Just tack on a zobel at the line's characteristic impedance at the speaker and be done with it. You can get away with a 150 or 200 ohmer without the cap if you wish..just don't use a wirewound. Whenever a load impedance rises well above line, the capacitive energy storage of the line dominates.
Sheesh, that was easy...🙂
John
Absolutely not a denigration of your designs - they are excellent.
You don't think I have not simulated them ?
My point is -
10 devices = <50ppm 16 = 20 and 20+ = maybe 1-2.
Double that with the layout , and double again with whatever semi's you
must use.
It's very good to start at 1ppm.
OS
marce,
Thank you for that message, I was wondering if any of my email was getting through to you. Takes one of those questions out of my mind.
Thank you for that message, I was wondering if any of my email was getting through to you. Takes one of those questions out of my mind.
It is often hard to convey in words what we are trying to say on a forum such as this. What I call art someone else calls competence and another calls experience. Experiences drive our thoughts and the art is in how we use that knowledge, whether it is Brad or Scott or Ed or JN or Dadood or ostripper, we all have a base of knowledge from our experiences that we draw on to do our work and as long as we really care about what we are doing draw on that experience to create the best whatever we do to the desired, required or expected level. Sometimes the powers that be just don't let you do that for financial reasons and in my eyes they are the losers for that, but when we have passion for what we do we use all the art in our brains to pull our experiences and thoughts together to create the best solutions.
When machines can think like humans, can create artfully and not just by a set of predetermined rules we will truly be in trouble, when AI thinks artfully then what will the need for any of us be besides building our replacements!
What we all have to do is continue to learn, to push or limits, to explore alternate design, not to think what we have known from the past is all there is, there is always room to improve or artfully create something elegant, something we just never considered before. When we stop at some point and think what we learned in the past is all there is,all that is important and anything new is inferior to what we already know we have lost the edge, we are just riding on past accomplishments.
When machines can think like humans, can create artfully and not just by a set of predetermined rules we will truly be in trouble, when AI thinks artfully then what will the need for any of us be besides building our replacements!
What we all have to do is continue to learn, to push or limits, to explore alternate design, not to think what we have known from the past is all there is, there is always room to improve or artfully create something elegant, something we just never considered before. When we stop at some point and think what we learned in the past is all there is,all that is important and anything new is inferior to what we already know we have lost the edge, we are just riding on past accomplishments.
That really surprises me. Good placement follows from a few sensible rules. I would think we would be able to code that sensibly - not rocket science at all.
Why would anybody pay kilobucks for a defect autoplacer? Or maybe all those who buy it are incompetent placers? That's not likely - incompetence and large budgets normally don't go together.
Hmmm.
I don't have an autoplacer because I can't afford one, but would love to try.
Jan
Auto-route
Is not always useless.
If you get placements right , it will just take the most logical shortest path.
The use of "nets" mean that you won't make a mistake on the traces.
"art" (below 1) is getting the grounding right.
Auto will not make an extra via to get the shortest route to a central
ground star , it will just get there any way it can. 🙁
Diptrace could not even figure out that U3 to U5 trace (green arrows - below 2).
I guess it needs a human "artist" to help.
PS - I'm learning too .... I did this "art" manually prior to this.
OS
Is not always useless.
If you get placements right , it will just take the most logical shortest path.
The use of "nets" mean that you won't make a mistake on the traces.
"art" (below 1) is getting the grounding right.
Auto will not make an extra via to get the shortest route to a central
ground star , it will just get there any way it can. 🙁
Diptrace could not even figure out that U3 to U5 trace (green arrows - below 2).
I guess it needs a human "artist" to help.
PS - I'm learning too .... I did this "art" manually prior to this.
OS
Attachments
If you can lay out analogue properly and digital you can do audio... RF is not for the faint hearted... Basic RF is within the realms of PCB design, when you get onto the esoteric stuff with components sculptured out of copper shapes you are in a different league an area not for us mere mortals, I have worked with some of those guys and I am convinced they are from another planet...🙂
There are more than a few sensible rules.... each board is different, the topography differs, do you go for power supply lines over signal lines, which signal lines do you give preference to...
Most CAD packages include autorouters and placers as standard its a selling point that those in the front line of the business of laying out PCBs hate... and as said I know no one in the game that uses either the autoplacers or autorouters.
Humans are good at pattern recognition and visual spatial abilities, both of these are crucial to looking at a design and getting the correct placement...
The majority of boards we do are not like the DIY designs we see on here, quite often we are talking a few thousand connections and depending on the basic design hundreds of components....
Last edited:
I am going to plod through the spam tonight, I normally do Sun-Thursday away but been doing Fridays so only had Saturday at home for the last few weeks, so no time for anything but work and driving.... Fed up of it now as I don't get a penny fro travelling time!
I'm very disappointed if Stuart actually said something that misleading and simplistic about distortion etc. This is almost as silly as the guy who told a friend that using a plurality of the same tube in a design would lead to what the friend translated into "sonic signature buildup", which I said sounded a bit like an ad for laundry detergent. What's a mother to do??
Of course as few as possible to meet design goals is unassailable. It should also be understood that most transistors are cheap. No one pushed back when I used a bunch for a portable product some years ago, as (1) it was overall a good deal cheaper than using sufficient-quality low-power op amps, and (2) it worked and sounded pretty decent. This was at a time that doing things in the symbol domain was both prohibitively expensive and far too consumptive of power.
I also had the satisfaction of finding out, from the manufacturer of the filterless class D amp chips in the unit, that a Far East outfit that attempted to reverse-engineer the product gave up after a while 😀
Sadly this keeps turning up, a lot of people who should know better feel compelled to say these things to certain audiences that expect to hear them.
Yes but the rules are reasonably universal. It's not that when I lay out a board that I all of a sudden develop new rules.
The preferences you mention could be easily coded in a selection list.
I'm not saying it would be easy but I believe the biggest hurdle is that we cannot or never bothered to reflect on how we actually do it.
If we can write programs that make a computer win the world chess championship, an autorouter/placement should be possible.
Thing is, millions have thought and written and studied chess, but on PCB routing it's all seat of the pants stuff.
Jan
The routing problem is a classic programming example. The first version was just for fleet management. If you have ten ships and x ports how do you route them optimally.
The PC routing is far more complicated. Then there is the issue of who would want to pay for such a program of limited sales.
JN,
Not quite that simple, if I have 70 pF per foot (typical) 4 twisted pairs, and 1500 feet of cable. Just getting the guys to twisted each pair with a different twist seems to be rocket science to them. Asking for concentric shields went beyond that.
Color codes they can do. Jacket color was an issue! But the characteristic impedance of the twisted pairs is in the very right range.
Meets CL3, that they understand.
Ed
I haven’t tested it yet but you may try to test JN’s advice
George
30 or 40 years ago there was an economic incentive to use fewer transistors - they were expensive.
Nowadays they are cheap and in volume, amazingly cheap.
Nowadays they are cheap and in volume, amazingly cheap.
You say seat of the pants. Most would say 'experience'. The whole 10,000 hours bit. Now neither IBM nor google seem to want to admit how much it cost to beat a chess master or a go master, but we are talking a similar sort of parallel processing problem with layout. Often things that seem easy to us are really hard for computers (or perhaps their programmers 😛 ).
When I started in at UCLA about 48 years ago, I had supervision from a electronisher grad student, who actually prescribed the transistor count on one project. The parts were in TO-5 cases, usually, or sometimes TO-18, and yes they were expensive.
Although I respected the guy and got along with him fairly well, I was not unhappy when he managed to get the PhD and moved on, leaving me essentially autonomous---except for having to explain why I was spending money on parts.
You're right. If you can only do something after lots of experience, that means there are no well-developed rules.
And yes it's a budget problem in the final analysis.
Edit: I'm not a chess player, but are there any good rules for a game? I don't mean the rules of the moves and such, but like known-good strategies?
Jan
The coding has been done its implementing it on each design, it requires original though and computers are not good at that...
I disagree on the chess PCB analogy, chess is always the same, same board, same pieces, same rules PCBs aren't, connectivity is different, different rules, different trace widths, different topography and that's just the basics, then you get spacing classes for higher voltages, current capacity of traces, thermal issues etc. etc. Rant time🙂 Even though most people think anyone can do a PCB design to do todays modern complex designs properly takes a lot of skill and learning, not only do we have to understand the circuit, but also PCB fabrication, circuit assembly etc. etc. Have a look at the IPC C-1000 series of basic specification regarding PCBs or the CID and CID+ courses and the range of topics they cover, there is more to PCB design than most think as I have said...
With todays digital you also get a constraint browser as well so you can control skew groups delays, diff pair matching, trace impedance as you change layers and thus the distance to return paths have a look at DDR2 routing for 2 devices (single DDR chips not dims) on a T bus, great fun (DDR3 is easier to lay out with fly by routing). Micron have some good guides...
Last edited:
But there are lots of rules for PCBs that are always the same.
For instance: 'if you hit an obstacle, go around, or use a via to go beneath it. Unless the strategy is not to change a layer on this track, then move the obstacle and start again'.
I think you would be surprised how often you use that in your work, if you would tally up everytime you would consciously or not, apply it.
Last edited:
- Status
- Not open for further replies.