Old School Hack

[ClausB]

I'm stuck in the 80s. I upgraded an Atari 8K RAM board to 48K. Not as interesting as some modern upgrades, no, but fun to do.

 

Had two problems along the way. The old 1978 74LS20 chip had outputs stuck at 1.1 V. Had to buy a new one. Second, there was a solder splash shorting two signals under the rat's nest. Old hands and eyes are not so steady anymore.

 

Will post a schematic later.

 

Edited November 17, 2019 by ClausB

[Mclaneinc]

Nothing wrong with the old ways but yes as you say our old eyes and hands are an issue

 

Nice to see good old fashioned stuff still going on..

[TGB1718]

I just love this sort of thing ?

[jacobus]

Is it a trick of the light or are those tin (as opposed to gold) connectors?

[ClausB]

They are indeed tin. Early 800s and 400s lacked gold-plated edge connectors. I would expect most 8K RAM boards were early enough to lack gold.

 

In the day I had to reseat my early 400's boards and chips every few months!

[ClausB]

Schematic for this mod. I started with a depopulated 8K board (which had been converted for 16K), and I had no spare 74LS10 (the stock chip at Z501), so I substituted a 74LS20. This board plugs into RAM slot 2 (middle RAM slot), leaving slot 1 empty and slot 3 free for peripherals. A wire to the OS board in slot 0 carries the required signal, A15 (Z403 pin 9).

 

 

An alternative design uses the original 74LS10 and adds a 74LS08, and needs fewer wires.

Edited November 18, 2019 by ClausB

[ochen]

Hi,

 

Just wanna also say that I also enjoy these old school hacks. The modern stuff is easier, but it's always fun to bust out the soldering iron. I might have to attempt this on either the 400 or 800 that I got recently, but are being neglected. Can you post the alternate schematic if possible? Which reminds me that I got an 800xl a bit ago and I think it included a memory upgrade schematic with a name similar to yours printed on it. Maybe it was you? Dated early or mid 80s... just curious.

[Faicuai]

On 11/16/2019 at 8:20 PM, ClausB said:

I'm stuck in the 80s. I upgraded an Atari 8K RAM board to 48K. Not as interesting as some modern upgrades, no, but fun to do.

 

Had two problems along the way. The old 1978 74LS20 chip had outputs stuck at 1.1 V. Had to buy a new one. Second, there was a solder splash shorting two signals under the rat's nest. Old hands and eyes are not so steady anymore.

 

Will post a schematic later.

 

 

Practically speaking, I may not enjoy the HW-upgrade that much per-se, but I really enjoy (and to a MUCH larger degree) the skills, know-how and dexterity displayed in crafting these cute inventions... Talk about a REAL, TRUE 800 "connoisseur" and user, clearly beyond anything I remember knowing back at the time, and even today!

 

I am also surprised that the power-consumption has not been even raised, here... I've always wondered how much a 8K, 16K (or even this board) really consume when powered, based on the chips used on-board (!) What about 3x16K + ROM board? Have you ever had a chance to check it?

 

All I know (since I am a bit lazy in reaching back my original ROM+RAM set), is that the machine, powered-on, with NO boards on the expansion bays, chews up about 9.1 watts (minus 1.8w parasitic, from just having the Power Supply on), so that equates to a 7.3w. net). At least I know that's where it all starts...

 

Just wondering....

 

 

Edited November 19, 2019 by Faicuai

[ClausB]

Thanks. I haven't measured the power, but going by datasheets: A 4116 draws 24 mA of 12 V at 1.8 MHz, so 48K worth, or 24 chips, draw 6.8 Watts. A 4164 draws 25 mA of 5 V, so 64K, 8 chips, draw 1.0 Watts.

[ClausB]

1 hour ago, ochen said:

Can you post the alternate schematic if possible? Which reminds me that I got an 800xl a bit ago and I think it included a memory upgrade schematic with a name similar to yours printed on it. Maybe it was you? Dated early or mid 80s... just curious.

I haven't built and tested the simpler alternate design yet, though it's equivalent logic, so it should work.

 

Yes, it was likely my DIY 256K upgrade for the 800XL. RAMbo and Wizztronics were commercial versions of it.

 

[Faicuai]

41 minutes ago, ClausB said:

Thanks. I haven't measured the power, but going by datasheets: A 4116 draws 24 mA of 12 V at 1.8 MHz, so 48K worth, or 24 chips, draw 6.8 Watts. A 4164 draws 25 mA of 5 V, so 64K, 8 chips, draw 1.0 Watts.

Thanks!

 

Your #s seem more reasonable, instead of the 12W+ I read around, some time ago... I will take yours, instead. So a stock 800 should be inhaling about 16watts of power, plus whatever the personality board (ROM) requires.

 

Just for fun, the lowest I have been able to bring it is 10.8W with Incognito, on any RAM / ROM config., as long as SDX is out of the way... So there is a bit of savings there, and less stress on the power-supply block (I've found those hunky / tank-like capacitors on the power-board appear to be running at 50C all day-long, per the infrared-gun...)

[BillC]

1 hour ago, ClausB said:

Thanks. I haven't measured the power, but going by datasheets: A 4116 draws 24 mA of 12 V at 1.8 MHz, so 48K worth, or 24 chips, draw 6.8 Watts. A 4164 draws 25 mA of 5 V, so 64K, 8 chips, draw 1.0 Watts.

You only listed the +12V current, the 4116 also uses +5V/-5V

According to several datasheets the 4116 consumes 462mW active, 24 @ 462mW = ~11.1W.

[ClausB]

The -5V bias supply is very low current. The +5V supply is just for the data out buffer, so it depends on the data bus load. The main chip power comes from the +12V supply, and it is very dependent on access frequency. My datasheet has graphs from which I estimated 24 mA at 1.8 MHz. What is the access frequency in your datasheets?

[ClausB]

I did that simplistic power analysis a while ago and pasted it above. Here's a more thorough one:

MOSTEK 4116 data sheet (https://console5.com/techwiki/images/8/85/MK4116.pdf) shows 462 mW max at max cycle rate of 2.7 MHz. Can't use that because the Atari's cycle rate is 1.8 MHz and not every chip is active every cycle.

Page 2 shows standby 12V current as 1.5 mA. Fig. 2 shows an equation for max active 12V current, giving 27 mA at 1.8 MHz.

 

Say each 16K bank is active at most 1/3 of the time on average (because there are 3 banks) during non-refresh cycles. All banks are active during refresh, which the Atari does 8% of the time. So each bank is active 39% of the time. Each chip then averages 0.39*27 + 0.61*1.5 = 11.5 mA.

 

At 12V that's 137 mW per chip, plus less than 1 mW at +5V and -5V, plus bus load. 24 chips then draw 3.3 W max. Bus load cancels in the comparison with the 4164, and the difference is only 2.3 W!

 

 

[Faicuai]

2 hours ago, ClausB said:

I did that simplistic power analysis a while ago and pasted it above. Here's a more thorough one:

MOSTEK 4116 data sheet (https://console5.com/techwiki/images/8/85/MK4116.pdf) shows 462 mW max at max cycle rate of 2.7 MHz. Can't use that because the Atari's cycle rate is 1.8 MHz and not every chip is active every cycle.

Page 2 shows standby 12V current as 1.5 mA. Fig. 2 shows an equation for max active 12V current, giving 27 mA at 1.8 MHz.

 

Say each 16K bank is active at most 1/3 of the time on average (because there are 3 banks) during non-refresh cycles. All banks are active during refresh, which the Atari does 8% of the time. So each bank is active 39% of the time. Each chip then averages 0.39*27 + 0.61*1.5 = 11.5 mA.

 

At 12V that's 137 mW per chip, plus less than 1 mW at +5V and -5V, plus bus load. 24 chips then draw 3.3 W max. Bus load cancels in the comparison with the 4164, and the difference is only 2.3 W!

 

 

VERY nice, talk about some real numbers there (!)

 

So that means stock 48K (3x16K) consumption should be in the vicinity of 3.3 W (with 4116) + OS/ROM personality board (?)

 

If that is the case, that is much lower than I ever imagined... So 9.1W (no boards) + 3.3W = 12.4W (RAM representing about 25%-27% of total power budget...

Edited November 20, 2019 by Faicuai

[ClausB]

We've ignored the power draw of the TTL chips on the RAM boards, which is about 0.3 W. So one 48K board will save another 0.5 W or so over 3 16K boards. Total savings, rounded up, is about 3 W.

[ClausB]

Here is the simplified version which uses the original 74LS10 at Z501, so fewer wires and lifted pins. Built and tested.

 

Look for complete modification instructions in an upcoming issue of pro(c) ATARI magazine!

 

Edited November 22, 2019 by ClausB

[ClausB]

Well, the prototype became unstable, probably the marginal CAS timing. Revised it to use Phi2 like the later Atari boards. Working so far. More testing ahead...

[ClausB]

And yes, that's what I did on Thanksgiving morning.

[ClausB]

Latest schematic:

 

 

[ClausB]

Turns out the 800 was unstable even with a good 16K board, and so the 48K board was likely working all along. Happy accident though because the redesign has more solid CAS timing, based on Phi2 instead of a couple gate delays.

 

Reseating boards and chips in the 800 fixed it, and it and the 48K RAM have run flawlessly for 36 hours.

[ClausB]

On 11/21/2019 at 7:54 AM, ClausB said:

... So one 48K board will save another 0.5 W or so over 3 16K boards. Total savings, rounded up, is about 3 W.

Measured with Fluke clamp meter at 10.2 Vac:

1.8 A w/ 3 x 16K

1.5 A w/ 1 x 48K

 

So difference is indeed 3 W.

Edited December 20, 2019 by ClausB

[ClausB]

Correction: voltage under load is 9.4 Vac, so power savings is 2.8 W.

[Faicuai]

On 12/20/2019 at 1:29 PM, ClausB said:

Measured with Fluke clamp meter at 10.2 Vac:

1.8 A w/ 3 x 16K

1.5 A w/ 1 x 48K

 

So difference is indeed 3 W.

Just wanted to close the loop here, with more precision around the power-consumption:

 

 

So I prepared a special version of SuperSalt ROM that gets loaded as virtual-cart in RAM ($8000-$BFFF) by means of active "Dellta-Loader" header attached to disk-based ROM-image. And also (finally) got the freaking SuperSalt Test Assembly, which I meant to do so but neglected for ages.

 

So, theory goes, with Incognito Installed, NO carts on cart-ports, and NO ram-cards on any other slot, except 0 (Incognito itself), and being fed through SuperSalt's A/C-to-A/C power-through feed, the total watts-draw is (MC+) * (VI+) = 4.91 * 1.81 = 8.8871 watts (NET), which seems rock-solid stable along the burn-in. 

 

Combined with the (externally measured) 1.8watts parasitic draw from the P//S, that gives a gross total of 10.6871 W. So external watt-meter was pretty on the mark, after all.

 

Cheers!

 

[ClausB]

Here are the text and images from the article:

48K800.txt