Let's move the moulds!


Quick update:
I'll be in Greece Monday and Tuesday, 13th and 14th of November.
Please post pics and/or news.



This. There is no improvement for the 2GB by fixing the 4GB.

It is like building a bridge that can support a million tons when it can already support 200 tons, and will actually only see 100 tons.
Which bridge will last longer? Will the 200 ton bridge still be safe for 100 tons in 500 years?
 
This. There is no improvement for the 2GB by fixing the 4GB.

It is like building a bridge that can support a million tons when it can already support 200 tons, and will actually only see 100 tons.
Its more like:
The Bridge support 100.5 tons and you want to drive 100 tons on it.
Sure, it supports it now, but will the bridge support the 100tons in 5 years when the first parts start to rust a little bit?
Will you be able to add an additional layer of coating?

What happens when the wind starts to blow?
(Google for "dancing bridge")

Its like the new car model sets on fire as soon you hit 88.8mp/h.
But you don't fix it because the maximum allowed speed is 80mp/h

Lets wait what ED is writing next.
All this theoretical discussions does not help us.
 
What happens when the wind starts to blow?
That was what I was originally thinking, then my mind went to Ancient Rome for some reason.
[doublepost=1509960008,1509959944][/doublepost]Maybe it was the Greek influence.
 
Its more like:
The Bridge support 100.5 tons and you want to drive 100 tons on it.
Sure, it supports it now, but will the bridge support the 100tons in 5 years when the first parts start to rust a little bit?
Will you be able to add an additional layer of coating?

What happens when the wind starts to blow?
(Google for "dancing bridge")

Its like the new car model sets on fire as soon you hit 88.8mp/h.
But you don't fix it because the maximum allowed speed is 80mp/h

Lets wait what ED is writing next.
All this theoretical discussions does not help us.

The problem with this type of thinking is there is no "wind" using the 2GB pyra.

But like you said, doesn't matter until ED makes a call
 
Tolerances of components are things which need to be considered. A few months ago I was repairing another old 8-bit computer. After extensive testing and finding bits just falling off nowhere into the bus, I decided to check every TTL possible (around 70 sockets in). In external tester, everything was perfect.
I decided to replace all suspicious ICs with new ones - computer started properly and hanged a moment later. So, oscilloscope in hand and let's trace the failing bits. I found that one flip-flop developed a significant delay, and one passive circuit was not introducing needed delay to bus switch (8286). Maybe because of capacitance, flip-flop held its output state for a bit longer, what was totally needed to put the bit to devices in the switched bus.
Some circuits tend to drift in timings, but these timings usually are still in tolerances... Until two parts won't start to bend these tolerance regimes in opposite directions.
 
The problem with this type of thinking is there is no "wind" using the 2GB pyra.

But like you said, doesn't matter until ED makes a call
Do you knew the terminology "process tolerances" and "process safety"?
It means that your prototype works fine, but when you start mass producing, the devices start failing.

Its because of process tolerances.
Non process is able to produce exact identical parts.
There is only a tolerance area within the parts get produced.
And the tolerances of all parts add up.

The art of producing products is to use manufacture all parts the way that the sum of parts stay within the tolerance area.
The funny thing is, that the characteristic values changes with time and temperature.

Traditional its considered not that optimal when the prototype operates at the tolerance borders because of this effects.
The experience showed that the tolerance-combinations in mass production tends to go the worse side.
 
And the tolerances of all parts add up.
...
The experience showed that the tolerance-combinations in mass production tends to go the worse side.

When you sum up a bunch of random numbers, you get a normal distribution or "bell curve."

The majority will be near the median, within tolerances, but a significant minority will be outside that area.

upload_2017-11-7_9-39-45.png


Statistically, it's virtually certain that some units will be defective.
 
That's why, even if the 2gb are deemed stable, improvement would also be welcomed for them... Since there's no mass production we have no idea how stable they really are... Will 0.01% or will 5% be unstable? Hopefully the 2gb will be fine as is.
 
Well, that sucks. Hopefully it's not TOO many.
It is a truth of manufacturing. Some items will be defective. However, in electronics, the vast majority of defective units can be caught during testing as part of the procedure to load the OS onto the eMMC. A simple test suite to ensure that the components are all operation will catch the vast majority of 'hard defects'. On a board population basis, this may be 1-5% of those created - this is normal and expected. The company doing the pick & place will typically guarantee a given success rate - so that can be baked into the costs.

There will also be a small segment of 'soft defects' where the system tests appropriately prior to shipping but develops an issue later. That is what RMA is for. It is good to minimize this, but it is expected.

Neither is a particularly 'big deal'.
 
Tolerances of components are things which need to be considered. A few months ago I was repairing another old 8-bit computer. After extensive testing and finding bits just falling off nowhere into the bus, I decided to check every TTL possible (around 70 sockets in). In external tester, everything was perfect.
I decided to replace all suspicious ICs with new ones - computer started properly and hanged a moment later. So, oscilloscope in hand and let's trace the failing bits. I found that one flip-flop developed a significant delay, and one passive circuit was not introducing needed delay to bus switch (8286). Maybe because of capacitance, flip-flop held its output state for a bit longer, what was totally needed to put the bit to devices in the switched bus.
Some circuits tend to drift in timings, but these timings usually are still in tolerances... Until two parts won't start to bend these tolerance regimes in opposite directions.

Here is the thing: how old is that eight bit computer?

Right now I daily drive my eeepc 701, and Pandora. Both are 10+ years old I reckon your 8 bit computer is 20+ years old.

Thing is, the pyra build quality and tolerances are already better than the two 10 year old computers when they were new... And they are perfectly fine.
 
Last edited:
but then again, there are more components on these new fangled devices, and therefore more that can fail, too :).
 
Traditional its considered not that optimal when the prototype operates at the tolerance borders because of this effects.
The experience showed that the tolerance-combinations in mass production tends to go the worse side.

All prototypes worked with 2GB, the board was designed for 2GB. The load is doubled on something not designed for that(4GB) and it is STILL almost there and people think that "isn't enough," and "close to tolerances," for the original 2GB design... When it is adequate, again. A million ton capacity bridge for that 100ton load
 
Last edited:
Right now I daily drive my eeepc 701, and Pandora. Both are 10+ years old I reckon your 8 bit computer is 20+ years old.

You've had your Pandora since 2007? Nice work.

I'm currently rocking my eepc 901 which is 7-8 years old, and one of the PATA controllers has burnt out after I used a Runcore 64GB mini-PATA card in there for about 5-6 years, and then about a year ago the green ring around the 3.5mm headphone socket started to crack and fall out, and then the sound completely stopped working (even took out the internal speakers). It's all workroundable though - I bought a <5$ USB sound card, and a 128GB SD card and I have the machine upgraded by most metrics, and working fine again. If anything, its performance has improved in the time I've owned it - when I first got it, sleep was broken, and the internal SD card reader would regularly eat cards and the wifi was ropey, but that's all better now thanks to driver improvements in linux.
 
If you develop consumer electronics it's always a risk versus reward thing ... rushing is bad and over engineering is known to happen too, good arguments all around here


However as ED carries all the risks he is the one who has to be comfortable ... his risk, his choice and everyone else please be patient :-#, don't stress the dragon ;)


The Gighartz realm has it's own rules, for those interested I recommend the classic books :

"High Speed Digital Design: A Handbook of Black Magic"
"High Speed Signal Propagation: Advanced Black Magic"

The titles speak for themselves there ...


Also is you want to sell consumer electronics you have to have it meet the regulations of the country you want to sell them in which means EMC testing : https://www.emcfastpass.com/emc-testing-beginners-guide/

That means even when the Pyra is working, when it fails those tests another round of redesigns must happen (let's hope not) but there is a whole can of legal worms attached to manufacturing and selling stuff people often do not know about (and thats one of the reason why kickstarter projects have trouble with customs all the time) ... luckily those do not apply to stuff sold in electronic kit form so ED can always sell "Build your own Pyra" kits to us :D
 
Luckily ED has done this before and is intimately familiar with the process of bringing a handheld electronic device to market. Also, luckily, many of us have done this preorder and wait thing before and are intimately familiar with the process of bringing a handheld electronic device to market. Thanks to ED being so open and forthcoming with details.

It's a long, slow process full of challenges and whiners. I know ED will ship eventually and I'm in no hurry. I also know that for many many years after I get a Pyra I will still be using it every day, just like my trusty Pandora here.

Bottom line is this, if you can't wait however long it takes save yourself a ton of grief and get a competitor. I hear the GPD Win is nicely mediocre, if you can stand that OS. I'll just wait however long for the Pyra because it is exactly what I want.
 
Back
Top