On 30/01/2018 11:39, Recliner wrote:
Someone Somewhere wrote:
On 30/01/2018 10:06, Roland Perry wrote:
In message , at 09:08:18 on Tue, 30 Jan
2018, Someone Somewhere remarked:


Nothing that was manufactered is unmanufacturable - it may not be
reasonably economic to do so,Â* or in certain cases legislation may
prevent it (lead etc) but if it was built once, it could be built again.

There are whole generations of custom-chips which aren't manufacturable
any more. Either the company which made them originally has gone out of
business/disappeared within another that's not longer in the foundry
business, or the tools and machinery required to produce a new batch
have long since been consigned to the dustbin of history.

A handful of generic chips may still be available, so you could perhaps
get a brand-new Z80 equivalent/clone processor chip to build a replica
Amstrad CPC464, but good luck getting Ferranti or SGS to make you a
fresh one of the ULAs.

You could still recreate them with enough time and money - they aren't
made of unobtanium - so it's economics. Now to rebuild the Ferranti fab
may be a ludicrous amount of money, but it's theoretically possible.

Or of course you could use FPGAs to do the same thing these days.


Would it be feasible to simply emulate all the old electronics and computer
components in software, running on a standard modern commodity CPU? The
modern CPU would be so much faster that it might deliver enough performance
to be able to precisely emulate the timing as well as the behaviour of the
old stuff.


Presumably there are issues getting it to talk to all the other bits and
the outside world (ie enought I/O pins and level converters etc) but I'm
sure that could be resolved.

What might be harder is proving it is equivalent - particularly for
organisations with a high and detailed safety requirement. You keep
your certification if the thing is exactly the same, but if it changes
you need it to be re-certified which may actually be impossible if the
relevant certification organisation no longer exists or offers that
certification.

In message , at 11:19:55 on Tue, 30 Jan
2018, Someone Somewhere remarked:

Nothing that was manufactered is unmanufacturable - it may not be
reasonably economic to do so,* or in certain cases legislation may
prevent it (lead etc) but if it was built once, it could be built again.
There are whole generations of custom-chips which aren't
manufacturable any more. Either the company which made them
originally has gone out of business/disappeared within another that's
not longer in the foundry business, or the tools and machinery
required to produce a new batch have long since been consigned to the
dustbin of history.
A handful of generic chips may still be available, so you could
perhaps get a brand-new Z80 equivalent/clone processor chip to build
a replica Amstrad CPC464, but good luck getting Ferranti or SGS to
make you a fresh one of the ULAs.

You could still recreate them with enough time and money - they aren't
made of unobtanium - so it's economics. Now to rebuild the Ferranti
fab may be a ludicrous amount of money, but it's theoretically possible.

I thought we were discussing things which *were* [claimed to be]
economically feasible. Like restarting Electrostar production.

Or of course you could use FPGAs to do the same thing these days.

If you can reverse engineer the circuitry inside the ULA.

Can you make a radio transceiver out of FPGA's?
--
Roland Perry

In message
-sept
ember.org, at 11:39:49 on Tue, 30 Jan 2018, Recliner
remarked:
Nothing that was manufactered is unmanufacturable - it may not be
reasonably economic to do so,* or in certain cases legislation may
prevent it (lead etc) but if it was built once, it could be built again.

There are whole generations of custom-chips which aren't manufacturable
any more. Either the company which made them originally has gone out of
business/disappeared within another that's not longer in the foundry
business, or the tools and machinery required to produce a new batch
have long since been consigned to the dustbin of history.

A handful of generic chips may still be available, so you could perhaps
get a brand-new Z80 equivalent/clone processor chip to build a replica
Amstrad CPC464, but good luck getting Ferranti or SGS to make you a
fresh one of the ULAs.

You could still recreate them with enough time and money - they aren't
made of unobtanium - so it's economics. Now to rebuild the Ferranti fab
may be a ludicrous amount of money, but it's theoretically possible.

Or of course you could use FPGAs to do the same thing these days.

Would it be feasible to simply emulate all the old electronics and computer
components in software, running on a standard modern commodity CPU? The
modern CPU would be so much faster that it might deliver enough performance
to be able to precisely emulate the timing as well as the behaviour of the
old stuff.

How plausible is a software emulation of an NRN radio, and how would one
get it approved for use?
--
Roland Perry

On 30/01/2018 12:41, Roland Perry wrote:

Can you make a radio transceiver out of FPGA's?

Probably not, but you can get SDRs which presumably you can make a radio
tranceiver out of.

However as I said, and as I see you said, getting type approval for such
a thing may be the unobtanium in this sauce.

Someone Somewhere wrote:
On 30/01/2018 12:41, Roland Perry wrote:

Can you make a radio transceiver out of FPGA's?

Probably not, but you can get SDRs which presumably you can make a radio
tranceiver out of.

And you can make an SDR out of an FPGA. For instance:
http://kiwisdr.com/

You do need a little bit of help with the frontend - ie an analogue
amplifier and ADC. Once you have I/Q samples you can do the rest in
soft-logic (or software). A lot of mobile base stations operate this way.

(You could, in principle, try and do a one-bit ADC using a bare FPGA pin
with an antenna. The gain is such that it won't receive anything but a very
strong signal, which is more likely to be local noise than anything else)

However as I said, and as I see you said, getting type approval for such
a thing may be the unobtanium in this sauce.

Indeed.

Theo

In article , (Roland Perry)
wrote:

In message
-
september.org, at 11:39:49 on Tue, 30 Jan 2018, Recliner
remarked:
Nothing that was manufactered is unmanufacturable - it may not be
reasonably economic to do so,* or in certain cases legislation may
prevent it (lead etc) but if it was built once, it could be built
again.

There are whole generations of custom-chips which aren't
manufacturable any more. Either the company which made them originally
has gone out of business/disappeared within another that's not longer
in the foundry business, or the tools and machinery required to
produce a new batch have long since been consigned to the dustbin of
history.

A handful of generic chips may still be available, so you could
perhaps get a brand-new Z80 equivalent/clone processor chip to build
a replica Amstrad CPC464, but good luck getting Ferranti or SGS to
make you a fresh one of the ULAs.

You could still recreate them with enough time and money - they aren't
made of unobtanium - so it's economics. Now to rebuild the Ferranti
fab may be a ludicrous amount of money, but it's theoretically
possible.

Or of course you could use FPGAs to do the same thing these days.

Would it be feasible to simply emulate all the old electronics and
computer components in software, running on a standard modern commodity
CPU? The modern CPU would be so much faster that it might deliver enough
performance to be able to precisely emulate the timing as well as the
behaviour of the old stuff.

How plausible is a software emulation of an NRN radio, and how would
one get it approved for use?

NRN radios weren't the problem. First of all, the later technology FM1000
type was type approved and in widespread use and they were standard enough
for a cottage industry to spring up keeping them going until NRNJ was
switched off recently. The problem was with the more specialised RETB radios
which were never made in such numbers and only used on a few lines, mainly
in the Highlands. Making a scratch-built CPC464 isn't a problem because it
doesn't have to be certified as in conformance to safety standards. RETB
radios are vital railway signalling components with a very strict approvals
regime.

--
Colin Rosenstiel

In article , (Someone
Somewhere) wrote:

On 30/01/2018 12:41, Roland Perry wrote:

Can you make a radio transceiver out of FPGA's?

Probably not, but you can get SDRs which presumably you can make a
radio tranceiver out of.

However as I said, and as I see you said, getting type approval for
such a thing may be the unobtanium in this sauce.

Even if you get a working radio that way you wouldn't have one type approved
to use on the railway. As far as type approval was concerned it would be a
new radio which would require to go through the approval process from
scratch.

If you were going to do that, you might as well use the modern design radio
which would also need type approval but if approved would allow more new
radios readily to be manufactured. In 1998, the type approval process was
too onerous and also unaffordable for the small additional RETB requirement.

--
Colin Rosenstiel

In message , at 14:55:47
on Tue, 30 Jan 2018, remarked:
Nothing that was manufactered is unmanufacturable - it may not be
reasonably economic to do so,* or in certain cases legislation may
prevent it (lead etc) but if it was built once, it could be built
again.

There are whole generations of custom-chips which aren't
manufacturable any more. Either the company which made them originally
has gone out of business/disappeared within another that's not longer
in the foundry business, or the tools and machinery required to
produce a new batch have long since been consigned to the dustbin of
history.

A handful of generic chips may still be available, so you could
perhaps get a brand-new Z80 equivalent/clone processor chip to build
a replica Amstrad CPC464, but good luck getting Ferranti or SGS to
make you a fresh one of the ULAs.

You could still recreate them with enough time and money - they aren't
made of unobtanium - so it's economics. Now to rebuild the Ferranti
fab may be a ludicrous amount of money, but it's theoretically
possible.

Or of course you could use FPGAs to do the same thing these days.

Would it be feasible to simply emulate all the old electronics and
computer components in software, running on a standard modern commodity
CPU? The modern CPU would be so much faster that it might deliver enough
performance to be able to precisely emulate the timing as well as the
behaviour of the old stuff.

How plausible is a software emulation of an NRN radio, and how would
one get it approved for use?

NRN radios

Typo, I meant RETB

weren't the problem. First of all, the later technology FM1000
type was type approved and in widespread use and they were standard enough
for a cottage industry to spring up keeping them going until NRNJ was
switched off recently. The problem was with the more specialised RETB radios
which were never made in such numbers and only used on a few lines, mainly
in the Highlands. Making a scratch-built CPC464 isn't a problem because it
doesn't have to be certified as in conformance to safety standards.

It *is* a problem, because the custom chip can't be made any more.

RETB radios are vital railway signalling components with a very strict
approvals regime.

Which both of us have mentioned earlier in the thread.
--
Roland Perry

On 30/01/2018 20:55, wrote:
In article , (Someone
Somewhere) wrote:

On 30/01/2018 12:41, Roland Perry wrote:

Can you make a radio transceiver out of FPGA's?

Probably not, but you can get SDRs which presumably you can make a
radio tranceiver out of.

However as I said, and as I see you said, getting type approval for
such a thing may be the unobtanium in this sauce.

Even if you get a working radio that way you wouldn't have one type approved
to use on the railway. As far as type approval was concerned it would be a
new radio which would require to go through the approval process from
scratch.

If you were going to do that, you might as well use the modern design radio
which would also need type approval but if approved would allow more new
radios readily to be manufactured. In 1998, the type approval process was
too onerous and also unaffordable for the small additional RETB requirement.


Which is what I said and what Roland said....