dfidler:Notes

[dfidler]

This page contains notes that I have taken either though the course of my investigations or of things that will go into a wiki article one day.

[h=Misc Electronics]%1[/h]

• Ground heatsinks; if not they can act like antennae and destabilize the circuit.

[h=Amplifier Design; Rules of Thumb]%1[/h]

• impedance at input should be minimally 10K
• putting a pot in parallel with the input signal allows for volume control
• most amplifiers will output full power with an input/source of 1v RMS
• input sensitivity - the required voltage for a specified output power (typical is 0.7-0.8)
• should always test amps with both loads of 4R and 8R; avoid extended use of 2R loads
• high-frequency limits of designs should be between 80-200kHz
• Harmonic Distortion at 20kHz is often ~10x that of the THD at 1kHz
• Damping Factor - DF equals Load Impedance / output Impedance
• Impedance - Resistance, measured in ohms
• Get Damping factor as high as possible but don't stress about it too much as it has little-to-no audible effect on sound as the amps total contribution to the dynamics of the speaker is negligible
• Types of Noise
  • Harmonic Distortion
  • Crossover Distortion
  • Intermodulation Distortion
  • Clipping Distortion, etc
• When testing the amp, short the input section; if you hear a hum then your amp has definite problems
• SNR (Signal to Noise Ratio) - volts RMS averaged across the audio range and designated in negative dB levels relative to 0dBr (reference full output of the amp); normally the amp is shorted for noise measurements
• Amplifier efficiency equals % Efficiency of PSU * Efficiency Output Stage (aka OPS)
• Common efficiency of class B is ~70% for PSU and OPS meaing total efficiency of about 49% (ouch!)
• Highest theoretical efficiency of class A is 50%, real life sees about between 25-40% meaning that a 200W class A amp could need to dissipate 600W of wasted power ((200W / 0.25) - 200W equals 600W) needs to be able to dissipate ~500W!!!!! Ouch!
• Class D amps often achieve 90% efficiency
• MOSFETS tend to show slightly lower efficiency than BJTs but FETs tend to have other advantages in high power situations
• ICs, resistors, etc expand/contract internally due to differentials in temperatures. Because they're made of different types of materials, expansions/contractions are not uniform which leads to internal wear of the component.
• Reliability of transistors can be improved by reducing the total dissipate that they must do per thermal cycle (heating up the transistor and cooling it down); MTBF is usually stated on ICs as a function of thermal cycles vs dissipation..
• More ICs in the OPS usually means better THD but worse statistical reliability (ie - the chances of failure of one of them is higher)
• 99% of solid state amps are 3-stage amps; most predictable topology
• 3 Stage amp
  1. Input Stage (IPS) - transconductance (voltage-to-current) amplifier; receives current signal from source and amplifies the current applied to the low-impedance of the second stage; a virtually always differential amplifiers due to the convenience of having an inverting input for negative feedback
  2. Voltage Amp Stage (VAS) - transimpedance (current-to-voltgage) amplifier; receives current signal from IPS and converts it to a high-level voltage signal, providing the gain compensation for optimum stability; generally contains high local negative-feedback
  3. Output Stage (OPS) - current amplifier; receives high voltage signal from VAS and provides near-unity voltage gain, high-current output to the speaker load.

[h=Input Stage]%2[/h]

• must buffer the input signal
• be immune from power supply variations
• linearity should be as high as possible
• noise generation should be as low as possible

[h=Speaker Design; Rules of Thumb]%1[/h]

This will get put into an article about getting started on speaker design. Most of it comes from my thread Rules of Thumb for Enclosure Designs - Questions, and right now it's just a bunch of random points.


[h=Getting Started]%2[/h]

• Speaker design is FAR more complex than you think it is; in fact it's a lifestyle. If you're not thinking about designing them because you have a keen interest in subjects like mathematics, physics and biology then your interest in this subject is likely to be short lived. Consider browsing the many different designs on the web and constructing those instead as the designers will have already worked out the majority of the sonic issues already.
• Your listening space is just as important as your electronics and speakers; does it need room treatments?
• speaker design is about compromise; pick your compromises carefully because you can't build a perfect speaker
• audition lots of stuff to find out what you like;
  • try to audition many amps with the same speakers in the same listening environment
  • try to audition speakers with an amp that appeals to your style of music
• it's easier to get it wrong than right
• some seem to view the likes of Linkwitz, Burnett and Troels as deities to be worshiped
• your second design will be remarkably better than your first
• Simulate first and try lots of simulations to get a feel for how the response will change as you adjust the componets. This is especially important for passive designs as when you come to tuning you will be able to tune better if you know what components do what.
• Start small, maybe with a satellite project and a sub project
• Subs are generally easier to construct for first timers
• Do the following:
  

  'What compromise best suits my specific needs'.
  ie, Does footprint size matter to you?
  Do you prefer BIG woofers?
  Reflex, sealed or transmission line?

• A note on digital XOs:
  

  You can always keep all music in the digital realm as much as possible. Rip it all to a lossless format, store it on a PC and play it back through digital outputs right into a DSP that accepts digital inputs.

[h=Sonics]%2[/h]

• flat frequency response = good
• human ear is most sensitive between 2-4Khz; try to avoid crossovers in this critical range or take extra care to ensure that this frequency region remains flat through the XO (which appears to be nearly impossible)
• spend as much as you can on your mids; they generate the majority of the sound that your ear is most sensitive to
• crossovers are the hardest part to 'get right'
• active crossovers are easier for first timers to build
• passive XOs are very difficult to get right (took <name> 18 months, full time, to design his first commercially viable passive XO)
• active speakers offer less distortion because they don't need to go through passive electronics (amp has direct control)
• open spaces in the enclosure should never be square (standing waves)
• cylinder/spherical enclosures (for bass) are most efficient, but carry some drawbacks (I don't yet know *why
• but I'm reading as much as I can, in the spare time that I have)
• shape of enclosure is less emportant than the damping of the panel resonances and internal sound waves
• putting mid + tweeter closer together makes it easier to unify their images
• off axis performance is nearly as important as on-axis performance
• active XOs often offer less distortion than passive XOs; however, they're also generally more expensive
• Measure your drive units on and off axis to get some idea of the off axis response of your speakers. Consider at what frequency they will integrate best taking into account both the flatness of the response on and off axis.
• gfiandy - annecdotal info:
  

  Here is a bit of my experience in tuning speakers although it is all only generalised and some drive units may respond differently.
  
  Speakers with alot of energy arround 5KHz tend to sound very detailed but a bit harsh. Energy arround 8KHz tends to make the speaker sound detailed and tends to make the instruments highly defined in the image. This is not natural but some people like it. More energy through the 200 to 80Hz region will tend to make the speaker sound warm, too much and it will sound thick and boomy. The 2-4K region affects the vocal significantly and adjusting the phase of the crossove even if it doesn't have much affect on frequency response can adjust the presentation of the vocal.

• Active XO vs Passive:
  

  low level circuits and active circuits create much lower levels of distortion than those that occur in passive crossovers and the distortion scales with the level.
  
  It is not hard to create low level active circuits that have -100dB or greater THD. It is very diffcult to achive this for the large inductors that are used in a passive crossover. If you use large air cored inductors you can minimise it but the magnetic field escaping from them can easly couple into other inductors in the crossover creating a little transformer and causing crostalk from one part of the circuit to another. As the power level increases these inductors will get hot, something low level ciruits if properly designed should never do. When an inductor gets hot its charcteristics shift so its value will start to change.
  
  But none of this is the main advantage of an active design. The main advantage is that the amplifer is directly coupled to the terminals of the speaker. This means that after a transient occurs and the speaker is returning to it resting position, it is a coil moving in a magnetic field so it will create a electrical signal. In a passive design this signal has to pass back through the crossover before it is damped by the amplifers voltage feedback. In an active design the electrical signal (back emf) is damped directly by the amplifier.
  
  I could go on about this but much more is described on Rod Elliots site:-
  http://sound.westhost.com/bi-amp.htm
  
  His rational for using active crossovers is good but I think his design methodology is not. I don't agree with using restance in the output to adjust the roll off of the driver as this reduces the available damping of the back EMF. I would either design the box the right size for the response I wanted in the first place or use a electronic filter to adjust the response.
  
  ...
  
  Designing good passive crossovers is very diffciult. There may only be a few components but just about every component intereacts with all the others in poorly predictable ways (this includes the speaker drive units). So none of the variables are independant, this make it difficult to tune. Due to the interaction of the speakers inductance with the crossover changing values somtimes has very different affects to what you are expecting if you think of it as a resistor.

• omnis and dipoles don't do well if close to a back wall; (wtf? omni, sure, but dipole?)
• At low frequencies (below about 500Hz, but particularly below 100Hz), the position of the woofer/subwoofer in the room dominates just about every other factor. Way beyond cabinet shape, cone material, box damping, ported or sealed
• For speakers that aren't total garbage, a cheap woofer in the right place will work a lot better than an expensive fancy one tossed in any old place. Two or more cheap subs in the right places will also usually work a lot better than one sub of high price. For heavens sake, don't restrict your low bass to have to come from the same place as your mids and highs! They have different acoustical placement requirements and the odds of the best place for subs being the same place as for tweeters is very nearly zero. Putting them in the same cabinet is a compromise right from the start. Get or build a couple of small active subs and experiment a little with room positions for them (use them as plant stands or hide them around the room to keep the SO happy). If done right, they will sound like nothing at all -- all sound will still seem to come from the mains but the mains will seem to have smooth tight bass instead of the usual lumpy mess.
• There's a lot of agreement (not universal, but almost) that off-axis radiation performance of the speakers is about as important as on-axis, in that the response off-axis should be similar to on-axis, but preferably dropping with off-axis angle. This is the failing of a large number of speakers, including many high priced ones. Floyd Toole's book would be well worth reading about this. Radiation pattern should be one of the highest design priorities.
• Crowhurst and Shorter - did work on correlating differences in amps to human sonics

[h=Racks]%1[/h]

Wikipedia

• Wdith: Outside=19" (482.6 mm); Inside=17" (431.8 mm)
• Rack "Unit" Height: 1.75" (44.45 mm)
• Rack "Unit" Spacer: 1/32"(0.031" or .079 mm) - to allow for space between adjacent components (this a 1U front panel would be 1.719" high)
• Avg Dimensions of Component: 434x171x381 mm (WxHxD)

[h=Heat Sinks]%1[/h]
[h=Materials]%2[/h]

Flewis 01-08-2007 at 08:40:57 PM
Ok, lets apply some science:

It is true that copper conducts heat better, higher Cp value (as mentioned by someone else) but no-one has mentioned the specific heat capacities (SHC).

The SHC is how much energy one kilogram of substance can absorb to raise it by 1 degree.

Copper has a SHC of 0.31 KJ/Kg/K
Aluminium has a SHC of 0.91KJ/Kg/K (more than double!)

This means that for the same weight of metal Aluminium contains twice as much energy which means it can transfer more energy to the air that passes over it thus keep the processor cooler.

An ideal heatsink i would think is one that has copper nearest to the cpu to transfer the heat away quickly (copper heat pipes) but then changes to aluminium to gain highest transfer of heat to the air (big aluminium fins).

Aluminium is also less dense which means that it can have the same surface area as copper and will weigh less. Higher surface area=more heat transfer.

I think this should clarify the issue.

(I am a chemistry student at oxford university btw)

----
JMecc 01-08-2007 at 09:10:39 PM

A key thing you have to remember is that this is all in STEADY-STATE:

1) One material absorbing heat into itself better than another means the cpu will stay colder when you first start up but in general operation the temperature in a given spot on the heat sink remains the same over time. i.e. don't count on copper's specific heat capacity to absorb your cpu's heat while you use it, the copper instead has to transport the heat to the fins where it is dissipated by the cooler air.

2) The heat drawn away from the cpu is at the exact same rate as the heat drawn away from the fins by the air (conservation of energy for steady-state systems means conservation of power).

I calculated this for a friend a few years ago and it is very design-dependent, but generally copper will be 5-10% better than aluminum for the same design. So CU is better, just not that much. If you want to get into a real sciencish discussion of how aheatsink should be designed, PM hotfoot.

Jo

Edit: I should also say that the transfer to air is the hardest part since both Cu & Al have very high conductivity, so even though Cu's conductivity is better than Al's by a decent amount (~69%), it does not translate into much extra heat dissipation. Designing a better shape, larger heatsink, thinner fins, better airflow though it are much much more important than Cu vs Al.
-- http://www.tomshardware.co.uk/forum/217118-11-copper-aluminum

[h=Thermal Conductivity]%2[/h]

Source: http://www.frostytech.com/articleview.cfm?articleID=233

border="1"
|-
| class="tcat" | Material
| class="tcat" | Conductivity
(W/m*K)
| class="tcat" | Density
(g/cm(3)
|-
| Aluminum || 247 || 2.71
|-
| Aluminum (6061)|| 171 || 2.6-2.9
|-
| Aluminum (6063) || 193 || 2.6-2.9
|-
| Aluminum (7075-T6) || 130 || 2.6-2.9
|-
| Brass (70Cu-30Zn) || 115 || n/a
|-
| Copper || 398 ||
|-
| Gold || 315 || 19.32
|-
| Magnesium || 170 || 1.74
|-
| Magnesium alloy ZK60A || 117 || 1.74-1.87
|-
| Silver || 428 || 10.49
|-
| Tungsten || 178 || 19.3
|-
| Zinc || 113 || 7.13
|-
| Diamond || 2500 || 3.51
|-
| Graphite || 25-470 || 1.3-1.95
|-
| Silicon || 141 || 2.33
|-
| Epoxy || 0.19 || 1.11-1.4
|-
| Anodize coating || 7 || n/a
|-
| Air (not moving) || 0.026 || n/a
|-
| Mica || 0.7 || n/a
|-
| Berquist sil-pad 2000 || 3.5 || n/a
|-
| Berquist sil-pad k-10 || 1.3 || n/a
|-
| Berquist sil-pad 400 || 0.9 || n/a
|-
| Grey thermal compound (AOS52031) || 2.51 || n/a
|-
| White thermal compound (AOS52022) || 0.7 || n/a
|-
| Solder (63Sn-37Pb) || 50 || n/a
|-