Bear with me here... Okay, the 1227727 and 1227730 are equals except for a quad driver and a weather proof case. You can use binaries for one in the other, and vice versa with limited problems. The 16149396 is supposedly a 7727 with extra sram (for their e transmission) in a space that is not populated in the 7727. I believe somebody here translated that to the 7730 and found that it too had several chip pads that weren't populated... so then the 4k ram modification was born by populating these, or later with an add-on board.

My question is this - WHY can't or hasn't anybody used the 16149396 which need no modifications to have that 4k of extra ram? Is there some internal difference that nobody mentions between the '96 and '27/30?

 

The extra 2kb of memory in the 16149396 looses it's contents when the power goes off. And secondly, 2kb of ram is not big enough to hold the calibration data. An extra 4kb of nvram makes it possibly to store and retain the calibration data outside of the eprom in the memcal (because nv ram is battery backed), and because it's ram instead of rom, the ECM can alter that area of memory. By modifying the mode 4 ALDL command to write to any desired memory location, we can use the ALDL interface to update the contents of the 4kb nvram. Instead of using an osterich or prominator or whatever to "emulate" an eprom, the ECM is doing it's own updates to it's calibration. If you have the ability to do surface mount soldering, the hack can be done for under $20 by using the un-populated 2k sram pins and a few other empty parts that aren't populated. The nvram module only exists to make it easier to install the hack because it's built with through-hole parts.

The only down side at the moment is that there is no support for this particular "emulator" in TunerPro R/T or CATS. I have a DOS based utility that uploads the calibration section of a 32kb bin into the nvram. I've also got a working windows GUI written in TCL/Tk scripting language that I'll probably give out soon to help out the folks that have the hack and want to try it out.

And finally, this is only one example of what could be done with an extra 4kb of memory in the ECM. You are only limited by your source code and imagination. This hack wouldn't even be possible if it wasn't for other folks donating commented source code. (thanks guys!)

 

okay, so it's 2k.

As I understand it, it's used as the stack in the stock configuration. How was the 4k ram modification possible? As I understand it, $1800-$1FFF is allocated to this ram on the 16149396 and there are limited free addresses.

 

It's mapped in at 0x2000-0x2FFF.

 

Can you slap in a 4k ram in place of the '9396s 2k ram and use the 1800-2FFF address range?

What I don't understand is why you can't use some kind of EEPROM in place of the RAM... Granted, it's slower, but as long as you change/upload code (if it's necessary) before it's running I don't see the big deal. Then again, I have never dealt with any kind of memory before, and still have a lot of learning to do.

 

You could use eeprom instead if you are willing to write the code to implement the write/poll method. In a sense, nvram is an eeprom without the hassles of the programming process. Keep in mind eeprom is not pin compatible with nvram I used. You would need to swap some wires around (read the device datasheets)

 

An NVSRAM typically has a minimum data retention life of about 10 years. It allows you to read & write it like a normal RAM. It retains the memory data when you remove the power to it just like an EEPROM.

An EERPOM typically has a minimum data retention life of about 20-30 years. It is read like a normal RAM and can be read just as fast as a normal RAM. The writing of the device is slower. It requires a sequence of codes to start the write process. It takes time for the write to occur. It is not instantanious like a RAM. During that write time, it can't be read properly. Therefore, you would need two EEPROMs. You need one from the ECM to use while the other is being written. That means you need to design a circuit board to connect to the ECM RAM pin/holes/footprint.

NVSRAM can be written to indefinitely. EEPROM can be typically written to 10,000 times at a minimum. Some are rated for a min. of 100,000 write cycles. So you technically wear out an EEPROM every time you write a memory location.

The end result is that you aren't going to be installing a PCB with two EEPROMS & other logic if you don't know much about memory devices. Instead, install the NVSRAM and call it done.

Why are you so set on using an EEPROM?

 

I have no idea... the battery backup just rubs me the wrong way.

I also like using what's given vs manufacturing new boards, etc... If I can get by by using an already available ECM... Are there any faster ECMs with more space out there that have been hacked and commented? (That still use assembler vs compiled code)

 

Don't get too hung up on the battery back part. Its part of the chip, and well designed. Looks like a 'fat' IC, so there is no battery or anything externally.

in all reality, the NVSRAM is really THE hot ticket, but not many people have picked up on it.

if it had support, (Montecarslow has dos commands, and sound like some windows stuff is in the works), it would be hugely popular. becuase it works right through the aldl, no extra hookups, and like was mentioned, parts cost around 20 bucks.(requires some soldering skills) Montecarslow's board is worth the money though, becuase it can be installed from ecm to ecm easily (three wires soldered)

the whole sram thing is actually, the factory designed way to alter contents and was designed when the whole ecm was being calibrated and troubleshot, in all reality, its like a factory 'emulator' built in.

I've gotten it wired into a test bench ecm, but haven't had the time to fire it up and get playing yet (just about a thousand other things going on, thats all)

Ernst, when you get the tck stuff up and running, I can run some tests for you, if you need it.