FlexMem: <$25 Memory expansion for ANY propeller board with a 10 pin header! (F
Bill Henning
Posts: 6,445
Hi guys,
In my PropCade thread, heater asked about a low pin count parallel SPI ram solution... I was not going to reveal FlexMem until it was tested, but since he was asking...
Like PropCade, FlexMem was inspired by my work with VMCOG. I have not had a chance to write the driver yet, however that is high on my list of priorities... until then, the driver from obex.parallax.com/objects/346/ Tim Moore's "N256S0830HDA SPI SRAM driver" should work, once the pins are adjusted to FlexMem.
After sending the design to Sapieha, he made another amazing layout!
FlexMem Features:
- Works on ANY Propeller board that supports the 10 pin ProtoBoard style header
- up to THREE FlexMem's can share the header
- works best on P0-P7 with the driver I have designed
- 3.3MB/sec burst read/write with one cog for RAM or FRAM
- 6.6MB/sec burst read/write with two cogs for RAM or FRAM
- four identical devices on the board at the same time
- supports SOIC-8 devices
- supports DIP-8 devices
- works with 23K256 32KB Microchip SPI ram's (128KB/board)
- works with FRAM's (up to 512KB/board)
- works with FLASH (up to 256MB/board)
- jumper selectable for one of three chip selects
- approx. 1 3/16" x 1 1/16" and 15/16" tall (including stacking header)
FlexMem will work with ANY Propeller board with 8 pins available (p0-p7 is ideal, p8-p15 or p16-p23 will work but will be slower)
This thread with be the "Official" FlexMem thread, and I will be uploading photo's, and documenting my progress in debugging the prototype.
Frequently Asked Questions
Q1) Why is FlexMem faster on P0-P7?
A1) The driver needs the four data bits to be on P0-P3 so that the MOVI instruction can be taken advantage of in the driver to assemble larger units of data
*** Initially only P0-P7 will be supported, as the performance hit makes using other pins unattractive
Q2) Why are you not using eight SPI ram's for a byte wide memory?
A2) A byte wide FlexMem is certainly possible, however economically it makes far less sense. Eight SPI 32Kx8 memory chips cost significantly more than a single 512Kx8 parallel sram chip. It would be easy to use two FlexMem's to provide an eight bit wide data bus, however it would require ten Propeller pins (P0-P7, and two other pins)
Q3) Why will the counter-assisted single cog driver burst at 3.3MB/sec when the bit-banging one can only achieve 2MB/sec?
A3) The bit banging driver requires two more instructions for every four bits read, which is an overhead of 16 instructions for reading a LONG, whereas the counter based driver only requires a two instruction overhead for reading a LONG - a savings of 14 instructions per LONG.
Q4) When will FlexMem be available for sale?
A4) As soon as the drivers are debugged, I will make a limited number of FlexMem kits available for sale. Large quantities are expected to be available for sale in early May, 2010.
Please feel free to ask any questions you may have!
(photo's will be added in a few hours)
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
www.mikronauts.com E-mail: mikronauts _at_ gmail _dot_ com
My products: Morpheus / Mem+ / PropCade / FlexMem / VMCOG / Propteus / Proteus / SerPlug
and 6.250MHz Crystals to run Propellers at 100MHz & 5.0" OEM TFT VGA LCD modules
Las - Large model assembler Largos - upcoming nano operating system
Post Edited (Bill Henning) : 4/16/2010 7:43:44 PM GMT
In my PropCade thread, heater asked about a low pin count parallel SPI ram solution... I was not going to reveal FlexMem until it was tested, but since he was asking...
Like PropCade, FlexMem was inspired by my work with VMCOG. I have not had a chance to write the driver yet, however that is high on my list of priorities... until then, the driver from obex.parallax.com/objects/346/ Tim Moore's "N256S0830HDA SPI SRAM driver" should work, once the pins are adjusted to FlexMem.
After sending the design to Sapieha, he made another amazing layout!
FlexMem Features:
- Works on ANY Propeller board that supports the 10 pin ProtoBoard style header
- up to THREE FlexMem's can share the header
- works best on P0-P7 with the driver I have designed
- 3.3MB/sec burst read/write with one cog for RAM or FRAM
- 6.6MB/sec burst read/write with two cogs for RAM or FRAM
- four identical devices on the board at the same time
- supports SOIC-8 devices
- supports DIP-8 devices
- works with 23K256 32KB Microchip SPI ram's (128KB/board)
- works with FRAM's (up to 512KB/board)
- works with FLASH (up to 256MB/board)
- jumper selectable for one of three chip selects
- approx. 1 3/16" x 1 1/16" and 15/16" tall (including stacking header)
FlexMem will work with ANY Propeller board with 8 pins available (p0-p7 is ideal, p8-p15 or p16-p23 will work but will be slower)
This thread with be the "Official" FlexMem thread, and I will be uploading photo's, and documenting my progress in debugging the prototype.
Frequently Asked Questions
Q1) Why is FlexMem faster on P0-P7?
A1) The driver needs the four data bits to be on P0-P3 so that the MOVI instruction can be taken advantage of in the driver to assemble larger units of data
*** Initially only P0-P7 will be supported, as the performance hit makes using other pins unattractive
Q2) Why are you not using eight SPI ram's for a byte wide memory?
A2) A byte wide FlexMem is certainly possible, however economically it makes far less sense. Eight SPI 32Kx8 memory chips cost significantly more than a single 512Kx8 parallel sram chip. It would be easy to use two FlexMem's to provide an eight bit wide data bus, however it would require ten Propeller pins (P0-P7, and two other pins)
Q3) Why will the counter-assisted single cog driver burst at 3.3MB/sec when the bit-banging one can only achieve 2MB/sec?
A3) The bit banging driver requires two more instructions for every four bits read, which is an overhead of 16 instructions for reading a LONG, whereas the counter based driver only requires a two instruction overhead for reading a LONG - a savings of 14 instructions per LONG.
Q4) When will FlexMem be available for sale?
A4) As soon as the drivers are debugged, I will make a limited number of FlexMem kits available for sale. Large quantities are expected to be available for sale in early May, 2010.
Please feel free to ask any questions you may have!
(photo's will be added in a few hours)
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
www.mikronauts.com E-mail: mikronauts _at_ gmail _dot_ com
My products: Morpheus / Mem+ / PropCade / FlexMem / VMCOG / Propteus / Proteus / SerPlug
and 6.250MHz Crystals to run Propellers at 100MHz & 5.0" OEM TFT VGA LCD modules
Las - Large model assembler Largos - upcoming nano operating system
Post Edited (Bill Henning) : 4/16/2010 7:43:44 PM GMT
800 x 391 - 198K
Comments
More pics!
For sale where? Price?
Question: Why is the module slower on pins other than 0-7?
OBC
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Are you Propeller Powered? PropellerPowered.com
Visit the: PROPELLERPOWERED SIG forum kindly hosted by Savage Circuits.
I will post more pics later. Now to your questions:
1) For sale where? (I think you meant when...)
After I verify functionality, I will have a very limited number of boards available immediately, with large numbers of boards being available about three weeks after that. The boards will be available from my site, orders will be taken by email. I will also make them available through Ebay, at a higher price.
2) Price?
MSRP for the kit including 128KB of SPI RAM will be $24.95USD+s/h
The forum special will likely be $19USD+s/h for the same kit (with 128KB of SRAM)... but the order will have to include your forum handle for that price.
3) Why is the module slower on pins other than 0-7?
My super-duper double-secret special sauce driver is much faster (due to the Prop's architecture) on P0-P7
Using four cogs 8MB/sec is not out of the question for burst reads.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
www.mikronauts.com E-mail: mikronauts _at_ gmail _dot_ com 5.0" VGA LCD in stock!
Morpheus dual Prop SBC w/ 512KB kit $119.95, Mem+2MB memory/IO kit $89.95, both kits $189.95 SerPlug $9.95
Propteus and Proteus for Propeller prototyping 6.250MHz custom Crystals run Propellers at 100MHz
Las - Large model assembler Largos - upcoming nano operating system
Hanno
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Co-author of the official Propeller Guide- available at Amazon
Developer of ViewPort, the premier visual debugger for the Propeller (read the review here, thread here),
12Blocks, the block-based programming environment (thread here)
and PropScope, the multi-function USB oscilloscope/function generator/logic analyzer
I people have been asking for an inexpensive memory expander for quite a while. I'd considered SPI ram's before, but before I started working on VMCOG I did not see a market for a four bit memory card - but it becomes quite useful when combined with VMCOG and my high speed four bit driver (in the works).
Talking about innovation... I can't wait to get my hands on Catalina with Black Cat and ViewPort support...
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
www.mikronauts.com E-mail: mikronauts _at_ gmail _dot_ com 5.0" VGA LCD in stock!
Morpheus dual Prop SBC w/ 512KB kit $119.95, Mem+2MB memory/IO kit $89.95, both kits $189.95 SerPlug $9.95
Propteus and Proteus for Propeller prototyping 6.250MHz custom Crystals run Propellers at 100MHz
Las - Large model assembler Largos - upcoming nano operating system
Now that I have the bit-banging single channel SPI ram driver running on PropCade, I am integrating it with VMCOG. As soon as that works, I will make a FlexMem version of VMCOG that uses a bit-banging quad channel SPI ram driver.
The PropCade bit-banging driver provides approximately 0.45MB/sec burst read/write speeds. A counter based future version is expected to provide approximately 1MB/sec burst speeds.
The initial FlexMem bit-banging driver will provide approx. 2MB/sec burst read/write speeds. A counter based future version is expected to provide approximately 3.3MB/sec burst speeds with one cog.
I expect to have the initial bit-banging FlexMem VMCOG running early next week.
If there is enough interest (ie enough FlexMem's sold) I have a paper design for a two-cog FlexMem driver that can achieve 6.6MB/sec read/write bursts.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
www.mikronauts.com E-mail: mikronauts _at_ gmail _dot_ com
My products: Morpheus / Mem+ / PropCade / FlexMem / VMCOG / Propteus / Proteus SerPlug
and 6.250MHz Crystals to run Propellers at 100MHz & 5.0" OEM TFT VGA LCD modules
Las - Large model assembler Largos - upcoming nano operating system
And if I read correctly you can have upto 512k per board ( FRAM ).. How many boards can be stacked and addressable at the same time?
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
New ICON coming, gotta wait for the INK to heal, now we have colour!
FlexMem boards must always be populated with four identical devices. With currently available SOIC-8 RAMTRON parts the following configurations are possible: (per board)
256KB with 4 x FM25V05
512KB with 4 x FM25V10
If RAMTRON releases a FM25V20, 1MB will be possible with four FM25V20 devices.
Up to three FlexMem's can share an EXP slot. My drivers will only support FlexMem's on P0-P7 as it is MUCH faster there than on other pins.
Please note, RAMTRON populated FlexMem boards will be fairly expensive, RAMTRON charges a LOT for their chips.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
www.mikronauts.com E-mail: mikronauts _at_ gmail _dot_ com
My products: Morpheus / Mem+ / PropCade / FlexMem / VMCOG / Propteus / Proteus / SerPlug
and 6.250MHz Crystals to run Propellers at 100MHz & 5.0" OEM TFT VGA LCD modules
Las - Large model assembler Largos - upcoming nano operating system
Pardon my ignorance... Why do you call it a "parallel" SPI SRAM interface? I'm sure you stated it correctly, but to my limited comprehension it seems oxymoronic.
Bill is correct - Them are 4x1 Parallel interfaced IC's that reads all 4 Bytes parallel
This give You one long with one read. 4 Bytes parallel = One long.
Regards
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Nothing is impossible, there are only different degrees of difficulty.
For every stupid question there is at least one intelligent answer.
Don't guess - ask instead.
If you don't ask you won't know.
If your gonna construct something, make it·as simple as·possible yet as versatile as posible.
Sapieha
The reason I call it "parallel" SPI ram interface is because I am using four SPI ram devices with common common /CS and CLK signals but separate data lines (DO and DI tied together at each device) - thus instead of reading/writing one bit per SPI clock cycle, I am reading/writing four bits per SPI clock cycle.
Mind you, I have to make the address and command bits identical to all four devices [noparse]:)[/noparse] but that is easy enough.
Tim Moore did something similar, however I have designed a more efficient driver that should allow my design to perform roughly 1.5x as fast as his driver with one cog, and 3x as fast with two cogs.
And yes, it is oxymoronic... but it should work just fine [noparse]:)[/noparse]
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
www.mikronauts.com E-mail: mikronauts _at_ gmail _dot_ com
My products: Morpheus / Mem+ / PropCade / FlexMem / VMCOG / Propteus / Proteus / SerPlug
and 6.250MHz Crystals to run Propellers at 100MHz & 5.0" OEM TFT VGA LCD modules
Las - Large model assembler Largos - upcoming nano operating system
Post Edited (Bill Henning) : 4/16/2010 8:20:06 PM GMT
Chances I'll buy one? Ca. 100%.
VMCOG 0.92 works, so I will now start making a FlexMem version of VMCOG!
I should have something running next week (consulting work permitting)
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
www.mikronauts.com E-mail: mikronauts _at_ gmail _dot_ com
My products: Morpheus / Mem+ / PropCade / FlexMem / VMCOG / Propteus / Proteus / SerPlug
and 6.250MHz Crystals to run Propellers at 100MHz & 5.0" OEM TFT VGA LCD modules
Las - Large model assembler Largos - upcoming nano operating system
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Propeller Pages: Propeller JVM
The good news is that (on paper) it looks like my driver ought to be 50% faster than his.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
www.mikronauts.com E-mail: mikronauts _at_ gmail _dot_ com
My products: Morpheus / Mem+ / PropCade / FlexMem / VMCOG / Propteus / Proteus / SerPlug
and 6.250MHz Crystals to run Propellers at 100MHz & 5.0" OEM TFT VGA LCD modules
Las - Large model assembler Largos - upcoming nano operating system
I even tried changing the clock polarity. It's really weird!
I hooked up two chips as per the comments in your driver, using 4 pins: /CS, CLK, DIO1, DIO2
Now that I am getting caught up, I may try your driver again - it is bugging me because your driver should work fine, I checked the data sheets to both spi devices carefully.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
www.mikronauts.com E-mail: mikronauts _at_ gmail _dot_ com
My products: Morpheus / Mem+ / PropCade / FlexMem / VMCOG / Propteus / Proteus / SerPlug
and 6.250MHz Crystals to run Propellers at 100MHz & 5.0" OEM TFT VGA LCD modules
Las - Large model assembler Largos - upcoming nano operating system
I have now also made another driver with 10 MHz, which combines the technik of the older two, to spare a lot of code size.
All 3 drivers works well with 2 x 23K256 from Microchip here. I don't know why they not work on the other side of the world [noparse];)[/noparse]
I use RAMs in SO-8 and have not tried it with DIP versions, maybe that can be a problem. But I have tested it with TQFP and DIP Propellers on different platforms.
Andy
Thanks. I will try again - perhaps I had a bad 23K256? I also tried the files from Tim's object, I quoted these filenames as they were the last ones I tried.
Based on your and Tim's responses, the problem was probably my wiring or a bad 23K256.
Bill
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
www.mikronauts.com E-mail: mikronauts _at_ gmail _dot_ com
My products: Morpheus / Mem+ / PropCade / FlexMem / VMCOG / Propteus / Proteus / SerPlug
and 6.250MHz Crystals to run Propellers at 100MHz & 5.0" OEM TFT VGA LCD modules
Las - Large model assembler Largos - upcoming nano operating system
The test also has an assembly loop which reads 64000 bytes as fast as possible to compare the different drivers.
This version expects another wireing than the others: common clock and data pins and 2 separate cs pins for both rams.
I have testet it also with 96 MHz Prop clock and had no problems (= 12 MHz SPI clock).
Andy
It would be great if it works for me this time; the SPI code I wrote is very simple, and runs only at 4Mbps (@ 80Mhz)
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
www.mikronauts.com E-mail: mikronauts _at_ gmail _dot_ com
My products: Morpheus / Mem+ / PropCade / FlexMem / VMCOG / Propteus / Proteus / SerPlug
and 6.250MHz Crystals to run Propellers at 100MHz & 5.0" OEM TFT VGA LCD modules
Las - Large model assembler Largos - upcoming nano operating system
I've started integrating FlexMem into VMCOG
I hope to release a version of VMCOG with FlexMem integrated into it on Monday next week.
(Obc, I've reserved a board for you)
Now that I fixed that nasty VMCOG bug, I am continuing to debug the initial FlexMem code.
Currently Working:
- Read Status
- Read Long
Currently NOT Working:
- Write Status - only seems to work for one of the four memory chips
- Write Long - does not even seem to get executed
I am using a combination of scope, ViewPort, debugging code, and LED's to try to figure it out.
- the scope shows that I have a good ground and good signals at all required pins
- ViewPort shows that I am generating the right SPI packets except for Write Long
- LED's show the write long code is never executed (VERY STRANGE)
I will keep this thread updated with my progress.
I have made significant progress today...
Currently Working:
- Read Status
- Read Long
- Write Long
Currently NOT Working:
- Write Status - only seems to work for one of the four memory chips
Need to add:
- burst read/write support for VMCOG, this requires a working Write Status routine
The is some good news - my calculations show that I should be able to make:
- 640x480 4 color per pixel bitmapped VGA driver (1 video page per FlexMem)
- 160x192 136 color per pixel bitmapped TV driver (4 video pages per FlexMem)
for boards equipped with a FlexMem once everything is working!
The above requires a counter-driven version of the driver I don't quite have working yet.
My current "higher priority" list that I have to finish before returning to FlexMem and VMCOG is:
1) pPLC (formerly known as PLC-G)
2) Morpheus+
3) Mem*
4) CPUModule and a few small I/O modules
I estimate the above will take until mid to end of November to finish.
Sorry,
Bill
It is possible to simple adapt this driver to 4-lines --->/CS, CLK, DI, DO.
As it is to little text info in driver for me to work on it alone.
And I have some Question ---> LOOK on ViewPort timing diagram picture
That clock pulses are even on DI/DO lines some times.
FlexMem works!!!
After some weeks away from the problem I've managed to find three separate problems, and after I found the third... it worked.
No wonder I had trouble getting it running:
1) one ram chip had an intermittent contact on MISO
2) there was a bug in my debugging code (oops)
3) I had a bad constant for the number of places to shift the address
Now I can go on to get the counter driven clock going to get the approx. 3.6MB/sec (mega bytes not bits) for burst reads/writes that my calculations show I can get...
This was a tremendous pain in the posterior...
(OBC: I have a FlexMem board with your name on it...)
Does that mean that it started working when you held the scope probe on it?
Frankly, I am sure it was seated properly for some of the past tests, but the other two problems I found today would have caused errors as well.
I hate multi-point failures!
The 4 bit wide pasm driver works; I am working on the high speed version with counter assist
I expect >3MB/sec burst read/write speeds, with a possibility for >7MB/sec using a multi-cog driver!
News:
I've ordered a production run, and will have blue production boards in approx. one week.
@Bill, I'll take one of those as soon as they are ready to be shipped. Just post your purchasing info for them.
OBC