Expansion board for the CD32

[plasmab]

Originally Posted by kipper2k

for the 3.3v ram, if this had not worked the next step for me would have been to cut the 5v rail feeding the ram and inserting a regular diode with a 0.7v voltage drop across it, that would reduce the stress, i also think that a lot of these chips were prob made when the 3.3v standard was entering the market and companies done a quick check to see if they were 3.3v compatible and relabelled them

Tend to agree but dont use as big a drop on the diode as 0.7v.. 0.3-0.5v is fine. We dont want to have the data lines sitting more than 0.5v above the power rails on a chip that we dont know whether or not it has protection diodes.

Jumper is a bad idea.. people tend to try all the options BOOM

[supaduper]

So made a TF328 with the K4E151612C-TC60 3v rams and flashed it with 15_9 firware that worked with the other board and it would not work after much head scratching checking solder joints blah blah blah, I then started flashing it with the different firmware`s Steve has made us...and blow me it was the one made on the.... 14_9_slow_Ram....that bloody worked !

so as one guy has said in an email to me It could the different manufacturing processes like all chips have ,small timing differences and the like .

Anyway thought I will mention it incase anyone else comes across the same problem...if one Firmware does not work try another

[plasmab]

Originally Posted by supaduper

So made a TF328 with the K4E151612C-TC60 3v rams and flashed it with 15_9 firware that worked with the other board and it would not work after much head scratching checking solder joints blah blah blah, I then started flashing it with the different firmware`s Steve has made us...and blow me it was the one made on the.... 14_9_slow_Ram....that bloody worked !

so as one guy has said in an email to me It could the different manufacturing processes like all chips have ,small timing differences and the like .

Anyway thought I will mention it incase anyone else comes across the same problem...if one Firmware does not work try another

My changes will involve 35ns changes min. I dont think any of those ram chips will have that sort of manufacturing difference.

[plasmab]

I have a large number of TF328 boards arriving tuesday. They'll go straight on eBay.

[supaduper]

Originally Posted by plasmab

My changes will involve 35ns changes min. I dont think any of those ram chips will have that sort of manufacturing difference.

Thanks Steve,
hey buddy I hope you a fantastic holiday mate, now its back to the grey wet skies of the UK

What other ram chips are you going to test bud ?

[alenppc]

Hi guys, these chips seem like a perfect candidate, they are in the right form, 1x16, EDO and 5V!

Look for them on UTSource: AS4C1M16E5-60TC

Also cheaper than the 3V K4E151612C-TC60

[plasmab]

Originally Posted by alenppc

Hi guys, these chips seem like a perfect candidate, they are in the right form, 1x16, EDO and 5V!

Look for them on UTSource: AS4C1M16E5-60TC

Also cheaper than the 3V K4E151612C-TC60

Ordered 8 chips.

[supaduper]

Originally Posted by alenppc

Hi guys, these chips seem like a perfect candidate, they are in the right form, 1x16, EDO and 5V!

Look for them on UTSource: AS4C1M16E5-60TC

Also cheaper than the 3V K4E151612C-TC60

Great find Alen, I ordered some

[Shadowfire]

Directly from the datasheet for the IS41LV16100B-50TLI-TR.

ABSOLUTE MAXIMUM RATINGS(1)
Symbol Parameters                                   Rating        Unit
VT      Voltage on Any Pin Relative to GND 3.3V     –0.5 to +4.6 V
VDD     Supply Voltage 3.3V                         –0.5 to +4.6 V

RECOMMENDED OPERATING CONDITIONS (Voltages are referenced to GND.)
Symbol Parameter                  Min. Typ. Max.      Unit
VDD    Supply Voltage     3.3V    3.0  3.3  3.6       V
VIH    Input High Voltage 3.3V    2.0  —    VDD + 0.3 V 

The IC is clearly only rated for 3.3V operation. There is tolerance up to 4.6V on the supply rail and input pins, but it is explicitly stated that

1. Stress greater than those listed under ABSOLUTE MAXIMUM RATINGS may cause permanent damage to the device. This is a stress rating only and functional operation of the device at these or any other conditions above those indicated in the operational sections of this specification is not implied. Exposure to absolute maximum rating conditions for extended periods may affect reliability.

This means that they won't guarantee that the device works continuously above 3.6V (this is primarily a thermal consideration), and that voltage surges due to power on or signal switching will eventually destroy the device if they rise above 4.6V (this is an upper limit imposed by the construction of the die, typically (but not always) the mosfet gate insulation barrier).

This particular IC is not suited for direct hookup to a 5V bus or power supply. It is going to die, and it isn't a case of if, it's when.

The further limit on VIH of VDD+0.3, and VIL to -0.3V, strongly suggests that there are indeed ESD diodes on the input lines to VDD. You can violate this rating (at your own peril) if you do something to limit the current (typically a series resistor) to a safe level as defined elsewhere in the data sheet; this actually brings the voltage down to VDD+0.3V. This particular part has a 50mA output current maximum, but there is also a 1W package power dissipation maximum which needs to be taken into account.

For instance, if VDD is 3.3V, and you are running a 5V bus, you would apply ohms law to find a series resistance.
The voltage drop across the resistor is 5V - 3.3V (its technically 3.6V but we're building in a 20% safety margin here) = 1.7V drop.
You want the current to be a maximum of 50mA, so R=E/I, R=34 ohms will limit the current to 50mA or less.
BUT you also need to check power dissipation, P=I^2*R, P = 0.085W. You typically would want to double this and select a resistor which can dissipate at least 170mW, which means 0805's (which typically dissipate 125mW) and smaller are out. 1005's are good for 200mW, so now you're looking at putting a 1005 resistor in series with every signal which could potentially be driven to 5V.
Now lets consider the power dissipation through that ESD diode, which is taking away from that 1W power envelope of the IC. P=I*E, = 0.050A*0.3V = 15mW. But, you have to protect 31 I/O lines, which brings the total power potentially dissipated (if every line is switched high at once) to 465mW. The data sheet says that the IC can consume 180mA of current (@3.3V) during active operation, but if you don't want to deal with product return issues, you calculates it at VDDMAX, which is 3.6V * 0.180A = 648mW of power.
648mW IC internal operation + 465mW extra ESD diode dissipation = 1113mW of potential dissipation - this takes all the manufacturer's safety margin, throws it out the window, and drives a car over it. You will need to raise the resistor value if you don't want to use dedicated voltage translators.

But the ugliness doesn't necessarily end there! You now have put a resistance of 34 ohms (possibly higher if you didn't want to jam 50mA of current through each ESD diode) in series with the input capacitance of the pin. Further examination of the datasheet reveals that ALL of the non-address lines have an input capacitance of 7pf. Lets assume that you're using just 1 memory IC, not several wired in parallel. You now have a lowpass filter on any input signals with a 10-90 rise time of approximately 2.2*R*C, or 523.6pS.

This kind of delay is background noise if you're driving the bus at 7mHz, but what if you only want 5mA of current through the diode, have two RAM IC's wired in parallel for a 32-bit bus, and are running a 100mHz state machine to try and minimize the delay?
R=1.7/0.005 = 340 ohms
P(resistor) = 0.005 ^2 * 340 = 8.5mW (ok to use smaller resistor packages now!)
P(esd)=.005*0.3 = 1.5mW, * 31 lines = 46.5mW
P(ic) = 648mW + 46.5mW = 694.5mW, or a 30% safety margin (GOOD! because that IC power dissipation doesn't take into account any power dissipated by the output drivers during a read operation, which is implementation-dependent.)
2.2*R*C(C=14pf, since you have two IC's loading down almost all of the control lines) = 10.47nS RC delay on control/data lines, 7.48nS on the address lines.
Your state machine is flipping pins at 100Mhz, with a tCycle = 10nS... but any inputs received from devices driving the data on the bus are delayed 10.47nS You now need to account for (in addition to PCB induced delays) a full extra clock cycle+ of RC delay on any input lines when trying to determine if you are violating any signal timing.
You also now have your RAM output drivers trying to charge the input capacitance of all the other (non-ram) devices on the bus through a 340 ohm resistor on your data lines, with the same effect, except you add together all those pin input capacitances. Assuming you have another 14pf of capacitance for everything else on that bus, that is an additional 10.47nS delay on outputs reaching the rest of the bus.

You can see that as you raise the clock speed, series resistor protection quickly becomes performance-limiting.

Now lets take one last look at the data sheet. What happens to power dissipation if you run at VDD=5V? P(ic) = 0.180mA * 5V = 900mW. That's right, disregarding the damage you'll do to the IC, 90% of the IC power budget (which is supposed to have a decent safety margin) is taken.

Anyways, the real solution here is, bring VCC down to below 3.6V, and clip bus signals below 3.6V.

[plasmab]

Originally Posted by Shadowfire

Directly from the datasheet for the IS41LV16100B-50TLI-TR.

ABSOLUTE MAXIMUM RATINGS(1)
Symbol Parameters                                   Rating        Unit
VT      Voltage on Any Pin Relative to GND 3.3V     –0.5 to +4.6 V
VDD     Supply Voltage 3.3V                         –0.5 to +4.6 V

RECOMMENDED OPERATING CONDITIONS (Voltages are referenced to GND.)
Symbol Parameter                  Min. Typ. Max.      Unit
VDD    Supply Voltage     3.3V    3.0  3.3  3.6       V
VIH    Input High Voltage 3.3V    2.0  —    VDD + 0.3 V 

The IC is clearly only rated for 3.3V operation. There is tolerance up to 4.6V on the supply rail and input pins, but it is explicitly stated that

1. Stress greater than those listed under ABSOLUTE MAXIMUM RATINGS may cause permanent damage to the device. This is a stress rating only and functional operation of the device at these or any other conditions above those indicated in the operational sections of this specification is not implied. Exposure to absolute maximum rating conditions for extended periods may affect reliability.

This means that they won't guarantee that the device works continuously above 3.6V (this is primarily a thermal consideration), and that voltage surges due to power on or signal switching will eventually destroy the device if they rise above 4.6V (this is an upper limit imposed by the construction of the die, typically (but not always) the mosfet gate insulation barrier).

The further limit on VIH of VDD+0.3, and VIL to -0.3V, strongly suggests that there are indeed ESD diodes on the input lines to VDD.

This particular IC is not suited for direct hookup to a 5V bus or power supply. It is going to die, and it isn't a case of if, it's when.

Whats your point? We know this. we expected failure. We're measuring the rate of failure. The excess voltage causes breakdown across the gates. However its DRAM so the pressure isnt continuous. Only when refresh/read/write happens. SRAM would be worse.

A suitable diode will bring the power rail down to 4.6V (small fluctuations above arent too much of an issue with DRAM) and keep the protection diodes from getting too stressed. However keep in mind that just because *THAT* particular chip may (and i stress MAY - because i've seen stuff documented on products that was pure fantasy) have ESD diodes not all the 3.3V ones we tested will. Hence the BOM should probably spec a 0.3 to 0.5V.

EDIT: I planned to run various 6V RAMs at 6V for a bit and see how quickly that cooked. If you get 100+ hours out of it at that voltage i suspect the documentation may be arse covering.

[plasmab]

Originally Posted by plasmab

Whats your point? We know this. we expected failure. We're measuring the rate of failure. The excess voltage causes breakdown across the gates. However its DRAM so the pressure isnt continuous. Only when refresh/read/write happens. SRAM would be worse.

A suitable diode will bring the power rail down to 4.6V (small fluctuations above arent too much of an issue with DRAM) and keep the protection diodes from getting too stressed. However keep in mind that just because *THAT* particular chip may (and i stress MAY - because i've seen stuff documented on products that was pure fantasy) have ESD diodes not all the 3.3V ones we tested will. Hence the BOM should probably spec a 0.3 to 0.5V.

EDIT: I planned to run various 6V RAMs at 6V for a bit and see how quickly that cooked. If you get 100+ hours out of it at that voltage i suspect the documentation may be arse covering.

EDIT2: the original design brief for the TF328 was no buffers and as simple as possible. If you add 3.3V buffers then you may as well just put SDRAM or SRAM on there too


Sent from my iPhone using Tapatalk

[supaduper]

2x tf328 going to new homes, donated by me to two users for services rendered ,you guys know who you are.
So more people to test these bad boys
Their on the way to you.

[kgc210]

Just want to recap on RAM that works.
So far the cheapest to buy is the K4E151612C-TC60 although it states 3v.
https://www.utsource.net/itm/p/67395...p/6739516.html

Is that correct?

Will be interesting to see if some of that really cheap stuff works. Weren't they about 64cent each or something?

[plasmab]

They will die eventually but at that price its worth investigating.


Sent from my iPhone using Tapatalk

[supaduper]

Originally Posted by plasmab

They will die eventually but at that price its worth investigating.


Sent from my iPhone using Tapatalk

True dat Steve and it only take a short time to take them off and put some more on..I know I`ve done quite on the tf328 loads of times , the board is very well made and tolerant..from Dirtypcs

[plasmab]

I have A LOT of TF328 boards...


Sent from my iPhone using Tapatalk

[cpiac64]

i want a tf328 pcb, if possible

[plasmab]

I'm going to verify one at random then I'll offer them on eBay again. I know eBay is a pain but it lets me track orders so I don't forget anyone


Sent from my iPhone using Tapatalk

[cpiac64]

ok, thanks for info

[FREEMILK]

you guys are awesome