I'm heading out for a long-postponed offsite with colleagues
(and dreading the travel aspect even more than usual!).

The "issue" I'll be bringing to discuss is design practices
to make largely analog designs "future safe" -- to guard
against component offerings going obsolete, etc. (I need
at least a 10 year window of continued availability).

Digital designs can usually be "refreshed" with relative ease.
Often considerably easier than the original design effort!

And, if you are prudent, software ports to new platforms and
architectures can similarly be straightforward.

SOME "analog" designs can be easy to "redo" -- but usually
only ones that have black-box type interfaces: power supplies,
converters, appliances with fixed functionality, etc.
A lot of an analog design's performance is tied up in the
actual implementation -- and, if not careful, you can slip
backwards with the next turn of the crank (a lot harder to
do that with digital/software designs).

I don't see anyway to pick devices for longterm availability
other than settling for run-of-the-mill devices that have
work-alikes available from other vendors. Anything bleeding
edge is bound to lead to issues down the road (availability,
support, etc.)

So, aim for banal designs? Do lifetime buys of the components
that you're most exposed to? Or, just resign yourself to
redesigning when the problem manifests (and hope it's
someone else's problem, then)?

[design of interest is an SDR; a COTS purchase would lead to
vulnerability for the entire design!]

I will try to check back but not sure how easy that will
be while traveling...

Don Y wrote:
> I'm heading out for a long-postponed offsite with colleagues
> (and dreading the travel aspect even more than usual!).
>
> The "issue" I'll be bringing to discuss is design practices
> to make largely analog designs "future safe" -- to guard
> against component offerings going obsolete, etc.  (I need
> at least a 10 year window of continued availability).
>
> Digital designs can usually be "refreshed" with relative ease.
> Often considerably easier than the original design effort!
>
> And, if you are prudent, software ports to new platforms and
> architectures can similarly be straightforward.
>
> SOME "analog" designs can be easy to "redo" -- but usually
> only ones that have black-box type interfaces:  power supplies,
> converters, appliances with fixed functionality, etc.
> A lot of an analog design's performance is tied up in the
> actual implementation -- and, if not careful, you can slip
> backwards with the next turn of the crank (a lot harder to
> do that with digital/software designs).
>
> I don't see anyway to pick devices for longterm availability
> other than settling for run-of-the-mill devices that have
> work-alikes available from other vendors.  Anything bleeding
> edge is bound to lead to issues down the road (availability,
> support, etc.)
>
> So, aim for banal designs?  Do lifetime buys of the components
> that you're most exposed to?  Or, just resign yourself to
> redesigning when the problem manifests (and hope it's
> someone else's problem, then)?

Depends. What's in it?

Cheers

Phil Hobbs

(Who has about 30k BF862s and CPH3910s on the shelf, among other things
such as a reel or two each of BFG31 and BFT92 5-GHz PNPs, BFT25A small
fast NPNs, quantities various pHEMTs and SiGe:C NPNs, and so forth.)

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

Hopefully this post makes it out; I'm using a phone (now THAT was
a mistake! I should have brought a small laptop or a tablet!)
via wifi to remotely access *my* server which, in turn,
accesses the NNTP service) <frown> I can see this should
have been thought out, better!

On 6/14/2021 5:48 PM, Phil Hobbs wrote:

>> The "issue" I'll be bringing to discuss is design practices
>> to make largely analog designs "future safe" -- to guard
>> against component offerings going obsolete, etc. (I need
>> at least a 10 year window of continued availability).

>> I don't see anyway to pick devices for longterm availability
>> other than settling for run-of-the-mill devices that have
>> work-alikes available from other vendors. Anything bleeding
>> edge is bound to lead to issues down the road (availability,
>> support, etc.)
>>
>> So, aim for banal designs? Do lifetime buys of the components
>> that you're most exposed to? Or, just resign yourself to
>> redesigning when the problem manifests (and hope it's
>> someone else's problem, then)?
>
> Depends. What's in it?

There's a fair bit of popcorn parts that I'm not worried about
(discretes, etc.).

But, most of the designs I've explored rely on high levels of
integration to bridge the analog-digital divide. Usually,
a digital tuner: high speed A/DC's (typ 60MSPS @ 12-16b),
AGC, digital filters, digital VGAs, etc. It's not just "active"
but actually can be seen as a hard-wired DSP (e.g., look at the
A/DC outputs and digitally adjust the gain of the VGA to
maximize dynamic range).

While different offerings may provide similar functionality,
they are usually completely different beasts.

I can opt for a less integrated approach ("more banal") but
that increases real estate (I am really constrained on space)
and the number of such potential component dependencies.
It's a question of which approach has the potential to screw
one the least! :-/

Once things are in the digital domain, I can easily accommodate
changes. But, getting to that point is the problem (in the
long term)

One of my buddies is the "goto RF guy" so I may just drop this
in his lap (he owes me favors). But, I'd like to be able to
intelligently argue with any initial ideas he presents (so he
doesn't head down a blind alley)

> (Who has about 30k BF862s and CPH3910s on the shelf, among other things such as
> a reel or two each of BFG31 and BFT92 5-GHz PNPs, BFT25A small fast NPNs,
> quantities various pHEMTs and SiGe:C NPNs, and so forth.)

So you've gone for "stockpile everything that you think you might need
that you think might become hard to acquire..." I'd like not to do that
as I don't intend to build the damn things (so why should I incur the
costs of ensuring parts remain available?)

[Now, lets see if I can actually SEND this... :< ]

Don Y wrote:
> Hopefully this post makes it out; I'm using a phone (now THAT was
> a mistake!  I should have brought a small laptop or a tablet!)
> via wifi to remotely access *my* server which, in turn,
> accesses the NNTP service)   <frown>  I can see this should
> have been thought out, better!
>
> On 6/14/2021 5:48 PM, Phil Hobbs wrote:
>
>>> The "issue" I'll be bringing to discuss is design practices
>>> to make largely analog designs "future safe" -- to guard
>>> against component offerings going obsolete, etc.  (I need
>>> at least a 10 year window of continued availability).
>
>>> I don't see anyway to pick devices for longterm availability
>>> other than settling for run-of-the-mill devices that have
>>> work-alikes available from other vendors.  Anything bleeding
>>> edge is bound to lead to issues down the road (availability,
>>> support, etc.)
>>>
>>> So, aim for banal designs?  Do lifetime buys of the components
>>> that you're most exposed to?  Or, just resign yourself to
>>> redesigning when the problem manifests (and hope it's
>>> someone else's problem, then)?
>>
>> Depends.  What's in it?
>
> There's a fair bit of popcorn parts that I'm not worried about
> (discretes, etc.).

Depending on what they are, they might be the hardest to replace. See
"5 GHz PNP" below.

>
> But, most of the designs I've explored rely on high levels of
> integration to bridge the analog-digital divide.  Usually,
> a digital tuner: high speed A/DC's (typ 60MSPS @ 12-16b),
> AGC, digital filters, digital VGAs, etc.  It's not just "active"
> but actually can be seen as a hard-wired DSP (e.g., look at the
> A/DC outputs and digitally adjust the gain of the VGA to
> maximize dynamic range).

Depends a lot on who makes them. If it's Maxim or NXP, yer scrooed.

<snip>
>
>> (Who has about 30k BF862s and CPH3910s on the shelf, among other
>> things such as a reel or two each of BFG31 and BFT92 5-GHz PNPs,
>> BFT25A small fast NPNs, quantities various pHEMTs and SiGe:C NPNs, and
>> so forth.)
>
> So you've gone for "stockpile everything that you think you might need
> that you think might become hard to acquire..."

No, just the ones that offer unique capabilities, enable a wide range of
design space, and are cheap enough to take a flyer on.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

"future proofing"

That's much more common.

About 2,440,000 results compared to about 120 results.

On 6/14/2021 8:18 PM, Phil Hobbs wrote:
>>>> So, aim for banal designs? Do lifetime buys of the components
>>>> that you're most exposed to? Or, just resign yourself to
>>>> redesigning when the problem manifests (and hope it's
>>>> someone else's problem, then)?
>>>
>>> Depends. What's in it?
>>
>> There's a fair bit of popcorn parts that I'm not worried about
>> (discretes, etc.).
>
> Depending on what they are, they might be the hardest to replace. See "5 GHz
> PNP" below.

If I stay in the < 1-1.5GHz region, I think there are COTS "integrated"
options that will be just as easy/difficult to acquire. Most of
the designs I've explored don't have any active *discretes*.
And, I think the passives should be findable from multiple vendors
without too much hairpulling.

Phones make lots of things affordable/possible.

>> But, most of the designs I've explored rely on high levels of
>> integration to bridge the analog-digital divide. Usually,
>> a digital tuner: high speed A/DC's (typ 60MSPS @ 12-16b),
>> AGC, digital filters, digital VGAs, etc. It's not just "active"
>> but actually can be seen as a hard-wired DSP (e.g., look at the
>> A/DC outputs and digitally adjust the gain of the VGA to
>> maximize dynamic range).
>
> Depends a lot on who makes them. If it's Maxim or NXP, yer scrooed.

Ah, I will make a note of that. OTOH, I've been screwed by lots
of different manufacturers over the years. Pick something "exotic"
(or just off-the-beaten-track) and you're likely to find it canceled
long before you're done with it!

>>> (Who has about 30k BF862s and CPH3910s on the shelf, among other things such
>>> as a reel or two each of BFG31 and BFT92 5-GHz PNPs, BFT25A small fast NPNs,
>>> quantities various pHEMTs and SiGe:C NPNs, and so forth.)
>>
>> So you've gone for "stockpile everything that you think you might need
>> that you think might become hard to acquire..."
>
> No, just the ones that offer unique capabilities, enable a wide range of design
> space, and are cheap enough to take a flyer on.

Of course. It would be silly to stockpile popcorn parts. (OTOH, damn near
anything can end up on allocation!)

Well, let's see what happens at breakfast. (and put this frigging phone away!
I don't see how people can lug such things around with them!!!)

Thx!