Altium signal integrity

[kvresto]

Hi All.

I’m trying to simulate with Altiums signal integrity feature, some clock traces routed to an ADC and FIFO clocked from a CLPD.

I have just routed the CLK track of interest and ignored all other components on my schematic, and also other components currently imported to PCB and it seems to work. Can I trust signal integrity to give me valid results used this way.

Can I just route the track of interest and ignore all other components on my schematic, and also other components currently imported to PCB, such as caps, resistors other IC’s, and tracks?

Has anyone had some experience in this area?

Cheers
kvresto

[kvresto]

How about Dave, care to add some expert comments, this thing is not exactly intuitive.

Cheers

[yanir]

I too would like to know if people rely on this feature. I'm doing a design with DDR1 and would like to use the SI tools to make sure the layout is ok. Has anyone done this before?

[EEVblog]

How about Dave, care to add some expert comments, this thing is not exactly intuitive.

My comment is the same as most people in the industry. Altium's signal integrity tools suck, just like their simulation tools and autorouting tool.
They don't seem to have improved these for quite a few years, so most people have just given up on trying to use them for anything useful. But they still like to market them to show that their tool does everything.

Dave.

[yanir]

How about Dave, care to add some expert comments, this thing is not exactly intuitive.

My comment is the same as most people in the industry. Altium's signal integrity tools suck, just like their simulation tools and autorouting tool.
They don't seem to have improved these for quite a few years, so most people have just given up on trying to use them for anything useful. But they still like to market them to show that their tool does everything.

Dave.

That being said, are there any tools available to perform SI on an altium design that won't break the bank/ open source?

[free_electron]

CST , hyperlinx. but none open source and it will break the bank...

and even then , you need to know what you are doing. you can't just shoot blind and hope the tool will tell you what to do.

[yanir]

CST , hyperlinx. but none open source and it will break the bank...

and even then , you need to know what you are doing. you can't just shoot blind and hope the tool will tell you what to do.

Well that's lame.

[poorchava]

I don't think it's lame. There are some areas, where open-source approach simply won't do. Those include:
-pcb cad
-mechanical cad
-signal integrity
-finite elements method modelling
-spice software

This type of software needs enormous amounts of experience, experiments, verification and user feedback. Being a good code developer is not enough.

[yanir]

I don't think it's lame. There are some areas, where open-source approach simply won't do. Those include:
-pcb cad
-mechanical cad
-signal integrity
-finite elements method modelling
-spice software

This type of software needs enormous amounts of experience, experiments, verification and user feedback. Being a good code developer is not enough.

Don't get me wrong, my facetious answer was more directed to not having any usable options at entry level prices.

I'm a hobbyist turned semi professional in this field and learning as I go. I use altium for all of my work. I tried the open source route but it wasn't cutting it for me and paid out of pocket for my license.  As a software developer myself I value the efforts involved to produce software, especially software as complex as we are discussing.

However pcb designs are getting more complex across the board. People want high end GUIs and wireless this and that. That means building small single board computers that use decent processors and not just simple microcontrollers hooked up to an lcd. These boards are way more complicated than what I typically do. I want to be sure that my design will work and not just have to cross my fingers.

When I first started to learn pcb design I was shocked at how expensive and outdated the software was. I evaluated quite a few at the lower end (2500 to 5K) and the options are few. I was also shocked to discover that I had to design my own footprints for practically everything. And If I didn't want to I had to pay through the nose for libraries. The EDA industry is still very much in the stone age.  Which is counter to the high end designs it supports. This however is going to change. Hardware is becoming more attractive to VCs and it is spurring a lot of development. The open source hardware movement is creating a lot of new hardware designers and they need tools. These tools will come from a new group of companies, and the old world of EDA will get caught with it's pants down if they do not do something to address it. Two new companies come to mind, Upverter, a webbased (crazy right?) schematic capture and layout tool, and circuithub, an community driven footprint repository (recently discussed on the amp hour). These two solutions are not ready for prime time but it's a start. The next 5 to 10 years are going to be pretty interesting.

[free_electron]

Don't get me wrong, my facetious answer was more directed to not having any usable options at entry level prices.

So you want cutting edge mathematical calculation software designed to run on compute clusters but it's got to be zero paid ? dream on.
nothing prevents you from solving maxwells equations ona piece of paper. good luck spending 10 years of your life really understanding them first...

I want to be sure that my design will work and not just have to cross my fingers.

so spend the money. it's either psend money on try-wash-rinsce-repeat , or tools.

- you build the box using a handsaw, lots of scrap wood and time
- or you get a fully kitted workshop with compound miter saw , shopbot , planer, bandsaw, router and all the other woodworking tools needed.
- you pay someone else with a workshop to do it for you

there is no other option.

Upverter,

mandatory eyeroll ...

i played with that thing a couple of weeks ago. TOTALLY USELESS.
the gripes

1 schematic
- half of the parts in the library are incomplete or contain mistakes. Made by monkeys comes to mind...
- the other half are 'arduino level' parts. look for anytinhg slightly more advanced and it's silzh
- the library manager is slow , cumbersome and a mess to find anything. there are no subcategories , its all one huge lump. 157 pages and all you can do is click next , next , next next in the hope you will find something. the searchbox is useless as it is dumb as a box of rocks if you type 7400 it does not show you 74hc00 or m7400. it has no concept of what parts are
- no subsheets
- no wiring harnesses
- no buses

in other words 'arduino schematics. Bleehh.

the PCB

- no interactive routing it doesnt even stay on grid. no obstacle avoidance. you may as well use MSpaint to draw the bloody board.
- pushing/pulling of traces.

the other gripes :

- runs in a browser
- needs javascript (security leaks anyone ?)

they sent 2 years of effort on this. yikes !

There is another project like that :Circuits.io
Those guys are a bit further along as you can actually order the boards once the design is done ( you can also just get the gerber. no need to buy anything)

While the idea is sound this is also another 'arduino style' schematics pool.
just look at this one :

http://www.circuits.io/circuits/1738

or this

http://www.circuits.io/circuits/1878

or this

http://www.circuits.io/circuits/673

it makes my eyes water and toes curl just looking at that. Bloody hell. doesn't anyone know how to make schematics anymore ? its either a pile of wiring spaghetti or a lump of netnames with stubs. it makes me want to vomit.

but then again : the tools does not let you place your own wires. it wants to 'autowire' and you end up with a buttugly schematic that is unreadable.
Not where it needs to be.

[yanir]

So you want cutting edge mathematical calculation software designed to run on compute clusters but it's got to be zero paid ? dream on.

Whats cutting edge today is commodity tomorrow. There is no reason this can't apply to EDA, The cost of software (or anything for that matter) comes down to volume. You have plenty of ph'ds and engineers working on 1 dollar parts because they can sell millions of them.

Your comments on upverter are true (if not a bit harsh). Again, my point is there are examples of new school solutions. You have furthered my point by providing more examples. Give it time, they will get better.

I don't want to use toys, i want to use tools. And I am willing to pay for them. But if you know anything about EDA you would agree that the software is archaic and so is the pricing.

[marshallh]

I know some guys who are hell bent on using open sores tools. It is sad to see them waste so much effort and time on stupid issues related to Kicad. They are some sharp folks, and their time would be much better served by tools that actually get things done instead of breaking down. They're trying to do a 6 layer multi-hundred ball BGA design in this software.

I see that you at least have come to grasp that it's worth it to pay for tools that are well capable instead of subscribing to a religion.

The open source hardware movement is creating a lot of new hardware designers and they need tools.

Well, if you count arduinos and endless variations of AVR led blinkers. Sure, there will be a small amount of actually competent HW engineers resulting from it. But they will be successful because they will do what is needed to get the job done.

[free_electron]

ah.yes 1 dollar parts if you can sell a million a day . bingo profit
users for field solvers? good luck finding 100 customers ... so spread the cost.and you end up where you end up.

Do you have any idea what real cutting edge CAD software costs ? Chip design software for example is custom tailored to a specific customer. You can get run-of-the mill software for a few tenthousands of dollars (Tanner Eda for example) .   Want it to do the latest deep submicron technologies ? it shoots to millions of dollars. YEARLY - PER LICENSE ... you don't own a licence. you lease it. Want 'custom builds' ? yiu have no idea.  If you look at a Cadence install including the place and route tools you ar elooking at over a million dollar per licence. and that is just for a base platform. IF you are a name like Intel or IBM you can tack on another digit to that price... and those are DIFFERENT tools. the Cadence sotware delivered to Intel is for Intel ONLY. there are modules in there that you will not find at IBM or TI's installations of the same software ...
That stuff is so cutting edge it is finely tuned to the fab chemistry in use and not transportable.


As for field solvers. they are limited in use. the runtimes are simply too long.
have a single agressor wire attack a susceptor. and yo look at 5 minute runtime on an average pc.
have two agressors attack and you look at half an hour
have three agressor attack and its hours...
Have a real life board with a for or 5 agressors attack a few susceptors and you can pull out the Cray and you will still be looking at days of runtime...

Those tools are good to learn a few thing , get away with some rules of thumb. do the board layout , build it , send it in an anechoic chamber do the eye diagrams. if it fails : focus on the failing parts. it is not feasible to do the whole board upfront through the tool. not yet. even with supercomputing clusters.

People have tried for decades trying to bring the 'pcb layout' part into the fully automated level... autorouters , field solvers etc.. you can't replace the human.
There is a reason it is still called 'artwork' andnot 'a collection of lines drawn by monkeys'