I also spent $7 per chip (shipping and sales tax included) for 4 pieces of NOS LT1585A-1.5 (part name LT1585ACM-1.5) LDO voltage regulator.
Linear Technology stuff is always really expensive. . .
For both reference designs, they use Linear Technology devices to generate VTT for GTL+ signalling, and costs so much money for what it does.

The current political climate has, unfortunately, had some negative impact on this project.
Trump administration illegally ending the $800 de minimis exemption has really messed up my ability to acquire NOS components and manufacturing equipment via AliExpress.
Not only the prices have gone up only for us US residents importing items from China, but also the shipping speed has slowed down considerably.
I used to get items in less than 10 days in March and April 2025, now take 2 to 3 weeks.
Worse yet, some items are no longer available to us US residents, but still available to Canadians (this can be easily figured out by setting your shipping destination to Canada, and notice that some items unavailable to US residents are now available).
2 1/2 weeks before de minimis exemption ended, I felt like I had to prematurely (honestly wanted to wait until the PCB was fabricated) order Puhui T-962A reflow oven way ahead of schedule.
It took 2 1/2 months to somehow receive the item, but it appears that the seller had to eat the Trump illegal regressive tariff tax .
I actually ordered the newer T-962A v2.0 model, but ended up getting the older model (T-962A) instead.
In the current political climate, it is not easy to return the item, so I will have to keep the older model.
I did not pay the tariff tax.
I paid the agreed upon price prior to the de minimis exemption illegally going away.
Some components on AliExpress now cost 50% more if the seller does not eat the tariff tax, and this has definitely discouraged me from purchasing more components from China, unfortunately.
On Dollar Express and Select Shipping items, it appears that seller and / or AliExpress is eating the tariff tax, but not all items sold there are Dollar Express or Select Shipping category.
That's where things stand.
I am starting to get good at creating symbols and footprints, so I will continue adding more parts to my custom component library.

Reply 1 of 2, by Nehalem501

User metadata
Rank Newbie
Rank
Newbie

Why do you need to use old components beside the few specific chips?
Couldn’t a modern LDO with the same output voltage be ok? Or even better, some kind of switching step down converter (better efficiency / less heat)

Reply 2 of 2, by bracecomputerlab

User metadata
Rank Newbie
Rank
Newbie
Nehalem501 wrote on 2025-07-10, 13:11:

Why do you need to use old components beside the few specific chips?
Couldn’t a modern LDO with the same output voltage be ok? Or even better, some kind of switching step down converter (better efficiency / less heat)

The first 2 mainboards I am working on are strictly replica of the original reference design.
I do not plan to deviate from the original circuitry, and this is intentional.
The goal of the project is to gain experience designing with KiCad and to also gain experience assembling a predominantly SMD (Surface Mount Device) PCB, including two BGA packages.
I already purchased Puhui T-962A that can bake a full size ATX (305 mm x 244 mm) mainboard for this project.
On the third mainboard and beyond, I will start to deviate from the reference design, and start using components not found in the reference design.
For example, I purchased NOS National Semiconductor LM2636, LM2637, LM2638, and LM2639 for $1 / chip (some of them for around $0.60 / chip), and I intend to use them with Socket 7, Socket 8, Slot 1, and Socket 370 mainboards I am planning to design.
LM2637 and LM2638 appears to generate GTL+ signaling VTT voltage more cheaply than Linear Technology LDO voltage regulator at the microprocessor subsystem BOM (Bill of Materials) level, so I plan to substitute HIP6004 (PMIC of the microprocessor) and two LT1587 (to generate VTT for GTL+ signaling) with LM2637 + miscellaneous cheaper components.
LM2637 is meant for uni-processor GTL+ signaling mainboard and LM2638 appears to be meant for dual processor GTL+ signaling mainboard.