It could be we're there!


It shouldn't, as the heat will always move where it's coldest - and that would be the heatsink.
That's only half the story. It goes everywhere, where it's colder. Dunno, if that's a problem in this scenario, though.
[doublepost=1513691566,1513691379][/doublepost]Iff the SoC is the warmest, it only means you connect it to more sinks - plus with more heatpipe material you get more buffer.
 
I wasn't actually thinking about the heat looping back to the other chips per-say, more about the sinks saturating faster because it also accumulates the heat from the other chips, causing the CPU to throttle faster than without the added heat from the other chips.

Well, if we DON'T do that, then the heat will just stay in that small CPU board area and heat up the CPU as well.
 
Well, if we DON'T do that, then the heat will just stay in that small CPU board area and heat up the CPU as well.
The CPU's surroundings can only heat up the CPU, if they are warmer than the CPU. For that you'd need a heat source warmer than the CPU. Got one?
 
Does it have an effect on the audio system?
Not directly, because it is not an electric noise going into the headset or speaker wires. But unintended tiny buzzers on the mainboard... The noise level is below what you can detect with a typical microphone but loud enough to be annoying if you are alone in some room. Unfortunately the frequency range is so that the human ear has highest sensitivity. But you won't notice in a commuter train :) Or if you unplug all USB devices.
Obviously I want the best Pyra possible, and I really don't think the quantities we're talking about are going to make much of a difference, but I don't plan to use USB much. The Pyra for me is going to be mostly used as a handheld games console. So, the noise is a non-issue for me.
 
Obviously I want the best Pyra possible, and I really don't think the quantities we're talking about are going to make much of a difference, but I don't plan to use USB much. The Pyra for me is going to be mostly used as a handheld games console. So, the noise is a non-issue for me.

For many others it is.
I thought the same until I once plugged in a USB device.
It's loud, really loud. Much louder as I ever expected.

That's only half the story. It goes everywhere, where it's colder. Dunno, if that's a problem in this scenario, though.
[doublepost=1513691566,1513691379][/doublepost]Iff the SoC is the warmest, it only means you connect it to more sinks - plus with more heatpipe material you get more buffer.

We tested it at the Gamescom with PtitSeb's Pyra and the difference was obvious.
And everyone is free to drill some holes into the the plastic if he feels to do so to get the heat out of the Case.
Maybe also use some sort of thermal paste and connect it to an external cooling mechanism.
 
Update time!

Thank you Michael, this is a promising update and an great yuletide gift. (Not as great as a finished Pyra would have been, of course, but very nice nonetheless!)

Here's a summary of what's left to do:

  1. Finish the tests of the mainboards and order them (PCB production takes 4 weeks)
  2. Check if the CPU boards can be further improved and if not, order prototypes for them (PCB production takes 6 weeks).
  3. Check that the cases are fine and if they are, order them (production shouldn't take longer than 2 weeks)
  4. Use the current prototypes for CE / FCC testing (planned for January 2018)
  5. WHILE the bare PCBs are being produced: Check for a whirring noise fix that does not involve using conflict material
  6. WHILE the bare PCBs are being produced: Design the packaging and material
  7. Do a test run with the newly produced PCBs and the transparent cases to produce prototypes for the prototype preorders.
  8. Recheck that everything works fine - then setup the mass production (I'm planning at least 2 - 3 weeks to recheck all of that).
  9. While that final check happens (or directly after the production of the transparent cases), move the molds to the new company to produce the cases for the mass production.

If I remember correctly, the last time you had PCBs produced they took 2 weeks longer (or so) than expected. I was wondering if the estimates above anticipate a similar delay?

In any case, it looks like Item 2 is the bottle-neck here. After the redesign work for the CPU board is finalised, it will be 1.5 months to produce the prototypes, 1 month for board population, prototype build, and shipping prototypes to testers; x months for testing, and 1.5 months for mass production of the boards (assuming no more prototypes are required.)

The last round of testing of the CPU board took at least 7 months. Supposing that this round of testing could be achieved in a day (zero months), rather than 7 months, it'll be a minimum of 4 months from the time the redesign is finalised until final boards are ready for assembly into Pyras. [And a minimum of 11 months if the testing phase drags out to 7 months again.]

Given this bottle-neck, would it not be better if Item 2 were worked on before Item 1. That is, would it not be better if Nikolaus finished the CPU board redesign work so that new CPU board prototypes can be ordered before he works on populating/testing the main board etc. (or anything else). Or am I missing something here?

Worst case scenario would be we'd have to use a few tantalum capacitors. It would still be A LOT less than used in any smartphone, but I'd still like to find a conflict free solution.

What's your opinion on this?

With regards to the noisy capacitors, I have one question and one comment:

Question: How certain are you that the noise is coming from one of the capacitors for the USB port power? And if so, is it the 100 uF cap or the 22 uF one?

Comment: It seems there are two potential approaches to the problem. Either

1) (If it is even possible, ) change the frequency of the ripple in the circuit (maybe by adding a bit of inductance?) so that it is different from the frequency at which the capacitor vibrates;​

or

2) (much easier) use a capacitor with different mechanical properties so that it resonates at a frequency different from that which it experiences in the circuit -- either one of the same physical size but of a different design (i.e. from a different manufacturer), or of a different physical size. Any such mechanical change should prevent the vibration, unless the circuit exhibits a wide range of different ripple frequencies (dependent on load?).​

It seems premature to jump to tantalum capacitors when changing to a different ceramic capacitor (or gang of capacitors) will probably work at least as well.
 
The last round of testing of the CPU board took at least 7 months. Supposing that this round of testing could be achieved in a day (zero months), rather than 7 months, it'll be a minimum of 4 months from the time the redesign is finalised until final boards are ready for assembly into Pyras. [And a minimum of 11 months if the testing phase drags out to 7 months again.]

That's a funny way to say Two Months.
 
You better watch out
You better not cry
Better not pout
I'm telling you why
Pyra 's is coming next year

ED's making a todo list
And checking it twice;
Gonna find out if 4GB works right
Pyra 's is coming next year
 
You better watch out
You better not cry
Better not pout
I'm telling you why
Pyra 's is coming next year

ED's making a todo list
And checking it twice;
Gonna find out if 4GB works right
Pyra 's is coming next year
He sees you when you're posting
He knows that you can't wait
He knows those damn molds will get done
So hang on for two months' sake!

Oh, you better strap up
You better pack heat
This better heat sink will make it complete --
Finished Pyra's coming to town.
 
Some background on the whining, audio and tantalum.

Unfortunately the frequency range is so that the human ear has highest sensitivity.
So the "several caps in parallel" will avoid the noise?
Or is there a way to change the frequency of the whining (to over the human hearable noise, like 40 KHz?) os is that what several cpas in parallel do?
 
Maybe try to put 2 caps not parallel, but in serial, then the voltage they use is just the half -> also half load, maybe no sirring anymore?
But this needs for both caps the double capacity to reach the same capacity than before. For using the existing caps use 2 serial and 2 parallel to get the same capacity.
As its cheap to test maybe give a try.
I boringly have no sirring electronics here for testing.
 
It looks like avoiding tantalum caps is not only the matter of conflict materials.
Recently I picked up another dead 8-bit micro (this time something I have no idea what it is, it's Z80 powered and has MS BASIC). Cleaned, because it was stored in high humidity conditions (read: probably a week in the lake). Desoldered and verified 3 random electrolytic caps - surprisingly OK. Verified current consumption - OK. Power on, after about minute of diagnosing I heard a series of loud bangs and all 4 tantalum caps installed in 5V and 12V power lines literally exploded - not pressurized like electrolytic but exploded, with a little fire from the mainboard.
I don't think we'll ever have 12 amps of current in Pyra (what caused my caps to burst), but minimizing their usage is a good step.
 
I don't think we'll ever have 12 amps of current in Pyra (what caused my caps to burst)
Limiting the current actually makes them repair themselves, the situation only escalated if the damaged spot heats up to over 450°C.

A wet tantalum cap heals itself without ever reaching any critical condition - but 50€ per cap is a bit expensive...
 
That's it, I'm out, cancelling my order. In this house we obey the laws of thermodynamics! Entropy shall always increase, never decrease, and that's final!
Thought, you could decrease entropy locally, if you bring in some energy from outside.
 
It's nice to see that an end is in sight!
Although the sound issue is defiantly something that needs to be fixed, it will drive me nuts
I already had experienced a noisy power supply with my Think pad charger replacement & I ended up buying a new (stock) charger to fix this problem.
So I don't need no noisy Pyra! :D
 
Last edited:
Back
Top