...

[cleaningOut]

...

[ebclr]

He accept offers also

[technix]

Seems so. Those are "leftover stock" carrying a 1309 date code, so it is likely that those chips are long past their expiry date and are forced to write off.

If you have the right tools for the BGA you can buy some, bake them, and build a DIY experimental board from them. You will need at least some DRAM for it though.

[technix]

So they're almost four years old... wow. I wonder where this guy got them.

The original company that bought those chips probably dumped them as they wrote them off as expired. Whoever received the dump decided to treat them as NOS and resell them.

This is the same reason why there was a time when Xeon E5-2670 and E5-2680 processors was dirt cheap - Facebook and Amazon decommissioned datacenters worth of servers and dumped them to e-waste processors. The e-waste processors treated the processors and memory sticks that are standard products as perfectly resalable second hand and dumped them to the second hand market, causing the street price of the chips to plummet. I grabbed two of them at the cost of 1/3 of a single E5-2620v2 that was used in the same server.

[andyturk]

How would you go about baking these?

[KE5FX]

build a DIY experimental board from them.

Easier said than done... these are 0.8mm pitch BGAs! Any boards you build to use these are going to be wildly expensive in low quantities.

Meh.

[technix]

It's just 0.8mm BGA, not that hard. 0.5mm is like a bitch, and 0.4mm is much harder.

For 0.8mm, you can still use dog bone fanout.

Sure, but going from 4 to 6 layers quadruples the price of the PCB. Any fabs you recommend?

I don't know if Zynq requires 32-bit DDRIII, but I've done 16-bit DDRIII for iMX6ULL, and that's doable with 4-layer, 0.2mm drill and 0.1mm track process without any advanced tech such as blind/buried vias, HDI and so on.
Top: trace
M1: GND/PWR
M2: GND
Bot: trace

It can do 8-, 16-, 18- and 32-bit DDR3. (18-bit DDR3 is 16-bit DDR3 with ECC.)

[eeviking]

Hi guys. I am the ebay seller of the Zynqs.
As technix have correctly guessed they are legit and were rescued from going to the dump due to expired shelf life.
If you want me to change the description I will be happy to do that.
They are MSL3 and should be baked before reflow. (The standard says 9h at 125C, 33h at 90C or 13days at 40C <5%RH)
I will try to ship air tight with silica gel, but I do not have access to vacuum pack.

If anyone have a simple open source board layout for them it would be nice ?

[Sal Ammoniac]

The original company that bought those chips probably dumped them as they wrote them off as expired. Whoever received the dump decided to treat them as NOS and resell them.

Expired? How the hell do chips "expire"?

[tszaboo]

How many do you have? If it is in the 100s it might be even worth to design something for it. Like couple it with DDR and some flash, route the rest to castellations. Selling that many otherwise might be hard.

[technix]

The original company that bought those chips probably dumped them as they wrote them off as expired. Whoever received the dump decided to treat them as NOS and resell them.

Expired? How the hell do chips "expire"?

Unsoldered chips have an expiry date after which the pins may become too oxidized to solder using normal thermal profile, flux and solder paste. Either the chips have to be reballed/reworked or new chips have to be bought. Most company would not bother hiring someone to rework the chips, so they write them off as expired and buy new chips.

For us tinkerers the reballing is a price to pay for buying the chips on the cheap, but with an hour or two of our work we can make use of a super expensive chip without actually paying a lot of the money.

[technix]

How many do you have? If it is in the 100s it might be even worth to design something for it. Like couple it with DDR and some flash, route the rest to castellations. Selling that many otherwise might be hard.

I would definitely like to buy a few of that. Oh you also need to put the PMIC on the module too.

Also you may want to provide an evaluation board for the castellated module so people can get one and start playing with it. For such modules the Raspberry Pi pinout may be a good idea.

[NorthGuy]

If anyone have a simple open source board layout for them it would be nice ?

It's about $500, if not more, to print a minimum amount of 6-layer ENIG boards with 4mil traces (5mil might be Ok though) and 8mil drills. This will kill all the savings from buying your chips.

[technix]

Hi guys. I am the ebay seller of the Zynqs.
As technix have correctly guessed they are legit and were rescued from going to the dump due to expired shelf life.
If you want me to change the description I will be happy to do that.
They are MSL3 and should be baked before reflow. (The standard says 9h at 125C, 33h at 90C or 13days at 40C <5%RH)
I will try to ship air tight with silica gel, but I do not have access to vacuum pack.

For now store them in a dry hot room with bags of rice surrounding them. The rice grains are very hydrophilic and can pull the moisture out of the chips.

If anyone have a simple open source board layout for them it would be nice ?

Is there anything else you have rescued from the dumpster? I do like to stock up some FPGA in QFP packages. (hand soldering does not mind the moisture level, and I can use some sand paper to quickly dispatch the corrosion.)

[eeviking]

How many do you have? If it is in the 100s it might be even worth to design something for it. Like couple it with DDR and some flash, route the rest to castellations. Selling that many otherwise might be hard.

I only have about 50 left.

[eeviking]

The original company that bought those chips probably dumped them as they wrote them off as expired. Whoever received the dump decided to treat them as NOS and resell them.

Expired? How the hell do chips "expire"?

Unsoldered chips have an expiry date after which the pins may become too oxidized to solder using normal thermal profile, flux and solder paste. Either the chips have to be reballed/reworked or new chips have to be bought. Most company would not bother hiring someone to rework the chips, so they write them off as expired and buy new chips.

For us tinkerers the reballing is a price to pay for buying the chips on the cheap, but with an hour or two of our work we can make use of a super expensive chip without actually paying a lot of the money.

I don't think the solderability will be a problem, they have been stored at <5%RH
The expiration date is set to ensure 100% yield witch is important if you produce stuff like offshore electronics and you don't want to take any risks. For hobby/prototype use it's not a problem.
They are stored in sealed bag with silica gel packs. I think they are better than rice 

[technix]

The original company that bought those chips probably dumped them as they wrote them off as expired. Whoever received the dump decided to treat them as NOS and resell them.

Expired? How the hell do chips "expire"?

Unsoldered chips have an expiry date after which the pins may become too oxidized to solder using normal thermal profile, flux and solder paste. Either the chips have to be reballed/reworked or new chips have to be bought. Most company would not bother hiring someone to rework the chips, so they write them off as expired and buy new chips.

For us tinkerers the reballing is a price to pay for buying the chips on the cheap, but with an hour or two of our work we can make use of a super expensive chip without actually paying a lot of the money.

I don't think the solderability will be a problem, they have been stored at <5%RH
The expiration date is set to ensure 100% yield witch is important if you produce stuff like offshore electronics and you don't want to take any risks. For hobby/prototype use it's not a problem.
They are stored in sealed bag with silica gel packs. I think they are better than rice 

It is the oxygen in the air that killed the solderability, not the humidity.

Big bags of rice would work much better than a small packet of silica gel.

[eeviking]

The original company that bought those chips probably dumped them as they wrote them off as expired. Whoever received the dump decided to treat them as NOS and resell them.

Expired? How the hell do chips "expire"?

Unsoldered chips have an expiry date after which the pins may become too oxidized to solder using normal thermal profile, flux and solder paste. Either the chips have to be reballed/reworked or new chips have to be bought. Most company would not bother hiring someone to rework the chips, so they write them off as expired and buy new chips.

For us tinkerers the reballing is a price to pay for buying the chips on the cheap, but with an hour or two of our work we can make use of a super expensive chip without actually paying a lot of the money.

I don't think the solderability will be a problem, they have been stored at <5%RH
The expiration date is set to ensure 100% yield witch is important if you produce stuff like offshore electronics and you don't want to take any risks. For hobby/prototype use it's not a problem.
They are stored in sealed bag with silica gel packs. I think they are better than rice 

It is the oxygen in the air that killed the solderability, not the humidity.

Big bags of rice would work much better than a small packet of silica gel.

Yes, but high temperature and humidity will accelerate the oxidation.

I use big bags 

[technix]

The original company that bought those chips probably dumped them as they wrote them off as expired. Whoever received the dump decided to treat them as NOS and resell them.

Expired? How the hell do chips "expire"?

Unsoldered chips have an expiry date after which the pins may become too oxidized to solder using normal thermal profile, flux and solder paste. Either the chips have to be reballed/reworked or new chips have to be bought. Most company would not bother hiring someone to rework the chips, so they write them off as expired and buy new chips.

For us tinkerers the reballing is a price to pay for buying the chips on the cheap, but with an hour or two of our work we can make use of a super expensive chip without actually paying a lot of the money.

I don't think the solderability will be a problem, they have been stored at <5%RH
The expiration date is set to ensure 100% yield witch is important if you produce stuff like offshore electronics and you don't want to take any risks. For hobby/prototype use it's not a problem.
They are stored in sealed bag with silica gel packs. I think they are better than rice 

It is the oxygen in the air that killed the solderability, not the humidity.

Big bags of rice would work much better than a small packet of silica gel.

Yes, but high temperature and humidity will accelerate the oxidation.

I use big bags 

Even in cool dry air the oxidation will ha;pen, albeit much slower. But those chips have been sitting on a shelf for more than 4 years...

Have you rescued other chips from the dumpster?

[KE5FX]

Even in cool dry air the oxidation will ha;pen, albeit much slower. But those chips have been sitting on a shelf for more than 4 years...

This kind of thing matters in ISO-compliant processes where procedures must be documented in detail and perfectly reproducible, where the difference between five-sigma quality and six-sigma matters.  Or if you're sending hardware to the Kuiper Belt.  For every other application, from prototyping to limited production runs, they'll be OK.

[Koen]

AllPCB have a special on leaded HASL 6-layers at 70 USD for 10 boards of 100x100.

[NorthGuy]

AllPCB have a special on leaded HASL 6-layers at 70 USD for 10 boards of 100x100.

You cannot really fan it out with 6mil wires and 12mil holes. They don't publish a separate "hole to copper" specs, but assuming it's the same as with annular ring: 12mil hole + 2*6mil annular ring + 2*6mil gaps + 6mil trace = 42mil, but the distance between ball centers is only 31 mil. So you only will be able to do the outer row of balls (perhaps the second row if you undersize ball pads). I don't think HASL would work neither.

[eeviking]

Did anyone get boards made with these?
Any problems with solderability?

I still have some left I plan to dump for a good price in the buy/sell section.

[asmi]

I've been thinking about designing a board for this SoC. Recently I found a fab that quoted me $234 for 10 6-layer boards with 0.1/0.1 mm traces and 0.2 mm drills and CI (and that price includes DHL delivery to my door), but I would like to see them deliver before I can formulate my opinion. If these parts would still be available at the time, I will definitely buy several - don't mind baking them at all.

[Tomorokoshi]

Or if you're sending hardware to the Kuiper Belt.

Now that sounds like an interesting kickstarter campaign!

[asmi]

I'm having a strange issue with the schematic for wiring up an 8-bit DDR3 to this chip, since I want to stay at 4 layers. The 512x8 DDR3 chip has 16 address bits [15:0], but the 256x16 DDR3s have only 15 [14:0], and apparently the Zynq bank 502 doesn't have a 16th address bit either. I can't find any documentation on what to do with the DDR's A15 pin - any suggestions?

Options are to let it float or tie to 1V5/ground.

If you don't want to use smaller density ICs with 15 address lines (I personally would've chosen that option), you can tie unused address line high or low (it doesn't really matter either way), but don't let it float! Keep in mind that different density ICs have different refresh times, so make sure you specify correct parameters when setting up DDR3 controller.

Also - I would love to see how you managed to breakout DDR3 lines using only 2 layers (as other two will need to be reference planes to maintain signal integrity). DDR3 traces NEED to have UNBROKEN reference plane along their entire length. This is even more so if you want to use DDR3L as it has even smaller margin than regular DDR3. Even with 0.1 mm traces and 0.2 mm drills process that would not be easy as Zynq is in 0.8mm pitch BGA package so you can't fit two traces between balls/vias.

[asmi]

I should have realized the lack of a 16th address bit means I can only access half of the actual depth (0-256 of 512M), so there's no benefit to using the DDR3-x8 ICs. I considered using DDR3L, but minimizing voltage bus count feels more useful than reducing cost/power. On this board, at least.

Make sure your 1.5 V rail can provide enough current - DDR3 can have huge transient currents (in the order of few Amps!), and if you PMIC (or power source) isn't quite up to task, you could get hard-to-debug issues caused by 1V5 rail going places. Also Xilinx recommends using termination to Vtt, so you will need tracking regulator for that. Read Ch.5 of UG933 very carefully and do as they say. Unless you are willing to experiment (read - got money to burn on possible board respins) that is.

I'm not even at the stage where I can work on the layout yet - still have to get an Ethernet PHY into the schematic. I may livestream the layout towards the end of this week if anyone's interested.

I doubt I can make it on the livestream (as my spare time is always sketchy), but would love to watch full-length recording at my own schedule. Don't mind 5+ hrs long YT videos at all.
But like I said, I seriously doubt you would be able to make it work on 4 layers. It least I couldn't make even much simpler Artix-7 + DDR3L x8 work on 4 layers without violating DDR3 layout requirements.
My first 6-layer boards are now on their way to me - looking forward to get my hands on them

[ThomasDK]

I'm having a strange issue with the schematic for wiring up an 8-bit DDR3 to this chip, since I want to stay at 4 layers. The 512x8 DDR3 chip has 16 address bits [15:0], but the 256x16 DDR3s have only 15 [14:0], and apparently the Zynq bank 502 doesn't have a 16th address bit either. I can't find any documentation on what to do with the DDR's A15 pin - any suggestions?

Options are to let it float or tie to 1V5/ground.

The Zynq PS only supports 16 and 32 bit databus, so you need at least 16 bits.

The address bus is behind the data lines, stretching all the way in to the 7th row    (see attached)

I doubt it can be fanned out in 4 layers, I would go for 6. This would also allow a 32 bit databus, doubling the memory bandwidth  .

Thomas

[Someone]

I'm having a strange issue with the schematic for wiring up an 8-bit DDR3 to this chip, since I want to stay at 4 layers. The 512x8 DDR3 chip has 16 address bits [15:0], but the 256x16 DDR3s have only 15 [14:0], and apparently the Zynq bank 502 doesn't have a 16th address bit either. I can't find any documentation on what to do with the DDR's A15 pin - any suggestions?

Options are to let it float or tie to 1V5/ground.

The Zynq PS only supports 16 and 32 bit databus, so you need at least 16 bits.

The address bus is behind the data lines, stretching all the way in to the 7th row    (see attached)

I doubt it can be fanned out in 4 layers, I would go for 6. This would also allow a 32 bit databus, doubling the memory bandwidth  .

Thomas

The TRM states (section 10.1.15) that the PS supports from 8- to -32 on all Artix-7 devices except the z7007s and z7010. Did you find a different source somewhere, or am I reading it incorrectly?

Components of minimum x8, for a total bus width of 16 or 32 bits.

[ThomasDK]

The TRM states (section 10.1.15) that the PS supports from 8- to -32 on all Artix-7 devices except the z7007s and z7010. Did you find a different source somewhere, or am I reading it incorrectly?

https://www.xilinx.com/support/documentation/sw_manuals/xilinx14_1/platform_studio/ps_r_gui_zynq_tab.htm

Note 8 bit interfaces are not supported; however, 8-bit parts can be used to create 16/32-bit interfaces.

[eeviking]

I also have a bunch of DDR3L MT41K128M16JT-125 IT:K FBGA's if anyone is interested
Will they work with the Zynq?

[asmi]

I also have a bunch of DDR3L MT41K128M16JT-125 IT:K FBGA's if anyone is interested
Will they work with the Zynq?

I've just ordered some of those exact modules from Arrow, as I couldn't find them anywhere else.
I don't see any reason why they won't work with Zynq.