Klondike - 16 chip ASIC Open Source Board - Preliminary

[jimrome]

Ah yes, well I assume the v1 board is released when Bkk thinks its ready. Also assuming he'd release it when he thinks it is production ready while considering overclocking potential. When it would be very easy to just replace a regulator, he'd do that before the v1 release. I've not checked intentions of other board manufacturers (like terrahash).
'Substantial user experience' refers also to hardware; temperature, stability, possible overclock on current hardware (etc). My main point is that it would be strange if Bkk would release a board that could be substantially improved (to a stable v2 board) within 2 weeks, I trust him to strike a good balance between quick release and 'maturity' (being overclockability). Maybe this boundary is around 2 weeks?

Personally I think that the IR3847MTRPBF (suggested replacement regulator) may be be a non-starter as quantities are essentially zero (okay 25) from what I can see. It might be more feasible to simply drop two of the existing IR3895 parts down and have one drive 8 chips each? I'm sure Bkk will come up with something.

EDIT: two more regulators and peripheral parts would be required, in addition to the two which are already present. Each would drive a bank of 4 ICs each.

Spacing would be tight - but likely quite possible.

[Vigil]

Why would these KiCAD files just give errors when trying to open?

[TomKeddie]

Why would these KiCAD files just give errors when trying to open?

It's something about the way you download them, go to the head of the repository and pull the whole thing as a zip file then extract the schematic.

[Bicknellski]

No need to speculate on V2 or V1. BKKCoins clearly pointed out he will go through V1 and test and if time warrants a V2 solution is already planned. It really doesn't matter on time as V2 will go through testing and will only be released if it works. So now that we know there is no conflict between those who want a V1 version and won't want to wait for O/C beyond 300 Mh/s. I sense there are plenty of people willing to wait the extra time and put faith in BKKCoins being able to test and put out a V2 in good time. Considering the scheduling I might be under here in Indonesia and when the chips actually get into my hands I can wait even as late as Mid July to order parts for a end of July early August build just after Ramadan here.

So no one needs to get their panties in a knot considering this was asked, it was answered. Those who jumped the gun like Terrahash and others are not going to be burned I hope if BKKCoins does still need to make revision in the prototype but it seems all is good so far. Fingers crossed nothing major that might make a complete board revamp necessary.

PLENTY of REGULATORS.

https://ec.irf.com/v6/en/US/adirect/ir?cmd=catProductDetailFrame&productID=IR3847MTRPBF


IR3847MTR1PBF   1725 available

[siran]

PLENTY of REGULATORS.

https://ec.irf.com/v6/en/US/adirect/ir?cmd=catProductDetailFrame&productID=IR3847MTRPBF


IR3847MTR1PBF   1725 available

"This part is not on the list of items that can be ordered online at this time.

Please contact ccsales@irf.com for further assistance."

[Vigil]

Why would these KiCAD files just give errors when trying to open?

It's something about the way you download them, go to the head of the repository and pull the whole thing as a zip file then extract the schematic.

Nice, got it.

Can anyone give an estimate of the wattage through the various chips, the avalon, the ir3895s, etc.?

[BkkCoins]

Update for my work today:

I got the hash count working correctly now. I'm still finding a lot of instability with USB that I didn't see before. Random errors, write errors, disconnects that never happened when using serial mode. I haven't tried tracking that down as my attention has been on improving the result capture, which I think is still unreliable due to poor clock delay.

Someone requested scope traces before and I still haven't been able to get the scope USB save to work. It just errors out or write 0 byte files - useless. It says use FAT32, which I am, but still doesn't work with either 512MB or 16GB flash sticks. I'll probably have to get some support for that but in the mean time I used my camera to capture some screen shots.

These traces are with 1K pullups and 500 ohm pullups (no 470 on hand so doubled up the 1K) but not much difference shown. This is with 330pF capacitors now. I'm going to try adding a second NOR gate with RC between to try and get a better delay and more defined capture edges. It's not clear if the PIC has Schmitt inputs but the NOR gates do so this should give better definition. This would result in a small board change and part change from 1G02 to 2G02, only 5 cents in cost, but if it makes the result capture more reliable it's well worth it. Looking at the traces I'm a bit amazed it captures results at all. It does but I think it's the source of HW errors in cgminer as even one bit miscaptured would throw a nonce out. Whether it gets a 1 or 0 right now looks like a dice roll.

The whole wave


In a bit closer


500 ohm pullups, 330pF, 200nS/div


1K pullups, 330pF, 100nS/div

[intron]

Someone requested scope traces before and I still haven't been able to get the scope USB save to work. It just errors out or write 0 byte files - useless. It says use FAT32, which I am, but still doesn't work with either 512MB or 16GB flash sticks. I'll probably have to get some support for that but in the mean time I used my camera to capture some screen shots.

Many thanks:)

intron

[Vigil]

I am aware you are trying to keep this thread to the electronics and testing, but just wanted to give you guys a heads-up on a simplified model of the K16 I've made using only the Avalon chips. I have them generating 2W each. They are modeled as QFN chips with most of the details and majority of the dimensions correct. I have modeled the lands that go through the board but they may not be meshing correctly so I will continue to look at that. I can add other chips and more detail if I can get the measurements and power output. The more exact I make the model, the more realistic the results will be.

This model uses an aluminum heat-sink which covers the entire bottom surface, the fins and base are 2mm thick - it is based off the heat-sink posted about 20 pages ago. I can alter this to any type of design: change fin size and number, change to fingers, etc. Additionally, this model is cooling only from the heat-sink, i.e., there are no fans or any air movement except that from buoyant convection. It is actually a transient scenario but here I am only using a steady-state model, but I can look at transient.

The temps seem a little high to me, 148 C (300 F), but of course this is without any air movement. However, many GPUs can get to 120 C or more if heat-sinks aren't attached or no air. This was just a first shot and I am going try some other things and recheck this solution, etc.


Avalon chips wire-frame.


Simplified board model with heat-sink.


Heat-transfer analysis showing temps of chips with no forced convective flow air exchange. Temps of chips reach approx. 150 C while operating at 2W with this heat-sink.

[Bicknellski]

Nice to see this Vigil... make sure to add this to the heat sink Klondike thread that Steamboat started a while back https://bitcointalk.org/index.php?topic=208381.0 . Nice work. What are the exact dimensions of the heat sinks your are running in the model? Plate thickness, spaces between fins etc.

Here are some heat sinks I actually sent to BKKCoins. https://bitcointalk.org/index.php?topic=190731.msg2264180#msg2264180

[alle.sello]

Update for my work today:

I got the hash count working correctly now. I'm still finding a lot of instability with USB that I didn't see before. Random errors, write errors, disconnects that never happened when using serial mode. I haven't tried tracking that down as my attention has been on improving the result capture, which I think is still unreliable due to poor clock delay.

Someone requested scope traces before and I still haven't been able to get the scope USB save to work. It just errors out or write 0 byte files - useless. It says use FAT32, which I am, but still doesn't work with either 512MB or 16GB flash sticks. I'll probably have to get some support for that but in the mean time I used my camera to capture some screen shots.

These traces are with 1K pullups and 500 ohm pullups (no 470 on hand so doubled up the 1K) but not much difference shown. This is with 330pF capacitors now. I'm going to try adding a second NOR gate with RC between to try and get a better delay and more defined capture edges. It's not clear if the PIC has Schmitt inputs but the NOR gates do so this should give better definition. This would result in a small board change and part change from 1G02 to 2G02, only 5 cents in cost, but if it makes the result capture more reliable it's well worth it. Looking at the traces I'm a bit amazed it captures results at all. It does but I think it's the source of HW errors in cgminer as even one bit miscaptured would throw a nonce out. Whether it gets a 1 or 0 right now looks like a dice roll.

The whole wave
https://i.imgur.com/G208Dxd.jpg

In a bit closer
https://i.imgur.com/t7OMtZh.jpg

500 ohm pullups, 330pF, 200nS/div
https://i.imgur.com/IfJcaMy.jpg

1K pullups, 330pF, 100nS/div
https://i.imgur.com/Jfhd3cu.jpg

I think you have problem with RC OWON probe compensation.

[BkkCoins]

I think you have problem with RC OWON probe compensation.

No, probes are compensated fine. What you're seeing above is RC delay components altering the wave shape on NOR gate output, and the other trace is an open-collector output so you're seeing slow rise time due to undriven floating and line capacitance. I wasn't expecting square waves here as my intention was to delay the clock signal enough to capture the "1"s data. The main problem is that the RC delay causes uncertainty in capture edge on the PIC input. So adding another schmitt input gate before the PIC will clean up that edge and allow seeing where the edge occurs and adjusting the RC values to position correctly. I'm working on this right now.

[Bicknellski]

PLENTY of REGULATORS.

https://ec.irf.com/v6/en/US/adirect/ir?cmd=catProductDetailFrame&productID=IR3847MTRPBF


IR3847MTR1PBF   1725 available

"This part is not on the list of items that can be ordered online at this time.

Please contact ccsales@irf.com for further assistance."

Distributor Inventory
                                              
Distributor                          IR3847MTRPBF           Inv Date  
DIGI-KEY CORPORATION                25                01-Jul-2013
IR OVERNIGHT                              1725          01-Jul-2013

To obtain samples or receive additional inventory information, contact your Sales Representative or Distributor.

Distributors and Agents

Arrow Electronics Thailand Ltd.
Le Concorde Tower, Unit 803, 8/F
202 Ratchadapisek Road, Kwaenghuay-Kwang
Khet Huaykwang
Bangkok 10310
Tel: (662) 694-2332
Fax: (662) 694-2331

Avnet
Unit A3005 (Tower A) Fl. 30th
Cyber World Tower,
90 Ratchadaphisek Rd.,
Huai Khwang Bangkok 10310
Tel: (66 2) 645 3678
Fax: (66 2) 645 3681

Future Electronics
947 Thosapol Land 3 18th Floor
Bangna-Trad Road
K.M.3, Bangna, Bangkok 10260
Tel: 66 2 361 8400
Fax: 66 2 361 8433

International Rectifier
23 Serangoon North Avenue 5
#06-02 BTH Centre,
Singapore 554530
Tel: (65) 6506 2000
Fax: (65) 6506 2507

Mouser Electronics
202 Le Concorde Tower
20/F Ratchadapisek Rd
Huaykwang, Bangkok 10310
Local: +66 2694 2310
Fax: +66 2694 2276
thailand@mouser.com

WT Microelectronics (Thailand) Co., Ltd.
Le Concorde Tower,
Unit 2003 Floor 20th
202 Ratchadapisek Rd,
Huaykwang, Bangkok 10310
Tel : (66)2-6941 727
Fax : (66)2-6941 707

Weikeng Technology Pte Ltd
KRITS Building
1032/ 1-5, 7-9 Rama 4 Road
Thungmahamek, Sathorn
Bangkok 10120, Thailand
Tel: 66 2 679 9902
Fax: 66 2 679 9901

[siran]

@Bicknellski

nice.

I wonder, if you get some IR3845, may I ask you to send me one or two* for fair amount of BTC? I want to create some easy basic blocks for breadboard.
You know, the power supply, some avalon chip (or maybe next gen one) and so on to hobby-play with it.


And I don't want to buy ready eval board.

regards

[turtle83]

Question: How easy is it to mount QFN chips?

Saw i get K1 or K16 fabbed by someone with everything on it except for the avalon chip(s). Is it easy to get get the chip mounted on it by someone? Like a repair shop or something? Someone told me all mobile repair shops deal with reworking QFN daily... is this true? Any risks involved?

[Lollaskates]

is anyone capable of rendering the 3d model from multiple angles? I've been unsuccessful with getting Kicad to load it properly to allow me to render. I'm trying to see the bottom of the board specifically, and from the side.

Is it flat on the bottom? It the board designed for a flat aluminum heatsink on the bottom, or are there components poking out that need to be avoided?

[turtle83]

Is it flat on the bottom? It the board designed for a flat aluminum heatsink on the bottom?

Yes. Its flat, designed for flat aluminium, but IMHO thermal pad is recommended.

[LaserHorse]

Question: How easy is it to mount QFN chips?

Saw i get K1 or K16 fabbed by someone with everything on it except for the avalon chip(s). Is it easy to get get the chip mounted on it by someone? Like a repair shop or something? Someone told me all mobile repair shops deal with reworking QFN daily... is this true? Any risks involved?

http://store.curiousinventor.com/guides/Surface_Mount_Soldering/QFN/

[cp1]

is anyone capable of rendering the 3d model from multiple angles? I've been unsuccessful with getting Kicad to load it properly to allow me to render. I'm trying to see the bottom of the board specifically, and from the side.

Is it flat on the bottom? It the board designed for a flat aluminum heatsink on the bottom, or are there components poking out that need to be avoided?

Check the heatsink thread, I think there's mention of a cutout for the power connector.

[Bluestreak66]

Why is it that the Avalon reference design and the klondike both use a rc reset per asic? Why not use a single reset pulse generator for all asics? That would replace 32 components with 2 or 3 and would give a more stable reset. Also has anyone considered a multiphase power supply? If you have 2 or three phases working 180 or 120 degrees out of phase you could achieve a lower noise power supply along with plenty of overhead for overclocking at a higher efficiency. There are several POL controllers out there that go up to 12 or more phases that could operate at very high frequency reducing inductor and capacitor sizes. I only mention this because I see your looking for more current on rev 2. I understand you don't want to totally respin the design but it's something to consider. I was going to say something awhile back but got stuck in newb jail and couldn't post. It's good to see this is still progressing along nicely. Massive respect for BKK and all the others that have put their time into this.