On Tue, 22 Nov 2011 08:18:23 -0800, John Larkin
wrote:
On Tue, 22 Nov 2011 01:45:25 -0600, John Fields
wrote:
On Mon, 21 Nov 2011 15:17:43 -0800, John Larkin
wrote:
On Mon, 21 Nov 2011 13:16:30 -0600, John Fields
wrote:
On Sun, 20 Nov 2011 16:33:39 -0800, John Larkin
wrote:
On Sun, 20 Nov 2011 17:56:24 -0600, John Fields
wrote:
On Sun, 20 Nov 2011 15:06:52 -0800, John Larkin
m wrote:
On Sun, 20 Nov 2011 16:42:39 -0600, John Fields
wrote:
On Sun, 20 Nov 2011 09:25:29 -0800, John Larkin
m wrote:
On Sun, 20 Nov 2011 08:16:21 -0800, Fred Abse
wrote:
On Sat, 19 Nov 2011 14:14:45 -0800
John Larkin wrote:
Lots of online calculators are available.
I use this. I'm not sure how good or otherwise it looks in Excel, it was made using Sun Microsystems' Star Office, and exported as XLS. I hope the bitmap appears in the right place to overlay the value fields, I don't have Excel, to try it.
It produces theoretically image-matched attenuators for any combination of input and output resistances and any value of attenuation. Unrealizable combinations just result in negative values.
Derived from Radiotron Designer's Handbook, Chapter 4, "Theory of Networks", Sect. 8(v) "Image impedances and image transfer constant of four-terminal networks". Equations 17a thru 17f apply. There's a neat trick to get rid of the hyperbolic functions.
I wrote it about 30 years ago, as an HP41CV calculator app, and subsequently made a spreadsheet out of it.
See attached file.
Ooh, very nice. Thanks. JF can use that to fix up his design.
---
Actually, I'm close to fixing the math errors in my design, which seem
to have come from a trusted source, and which I'll post when I'm done.
You must mean formerly trusted source.
---
Nope, they're still trusted, by-and-large, because of previous
flawless performance.
If I get a bad hit again, though, they'll bear a little closer
watching.
Kinda like where you are now, where truth bows to face and I must
therefore consider you less than trustworthy.
All you have to do is look at the values of the first tee, which I
did, and it's obvious that it won't attenuate 32 dB. I don't trust
anything without giving it a sanity check.
---
PKB, since didn't your recent little resistor fiasco come about
because of excessive trust and inadequate sanity check?
Resistor fiasco? You mean the 0.05% thinfilms? Not the same thing at
all.
---
ISTR that you trusted your vendor enough to supply you with the right
stuff that you didn't do the sanity check of an incoming inspection
and, consequently, you got bit on the ass for your fox paws.
We have 5325 different parts in stock, 2.08 million pieces, so things
like that happen once in a while. We can't possibly incoming inspect
every part we buy.
---
Then you're substituting trust for incoming test and are willing to
take the production test fallout/repair cycle.
---
Imagine setting up test jigs for opamps,
microprocessors, FPGAs, bare PC boards, transformers, sheet matal, all
that.
---
Imagine?
It isn't hard to do, but for you it seems to be a question of
economics, and you're willing to initially trust vendors' spec's and
apply salve to the wounds when the parts fall short.
Or long.
---
I have done flight hardware for spacecraft, where every single
part is tested and certified and traceable, and that's absurdly
expensive.
---
Yeah, but just consider the alternative. :-(
---
We do have procedures for picking up on parts problems in production
test, or from field returns, and investigating any patterns. That's
how we caught the resistor problem. In fact, it was a BIST reutine run
on an RTD acquisition section that found it, namely a circuit that
uses two resistors and checks them against one another.
Which airplanes do you have stuff flying on, or you supply GSE for,
anyway?
We don't have much actually flying. Some on the U2, some heads-up
display stuff on some AH130s. Most of our stuff is used in engine and
FADEC development and test cells (United Airlines uses our gear to
test APUs). We do some ground test stuff for B52 radars. That's a good
mix, aerospace but no mil or FAA certifications, which are a lot of
work.
---
Agreed.
---
Trust is no substitute for checking. As systems get more complex, the
margin for risk goes down.
Spare me the pontification of platitudes, blowhard.
Check your math, doofus.
---
All in good time, and at my leisure.
Did you use one of those online calculators?
---
No, I used the formulas and the table, attached, from the 7th edition
of the ITT Handbook, ISBN 0-672-21563-2.
I just plucked the values for 1, 2, 4, and 8dB from the table, and
calculated the values of the resistances for 16 and 32 dB pads using
the formulas.
Using their notation for a tee pad:
.. o---[a]-+-[a]---o
.. |
.. [b]
.. |
.. o-------+-------o
with input and output resistances (R0) of one ohm, they state that:
dB/20 dB/20
a = (10^ -1) / (10^ +1) ohms
and
dB/20 dB/10
b = 2.1^ / (10^ -1)
and, that for values of R0 not equal to one ohm, the calculated values
of the resistances should be multiplied by the desired R0.
Since you found errors, I decided to run my numbers again, and
especially to run them at 10dB, where they could be double-checked
against the table and the formulas.
I also ran the numbers from the 10dB table entry through LTspice and
found them to be rock-solid for dB = 10 log P1/P2 and dB = 20 log
V1/V2.
I've attached my hand calculations, using the formulas from the book,
for your perusal.
If you don't want to get that deep into it, here's a summary:
ATTEN Ra Rb
dB ohms ohms
----|--------|------
10 25.973 8.05
16 36.318 2.22
32 47.5 0.1
Interesting to note is that for the 10dB entry, the values of Ra -
according to the table and the formula - are identical, but the values
of Rb are not.
Using LTspice as a referee, it turns out that the values given in the
table are perfect, but those calculated using the formulas are not.
So where's the error?
A typo in the book.
It turns out that in the formula for b,
dB/20 dB/10
b = 2.1^ / (10^ -1),
the decimal point between the 2 and the 1 should actually be a dot
indicating multiplication, and the formula for Rb in a 50 ohm pad
would then be:
dB/20 dB/10
Rb = 2 * 50^ / (10^ -1)
Using that formula gives:
ATTEN Ra Rb
dB ohms ohms
----|-------|------
10 25.97 35.14
16 36.32 16.26
32 47.55 2.51
BTW, here's where I found that the error was in the book:
http://www.microwaves101.com/encyclo...attenuator.cfm
A lot of them are plain wrong.
---
I'm surprised, then, that in an earlier post you didn't offer me a leg
up by referring me to one which you knew wasn't flawed.
---
I also suspect a lot of people are copying one anothers'
javascript, because I see the same wrong calculators in multiple
places.
A lot of microstrip-type calculators are wrong. Try a really wide
trace; the bad ones will report a negative impedance.
We could start a list of bad online calculators.
---
I don't care to, but YMMV.
--
JF