> Jerry Avins wrote:
>
>>Stan Pawlukiewicz wrote:
>>
>> ...
>>
>>
>>>At least you didn't have to mine and smelt the metal.
>>
>>Wise guy! To win a bet, I actually smelted lead once and extracted its
>>rather high (2%) content of silver. The bet was to prove I could do it
>>with stuff in my basement. (I did it outdoors; it made a lot of SO2.)
>>Galen (lead sulfide) is particularly easy to smelt and there's a neat
>>trick for extracting silver from metallic lead.) Details only on
>>request. I don't want to be boring.
>
>
> OK, I'll bite. I have fiddled around with metallurgy just enough to
> be interested...
Crush the galena. Divide it 1/3, 2/3. Heat ("roast") the 2/3 in air,
stirring occasionally.
2PbS + 3O2 -> 2PbO + 2SO2.
PbO is yellow. Pb2O3 is red. When the yellow starts to darken, stop.
Crush the PbO and mix with the PbS. Cover loosely to exclude air but
allow gas to escape and heat again.
2PbO + PbS -> 3Pb + SO2
Zinc and lead are immiscible, but silver dissolves in both. Zinc is
"hungrier" for silver than lead is, so when molten zinc and lead are
stirred together, most of the silver ends up in the zinc. Take "up"
literally; the zinc floats on the lead where it can be skimmed off as
the mix cools. (Zinc solidifies first.) Zinc sublimates readily. Heating
the zinc in an open crucible leaves the silver behind, partly oxidized.
The best way is heating the zinc in the closed bottom of a long tube.
There is practically no oxidation of the silver and the zinc
resublimates as dust in the upper end of the tube, where it can be
collected and reused.
WARNING! Breathing zinc vapor causes "welders disease". Not a cumulative
poison.
Jerry
--
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������
Reply by Rune Allnor●February 8, 20062006-02-08
Jerry Avins wrote:
> Stan Pawlukiewicz wrote:
>
> ...
>
> > At least you didn't have to mine and smelt the metal.
>
> Wise guy! To win a bet, I actually smelted lead once and extracted its
> rather high (2%) content of silver. The bet was to prove I could do it
> with stuff in my basement. (I did it outdoors; it made a lot of SO2.)
> Galen (lead sulfide) is particularly easy to smelt and there's a neat
> trick for extracting silver from metallic lead.) Details only on
> request. I don't want to be boring.
OK, I'll bite. I have fiddled around with metallurgy just enough to
be interested...
Rune
Reply by Jerry Avins●February 7, 20062006-02-07
Stan Pawlukiewicz wrote:
...
> At least you didn't have to mine and smelt the metal.
Wise guy! To win a bet, I actually smelted lead once and extracted its
rather high (2%) content of silver. The bet was to prove I could do it
with stuff in my basement. (I did it outdoors; it made a lot of SO2.)
Galen (lead sulfide) is particularly easy to smelt and there's a neat
trick for extracting silver from metallic lead.) Details only on
request. I don't want to be boring.
Jerry
--
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������
Reply by Stan Pawlukiewicz●February 7, 20062006-02-07
Jerry Avins wrote:
> Rune Allnor wrote:
>
>> Jerry Avins wrote:
>
>
> ...
>
>>> As an amateur silversmith, I make square wire from round with hammer and
>>> anvil block. One of the most impressive bit of blacksmithing I ever saw
>>> was the transformation of a 1" square, 8" long rod into a 2" cube by
>>> hammering it only on its ends. Much depends on how it is heated in the
>>> forge and with a torch.
>>
>>
>>
>> I never assigned that sort of "torture" to others, but at one point I
>> did try to implement an Singular Value Decomposition from scratch.
>> While I never quite got there, I did find out how much I am willing to
>> pay for
>> a good numerical library...
>
>
> There are some things one can't buy. The presents I made for my sisters'
> twelfth birthday were twisted copper and silver #8 wires (about 1/8"
> diameter). They were square for the middle inch only, the rest being
> round. The middle parts were set side by side to make a rectangular
> cross section 1/8 by 1/4, and the round parts twisted together. (The two
> ends have opposite-handed twists.) Then I hammered the twists to the
> same cross section as the middle, and shaped the bracelet to fit their
> wrists. To have part round and part square wire, you have to make it.
>
> Jerry
At least you didn't have to mine and smelt the metal.
Reply by Jerry Avins●February 7, 20062006-02-07
Rune Allnor wrote:
> Jerry Avins wrote:
...
>>As an amateur silversmith, I make square wire from round with hammer and
>>anvil block. One of the most impressive bit of blacksmithing I ever saw
>>was the transformation of a 1" square, 8" long rod into a 2" cube by
>>hammering it only on its ends. Much depends on how it is heated in the
>>forge and with a torch.
>
>
> I never assigned that sort of "torture" to others, but at one point I
> did try to implement an Singular Value Decomposition from scratch. While
> I never quite got there, I did find out how much I am willing to pay for
> a good numerical library...
There are some things one can't buy. The presents I made for my sisters'
twelfth birthday were twisted copper and silver #8 wires (about 1/8"
diameter). They were square for the middle inch only, the rest being
round. The middle parts were set side by side to make a rectangular
cross section 1/8 by 1/4, and the round parts twisted together. (The two
ends have opposite-handed twists.) Then I hammered the twists to the
same cross section as the middle, and shaped the bracelet to fit their
wrists. To have part round and part square wire, you have to make it.
Jerry
--
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������
Reply by Rune Allnor●February 7, 20062006-02-07
Jerry Avins wrote:
> Richard Dobson wrote:
> > Jerry Avins wrote:
> >
> > ..
> >
> >> I'm here to elaborate and to scold. I don't know who should be
> >> scolded, I presume it is the instructor. By the time one gets an
> >> assignment like yours, one should know that convolution in the time
> >> domain has the same effect as multiplication in the frequency domain
> >> (and vice versa). This means that it is possible to convolve in the
> >> time domain by using FFT to convert to frequency, multiplying, and
> >> then using IFFT co convert back to time. It may be surprising, but
> >> this procedure is faster than direct convolution for large enough
> >> sample sets.
> >>
> >
> > I can imagine a pedagogical purpose to the idea - how better to
> > appreciate the performance of the FFT than to first implement direct
> > convolution? And in any case, the dc algorithm is still appropriate for
> > small N. Otherwise, if one simply accepts ex cathedra "FFT is better",
> > one might never have a reason to learn how to code a dc.
> >
> > Its perhaps the equivalent of those torture tests apprentice engineers
> > used to be given in "the old days" - take a piece of round rod, and
> > square it up just using a file and hand measuring tools. Then file it
> > round again, ditto. Not a test I ever took, nor could pass.
>
> As an amateur silversmith, I make square wire from round with hammer ans
> anvil block. One of the most impressive bit of blacksmithing I ever saw
> was the transformation of a 1" square, 8" long rod into a 2" cube by
> hammering it only on its ends. Much depends on how it is heated in the
> forge and with a torch.
I never assigned that sort of "torture" to others, but at one point I
did try to
implement an Singular Value Decomposition from scratch. While I never
quite got there, I did find out how much I am willing to pay for a good
numerical library...
Rune
Reply by Jerry Avins●February 6, 20062006-02-06
Richard Dobson wrote:
> Jerry Avins wrote:
>
> ..
>
>> I'm here to elaborate and to scold. I don't know who should be
>> scolded, I presume it is the instructor. By the time one gets an
>> assignment like yours, one should know that convolution in the time
>> domain has the same effect as multiplication in the frequency domain
>> (and vice versa). This means that it is possible to convolve in the
>> time domain by using FFT to convert to frequency, multiplying, and
>> then using IFFT co convert back to time. It may be surprising, but
>> this procedure is faster than direct convolution for large enough
>> sample sets.
>>
>
> I can imagine a pedagogical purpose to the idea - how better to
> appreciate the performance of the FFT than to first implement direct
> convolution? And in any case, the dc algorithm is still appropriate for
> small N. Otherwise, if one simply accepts ex cathedra "FFT is better",
> one might never have a reason to learn how to code a dc.
>
> Its perhaps the equivalent of those torture tests apprentice engineers
> used to be given in "the old days" - take a piece of round rod, and
> square it up just using a file and hand measuring tools. Then file it
> round again, ditto. Not a test I ever took, nor could pass.
As an amateur silversmith, I make square wire from round with hammer ans
anvil block. One of the most impressive bit of blacksmithing I ever saw
was the transformation of a 1" square, 8" long rod into a 2" cube by
hammering it only on its ends. Much depends on how it is heated in the
forge and with a torch.
My scolding has to do with the OP's not knowing what convolution is
aside from a bit of code he can cut and paste.
Jerry
--
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������
Reply by Richard Dobson●February 6, 20062006-02-06
Jerry Avins wrote:
..
> I'm here to elaborate and to scold. I don't know who should be scolded,
> I presume it is the instructor. By the time one gets an assignment like
> yours, one should know that convolution in the time domain has the same
> effect as multiplication in the frequency domain (and vice versa). This
> means that it is possible to convolve in the time domain by using FFT to
> convert to frequency, multiplying, and then using IFFT co convert back
> to time. It may be surprising, but this procedure is faster than direct
> convolution for large enough sample sets.
>
I can imagine a pedagogical purpose to the idea - how better to appreciate the
performance of the FFT than to first implement direct convolution? And in any
case, the dc algorithm is still appropriate for small N. Otherwise, if one
simply accepts ex cathedra "FFT is better", one might never have a reason to
learn how to code a dc.
Its perhaps the equivalent of those torture tests apprentice engineers used to
be given in "the old days" - take a piece of round rod, and square it up just
using a file and hand measuring tools. Then file it round again, ditto. Not a
test I ever took, nor could pass.
Richard Dobson
Reply by James Van Buskirk●February 6, 20062006-02-06
"James Van Buskirk" <not_valid@comcast.net> wrote in message
news:XPOdnftiRZ48SnreRVn-tg@comcast.com...
> The product of the central 3 factors above is permutation-
> equivalent to a 2-block-diagonal matrix (this is the content
> of the convolution theorem.) If RF(N) were replaced by
> RF(N/2).tensor.RF(2), the central 3 factors would now be
> permutation equivalent to a 4-block-diagonal matrix. The middle
> part is now slightly harder to compute, but in terms of
> total arithmetic operation counts the overall algorithm is
> slightly shorter. Additionally, using two-way SIMD is a
> snap when computing RF(N/2).tensor.RF(2), and not as easily
> conceived with RF(N).
Oh my, did I really say RF(N/2).tensor.RF(2) in the above
paragraph? I should have said RF(N/2).tensor.identity(2)
instead. Sorry.
--
write(*,*) transfer((/17.392111325966148d0,6.5794487871554595D-85, &
6.0134700243160014d-154/),(/'x'/)); end