Posts tagged ball grid array
In this article we will discuss Power Delivery system in Motherboards . For more in depth training , join PCLR Course of chiptroniks or you can also buy our course materials with online support.
Power delivery—Why & How
Why: Motherboard components need one or multiple stable and clean DC power to work correctly
How: (1) Power Supply directly to motherboard components (2) for the power which Power Supply can not provide directly, DC to DC power converter on the motherboard converts the power and provide to components
Voltages type needed
Postive DC Voltage: generally between 0V to 12V, generated by DC-DC converter 0.75V, 1.5V, 1.1V… or directly from power supply, like 3.3V, 5V, 12V
Negative DC Voltage: typically -12V
Motherboard voltage normally ranges from -12V to 12V
Tips: General speaking
Higher speed component=> lower voltage needed
(especially for IO function)
Current types needed
Simple answer: Power/voltage=current needed
Low power device: <2A, example: Clock chip, LAN…
Medium power device : between 2A to 50A: example: Fan, DIMM, Chipset
High power device: >50A, example: processor, high power DIMM, high end Graphic card etc
The low/medium/high is just general category, no standard
Tips: High current device has higher requirements on the PCB
Space, layers, cost, copper thickness…, all in all, bigger current,
more design challenge for power designer and CAD engineer
Examples: components Voltage & Current
1.0V to 1.5V, 50A to 150A, 130W
1.8V/0.9V for DDR2, 1.5V/0.75V for DDR3, 20A to 40A, 50-100W
Chipset: 1.1V, 10-20A, 5W to 30W
Onboard device: 1.5A, 1-2A, 3.3V, 0.5W to 5W
PCI slot: PCI slot: 12V, 0.5A, 3.3V, 3A, 5V, 1A, 15W, 25W, 75W or more
Fan connector: Depends on fan used, ranges from 0.1A to 5A, 5W to 50W
Normally 1 Components need multiple voltage rails
depends on what function needed, such as ICH need
1.5V, 3.3V, 1.8V…, more function, more voltage rails needed
For example: ICH has more voltage rail than CPU
due to ICH has more functions
Voltage types by components function
Components may need several voltages by functions: below is general category
(CPU), VDD (DIMM), occupy most the power pin of the components
IO Voltage: Core Voltage: Main voltage for core logic, most of the power consumes on the main voltage) for the core function, example VCCP Voltage for BUS, example: CPU Vtt
Reference Voltage: voltage used for signal sampling
Analog voltage: Some components include analog function, so analog voltage needed, such as Video, PLL circuit, analog voltage require to be clean ! Need to be separated from normal voltage
Components may contain 1 or more type of voltages depends on
Function needed, such as ICH need all 3 above voltages
Voltage types by power state
Some voltage are only required for certain power state
Normal Voltage: Voltage existing when the system is at S0 to S2 state, which means system is at ON state, like CPU main power, fan power, which is main power for the system
Battery Voltage: Voltage existing when the system at AC OFF status, it is powered by onboard battery. Example RTC clock
Standby Voltage: voltage always exists at S0 to S5 state (DC OFF), which means system at DC off state, AC power code is plugged, it is used for board power on/off logic and wake up function and some management function and other functions need to be functional at main power off state, remember, when AC power cord inserted, standby voltage exists !!
Aux Voltage: Voltage switch by between Standby voltage and same Normal Voltage, the main reason of Aux voltage is the function is needed through S0 to S5 state, but standby power can not provide enough current at S0-S2 state due to the device consume more power at S0-S2 state then S3-S5 state, so voltage need switch from standby voltage to normal voltage to get enough current , example: DDR voltage 1.8V, when system is at S3, the Aux voltage comes from 1.8V standby power to keep DIMM refresh, after power on to S0 state, Aux voltage switch to 1.8V normal voltage to support DIMM normal read/write (which consume much more current)
Components may contain 1 or more type of voltages depends on
Function needed, such as ICH need all 4 above voltages
Let us take a look at a real sample-Chipset
G41 MCH (north bridge) function/power mapping
(not exactly correct, just for example)
Another example—ICH 10
ICH 10 has require more than 20 voltage rails !! due to lots of functions integrated in ICH 10
Refer to product EDS for pin definition and power requirement
Example 3—PCI-E slot Power requirement
This voltage supply to add in PCI-e card, Card is required to design within this limit
Overall Power Delivery Example–Thurley
Overall Power Delivery Example2—Romley
Motherboard Input Power
Now, we know what kind of power (voltage/Current) needed by components, but where does it come from? Answer: from Power Supply, directly or indirectly
Power Supply Output (motherboard input)
Power Supply output type:
Power supply has multiple DC output rail (NOT connector)
Popular 12V, 5V, 3.3V, -12V, 5VSB and other voltage
12V output may have separate rails, like 12V1, 12V2, etc for 240VA protection
Single output: 12V or other voltage only
Power supply has single DC output, 12V is most popular
Battery is single output example
Power Supply output interface:
Connector: board to board or board to cable connector
PCB gold finger: PCB to mating connector
Most of single output PSU also has standby output, like 5VSB
Power Supply Output example 1
Desktop ATX PSU : Multiple output, cable + connector
Server EPS12V : Multiple output, cable + connector
Power Supply Output example 2
19V Single output, connector, connect to motherboard directly
Hotswap module :
12V single output, gold finger and board to board connector
normally it also has 5VSB output
Motherboard side interface
General Rule: mate with power supply output
Gold finger mating connector
Board to Board connector
Motherboard power rails & Power supply rails
As we talked before, multiple-output power supply has multiple output, each rail will have current limit, and each rail are separated below is example
Same for motherboard, motherboard will also have multiple rails, like 3.3V, 5V, 12V1, 12V3a…, each rail has current requirement, so we need to mapping the power supply rails to motherboard rails to make sure both power supply & motherboard rails can be met
Next page is example
Rail mapping Example
Power supply connector/rail mapping
Power supply rail can be separate to support multiple
motherboard rail, but reverse is NOT allowed!, otherwise it will
Short power supply rails and cause protection
DC to DC converter
So far, we know how power supply provide voltage rail to motherboard, like 12V, 5V 3.3V, etc by connectors or PCB gold finger or other method, but for the other voltage power supply can not provide, like 1.1V, 1.5V, 0.8V, we need DC to DC converter on the motherboard to convert the power supply voltage to the voltage we needed
DC to DC converter also called Voltage regulator (VR)
DC to DC converter (VR) types
(1) Linear voltage regulator
-Clean (little noise)
Simple & Clean (little noise)
-Low voltage drop
(1) Why low current and low voltage drop?
vdrop on the VR= Vout-Vin, so the power loss = I x Vdrop, for example: Vin=3.3V, Vout=1.5V, 2A, so the power loss on converter is (3.3-1.5)x2=3.6W, assume 50C/W, so the temp rise will be 150C, which is burn the components, so only low current and low voltage is allowed, Linear VR only support low current requirement
(2) Why low efficiency?
The efficiency= output power/input power, obvious, it is low efficiency due to the power loss on the converter is big, the bigger difference between Vin and Vout, the lower efficiency is.
(3) Why simple & clean & low cost
It is simple & due to just a few components needed
It is clean due to no switch components, it is easier to place & layout the linear VR
Switching VR Types—Single Phase
Basic working principal is by control the mosfet PWM value to adjust the output voltage, Vout/Vin=PWM%, for example: 12V to 1.5V, PWM=12.5%
Switching VR efficiency is between 80 to 98% depends on VR design, the main power loss is VR Mosfet switching & conduct loss
It can handle high current due to high efficiency
High cost /complex is obvious: it need chip, mosfet, inductor, capacitor…
High noise: due to switching method and mosfet switching, it has much higher noise than linear regulator
We will NOT discuss how VR works here, refer to VR training slides
if you are interested, Overall speaking, VR is a complex technology
Switching VR Types—Multi Phase
Switching VR—single phase 12V to DDR 1.5V
Switching VR—multi phase 12V to CPU Vcore
Linear VR–3.3V to IOH 1.8V
Linear VR–3.3V to IOH 1.8V
VR placement & layout
CPU VCCP VR placement
CPU VCCP VR copper planar
In this article , we will show how to reball . The following video will describe the reballing process in great detail.
In this article , we will cover how to remove North Bridge with BGA REWORK Station . Since the best BGA Rework Station currently available in the Market is Jovy-Systems RE-7500 . We will show its working in the post .
In this tutorial , we will discuss simple procedure for reballing process. Hope you will enjoy it . Look at our complete BGA REWORK SOLUTIONS
CHIPTRONIKS , A Division of VD Intellisys is Authorized Distributor of JOVY SYSTEMS BGA REWORK STATIONS in India . For more details visit http://www.bga-rework.in or call 09971004998.
A guide to the BGA Package
- What is it?
- Why have they caught on?
- Are they difficult to place – by machine?
- Is it possible to place them by hand?
- How do you know if a BGA has fully soldered – and don’t you need an X-ray machine?
- Can a BGA be removed, reworked and replaced?
- Can the PCB design influence the manufacturability?
- What works best – printed paste or flux only?
- What ways of soldering are used – and can they be verified?
1. What is a BGA?
The B(all) G(rid) A(rray) or BGA package invented by Motorola, is now a mainstream packaging technology. The most common example consists of a thin substrate of PCB material onto which the chip is mounted. Under the substrate is an array of solder balls forming the terminations. During reflow these balls fuse with corresponding pads on the Main PCB and form the joints.
2. Why have they caught on?
The BGA excels when it comes to high pin count devices, putting all terminations underneath the package instead of around the edges as they are on a QFP saves a lot of space allowing smaller products to be made.
Using a 2-dimensional grid means that ball to ball spacing can be quite coarse compared to the lead pitch of a high pin count QFP – so less problems with solder shorts.
Consequently they are easier to solder, no legs to get damaged and they have a huge self centering effect due to the high solder surface tension effects caused by the array of solder balls.
High pin count QFP’s by contrast either have to be bigger to accommodate the same number of edge mounted pinouts or the legs have to be extremely fine and damage prone.
So they are easy to handle and give very high assembly yields – consequently they have started to supplant other package styles in mass production.
3. Are they difficult to place by machine?
From a manufacturing perspective; a BGA is designed to be machine placed using vision systems to align the device to the grid of pads on the PCB. During reflow it has a very strong self centering effect due to the surface tension of all the solder balls – consequently it is quite tolerant of placement errors – as much as half a pitch of misalignment will usually not cause problems. Most machine systems place far more accurately than this.
4. But what about hand placing?
However, every silver lining etc: A BGA is not designed for hand assembly. Of course there will be a very small need to do so – for e.g. a prototype. Whilst this is tricky, it is not, as we shall see, impossible
5. How do you know if a BGA has fully soldered – and don’t you need an X-ray machine?
Once a BGA has been soldered it is impossible to visually inspect the joint – the only viable method is to use x ray or possibly fibre optic endoscopes – so how do you know if it has soldered properly? Well the question should be “why do you want to inspect it anyway?” People feel they need to inspect because they can’t be sure of their soldering process. Most assemblers use convection ovens and despite all that the manufacturers claim, there is no way that hot air can penetrate fully under a BGA package that is sitting a millimetre or two from the board and heat every ball the same. The centre balls will inevitably be cooler than the outer ones. What actually has to happen is the package must heat through by conduction and often overheat the outer edge to ensure the centre sees the right temperature..
We use a different type of soldering (see 9) – and we can be certain of the soldering conditions – therefore we can say that since every ball has reached a known temperature – no matter where in the grid they are, the device will have soldered properly.
This is born out in practice – over 2000 BGA’s soldered, no reported failures – and we don’t have an x ray facility. One customer did x-ray a board and the results were perfect
6. Can a BGA be removed, reworked and replaced?
If a BGA has to be removed it cannot be done without destroying the balls beneath the device. Usually this means the device is scrapped although high value BGAs can be recovered by specialist companies who can re-ball the package so it can be used again. A typical cost of doing this may be £70 so clearly only worth doing on devices worth much more.
We are in the business of low volume manufacture so the BGA device initially presented us with some concerns. However, we have evolved methods of assembly that work and are viable. We have in fact, to date, (mid 2005) placed in excess of 2000 devices – all by hand and without defect. We now have fully automatic placement capability through our MYDATA machine and semi auto placement on our Fritsch MicroPlacer.
7. What can be done at the PCB design stage to make life easier for the assembly company?
As the package itself obscures the grid of connections it is impossible to see if the package is in the right place. For this reason alignment indicators are extremely useful – see photo. Note how they have made two chevron marks on opposite corners – they have used two marks per corner to allow for two different package dimensions – most people only use a single mark per corner. Even a simple dot at each corner will do – two corners minimum but three or all is better still – see photograph.
Please Note that these marks MUST be in the copper itself, silkscreen printing is nowhere near accurate enough for this purpose.
A pin 1 mark that is not obscured by the package – this can be done in silkscreen. It’s amazing how many pin one marks vanish once the package is down…
DOs and DON’Ts and things that are OK
Resist defined pads are OK
Don’t put vias in BGA pads – unless they are microvias. The solder ball will wick down the hole by capillary action and you WILL get an open. These are non-repairable and not covered by our warranty.
DO make sure vias on short stubs have a resist barrier between via and BGA pad – or the same thing will happen.
Wetting indicator pads (dog bone or tear-drop shaped) are OK if you want to use them to us – but not so popular now. They were intended to allow an x ray photo to reveal that the ball has wetted the pad by distorting its shape
8. What works best – printed paste or flux only?
Two main methods for fitting BGAs are in use:
Printed Solder Paste
The main method is to print paste to all BGA pads along with all the usual SMT parts, the device is placed onto the paste and reflowed with the rest of the parts.
Having solder paste is said to take up minor co planarity errors if the device or PCB is warped although this sis debatable. This method is fine for machine or vision assisted placing as any smears of the printed solder paste can lead to short circuits just where they are least wanted.
The real risk of using printed paste is that if it is too much – or gets smeared – a solder ball can become large enough to touch its neighbour and form a short that is impossible to remove – or see unless it is on the outer edge or you have x ray facilities.
Flux only method.
This is approved by Motorola (the inventor of the package) and is the method usually used if reworking a package onto an otherwise populated board. IF a board is already loaded with parts it is not usually possible to re-paste the BGA pads – although micro stencils are made for this purpose. Instead flux is applied to the pads or the BGA balls themselves and the joint is made during reflow by the solder from the ball flowing onto the pad. Some people think that the lack of solder paste may increase the likelihood of an open is the device or board is warped but in practice we have never had one in over 2000 parts.
A characteristic of this method is that the package will sit a little lower on the PCB, as the solder ball has not been increased in volume by the printed paste.
9. Thoughts on BGA soldering.
After much analysis – I decided that what worried people most about BGAs was not alignment – that turned out to be easier than everyone first thought. No, it was “has the flipping thing soldered?” You can’t have a quick squint under the microscope so how do I really know that the balls right underneath have gone?
So what methods are in use – and why are people uncertain?
Most people in the SMT world had switched to convection (hot air) ovens long before BGAs arrived. Unfortunately if they had stuck with the first generation infrared systems they’d have been better off.
Problem is a BGA has all its connections underneath – the BGA body to PCB gap is a millimetre or so. Now you have to get even heating right under the BGA – just one cold spot means a defective joint. Hot air – even if turbulent just is not going to penetrate that well into such a narrow gap. So what do solderers do? They either increase the temperature to ensure that every part of the device is hot enough or increase the heating time so as to allow the package time to heat through by conduction. Hence why IR is better – it heats the package rather than tries to blow heated air under a narrow opening.
However, there is a third heating method – the one we use. It is guaranteed to heat every part of every device evenly, it is impossible to overheat from its specified temperature, it completely surrounds the job in an oxygen free, totally inert environment which helps the flux do its job better still. What is this heating method: it is now called “condensation reflow” although many old hands at SMT know it as Vapour Phase soldering.
A quick description is that the process uses a special chemical (basically a fluorocarbon) that boils at a known temperature – we use a 230-degree BP. The boards to be soldered are placed in a chamber in the bottom of which is a sump of this fluid, which is heated. AS it heats up it produces steam – which just like it does in your kitchen condenses on any surface cooler than itself. As it condenses it give sup its heat to the cooler item. Steadily the cooler PCB gets hotter – until eventually (having passed the solder melting point) it reached the same temperature of the steam. At that point no more steam can condense – a special heat probe detects this point and shuts off the heat source. The liquid stops boiling, the PCB can be removed from the steam chamber to cool. The vapour blanket is totally inert and heavier than air so all oxygen is displaced from the joints – so the flux only has to clean the joint not also cope with the oxidation occurring during normal reflow in air. Less active fluxes can be used.
Also the vapour penetrates everywhere – around tall objects, down between things and crucially for us – under BGAs. The whole board and parts are evenly heated, all around and from both sides. In a conventional linear oven the hot front moves along the board so that at any time there is a wide variation in temperatures across the board – this can lead to distortion.
This content has been taken from http://www.allgoodtechnology.com/pages/bgaguide.htm
Condensation reflow is not without its critics (who often seem to make rival technology ovens surprisingly). The first generation VP systems plunged a board at ambient temp into a steam blanket – imagine walking into a Turkish baths – instant thermal shock. The current generation machines like ours start with the fluid at ambient too – it gently heats up and the temperature rises slowly – our system shows a rise of below 2 degrees a second – considered ideal by the component manufacturers.
CHIPTRONIKS , A Division of VD Intellisys is Authorized Distributor of JOVY SYSTEMS BGA REWORK STATIONS in India . For more details visit http://www.bga-rework.in or call 09971004998
BGA Rework Station
BGA is short for Ball Grid Array. It is a kind of package method which use organic carrier in IC. It has the following features:
1. Small package area.
2. Greater functions and more pins
3. Self-centerize while PCB puddle welding, easy to put on tin.
4. More reliable.
5. Good conductivity and low overall cost. Memory which applies BGA can enlarge the memory capacity by 2 to 3 times while the volume of memory remains the same. Compared with TSOP, BGA is much smaller and better at radiating and conducting electricity.
Types of BGA
According to the encapsulation material, BGA can be classified into the following types:
Features of BGA
Compared with QFP, BGA has the following features:
BGA Rework Process
Most of the semiconductor device’s heat-resistant temperatures are between 240°C and 600 °C. Therefore, the control of the temperature and uniformity are very important to BGA rework systems. BGA rework process as follows:
Mounting: The main purpose of mounting to make every BGA solder align to the PCB pad with special equipment.
Hot air reflows: Hot air reflow soldering is the key to the whole BGA Rework.
In order to ensure the validity of BGA Rework Station, the installation should meet the following requirements.