That does look like an older drawing from the days of vellum, ink and sleeve protectors. But doesn't have a date on it.
I remember when we designed our ultrasound machine the Asian competitors we knew of were still in thru-hole.
-- Regards, Joerg http://www.analogconsultants.com/
Pin numbers are not a standard requirement for SOT
I deal with LEDs a lot and for the 5050 RGB LEDs there are every combination of RGB,GBR,GBR etc and Anode or Cathodes on Pin 1 side as well as CCW topside pin numbers and even/odd pin numbers.
I think the Chinese want to make their copy design unique, just as Amphenol and Berg were different in edge connectors. They may alway see it as an advantage to make their products less easy to drop in cheaper replacements,
or there may be technical reasons why the layout is different at 20GHz.
Does it matter? These parts are not easy to match on Smith Charts, so consider them unique.
The 'original' was likely computer drawn, with Japanese characters added 'manually' (they never seem automated, even when they are), but the source of this image was digitized from a cheap commercial paper print, with the usual results.
Unless the ultrasound companies were vertically integrated, they wouldn't have had the resources to go surface mount. Early development and use was subsidized by other sectors buying into big ticket hardware, but the actual fab and packaging of small parts has always depended upon SE Asian branch plants, even before the transition.
Something in design and development is not actually in competition with hardware in the marketplace ;)
RL
When - at Cambridge Instruments in 1988 - I pointed out that our mad Techni cal Director's absurd timing resolution demand - 10psec granularity - could only be met with Gigabit Logic's GaAs parts, I turned out to have nominate d myself as the engineer who introduced surface mount parts to Cambridge In struments, and had to run around qualifying sub-contractors and so forth.
GaAs logic was only available in surface mount packages ... 100k ECL worked quite a lot better in surface mount packages too. I think that Fairchild r e-labled their 100K logic as 300K when they got around to putting it in sur face-mount packages
-- Bill Sloman, Sydney
It was just the opposite and that's what puzzled me. Our competitors were major corporations, huge. Think the size of GE. We were a small company with less than 1000 employees and at the European location where we did the SMT we had less than 100 employees. This included the whole sales force in the field.
The major suppliers of SMT in those days were Philips, Siemens and some US companies.
Our machines were out there in the marketplace. In hospitals, clinics and doctor's offices. I still see those machines pop up on the used equipment markets and they are now around 25 years old.
-- Regards, Joerg http://www.analogconsultants.com/
10psec ain't absurd. I am designing something to that tune right now. There is no ECL in there, just RF transistors and stuff. Ok, the Schottky diode costs over $10 and that irks me but what can ya do.
-- Regards, Joerg http://www.analogconsultants.com/
The Asians (actually just Japan) were the leaders in (consumer) surface mount. The US sort-of led the way before that in military flat-packs and such like, but they were never reasonably priced.
I don't think the small volume guys came along until later, but the big guys were behind it 100%, especially with the skyrocketing JPY after about 1977. China had not yet opened up at that time (Cultural revolution) and Taiwan could only do so much.
Probably before your time.
Some US companies were leaders in adoption too- Cincinnati Microwave and some of the big guys too.
I've got a Toshiba advert (in Japanese only) from the 1982 transistor manual that lists a whole pack of their their equivalent numbers for SMT and through-hole. Their sorta SOT-23 was called the EIAJ (Electronics Industry Association Japan) SC-59 at the time. At that time, SMT penetration was maybe 5-10%, as opposed to more like 50% in
1990 by which time everyone had gotten on the bandwagon.The SMTA wasn't even formed until 1984. To the extent the Europeans got involved relatively early, it was probably because they still had some consumer electronics indudustry left (protectionism) and were chasing the leaders at the time (the Japanese).
Best regards, Spehro Pefhany
-- "it's the network..." "The Journey is the reward" speff@interlog.com Info for manufacturers: http://www.trexon.com Embedded software/hardware/analog Info for designers: http://www.speff.com
It was only absurd in the application. Our sampling electron pulse was 500p sec wide. We had hopes of getting it down to 100psec, and might have made t hat if the project had lasted longer than the three years it took us to get it working (with 10psec granularity - it should have been 5psec, but since our local oscillator had about 60psec of jitter it was definitely a waste of development effort). Placing that sampling pulse to 10psec was pure spec manship, and it proved to be fatally expensive specmanship.
This was back in 1988. We threw in quite a few BFR92's, BFT93's and an occa sional BFR96. They were all 5 GHz bandwidth parts. You can do better today.
My most recent ECLinPs circuit - in 1996 - used BFR92s and BFT93s for buffe ring and occasional bits of level-shifting.
-- Bill Sloman, Sydney
Then BFR92 should have gotten you almost there.
60psec jitter? Why didn't you use a quality crystal oscillator? Even the cheap ones "in a can" for under a buck are less single-digit psec.If you have to do accurate threshold detection or zero crossing on the result it is not pure specsmanship. Why was it expensive?
Then I don't understand why you gave up at 500psec.
Meantime you can get faster stuff, 70GHz ft and up.
-- Regards, Joerg http://www.analogconsultants.com/
Not enough current to drive a 50R termination through several volts.
I did a Percival distributed amplifier with a couple of BFR96 parts effecti vely in parallel, and still couldn't get better than 800psec.
The 500psec circuit used a couple of dead expensive HP RF power transistors .
I was using an 800MHz local oscillator, locked to a 50MHz crystal reference .
We'd originally planned to use a VCO built around a GigaBit part to produce the 800Mz but couldn't get it to go faster than 750MHz, and I went off to a Microwave trade show and bought an cheap 800MHz VCO designed for the cell
-phone market. We always had plans to do better - probably with a purpose b uilt SAW oscillator - but never got around to it. With a 500psec wide sampl ing pulse 60psec of jitter didn't matter.
My next design - nearly ten years later - would have used a 500MHz crystal oscillator which did offer roughly 1psec jitter. It would have been closer to $100 in the small quantities I'd have been buying.
Using GaAs rather than ECL made the project more difficult, and it took us about 18 months to get the boards well-enough debugged to let us test the s ystem as whole. When I did a bit of digging a few years ago, it started loo king as if the mad Technical director had a suppressed a much less adventur ous system design, which wouldn't have had such impressive specifications a nd would consequently would have been harder to sell.
Not much harder, since this wasn't the unique selling point of the project, but the mad technical director was - at heart - a salesman, and wanted the machine to be as easy to sell as possible.
This turned out to be an expensive choice, because the machine turned out t o be competing with computer simulation of integrated circuit performance, and that picked up enough in the 18 months we lost to debugging to shrink o ur market below a viable size.
We were generating a balanced +/7.5V "blanking voltage" which pulled our el ectron beam off the blanking aperture, and let it through the blanking aper ture when it collapsed to 0V.
Our 100psec pulse would have been made by sweeping the electron beam across the aperture, but we'd have needed a second set of (slower) blanking elect rodes to make sure that it missed the aperture when it swept back - the tim ing on the two excursions couldn't be relied on to be close enough for our purposes.
We were convinced that the 500psec system should have been tweakable down t o about 200psec, but nothing we had time to do made it significantly faster than 500psec.
Sure. The husband of one of my wife's graduate students once offered me a c ouple of pre-production samples of a 50GHz Philips part, but this was long after I was anywhere where I could have used them.
-- Bill Sloman, Sydney
an
chute
an
Sounds like a personal problem to me. $20 to Astraweb or any of many similar providers will fix that for years to come. It has lasted me at least 5 years now.
?-)
That's odd, unless "several" means 5V or more. You can always parallel them, they are cheap.
Seriously, you don't need expensive hardware to do 500psec transitions.
True, but the designer of that 800MHz VCO should have his head washed :-)
I rarely pay over a buck for >> Placing that sampling pulse to 10psec was pure specmanship, and
Then he may have been a good salesman but not a good marketeer.
Ok, then you would have probably needed 2-3 BFR92 in parallel. But considering that they are around 10c it's not so bad.
Hire a consultant who knows how to do this :-)
In the 90s we saw some BJTs well above 10GHz. Not 50GHz but for a
100-200psec sampler you don't really need that. A downside of the fast ones is a greatly reduce voltage rating.-- Regards, Joerg http://www.analogconsultants.com/
When my IPS (AT&T) dumped all of Usenet I signed up with a server at a university in Germany for 10 Euros/year. I don't care much about what it costs but don't want to keep switching. Binaries are dying anyhow for schematic sharing and stuff because less and less people have access.
-- Regards, Joerg http://www.analogconsultants.com/
7.5V, 150mA.
At 150mA, at the time, we did.
He was a terrible marketeer. He'd previously bankrupted his own company wit h a aeries of similar technical choices.
The BFR92 has an absolute maximum collector current limit of 25mA. Gain ban dwidth peaks a little below 20mA. For our 150mA we would have need to paral lel at least six, and eight would have been the fewest anybody sensible wou ld have used.
The BFR96 - which I did try - had a collector current limit of 100mA and it 's gain-bandwidth product peaked at about 40mA with 10V collector voltage, so four of them should have done the job, but I couldn't get better than 80
0psec out of them - that was two 400psec 10% to 90% transitions times in qu ick succession so it wasn't too shabby.We did. He did tidy up a few defects in the final 1984 layout - I'd been on holiday in Australia when that was done, otherwise they wouldn't have been there - and he did suggest that we used Teflon/PTFE board material which t urned out to be pretty much a waste of effort - but his efforts didn't get us a significantly narrower pulse.
Which would have made them useless in our application.
We really did need a +/-7.5V swing. )V to plus +7.5V on one side and 0V to
-7.5V on the other. My BFR96 circuit had a cascode stage on one side, with about 1mA of "wetting current" so that it was never fully off. Very messy.
-- Bill Sloman, Sydney
Think you could email a scan this way? Every bit of cross-referencing documentation from this era helps nail down identifiction. Doesn't have to be in english (or in Joerg's memory) .......
RL
Here it is:
It's on the back cover of the 1982 Transistor Manual, which I bought in Tokyo (Akihabara district) back in '82.
At the time we were using A1015s and C1815s in volume, and this suggested the A1162 and C2712 as SMT subsitututes.
Best regards, Spehro Pefhany
-- "it's the network..." "The Journey is the reward" speff@interlog.com Info for manufacturers: http://www.trexon.com Embedded software/hardware/analog Info for designers: http://www.speff.com
Piece a cake :-)
Why? A BFR92 can swing 5V into a 50ohm load in under 200psec so if you parallel a few it should work at 7.5V and 150mA.
Or have you never tried the BFR92?
Ivory tower design :-)
It's amazing that such people are often able to obtain one round of financing after the other. Sometimes I am baffled with how little engineering due diligence some investors enter into "deals".
You can do it with less. Don't confuse DC values with pulse use. It is often better to consult older datasheets because they contain much better information than "modern" ones. See permissible pulse load on page 5:
Even if you needed six (which you don't), what's the big deal at 10c a piece? I'd probably do that job with 2-3. The important thing is to really drop the hammer on the base.
Should have tried the BFR92 :-)
Then he wasn't the right guy for this job.
Why give up so quickly? 4:1 baluns haven't been invented yesterday.
Did you hammer the base appropriately? How did you drive it? 7.5V in under 200psec with the BFR92 should be no sweat. One just has to leave behind all this small-signal theory from university. I never really believed in that stuff much.
-- Regards, Joerg http://www.analogconsultants.com/
That's what we thought, when we started.
Small production volume design. Electron microscopes - at around a 100 unit s per year - need a similar mind-set. Electron-beam microfabricators, selli ng for a couple of million dollar each at around ten units per year - neede d an exactly similar mindset. Nothing ivory tower about it. The hardware ha d to work and keep on working.
This guy got his financing from his wife's uncle - who made quite a lot of money on the deal until the guy managed to wreck an initially profitable pr oduct by not spending enough on making it reliable.
5:
Bipolar transistor parameters tend to decline abruptly at high collector cu rrents. Lots more stored charge.
Stray capacitance. The collector-base and collect-emitter capacitances of t he BFR92 aren't high, but soldering them onto printed circuit board traces adds extra capacitance.
The BFR96 was a lot more appealing.
Not a well-informed suggestion.
Obviously. But that was less obvious at the time. I didn't do the search th at found him. so I've got no idea who else was available. The impression I had was that not that many people were interested in generating very narrow pulses, and those that were were generating them in rather specialised con texts.
As I was well aware at the time
Ghiggino, K.P., Phillips, D., and Sloman, A.W. "Nanosecond pulse stretcher" ,Journal of Physics E: Scientific Instruments, 12, 686-687 (1979).
uses a 1:1 transmission line transformer, and Ken Ghiggino had amused me by trying - later - to get voltage gain out of a transmission line transforme r some time before 1988 by changing the number of turns. Our paper had cite d
Matick, Richard E. "Transmission line pulse transformers--Theory and applic ations." Proceedings of the IEEE 56.1 (1968): 47-62.
which does spell out how to get step-up and step-down, so he should have kn own better.
Yes.
Never having studied electronics at a university, I wasn't encumbered with it. We were thinking more along the lines of controlling stored base-charge.
In 1991 I started thinking about using BFG31 100mA 5GHz PNP to get a bit mo re base drive, but the project got canned before I could act on the idea.
-- Bill Sloman, Sydney
When it's possible with 5V into 50ohms, why should 7.5V/150mA be impossible?
[...]
The problem with that mindset is that people tend to approach problems with true-and-tried methods, usually by throwing insane amounts of money at the problem. Or trying to. That often leads to seriously less than stellar results, schedule slips, ultimate project failure or a whopping
18 months of debug time like in your case. That is where you need true outside-the-box thinkers who can do things McGyver style.I've had more situations than I can count where people said "Unless we have this, that and the other instrument we can't do it", the boss flew off the handle because that was high six-digit and then in the end we did it with very mundane gear that they already had.
Spending oneself out of a problem is normally the wrong approach, one must hire the right engineers. Real hands-on guys that know analog.
It does not matter what the transistor does after it reached saturation if the circuit is designed right. I often let mine deliberately come out of that state real slow because I don't want a 2nd pulse of unknown shape popping up with reverse polarity.
That's where the analog skills come in. You have to make sure that it doesn't add much.
Well, the fact is it didn't do it. It's also gone lalaland a long time ago while the BFR92 is still available. There's reasons for such trends.
How do you know that without having tried it? I know for a fact that these things work.
One problem was that masters projects were not recorded in electronic format. For example, I designed fast pulsers because I needed them for three samplers on my project. The whole publication with schematics and all then went into a bookshelf at RWTH Aachen University where it probably still sits. Plus some on my bookshelf. I presented at SPIE so it was known but it was nearly impossible for people from other sectors to see because there was no Internet.
Nowadays it's easy to find such people. But most are booked up heavily.
They could have asked any experienced ham radio operator how to do it. This kind of stuff is also explained the respective literature. My best print investment ever was the ARRL Handbook, cost me all of five bucks and had lots of this know-how in it.
Then I don't understand the sluggish performance of the pulser. Either the BFR96 was a dud or something was wrong with the circuitry. Possibly the layout.
Those are ok but the capacitances are a bit high. For a pulse application with not too high PRF smaller devices often perform better.
-- Regards, Joerg http://www.analogconsultants.com/
be an
rival
chute
was an
to
at
For someone renowned as a pinch-penny, you staying with a provider with inferior service and a higher price seems rather odd.
Binary groups for general use are fading due to all the picture hosting sites.
?-)
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.