Why would modern digital scopes only offer 2.5kpts memory depth?

[edavid]

Has Tektronix ever said that they still use a FISO architecture?  Who would they get to fab the chips?  No one has a non-sensor CCD process any more.

If they aren't then there is absolutely no reason to have the exact same 2.5KB memory as used 20 years ago.
Rumor has it have have huge stockpiles of those old chips.

So you are saying that they use 15 year old chips in new scopes?  That doesn't really add up either.

[aviphysics]

Low end ought to be important, because it will play a role in what students get trained on. When students grow up and file requisition forms for new scopes, I bet a lot of them are buying the brand they used as students.

I also bet that organisations ordering lots of units pay the list price.

[EEVblog]

So you are saying that they use 15 year old chips in new scopes?

Nope, more likely whenever the last buy was. Doesn't have to be 15 years ago.
The problem with verification is that no one is dumb enough to buy one and tear it down 
Happy to be proven wrong of course, but in either case Tek look like fools. Either using 20yo tech in their latest scope, or designing new hardware and only giving it the same 2.5K memory.
The thing is that 2.5K value, it just does not make sense using any modern design technology, you'd at least round it to 4K for example.

[mikeselectricstuff]

Has Tektronix ever said that they still use a FISO architecture?  Who would they get to fab the chips?  No one has a non-sensor CCD process any more.

If they aren't then there is absolutely no reason to have the exact same 2.5KB memory as used 20 years ago.
Rumor has it have have huge stockpiles of those old chips.

So you are saying that they use 15 year old chips in new scopes?  That doesn't really add up either.

Why not ? If the process was going obsolete it is entirely reasonable they'd do enough wafers for foreseeable future production and then a few more for luck.
Remember that a lot of the cost of a chip is packaging and test, and they only need to do this once they  actually need the chips for production.

[nfmax]

Just a guess, but maybe, when Tek sold their fab (to Maxim, was it?) part of the deal was a commitment to buy some large, over-optimistic number of the chip in question?

[nctnico]

So you are saying that they use 15 year old chips in new scopes?

Nope, more likely whenever the last buy was. Doesn't have to be 15 years ago.
The problem with verification is that no one is dumb enough to buy one and tear it down 
Happy to be proven wrong of course, but in either case Tek look like fools. Either using 20yo tech in their latest scope, or designing new hardware and only giving it the same 2.5K memory.
The thing is that 2.5K value, it just does not make sense using any modern design technology, you'd at least round it to 4K for example.

I also find it hard to believe Tek still uses a 20 year old chip in a new scope. Even if the chips are 10 years old they are not getting better in storage. I'm more inclined to believe the 2.5k memory is a limit which is tied into the software. In a DSO the software development costs are easely 50 times higher than the cost of developing hardware so it pays off quickly to keep the software as it is.

[mikeselectricstuff]

I also find it hard to believe Tek still uses a 20 year old chip in a new scope. Even if the chips are 10 years old they are not getting better in storage.

Like I said, they probably don't store complete chips, with issues like solderability degredation, they store wafers, and package & test them  as required.

[coppice]

I also find it hard to believe Tek still uses a 20 year old chip in a new scope. Even if the chips are 10 years old they are not getting better in storage.

Like I said, they probably don't store complete chips, with issues like solderability degredation, they store wafers, and package & test them  as required.

How long can you store bare die before they start to degrade? The question doesn't often come up, as people don't usually want to store them this long.

[Whales]

How long can you store bare die before they start to degrade? The question doesn't often come up, as people don't usually want to store them this long.

 It would be interesting to see how delicate they are to age.  Chuck them in an esky with a nitrogen atmosphere? Does anyone know anything about end-wafer storage?

[coppice]

How long can you store bare die before they start to degrade? The question doesn't often come up, as people don't usually want to store them this long.

 It would be interesting to see how delicate they are to age.  Chuck them in an esky with a nitrogen atmosphere? Does anyone know anything about end-wafer storage?

http://www.micross.com specialise is die banking, and claim "Bare die products can be safely stored for many years" with suitable storage in an inert atmosphere, but just how big is "many"?

[ivan747]

That sounds like a question for Vincent a.k.a free-electron.

[EEVblog]

I also find it hard to believe Tek still uses a 20 year old chip in a new scope.

They wouldn't be 20 year chips, but the design dates back to the original TDS200 series from 20 years ago.
IIRC is was the TDS200 series, then the TDS1000 series, then TDS1000B, then TDS1000C, and now this TBS1000, all with the same 2.5KB of memory.
There might have even been revision of the chip somewhere along the line, but it still retained the internal fixed 2.5KB of sample memory.
The TDS1000/B/C series scope are not that old in the scheme of things, so they must have been producing them until they released the TBS1000 at the end of 2012.

I'm more inclined to believe the 2.5k memory is a limit which is tied into the software.

Almost certainly not. The TDS200 series very famously had 2.5KB of sample memory in the ASIC, it is a pure hardware limit, just like agilent have their sample memory in their Megazoom ASIC's and they can't increase it.

In a DSO the software development costs are easely 50 times higher than the cost of developing hardware so it pays off quickly to keep the software as it is.

All evidence points toward them very substantially upgrading the processor and software O/S in the scope. From the 68000 in the old TDS200 to something new that can handle all the new features in the TBS, which are actually pretty extensive. Colour display, mask testing, educational info etc.
It seems inconceivable that the 2.5KB limit is now a software limitaton when it's always famously been a front end sampling ASIC hardware limitation with the TDS200/1000 series scopes.

[philpem]

All evidence points toward them very substantially upgrading the processor and software O/S in the scope. From the 68000 in the old TDS200 to something new that can handle all the new features in the TBS, which are actually pretty extensive. Colour display, mask testing, educational info etc.
It seems inconceivable that the 2.5KB limit is now a software limitaton when it's always famously been a front end sampling ASIC hardware limitation with the TDS200/1000 series scopes.

I've been inside a TDS2000B - it's still a 68K CPU. Can't remember which one, though, sadly.

Typical Danaher, milk it for all it's worth, and then milk it some more, watch it kick the bucket, then flog it a wee bit...