Moonsound SRAM

Door meits

Scribe (6530)

afbeelding van meits

02-06-2010, 17:17

Last night I found out MBWAVE could crash on my MSX with Moonsound...

So I decided to do a test:
I let the MSX play the same song all night long til next morning with a wavekit with custom samples.
I woke up and checked if it was still playing. It wasn't. The MSX was frozen...

Then I restarted the MSX, booted MBWAVE and loaded the same song, but this time without wavekit.
I went to my job and when I came home, like 9,5 hours later, the MSX was still playing the song...

Could my SRAM have gotten unstable?

My config is:
Sony HB-F700P
Gouda Slotexpander 5.0 with from back to front: Philips music module, Pana FM-Pac, Moonsound (1x128kB, 1x512kB), Padial keyboardinterface.

Remarkable thing was that MBWAVE 1.16 froze on the same place all the time with an Illegal transpose error message while MBWAVE 1.04 froze on a random place, mostly after the song had looped one or two times...

Anyone who can tell me what is wrong? Maybe the SRAM?

Aangemeld of registreer om reacties te plaatsen

Van Leo

Paragon (1236)

afbeelding van Leo

02-06-2010, 22:31

maybe a test on someone else 's card to double check ?

Van evulopah

Paladin (669)

afbeelding van evulopah

02-06-2010, 23:49

I think, mister Meits, you should keep your cat (mem Bunzing?) out @ night and do the test again Tongue

Van flyguille

Prophet (3031)

afbeelding van flyguille

03-06-2010, 04:44

most probably, a electric power fluctuation, just enought to tilt but not for a full reset.

Van meits

Scribe (6530)

afbeelding van meits

03-06-2010, 07:00

Last night I let it play again with just 128kB in it (the extra 512kB popped out)... It still plays...

It most probably isn't a power fluctuation, cuz it fluctuate more often with MBWAVE 1.16 and exactly on the same time...

I suspect my 6 year old extra 512kB SRAM...

BUT... Is that plausible for the hardware people?

Will double check soon with another moonsound...

Van Leo

Paragon (1236)

afbeelding van Leo

03-06-2010, 08:14

i d say it is more a temperature change that cause the failure of the sram.
could be from the sound chip as well , the upper address lines being used or not
depending on youe sram config.

maybe if you let the sametest but with better air cooling : like do the test removing the case of the moonsound and leaving some space around .

have you tried many songs or just one ?

Van meits

Scribe (6530)

afbeelding van meits

03-06-2010, 16:59

Temperature you say? I've never heard that... Not even experienced any oddities on extreme hot days with the pcb in its case...
I could give it a try with a naked moonsound in the slotexpander though...

I have just tried one song... And the peculiar thing is that with mbwave 1.16 the whole shebang crashes on the same place all the time... That can not be a temperature issue if you ask me... Just something bugging with something else...

Van Leo

Paragon (1236)

afbeelding van Leo

03-06-2010, 18:24

temperature is a typical stress for ic's in genral.
delays in ic's are made of currents flowing through a resistance that charges a capacitance.
so you have seconds = ohms * farads.
also ohms is a function of the square of temperature.
so each time you double the temperature ( from 25 to 50oC lets say ) your delays are 4x times longer.

normally commercial ic's can be used up to 85oC or 105oC , for junction temperature.
that means you can have 30oC outside package but much more inside chip.

As circuits get old they tend to stand less and less temperature stress, they are getting older.

So the temperature stress can reveal weakness of the circuit , the fact that it fails on same location in songs is compatible with temperature stress , the weak point is always the same to fail first.

Van RetroTechie

Paragon (1563)

afbeelding van RetroTechie

03-06-2010, 18:55

It's always possible a memory IC fails although in my experience, SRAMs rarely do... btw. are you sure it's not simply a loose contact on IC socket, or something like that?

What I don't understand is why a Moonsound would hang because of that, does the data in its RAM need to be valid in any way? Why would it get stuck even if that RAM would return random data? Or does that RAMchip somehow output its data directly to MSX databus? Question

delays in ic's are made of currents flowing through a resistance that charges a capacitance.
so you have seconds = ohms * farads.
also ohms is a function of the square of temperature.
so each time you double the temperature ( from 25 to 50oC lets say ) your delays are 4x times longer.

In Wikipedia terms: "citation needed!". Perhaps if you're talking about a single, 'ideal' part and absolute temperature (where the scale starts at around -273 Celsius), but for real-world IC's things are a lot more complicated (read: "so each time you double the temperature ( from 25 to 50oC lets say ) your delays are 4x times longer" is nonsense, IC's don't get 16 times times slower when you go from 25 to 100 degrees Celsius). And also depends on supply voltage, or what's connected to IC outputs. A datasheet usually has min/typical/max values for a specified temperature range.

You are right that IC's generally become slower at higher temperature - therefore overclocking, cooling mods & stress testing is best done in the midst of summer. LOL!

Van Leo

Paragon (1236)

afbeelding van Leo

03-06-2010, 21:27

i tried to be simple in my explanation , just to put emphasis that temperature can have lot of impact all other parameters being equal ( voltage , loading , ...)

of course this is about wire delay only.
ic 's which delays are made of transistor commutation and wire delay and loads + pvt conditions + blabla : some people do their phd on this so i cant be exact in 2 lines , i try just to give practical information that is to say "try to cool the device as a test".

a hair dryer is also helpfull in this case if the failure is related to temperature you may recreate conditions in 5mins : it is done like this even in big companies with $$$

...

maybe the error is just software like there is some error accumulation on the tempo with roundings or on sample length : usually this can be detected before the hard failure with sound distorsion measurement

Van dhau

Paragon (1570)

afbeelding van dhau

04-06-2010, 03:10

Thicker wires for +5V and GND (in addition to PCB traces if needed) and make sure decoupling capacitor is soldered directly to +5V and GND pins. If that doesn't help, see if your MSX outputs proper +5V. May be Moonsound with extra RAM chip draws too much current and the voltage drop is too big.