Date: Fri, 20 Aug 2004 22:03:36 +0200
From: "Stefan Trethan" <
stefan_trethan@...>
Subject: Re: RE: making vias (was: Re: Plating thruholes.)
On Fri, 20 Aug 2004 14:41:23 -0500, Thomas P. Gootee <
tomg@...>
wrote:
> Thanks, to everyone who replied so far, for the good suggestions and
> information!
>
> This re-design is still in the very early stages. I don't know if this
> is too off-topic. But I'll post a sort of summary of the different
> board/connector options I'm considering (mostly copied from an email
> that I sent to a friend, recently, about this stuff). Maybe some of you
> can set me straight, or offer some practical tips, or some ideas.
>
I didn't read all of your post but i will tell you a story a good teacher
told me some
time ago.
He was working for a company making TVs, and had a new design to make.
Recently they had hired a guy which should make the units more reliable,
by statistics and such.
He was to check all new designs.
Well, my teacher designed the new TV as usual with plug-in cards.
After he was done he showed it to the reliability guy to be checked.
This guy only asked him if he is mad using several hundred connections.
He told him he an forget about the plug-in cards because the statistics
say they
will fail too often. That meant a major redesign of the unit of course.
That was the last Plug-in type TV he built, from then they only used
single boards or
soldered connections.
so i wouldn't say connectors make things reliable.
Another issue is upgradeability. I expect every single one of your
customers
to be well able to solder and even design circuits.
Do you really believe a PCB for all the front-plate elements is a good
idea?
It would be MUCH harder to add an additional connector or something then.
If i were you i'd just put a solder pin in every wire pad. Then solder the
suitble pug to the wire and done. You can easily take it apart with no
redesign at all.
Cheapest, fastest, and IMO userfriendliest.
I would only use "combined" connectors where it makes sense.
E.g. a parallel communications port will not be wired with individual wires
or connectors. A power supply is on the border. A combined connector would
be nice to prevent reversing, but then it is not much hassle to plug 2
individual
connectors. I would not combine "unrelated" signals (e.g. a horror to
debug).
If you want my advise forget about that. Leave your design as it is with
solder pins/connectors
where now the wires are. Otherwise you make your PCBs HUGELY complicated,
with very little gain.
If I were your customer i would prefer it simple, so i can make changes.
Remember, your are
not making a consumer electronics device, you are dealing with
professionals.
By the way i'm also planning to make a curve tracer some day. But it is
very far down
on the list and i usually just crank up a circuit for the particular case
needed.
ST
ST,
Thanks for the thoughtful reply.
I agree that connectors, sockets, etc, don't usually make things more reliable. What I meant by saying that they might make the unit more-reliable was really just that there might be a lower probability of wiring errors at the time of construction. The multitudes of wires can also make the construction tasks difficult, which probably tends to cause lower quality, since it can get very difficult to reach some of the solder lugs, etc, after lots of the wires are in place.
I also already knew that card-edge connectors are probably less-reliable than other types. But I don't think that they are always unacceptable. (And the architecture, using them, just seems so darned attractive and "clean", to me, hehe...)
You make a good point, if I understood correctly, about using individually-pluggable wires, with a single-pin on the pcb for each one. I did read some posts in one of the pinball or arcade game newsgroups, or a website referenced there, about different types of connectors, where they were saying/claiming that crimp-type connectors for discrete wires are one of the most-reliable types (which includes wires with crimped-on pins that are then inserted/mounted into multi-connection sockets or plugs). They also claimed that soldering, in addition to crimping, is not necessary, and should probably even be avoided, since it might do some harm to the connector, or the wire/insulation, if not done perfectly.
I'm looking through my Mouser.com catalog, again, right now!
Regarding using a PCB for the front panel controls, etc: I see it as a ∗huge∗ benefit. The mass of wiring that results from the way it's done now is "a mess". Adding a new control or something to the front panel design, later, wouldn't be too difficult, or, at least not difficult-enough to significantly offset the benefits of using a PCB, in my opinion. And if some user wants to add their own control to the panel, later, in an existing instrument, well, they'll just have to find a way (There will also be some space between the panel and its pcb.). But that would be very rare, I think, and is not too much of a concern.
Regarding making a Curve Tracer (kind of a long story...):
MINE started out to be just for MY OWN USE, and was basically just one of the "quick and dirty" curve tracers, like the one described here:
http://www.repairfaq.org/REPAIR/F_semitest.html#stqdc , i.e. just a small signal source and a few resistors.
But I started adding "convenient" features, one by one, and then just kept on adding more and more to it:
e.g. I also REALLY wanted to be able to see the whole "family" of curves, for a transistor, all at once, like on a "real" curve tracer. So I eventually figured out how to have a ramp "sweep" signal for the DUT, with a synchronized staircase signal for the base/gate. And I needed multiple voltage rails, so I designed a boost-mode switching power supply board that takes 12 vdc and makes the plus and minus 18 vdc (variable, actually) needed by most of the ICs (and at up to 3 amps, total), and the +5 for CMOS ICs. With all of that working, I decided I also needed to be able to "push some more amps" into the device under test (DUT). So I designed a power amplifier. And I wanted calibrated excitation-voltage levels for the DUT so I made the amp's gain switchable. Then I decided that the resistor networks required for switching the gain were kind of "ugly", so I designed a coool feedback control loop that sets the gain.
I'm pretty proud of that feedback control loop gain-control circuit: It uses a Vactrol (VTL5C2) current-controlled resistor (basically just a photocell and LED encapsulated together). Since there are a discrete number of desired peak-to-peak output voltage levels, I used a TL1431IZ 0.4% 2.5v precision voltage reference to derive six DC voltages, equal to the six selectable peak output levels' 0-to-peak values. A frequency-compensated peak-detector/envelope-follower type of circuit continuously produces a quasi-DC voltage that's equal to the ACTUAL output's 0-to-peak value. That voltage is then subtracted from the selected DC reference voltage, to get an "error signal" voltage, which is used to drive current through the VTL5C2's LED, which sets the resistance that sets the amplifier's gain.
The VTL5C2 is a pretty nifty device: Its resistance varies (non-linearly, though) from about 2 Megohms with no current through its LED to about 200 Ohms with 40 mA through its LED. And it has none of the "quirks" that you have to worry about when using an FET as a voltage-controlled resistor. (And bgmicro.com has them for $0.50 each.)
The amplifier's gain can vary from 1 to about 101. It uses a National LM1875T "opamp", which can easily push 3 or 4 amps from its' output, although I limit it to 1.5 Amps, in the Curve Tracer product.
I also added lots of other stuff, including instrumentation amplifiers to sense the voltage and current used for the x-y scope display, and inverters, buffers, and switching, etc, so I could display the current through ANY of the three DUT leads versus the voltage across ANY two of the DUT leads, and flip the displayed polarity of either of them at any time. And I added a range of frequencies for the sweep signal. And I added twelve selectable current-limiter resistances. And I added an option for a triangle waveform instead of a ramp, and an integrator to produce a quasi-sine waveform from the triangle. And so on and so on. And on and on and on...
So, ∗eventually∗, I thought to myself, "Hey! OTHER people might like to have one of these, TOO!". Of course, when a hobbyist designs and builds a piece of electronic equipment, especially if it's essentially the first time they've done a large self-designed project, then when it's all "finished", and works "perfectly", they've still maybe done only about ∗∗5%∗∗ of the work that's needed to make it into an actual "commercial product".
Mine's also available in KIT form, which makes it even MORE work, in some ways, since I have to produce (AND keep ∗updated∗) all of the construction diagrams, for component placement, wiring, mechanical stuff, etc, and assembly and alignment instructions, plus complete schematics, detailed parts lists, instrument-panels' artwork, etc etc. (all in "presentable" forms). And I have to do all of the sourcing and supplier stuff, keep a large parts and supplies inventory, make circuit boards, make apply-able instrument panel artwork, count parts and supplies into nice little multi-compartmented plastic kit-boxes, market and sell them, pack and ship them, support them, etc etc etc. I'm also working on a "real" ops/service manual, which, eventually, is intended to be as good as the legendary older Tektronix manuals. (Hmmm... Maybe I ∗AM∗ crazy... Hehe...)
And I still have MANY, many things that I'd like to add, and change, in the Curve Tracer product. But, first, I want to "clean up" the current version, especially since I may eventually have it (or parts of it) mass-produced, maybe by a third party. I've also got some other great electronic products "in the pipeline". But I think that I need to "get to the next level", first, so I can hire some people, to hopefully allow ME to spend my time on things where I can contribute the most, i.e. where any specialized abilities that I have might make the most difference, instead of on things that almost anyone could be doing. I've been trying to "bootstrap" this business, starting with not much capital. But I may end up having to take on some investors, to be able to get where I want to be (quickly-enough, anyway).
Thank you, again, Stephan, very much, for your good suggestions!
And I am sorry that I blathered-on, for so long, ∗AGAIN∗. (This business has basically "taken over my life", as you can probably tell, hehe. But it IS ∗quite∗ enjoyable...)
Cheers,
Tom Gootee
http://www.fullnet.com/u/tomg-----------------------------------
[Non-text portions of this message have been removed]