I attended a seminar on the future of PCB production last year, some interesting stuff, for additive mainly inkjet based. My favourite PCB technology is currently HDI:
Microvia HDI PCBs | Wrth Elektronik: Printed Circuit Boards > Products > Microvia HDI
Wirelaid - partial high current solutions | Wrth Elektronik: Printed Circuit Boards > Products > High Current Wirelaid
for high current....
Printing PCBs...
http://cdn.intechopen.com/pdfs-wm/12292.pdf
The core voltages are catered for by the on board decoupling, the planar capacitance and the on silicon capacitance, we use this for really really critical designs where the pockets are deep.... Like all simulation software it takes time and careful setup to get good results, but for most designs good decoupling practice and data sheet reccomendations will suffice.
Microvia HDI PCBs | Wrth Elektronik: Printed Circuit Boards > Products > Microvia HDI
Wirelaid - partial high current solutions | Wrth Elektronik: Printed Circuit Boards > Products > High Current Wirelaid
for high current....
Printing PCBs...
http://cdn.intechopen.com/pdfs-wm/12292.pdf
The core voltages are catered for by the on board decoupling, the planar capacitance and the on silicon capacitance, we use this for really really critical designs where the pockets are deep.... Like all simulation software it takes time and careful setup to get good results, but for most designs good decoupling practice and data sheet reccomendations will suffice.
I still use wire wrap (my green Cooper w/w gun) for some proto (one of) stuff. It is fast and efficient esp. for the low speed stuff. It is prone to pickup noise however.
HDI (micro-vias) is now nearing 10-15+ year old pcb technology. Pcb fab technology is slow to evolve compared to other facets of electronics.
The last pro pcb design I did is now +10 years ago, using a 1152 Xilinx Virtex4 and the first time to use 0.8mm pitch BGA on a Intel 10Gb/s fibre channel serdes. Had to go with 8 mil FHS for the breakout pattern on that one. The limits of mechanical drilling, which I think still stands. I recall we used 2mil ZBC core material in a 16 layer controlled Z stackup, for 62mil finished.
HDI (micro-vias) is now nearing 10-15+ year old pcb technology. Pcb fab technology is slow to evolve compared to other facets of electronics.
The last pro pcb design I did is now +10 years ago, using a 1152 Xilinx Virtex4 and the first time to use 0.8mm pitch BGA on a Intel 10Gb/s fibre channel serdes. Had to go with 8 mil FHS for the breakout pattern on that one. The limits of mechanical drilling, which I think still stands. I recall we used 2mil ZBC core material in a 16 layer controlled Z stackup, for 62mil finished.
Last edited:
0.2mm (FHS) is still pretty much the minimum hole size that they can still do without getting into silly silly costs..... We have Just had a prototype board done with mechanical drilled vias .... 17,000 of them on one board and the proper board will be 6 times the size, went with mechanical drilled vias even though HDI was recommended, the proper board will be HDI though... Now with over 27,000 connections how long to wire wrap one wonders......
Sold all my wire wrap stuff to a student a couple of years ago, made a few quid and the student got a bargain
Sold all my wire wrap stuff to a student a couple of years ago, made a few quid and the student got a bargain
Most probably you are referring to this:
"Phase and amplitude sinusoidal dither adaptive control system"
Smyth, R. ; Nahi, N.E.
Automatic Control, IEEE Transactions on
Volume: 8 , Issue: 4
DOI: 10.1109/TAC.1963.1105589
Publication Year: 1963 , Page(s): 311 - 321
But the process of dithering has been suggested prior to that date.
See:B. Widrow, “A Study of Rough Amplitude Quantization by Means of Nyquist Sampling Theory”, Ph.D. thesis, M.I.T., Department of Electrical Engineering, June 1956. (*)
http://dspace.mit.edu/bitstream/handle/1721.1/12139/31905269.pdf?sequence=1
In that paper, he doesn’t call it dithering but ‘addition of random independent noise’
B. Widrow in his 2008 book “Quantization Noise”
Quantization Noise: Roundoff Error in Digital Computation, Signal Processing, Control, and Communications: Bernard Widrow, István Kollár: 9780521886710: Amazon.com: Books
says the following about the early adoption of dither:
Regarding Lipshitz – Vanderkooy
The first paper from them on dithering in audio, is from 1986:
S. P. Lipshitz and J. Vanderkooy, "Digital Dither," presented at the 81st Convention of the Audio Engineering Society, J. Audio Eng. Soc. (Abstracts), vol. 34, p. 1030 (1986 Dec.), preprint 2412.
J. Vanderkooy and S. P. Lipshitz, "Dither in Digital Audio," J. Audio Eng. Soc., vol. 35, pp. 966- 975 (1987 Dec.).
George
(*) The bare essentials of that great work can be read here:
http://www-isl.stanford.edu/~widrow/papers/j1956astudy.pdf
Additional historical data
From the book: Analog Dithering Techniques for Wireless Transmitters
http://www.springer.com/gp/book/9781461442165
Last edited:
That's a nice set of references George. It seems that Bernie Widrow pops up a lot in those days. I remember when he addressed a group---although I heard about it, I wasn't there---and admonished everyone for being slow on the uptake on a lot of matters he thought were clearly important. I'm sure it seemed obvious to him!
Brad
Brad
Actually it was one of Widrows students, defining the point mathematically where the dither becomes indistinguishable from uncorrelated additive noise. My choice of words and emphasis.
Yes Brad, when presented properly obvious to continue to attach fuzzy concepts like "coloration" to it would probably irritate Mr. Widrow no end.
L Schuchman - 1964
Abstract-In this paper, the conditions a dither signal must meet
So that the quantizer noise can be considered independent of the
signal are derived for a quantizer having a finite number of levels.
An infinite class of dither signals which satisfy these conditions is
given. It is seen that the most useful member of this class is one
whose probability density function is uniformly distributed over a
quantizing interval. It is also shown that a necessary and sufficient
condition for the noise n(t) to be independent of the signal x(t) is
that a measurable quantity, called the “D” factor, be zero.
Last edited:
0.2mm (FHS) is still pretty much the minimum hole size that they can still do without getting into silly silly costs..... We have Just had a prototype board done with mechanical drilled vias .... 17,000 of them on one board and the proper board will be 6 times the size, went with mechanical drilled vias even though HDI was recommended, the proper board will be HDI though... Now with over 27,000 connections how long to wire wrap one wonders......
Sold all my wire wrap stuff to a student a couple of years ago, made a few quid and the student got a bargain
Wire-wrap horror story --- An entire mini-computer interface to run an experiment, gather data and all the I/O was designed (not tested) and then wire-wrapped by some great engineering wizard who shall go unnamed....... 300 digital IC's on two planes each the entire inside of a 19" chassis and about 12" deep...... of course it didnt work right away.... I got it to put in the system and get it's 15 interfaces all running for the physicists. Ended up doing a lot of trouble-shooting and fixing design and mis-wired parts/pins etc..... set the project back a whole year. Design logic errors and wiring errors etc. Not easy to fix a wire-wrapped board when the wire to be removed and changed is not the wire on the top of the post adds to the trouble-shooting problems. I'd keep it to small size projects.
THx-RNMarsh
Last edited:
I just visited http://www.protechmn.com/index.php?id=9
to update myself on current high end pcb fab capabilities, 3/3 mil (.08mm) std line width/spacing, 6mil (0.15mm) FHS. They offer 5 oz Cu but of course the rules must be relaxed.
These capabilities cover about 99% of the designs out their. Anyone need more for audio?
Do you think you have enough materials to chose from?
Pro-Tech Interconnect Solutions | Materials And Finishes
This a new one for me, ENEPAG (Electroless Nickel / Electroless Palladium / Autocatalytic Gold)
Digital design with no simulation and wire-wrap fab = glad to be gone to the museum.
to update myself on current high end pcb fab capabilities, 3/3 mil (.08mm) std line width/spacing, 6mil (0.15mm) FHS. They offer 5 oz Cu but of course the rules must be relaxed.
These capabilities cover about 99% of the designs out their. Anyone need more for audio?
Do you think you have enough materials to chose from?
Pro-Tech Interconnect Solutions | Materials And Finishes
This a new one for me, ENEPAG (Electroless Nickel / Electroless Palladium / Autocatalytic Gold)
Digital design with no simulation and wire-wrap fab = glad to be gone to the museum.
Last edited:
Keentoken, Paul Klipsch wrote a number of good AES papers in the 1960's, mostly concerning FM distortion. Horns have almost no FM distortion, even though they have more 'horn' distortion due to their constrictive throats.
FM distortion can be computed to exist in most direct radiator based loudspeaker systems. As Klipsch preferred horns, whenever possible, he became attached to the reduction of this distortion in his loudspeaker designs, large and small. Today, we find that for most applications that direct radiators do well enough.
FM distortion can be computed to exist in most direct radiator based loudspeaker systems. As Klipsch preferred horns, whenever possible, he became attached to the reduction of this distortion in his loudspeaker designs, large and small. Today, we find that for most applications that direct radiators do well enough.
As a student intern I did the entire data I/O for an x-ray survey of the crab nebula sent up in a balloon to 120k feet in wire-wrap (no simulation available). Everything worked fine on the second assent, on the first no one checked to see if the battery technology worked at -55C (it didn't).
I did forget to write an EOF on the tapes, much to the chagrin of the computer techs since the drives were capable of spinning up to dangerous speeds.
If you look at this page it seems that it is easy to end up needing to keep 64 bits of precision:
IIR Filter Design, Software and Examples
Look at the 10 pole chebychev about 1/3rd or so down... in theory the bit depth does keep increasing every multiply, but careful analysis or trial and error will show where you can throw away the extra bits as not mattering to the end result.
I think they refer to 64bit floating point numbers. SoX uses 64 bit floats for all intermediate computations, I am interested in seeing how miniDSP does on some of these numerical tests.
If I remember correctly Paul Klipsch started out working in ballistic science. I have read many of his papers and actually used his basic ideas to develop horn designs. What I always found interesting was that though Paul knew very well the theory and application of these theories his speakers were far from those theoretical results. It was just not convenient or practical at the time to actually make a horn that followed the math that he espoused in his papers. They were really crude examples of his theories and he definitely overlooked those problems. I and my partner had a run in with him long ago at an AES convention, he was not to happy that we had made some designs based on his work that removed the inherent problems of building horns from flat sheets of material. We were the young guys who pissed off the big manufacturers as we had taken a leap in manufacturing that the big players would have found cost prohibitive with their current manufacturing methods. I think the closest bass horn design to compare with what we had done was a very early RCA Ubangi horn enclosure. We carried that design through to mid-range horns using smaller devices than others were trying to make work at the time. It was a real learning experience for sure.
Last edited:
Very interesting info on Klipsch, thanks.
I saw some of their floorstanders close to my price range and was wondering. They advertised "Linear Travel Suspension" which sounds nice. The cones were apparently made of copper! I don't know what benefit that has. Maybe the cone is stiffer, but you also have a lot of mass that must be difficult to move. At the same time I imagine pure inertia is more linear than most other forces that come into dominance at that point.
I saw some of their floorstanders close to my price range and was wondering. They advertised "Linear Travel Suspension" which sounds nice. The cones were apparently made of copper! I don't know what benefit that has. Maybe the cone is stiffer, but you also have a lot of mass that must be difficult to move. At the same time I imagine pure inertia is more linear than most other forces that come into dominance at that point.
wirewrapped computers
When I was working for UCLA circa mid-70's, a programmer I knew said Varian was donating a minicomputer to his Venice co-op something-or-other. I went with him to pick it up. Kirk B. at the time sported a huge beard, very long hair, and wore rather unconventional clothing, including sandals. We cut quite the figures as we went down the halls at the rather-stuffy Varian---people weren't quite fainting or falling out of door jambs, but it was a near thing.
I have forgotten the model, but the machine was full of socketed TTL SSI and MSI chips, and all wirewrapped. It emerged, as Kirk continued to try to get programs to run for very long, that the thing had a serious design defect and periodically crashed. Varian knew about this and had fixed it, and it only required a change in the wirewrapping---but they wouldn't tell Kirk how to do it. They had gotten a tax writeoff giving away a known-defective machine.
After a while I got the machine to use for parts. I extracted many of them and the sockets, and used them for state machines in instruments thereafter for years. I don't believe any of the ICs ever failed.
Brad
When I was working for UCLA circa mid-70's, a programmer I knew said Varian was donating a minicomputer to his Venice co-op something-or-other. I went with him to pick it up. Kirk B. at the time sported a huge beard, very long hair, and wore rather unconventional clothing, including sandals. We cut quite the figures as we went down the halls at the rather-stuffy Varian---people weren't quite fainting or falling out of door jambs, but it was a near thing.
I have forgotten the model, but the machine was full of socketed TTL SSI and MSI chips, and all wirewrapped. It emerged, as Kirk continued to try to get programs to run for very long, that the thing had a serious design defect and periodically crashed. Varian knew about this and had fixed it, and it only required a change in the wirewrapping---but they wouldn't tell Kirk how to do it. They had gotten a tax writeoff giving away a known-defective machine.
After a while I got the machine to use for parts. I extracted many of them and the sockets, and used them for state machines in instruments thereafter for years. I don't believe any of the ICs ever failed.
Brad
On Klipsch
After Harman threw me out over some tax problems I was having at the time, I was contacted to do some consulting work for Klipsch. There were a couple of i-centric powered speakers that a new management guy there believed could be "value-engineered" for cost savings without impairment of performance.
This attitude was generally viewed with immense suspicion by some of the Klipsch stalwarts. The interesting thing was, the latter could not explain the Paul Klipsch "secret sauce" that characterized the product line, even if they were equally adamant that it existed. It was as if you had to be inducted into the mysteries---I envisioned some sort of Masonic rituals.
So I set to see (a) how the existing amplifier in one product worked, and (b) where they might be opportunities for achieving the same behavior at lower parts cost. I replaced a class D amp with a linear one, and a switchmode a.c. adapter with a simple transformer. There were several additional changes. At the end I shipped a prototype to them, and they agreed that it was indistinguishable from the production unit (until they probed it internally and blew it up). They were shipping about 10k units a month at that point. I showed how the changes advocated would save about 8 dollars a unit, a pretty decent engineering salary---per month. They weren't interested. This told me something.
Brad
After Harman threw me out over some tax problems I was having at the time, I was contacted to do some consulting work for Klipsch. There were a couple of i-centric powered speakers that a new management guy there believed could be "value-engineered" for cost savings without impairment of performance.
This attitude was generally viewed with immense suspicion by some of the Klipsch stalwarts. The interesting thing was, the latter could not explain the Paul Klipsch "secret sauce" that characterized the product line, even if they were equally adamant that it existed. It was as if you had to be inducted into the mysteries---I envisioned some sort of Masonic rituals.
So I set to see (a) how the existing amplifier in one product worked, and (b) where they might be opportunities for achieving the same behavior at lower parts cost. I replaced a class D amp with a linear one, and a switchmode a.c. adapter with a simple transformer. There were several additional changes. At the end I shipped a prototype to them, and they agreed that it was indistinguishable from the production unit (until they probed it internally and blew it up). They were shipping about 10k units a month at that point. I showed how the changes advocated would save about 8 dollars a unit, a pretty decent engineering salary---per month. They weren't interested. This told me something.
Brad
I just visited http://www.protechmn.com/index.php?id=9
to update myself on current high end pcb fab capabilities, 3/3 mil (.08mm) std line width/spacing, 6mil (0.15mm) FHS. They offer 5 oz Cu but of course the rules must be relaxed.
These capabilities cover about 99% of the designs out their. Anyone need more for audio?
Do you think you have enough materials to chose from?
Pro-Tech Interconnect Solutions | Materials And Finishes
This a new one for me, ENEPAG (Electroless Nickel / Electroless Palladium / Autocatalytic Gold)
Digital design with no simulation and wire-wrap fab = glad to be gone to the museum.
ENEPIG/ENEPAG is used where wire bonding is carried out... The reason its being promoted instead of good old ENIG (my finish of choice for over 15 years now) is a reaction to the price of gold recently. As far as I know the uptake in the UK has been minor.
4/4 (0.1/0.1mm) with a 0.45/0.2mm via is pretty standard these days for BGA based boards, I like to go with 0.2/0.2mm 0.6/0.3mm for analogue sections as an absolute minimum. With HDI you can go down to 0.3/0.1mm via, but it will only span 1 layer (0.1mm) so you have to dog-bone or stack the vias.
Going below 0.2mm for drilling does incur costs due to the small size of the drill, breakages and drilling time, so again anything below 0.2mm and its time to think HDI and via in pad, when you use 0.5mm pitch BGAs this is the only way to route the board.
Flex circuit design is fun, page 12 of the attached below shows sculptured circuits (start with 7oz Cu) great for power a sod to solder.
For materials the various FR4s cover the requirements for most designs.
- Status
- Not open for further replies.