Openscshemes - Open-source schematic exchange

[electronrancher]

Hi all,

I am helping a buddy set up his site, and I wanted to get feedback from you guys. His idea is to set up an exchange for commonly used schematics, especially for IC cells like amps, oscillators, etc. It uses all scalable CMOS so you can easily transfer a schematic into your own process.

I am going to be drawing up a bunch of schematics from papers and books over the next few weeks as he sets up the site. Right now, the forum is up but is still very crude. Go ahead and take a look if you like - the site is

forum.openschemes.com

I think the full site is not yet uploaded, but the forum seems to be running stable for about a week now.

The "join forum" option works already, so feel free to join up and post any ideas or suggestions that you have, or post them here and I will ask him to join this board too.

If anyone wants to be a contributor, let me know. I have a few example files from him that I am using to build my cells. I can probably share these with you to get you started. Either post here or PM me - it's a busy month for me but I'll try to respond ASAP!


Think of it - one day if you need a class-AB op amp, you just go download it and scale it to your process and bias current.. No more pages of math equations just to get a result that's already been solved thousands of times, I love it!!

 

[pfd001]

it is a good idea

 

[bicave]

There's many professional designer in this forum And ask they post their work public is quite difficult. if i can, i will post my idea and basic circuits.

Thank you

 

[openschemes]

Hi Guys!

I have joined edaboard - this is my first post! To bicave: I agree entirely, the openschemes project is not for giving production designs, just for the basics like you said. I know it's a lot of work to even recalculate the class-AB op-amp for a new process, so it's generic blocks like this that I think are OK to share.

To all: Right now, I am just building the site - as electronrancher says the forum is OK already. If you want to log in to the forum, go ahead! It would be nice to hear your opinions, or circuit you have in mind that would be a good cell for the site!

Now that I am a member here, you can PM me or email at forum @ openschemes.com if you like. I will post more information here soon when I have the site's download manager set up. I think EDAboard is a great resource, I am hoping to provide the same!

Cheers,
Openschemes

 

[yankuangtu]

it is really great!In the software and digital disign field,there are lots of sucessful samples--such as linux and opencores.org
but there is some different for analog and RF,because it is too detial or somewhat else,so maybe we should try to think clearly that what kind of design is fit for open and reuse.

 

[electronrancher]

exactly! if you think about it, most students and a lot of professional users read the IEEE papers frequently when looking over new topologies. Most of the "classic" architectures of op-amps, comparators, bandgaps, oscillators, etc are seen here first.

Once they are published, the owner's name is "landmarked" to show their advancement, and unless previously patented this new cell is now offered to the community as a possible benefit for their own work.

As far as the types of cells, I am thinking about all sorts of general-purpose stuff that people commonly use. Other than the simple things listed above, a charge-pumped PLL is a good one, output drivers, current comparators and sense-fets, class-ab op amp, switched-cap filters. Most of these things you can calculate out of any ee textbook so it's no "trade secret"!

Let me know if you are interested in contributing - I have not yet started building cells, so it's a good time to agree on what tools to use, etc..

Added after 4 hours 12 minutes:

Now here's where things get interesting - A free MOSIS tapeout!

OS has earmarked some funds for startup costs which will hopefully include fabricating the first batch of cells that make it onto the site. This means a free MOSIS tapeout to verify and characterize the IP available on the site.

Right now, it's just one process but the intent is to keep fabbing out IP periodically to get it's data on multiple processes. The processes will of course be available from MOSIS, so designers can use their public-domain models (at first) to port the cells into another process with reasonable confidence..

I just got the OK to start talking about this, so that's why I've not said anything to you guys yet. The main threads will be here and on the Openschemes site at
**broken link removed**

Regards,
-er

 

[openschemes]

Hello!

I got some PM about the MOSIS tapeout, I will post my replies here and at openschemes forum.

1) Will this be one project per chip? How much area can I use?

The tapeout will be multiple cells in one chip to keep the cost down. Hopefully, these runs will occurr on a regular basis, but this first run is only one chip so far with all cells instanced on it. There are no area limits so far, but we must all agree if one project is taking a very large area that it is worth taking the area.

2) Will you distribute PDK or models to me?

No, sorry. PDK and fab models can only be obtained by an NDA with the fab, they are not public and we cannot out any proprietary info. Use the generic models on the MOSIS page and your circuit should be fine. Feel free to discuss if you have any concerns due to unique structures you use.

3) I use a lot of logic - can I have the standard cell library for SPR?

We are currently negotiating the use of some generic libraries, as of now we cannot give anything out.
A good start is the OSU libraries
**broken link removed**

or
**broken link removed**
for the new ones

or
**broken link removed**

or
**broken link removed**

 

[greenhand]

I think this is a great idea . But I hope the new forum don't ask for points
as this forum .

 

[electronrancher]

Probably not points, but the downloads will have to be controlled somehow to prevent unauthorized use. It's up to OS, but I think that agreeing to the license terms (no reselling, etc) is probably good enough. If you've ever used opencores, you know that you must log in and describe your usage before you get the link - maybe something like this. Note: Openschemes is not affiliated with opencores, although we do think their site is cool!

 

[pixel]

In the case of analog electronics is not easy to generalize circuits. In other way you would not have so much OP amps at the market, and there would be tools that can create circuits from ahdl code.

 

[jutek]

hi

i can't open this forum unfortunately. can you check it?

the idea is great, especially for students. good luck!!

regards

 

[electronrancher]

Yes, the server needed a new configuration so it all came down last night I think. It will probably be back up in a day or two. OS is transferring the board over to phpBB, so it will be better overall to get it done now.

 

[bauer]

forum.openschemes.com link down ?

 

[name_windy]

It is really a good ideal.
I hope for it...

 

[openschemes]

Hello all!

The server is set up, and phpBB (although a simple one) is running. You can go back to
forums.openschemes.com
to see it.

Sadly, I did not have time to re-load the users table so if you had an account it is missing now. You can re-register or PM me and I will load it manually.

Hi Pixel,
I agree with your point, but as you can see there are a lot of IP guys now providing analog IP. If you've spent much time retooling your circuits for a new process, you know it is not difficult - but more tedious. Starting from the right methodology makes it actually pretty easy.

As for ahdl - I don't think compiled circuits (unless using all standard cells) are suitable for analog either. A digital standard cell is the absolute minimum circuit needed to generate the AND function for example. These cells are indeed sized and designed for portability. They share the same constraints that analog do but can afford wide tolerances (I need to drive x gates in n nanoseconds at y volts Vdd). So I think the reuse is feasable, but perhaps not to compile ahdl - YET! Let us get some examples in to demostrate a few key concepts.. I think that will help explain things..

 

[openschemes]

Not sure what's wrong - I just took a look from a different location and it seems fine.

Maybe you need to refresh.

www.openschemes.com
forum.openschemes.com

Let me know if there is any problems.

 

[FLEXcertifydll]

I don't think it is a buiness way of the indusry, but a good idea for the beginner.

 

[electronrancher]

Aha! Finally we have decided on the first project, it's the good old Huijsing class-ab op amp. This one is easy so I think it's a good tutorial for people to start with. I'll let you know when it's ready for download!

Edited 7/7/06

The first revision of the op-amp is up in the releases area of the download manager, it requires login. This release is done in Microsim/PSPICE (sch/slb), but as always includes a spice subcircuit netlist and handwritten spice test file for use in other simulators.

Features:
1.8v-6v input range
100uA Quiescent current
Drives up to 10mA load
100dB DC gain
3.5MHz GBW
77dB PSRR

https://www.openschemes.com/modules/wfdownloads/singlefile.php?cid=7&lid=5

To Do:
1) I need to revise this cell to the full rail-to-rail input stage, but that will take a few days with my hectic schedule.
2) I need to build a tutorial on how to set up microsim/pspice for those unfamiliar, this will also come out in a few days.
3) I need to verify the cell over process and temp. So far, I have only done a few cases since it is such a well-known cell and came directly from Huijsing's paper (but using 0.75um lambda). It runs on the 0.5um models, as well as the 1.5um models (not up yet I think) that it is naturally ported to.

I got OS to agree to call it a "bronze" release, meaning it is verified in only a few cases. Once the corner sims are done and the schematic is finalized it will be updated to "silver", complete but not verified in silicon. And hopefully it will work on the MOSIS chip to become a "gold" release!