RAM current, power tests

Last updated June 4 2018. Edited by Herb Johnson, (c) Herb Johnson, except for content written by Lee Hart and others. Contact Herb at www.retrotechnology.com, an email address is on that page.


This Web page was about testing of .300-wide "narrow" DIP-packaged RAM for average DC current consumption on the 1802 Membership Card., which can be a low-power microcomputer. IN 2018 I added low-power information about wide .600 mil ROMs and RAMs. "Low power memory" means memory devices which draw only milliamps of currrent in use, and microamps when disabled.

Here's why this matters. The Membership Card has a power connector which, when removed, not only removes DC power, but also stops the CPU clock, resets the 1802, disables the LEDs. Thus the card is in "standby" mode; with RAM and ROM powered by a supercap. When the cap discharges to a few volts, some RAM models go into a "low-voltage standby" mode to save contents.

The standard kit comes with older "slow" all-CMOS parts, so power consumption is essentially zero in standby, and the supercapacitor will maintain RAM memory for days. Other parts, "faster" CMOS or non-CMOS, will only maintain RAM memory for hours or less. The other virtue of CMOS RAM and ROM is (or can be) low-power consumption in operation - longer battery life. An RS-232 interface in use, may also increase supercap rate of discharge; the RS-232 interface is also a power source for the low-power Membership Card.HEre's what Lee Hart means about drawing DC power from an RS-232 port.

Identifying low-power CMOS While some memory devices are identified as "CMOS", and some have a "C" in their model name like 27C256, those designations don't guarantee low-current consumption in use. Generally the faster-rated the device, the more current it consumes (in order to switch quickly). "600 mil" is simply the width of the IC package - 6/10 of an inch (like the CPU), or "300 mil" for 3/10 inch (like most of the chips). The device data sheets may be informative, but often you have to hand-test for low current. Another factor, is that IC's imported from "overseas" scrap, may be remarked - the markings may not match the chip.

Tests: Lee Hart tested in Jan 2017 and Mar 2018, several brands of .300-wide static RAM. Herb Johnson did similar tests in 2015, which follow Lee's account. These RAMs were often designed for fast-cache operation with Intel 80486 processors. It's clear, current consumption varies by RAM brand and rated CPU speed of the RAM. Current decreases in most cases, when the RAM's outputs are disabled, because they use high-current drivers for speed.

A low-power application considered in this report, is stealing current from a RS-232 port to DC-power the 1802 Membership Card.

Tests of 1802 Membership Cards at faster speeds and reports of current and power, are on this linked Web page.

Current consumption, memory retention - RAM and ROM

These dynamic tests below may not be consistent, in part because running power consumption varies a lot depending on the program. But they show there is quite a difference in standby and running current, among the RAMs and ROMS discussed. "Standby current" amounts to the voltage left on the supercap, to retain memory contents; typically about 2 volts. - Herb

Lee said in 2015-16: "What's the best choice so far, to run the 1802 Membership Card from current drawn from a RS-232 port? The Sony CXK58257ASP RAM, at about 15ma at 4v running a BASIC interpreter in [CMOS] ROM, at 1.8MHz clock. But that's about twice the current I can [draw upon to make DC power] from an RS-232 port. If I drop the clock speed and baud rate, I can get it to run at 3v at 10ma, which is getting close." Lee also reported the Hitachi HM62256BLSP is a low-current device, but more than the Sony device.

In June 2018, Lee Hart and I discussed low-power (low current) wide or 600 mil RAM and ROM available for the Membership Card. Some (not all!) CMOS EPROMS with "C" in the model name, are really low power. Slower speed ratings usually equals lower power. Lee found the National 27C256Q 450nsec is low-power. He also reported old 27C32's and 27C64's may be low-power, but small-capacity for some uses.

"[I ran an] 1802 M/S Card rev.J with National NMC27C256Q EPROM, Hitachi HM62256BLSP 0.3" RAM. Start with supercap charged to 5.30v (what my RS-232 dongle supplies). Unplugged Front Panel [power connector]. After 48 hours, supercap at 1.41v. Plugged Front Panel back in, and the BASIC program I left in memory was still intact and ran. Unplug P4 and DB25 connectors; after 48 hours, supercap at 1.30v, program still intact."

"I ran [another] test with a Cypress CY7C199 300 mil RAM. *Much* higher power consumption than the Hitachi HM62256BLSP. With just the CPU board alone, the voltage fell from 5.3v to 0.7v overnight (about 10 hours), and memory contents was lost. Peak power consumption is around 70ma when reading RAM continuously (such as in LOAD mode)." - Lee Hart, June 2018

Tests and considerations

300 mil RAM, Lee Hart, Jan 22-24 2017

I did some more testing on different brands of 32k 0.3" wide "skinny" RAMs [with the Rev H2 CPU]. I was looking for one low enough power to be able to run it from a serial port for both data *and* power. Milliamps of current can be "harvested" from RS-232 connections.

These narrow SRAMS were generally used as fast cache memory for 80486 processors. They are either all-CMOS chips, or "partial-CMOS chips. All-CMOS chips have essentially zero power consumption when all the inputs are static; at either 0v or VCC. Partial-CMOS chips have low (but not zero) power consumption when not selected; but their power consumption rises dramatically when they are chip-selected.

In addition to this static power consumption, there is a dynamic power consumption. The more inputs that are changing, and the higher the frequency of the changes, the more power they use. The power is consumed by high-current drivers, to get the 10-20 nanosecond data rates which cache needs.

Finally, of course the supply current varies with supply voltage. The 1802 and RAM can run at voltages below 5 volts. All this complicates power consumption measurements.

Here is some static power consumption data, again Rev H2 CPU, 1.8Mhz:

VCC milliamps:	/CE=/OE=1(not selected)		/CE=/OE=O (selected)
part number	 3v	 4v	 5v	 	3v	 4v	 5v
-----------	----	----	----		----	----	----
MCM6206NP45	2	2.2	3.1		25	43	69
CXK58257ASP-10L	0	0	0		0.14	0.25	0.4
CY7C199-35PC	0.7	1.2	1.5		12	35	60	
HM62256BLSP-12	0.06	0.08	0.1		0.3	0.56	0.9
UM61L256K-15	0.14	0.23	0.37		19	49	77

Next is a dynamic test: I put each RAM in an 1802 Membership Card, and measured the TOTAL power consumption running under ROM with a 1.8 MHz clock.

part number	 3v	 4v	 5v
-----------	----	----	----
MCM6206NP45	20	28	43
CXK58257ASP-10L	8	13	24
CY7C199-35PC	17	25	35
HM62256BLSP-12	10	19	28
UM61L256K-15	27	47	75

The standard 0.6" wide RAMs I'm using, have power consumptions very similar to the Sony CXK58257 and Hitachi HM62256. Their extremely low STATIC power consumption, is what determines how long the supercapacitor can maintain memory with power removed and of course the processor idle.

- Lee Hart, with edits and updates.

300 mil RAM, Lee Hart, March 2018

The highest current occurs in LOAD mode during the anomalous "before you press IN the first time" state. The data bus has a fight; the IN switch setting and 1802 D register are both on the bus at the same time. This can exceed 100ma when all 8 bits are different.

Normal LOAD mode power consumption is higher because the memory is chip-selected 100% of the time. With newer not-quite-CMOS memories, this results in high static power consumption. During RUN mode, supply current is less because the memory isn't selected with as high a duty cycle. - Lee Hart

Lee later added, that near 5 volts, the 5.1V Zener across Vcc starts to conduct and adds some milliamps. This varies with each diode.

Mar 27 Test setup:
    1802MC rev.I, standard 1.8MHz clock speed
    32k EPROM, National NMC27C256Q-120
    32k 0.3" RAM, Hitachi HM62256BLSP-12
    RS-232 serial
    power connected to P4
    RUN-to-VDD jumper (enables CPU to run)
    /ON-toVDD open (disables Front Panel LEDs)
    running my [display a] MENU program with RCA's BASIC3 

Running BASIC3 at VDD=3.6v, 2.98ma
Running BASIC3 at VDD=5v, 5.12ma - some Zener current
In LOAD mode after the first IN button press - Vdd=5V, 11.97mA

Mar28: 1802MC rev.G running a BASIC3 program, LEDs disabled:
with HM62256 32k RAM
	3.6v at 4.4ma
	5.0v at 8.75ma
CY7C199 32k RAM:
	3.6v at 13.5ma
	5.0v at 27ma

Lee found the lowest-current 300 mil RAMs to be Hitachi HM62256BLSP or Sony CXK58257ASP. These accounted for most of the narrow RAMs he's offered as of March 2018. Since some RAMs sold, particular as used parts, are "remarked", it's hard to know what you've bought until testing. - herb

300 mil RAM, Herb Johnson, Sept 2015

I performed total-power tests in Sept 2015. Rev G CPU modified to Rev H, 1.8MHz, no PROM, only narrow SRAM. See this Tech Note about the Rev G/H modification. DC voltage between 4.3V and 4.5V. At "idle" no execution state but RAM selected; versus running a simple program in RAM.

		mA	mA
SRAM	    	not-run	run	DC volts 	brand

AS7C256-20PC  	13.5	16.5	4.28V		Alliance
same	      	12.0	13.0
same	      	13.0	12.5
same	      	12.0	12.5
same	      	13.5	14.0	4.5V
CY7C199-15    	29.0	30.5			Cypress
same	      	28.5	30.0	4.33V
same		29.0	30.0
CY7C199-25    	50.0	53.0
CY7C199-25PC  	52.0	54.0
UM61L256K-15  	58.0	59.5
MS62256A-25NC 	23.0	24.0			MOSEL
W24257AK-15   	25.5	27.0
W24257AK-20   	26.5	27.5			Windbond

Similar total-power tests on May 2015, current varies +/-1mA:

SRAM			run mA	DC volts

AS7C256-20PC    	16  4.28V
CY7C199-25PC    	53  4.05V
MS62256A-35NC    	44  4.2V
W24257AK-15    		28  4.2V (Winbond)
LM61256FK-15    	36  4.15V
GLT7256M08-15T    	44  4.09V

Total-power tests, May 2015, among some SRAMs. Running a 27C128 ROM with monitor, with occasional access to RAM. This primarily measures current consumption by everything but RAM. Data suggests a "baseline" current consumption at 4.4V of under 14mA.

SRAM			run mA	brand
AS7C256-20PC    	15	Alliance
CY7C199-25PC    	17
MS62256A-35NC    	14.5	MOSEL
LM61256FK-15    	14
GLT7256M08-15T    	16
IS61C256-20N    	15

- Herb Johnson

This page and edited content is copyright Herb Johnson (c) 2018, execept content written by Lee Hart. Contact Herb at www.retrotechnology.com, an email address is available on that page..