Welcome Guest ( Log In | Register )

33 Pages « < 14 15 16 17 18 > » Bottom

Outline · [ Standard ] · Linear+

 Computer Engineering Thread, # 67 members already :D #

views
     
ikanayam
post Feb 23 2006, 03:44 AM

there are no pacts between fish and men
********
Senior Member
10,544 posts

Joined: Jan 2003
From: GMT +8:00

QUOTE(X10A Freedom @ Feb 21 2006, 05:48 PM)
using Design Architect bar the analogue library(apparently my college can't determine which analogue library we should use, so we have to use the digital nmos/pmos library)
the problem is, without putting a buffer at the input, the software apparently assume that the signal at the output of the transmission gate will feedback some signal to the input which in the end render the transmission gate not attractive anymore if i always need to put a buffer just to prevent the signal from feedbacking
*
Ok so unless your digital simulations can take care of signal degradation and all the funky stuff with pass logic, then it's pretty hard to actually be sure that it will work in a real circuit. Usually digital simulations can only output 1 and 0 so it does not accurately represent what happens in a real circuit.

I'm not sure what you mean by feedback but transmission gates are just like switches. They do not pass values in a single direction. A switch just completes a circuit, it does not care which direction the electrons flow. Transmission gates load the previous gates because they add capacitance and resistance. That's why stacking transmission gates is evil. Always ensure proper buffering when using them.
X10A Freedom
post Feb 23 2006, 07:24 AM

ZGMF-X20A Strike Freedom Gundam
*******
Senior Member
3,875 posts

Joined: Jan 2003
From: SJ


yes, i know transmission gates are bidirectional
but the let's say i supply 1 at point A
and it suppose to flow out at point B when the switch is open
but, in the end the signal at point B flow to point A which is pretty ridiculous as voltage potential flow from high to low, not low to high
sigh, hope my college can quickly solve the software issue on the analogue simulator

TScharge-n-go
post Feb 25 2006, 04:20 PM

Look at all my stars!!
*******
Senior Member
4,060 posts

Joined: Jan 2003
From: Penang / PJ

I've been working on a register design in multiplier more than 2 days, problem still unsolved sad.gif


M0 (from bit 0 to 6)

Either INIT, MUL0 or MUL1 signal can enable the register.

During INIT state, the register is reset to '0'.

During MUL0 or MUL1 state,
it shifts right by 1-bit when M20 = 0.
it accepts AO input when M20 = 1.


Anyone knows the problem? Thanx !

VHDL Code:
» Click to show Spoiler - click again to hide... «


This post has been edited by charge-n-go: Feb 25 2006, 04:21 PM


Attached thumbnail(s)
Attached Image Attached Image
harrychoo
post Feb 26 2006, 11:55 AM

Look at all my stars!!
*******
Senior Member
3,589 posts

Joined: Mar 2005
From: Bolehland


QUOTE(charge-n-go @ Feb 25 2006, 04:20 PM)
I've been working on a register design in multiplier more than 2 days, problem still unsolved sad.gif
M0 (from bit 0 to 6)

Either INIT, MUL0 or MUL1 signal can enable the register.

During INIT state, the register is reset to '0'.

During MUL0 or MUL1 state,
it shifts right by 1-bit when M20 = 0.
it accepts AO input when M20 = 1.
Anyone knows the problem? Thanx !

VHDL Code:
» Click to show Spoiler - click again to hide... «

*
hi there. i'm new here. degree for electronics. now working but my project now mostly in power electronics sweat.gif

i have been reading ur code and the schematic.

can u explain more details what u want to achieve and what problem u encountered? seems like from the code and schematic is nth wrong in ur explanation above.
TScharge-n-go
post Feb 28 2006, 09:50 AM

Look at all my stars!!
*******
Senior Member
4,060 posts

Joined: Jan 2003
From: Penang / PJ

QUOTE(harrychoo @ Feb 26 2006, 11:55 AM)
hi there. i'm new here. degree for electronics. now working but my project now mostly in power electronics  sweat.gif

i have been reading ur code and the schematic.

can u explain more details what u want to achieve and what problem u encountered? seems like from the code and schematic is nth wrong in ur explanation above.
*
The register should b able to take in the output from adder when M20 = '1'. It will shift right when M20 = '0'. However, my simulation shows some 'glitches' which affects the output during the clock edge.

Well, i think I've somehow solve the problem by using behavioral statement with some IF and ELSE statements. It only takes 2.4ns instead of 17.5ns. This MaxPlus2 really weird, it doesnt like structural coding for sequential circuits tongue.gif

Thanx for the help and welcome to this thread biggrin.gif

This post has been edited by charge-n-go: Feb 28 2006, 09:51 AM
X10A Freedom
post Feb 28 2006, 09:08 PM

ZGMF-X20A Strike Freedom Gundam
*******
Senior Member
3,875 posts

Joined: Jan 2003
From: SJ


most of the times, certain operations are embedded into the fpga's lut
so when u use behavioral instead of structural, it gives the software more flexibility when synthesizing it into the real hardware
TScharge-n-go
post Mar 1 2006, 02:51 PM

Look at all my stars!!
*******
Senior Member
4,060 posts

Joined: Jan 2003
From: Penang / PJ

But the device is a Max7000B CPLD, i thought PLD should have standard logic cells? weird eh tongue.gif
ikanayam
post Mar 2 2006, 10:17 AM

there are no pacts between fish and men
********
Senior Member
10,544 posts

Joined: Jan 2003
From: GMT +8:00

Making logic for FPGA is not the same as ASIC. With an ASIC you are pretty much free to do what you want within the design rules, so you always want to use inverting logic and minimize it.
I'm guessing (i have only one semester's experience with FPGA programming) with an FPGA the rules are quite different and more limiting. You probably have a limited number of a certain type of gate? And the layout of the logic cells are somewhat fixed so you can't put everything where and how you want which would be optimal for an ASIC. I guess with FPGAs it's more like hand optimizing your program for a specific architecture, since the hardware is already "fixed" so you have to know exactly how the hardware works to get optimal performance on your software.

Btw i just got my 37bit adder to clock at 770MHz (up from 320MHz in my old design). Tonight i can take a break to celebrate laugh.gif

This post has been edited by ikanayam: Mar 2 2006, 10:20 AM
harrychoo
post Mar 2 2006, 10:59 AM

Look at all my stars!!
*******
Senior Member
3,589 posts

Joined: Mar 2005
From: Bolehland


anyone of u had experience in doin USB interface? Setting up the USB interface between MCU and PC side.
ikanayam
post Mar 2 2006, 11:10 AM

there are no pacts between fish and men
********
Senior Member
10,544 posts

Joined: Jan 2003
From: GMT +8:00

QUOTE(harrychoo @ Mar 1 2006, 09:59 PM)
anyone of u had experience in doin USB interface? Setting up the USB interface between MCU and PC side.
*
What exactly do you want to know about it?
harrychoo
post Mar 2 2006, 12:45 PM

Look at all my stars!!
*******
Senior Member
3,589 posts

Joined: Mar 2005
From: Bolehland


QUOTE(ikanayam @ Mar 2 2006, 11:10 AM)
What exactly do you want to know about it?
*
Some informations about it regarding like how to develop the driver on the PC side. I think MCU side should not be a problem.

Do u have any good reference website or books regarding USB interface. Maybe some applications examples will help.

Thanks
X10A Freedom
post Mar 2 2006, 08:17 PM

ZGMF-X20A Strike Freedom Gundam
*******
Senior Member
3,875 posts

Joined: Jan 2003
From: SJ


QUOTE(charge-n-go @ Mar 1 2006, 02:51 PM)
But the device is a Max7000B CPLD, i thought PLD should have standard logic cells? weird eh tongue.gif
*
it still utilizes lut's and muxes to do the work
muxes are good stuffs in terms of implementing arithmetic and sequential devices
ikanayam
post Mar 5 2006, 12:48 PM

there are no pacts between fish and men
********
Senior Member
10,544 posts

Joined: Jan 2003
From: GMT +8:00

My adder is being changed again to a sparse tree based one. If i can hit 1ghz at 32bits i will sleep. Adders are so addictive (oooh the pun).

Btw chargey, you should look into CLA trees if you have time, you can get the 8 bit one to go at 3ghz i'm sure laugh.gif. Msg me and i'll send you a paper i have on different adder types including the CLA tree.

This post has been edited by ikanayam: Mar 5 2006, 12:53 PM
TScharge-n-go
post Mar 5 2006, 06:43 PM

Look at all my stars!!
*******
Senior Member
4,060 posts

Joined: Jan 2003
From: Penang / PJ

heh, that's cool, will msg u about it later biggrin.gif biggrin.gif

Well, i got an 8-bit Adder/Subtractor which theoretically can work at 3.5GHz, using CSA, CLA and RCA combination. However, it is quite large la, around 500+ transistors.


QUOTE
it still utilizes lut's and muxes to do the work
muxes are good stuffs in terms of implementing arithmetic and sequential devices

icic. I thought FPGA is the only one using LUT. tongue.gif

This post has been edited by charge-n-go: Mar 5 2006, 06:43 PM
X10A Freedom
post Mar 9 2006, 11:02 PM

ZGMF-X20A Strike Freedom Gundam
*******
Senior Member
3,875 posts

Joined: Jan 2003
From: SJ


just wondering, is it possible to have a via that connects from metal 2 or 3 straight to the polysilicon?
coz currently it needs to be done in a ladder form
ikanayam
post Mar 10 2006, 02:26 AM

there are no pacts between fish and men
********
Senior Member
10,544 posts

Joined: Jan 2003
From: GMT +8:00

QUOTE(X10A Freedom @ Mar 9 2006, 10:02 AM)
just wondering, is it possible to have a via that connects from metal 2 or 3 straight to the polysilicon?
coz currently it needs to be done in a ladder form
*
You need these vias: M3->M2, then M2->M1, then M1->poly. That's on TSMC 0.18micron rules. Maybe if you are using different design rules you might be able to make it but i don't see the difference since fabrication is done in layers.

What layout are you working on?

This post has been edited by ikanayam: Mar 10 2006, 02:34 AM
X10A Freedom
post Mar 10 2006, 06:55 AM

ZGMF-X20A Strike Freedom Gundam
*******
Senior Member
3,875 posts

Joined: Jan 2003
From: SJ


QUOTE(ikanayam @ Mar 10 2006, 02:26 AM)
You need these vias: M3->M2, then M2->M1, then M1->poly. That's on TSMC 0.18micron rules. Maybe if you are using different design rules you might be able to make it but i don't see the difference since fabrication is done in layers.

What layout are you working on?
*
ya, that's wat i'm doing now, but i was hoping to be able to make a via straight down to the poly from any metal layer tongue.gif
i'm doing the layout for the memory section of the viterbi decoder
trying to use only metal 1 but it's quite difficult to maintain it at metal 1
ikanayam
post Mar 10 2006, 08:28 AM

there are no pacts between fish and men
********
Senior Member
10,544 posts

Joined: Jan 2003
From: GMT +8:00

QUOTE(X10A Freedom @ Mar 9 2006, 05:55 PM)
ya, that's wat i'm doing now, but i was hoping to be able to make a via straight down to the poly from any metal layer tongue.gif
i'm doing the layout for the memory section of the viterbi decoder
trying to use only metal 1 but it's quite difficult to maintain it at metal 1
*
Hopefully you already know all this but i found this advice really useful: It may not be optimal to use all metal 1, or do so many local optimizations that global optimizations suffer. Also poly is evil, try to use it minimally and go to M2 if you need more wiggle space. I'm also using directional layers for the layout, so poly goes vertically, M1 goes horizontally, M2 vertically, M3 horizontally, M4 vertically. It's supposed to help with manufacturing also.

1st time doing layout? What memory is it? SRAM?

This post has been edited by ikanayam: Mar 10 2006, 08:35 AM
Annie
post Mar 10 2006, 08:35 AM

Getting Started
**
Junior Member
234 posts

Joined: Jan 2003
From: Petaling Jaya


QUOTE(X10A Freedom @ Mar 9 2006, 11:02 PM)
just wondering, is it possible to have a via that connects from metal 2 or 3 straight to the polysilicon?
coz currently it needs to be done in a ladder form
*
nope....u have to go thru poly contact to m1, then via1 to m2 and so on.
X10A Freedom
post Mar 10 2006, 07:24 PM

ZGMF-X20A Strike Freedom Gundam
*******
Senior Member
3,875 posts

Joined: Jan 2003
From: SJ


QUOTE(ikanayam @ Mar 10 2006, 08:28 AM)
Hopefully you already know all this but i found this advice really useful: It may not be optimal to use all metal 1, or do so many local optimizations that global optimizations suffer. Also poly is evil, try to use it minimally and go to M2 if you need more wiggle space. I'm also using directional layers for the layout, so poly goes vertically, M1 goes horizontally, M2 vertically, M3 horizontally, M4 vertically. It's supposed to help with manufacturing also.

1st time doing layout? What memory is it? SRAM?
*
any useful links or books on mask layout?
i seriously need one though my fyp due date is soon tongue.gif
poly is evil? y eh? my lecturer told me otherwise sweat.gif
actually, i did try doing my layout using multiple layers of metals and use lesser poly
but the result isn't good as in the end all my metals is on top of my gates which i think it isn't good especially thermal dissipation

yah, first time doing layout
not doing SRAM, only latches(though i wanted to use pulsed latch, but the analogue library and the simulator came a bit too late as my project due date is soon)
SRAM needs a precharge circuit which i for one can't find any info about it sad.gif

thanks for the advice

This post has been edited by X10A Freedom: Mar 10 2006, 10:22 PM

33 Pages « < 14 15 16 17 18 > » Top
 

Change to:
| Lo-Fi Version
0.0307sec    1.78    6 queries    GZIP Disabled
Time is now: 23rd December 2025 - 03:25 AM