LM3886 Amp - Transformer Question

[mydementia]

Hi guys.
I'm tired of hauling my tube amp to my buddy's house for practice so I'm going to finish my other speaker cab and build an LM3886 amp (got the chip with my last Futurlec order...) to leave there (that way, I can bring my pedal-du-juor or my SansampGT2 for a preamp). 
I started digging around for a transformer and wonder if this one will work:
http://www.partsexpress.com/pe/showdetl.cfm?&Partnumber=129-035

Also, regarding heat sinks... how big does it REALLY need to be?  The two in JoeCool's pictures (sticky thread in this forum) are HUGE!!!

Thanks for looking.
Mike

[joecool85]

One was huge.  The big black one here could support 4x LM3886 chips or so, even running at 4ohms shouldn't be a problem:


The minimum for a LM3886 running at 8 ohms would be like the other heatsink I posted:


Check eBay, you can normally find some good sinks, thats where I got mine.

One of these would work:
http://cgi.ebay.com/2-Aluminum-4-5-x-4-5-x-5-Heatsinks_W0QQitemZ7604653489QQihZ017QQcategoryZ66946QQrdZ1QQssPageNameZWD1VQQcmdZViewItem

But overkill is better than a heatsink that isn't big enough.  Ask RDV about that...

[mydementia]

Right on - but what about the transformer?  Do I need to buy a $50 toroidal, or can I get away with the 'el cheapo' one from Partsexpress?
Mike

[teemuk]

Unless you are designing a commercial product there is no such thing as a too big heatsink. The size of the sink is only half of the story, though. Basically, the physical size is not a very good measure: It matters how much fin area the sink has, how the fins are spaced, how wide and long the fins are, how thick the base plate is etc. For example, needle-like "fins" are more effective than large fin plates spaced close to each other. Also, the sink material matters since some metals are better heat conductors than others. Better conduction means better ability to remove heat from the source and spread it around the sink. If the conduction is poor you'll be just heating a certain point in the sink. Massive objects conduct better so large baseplate or use of external copper "spreader" plate is beneficial. Thick baseplate also increases mass, which is good since larger mass allows to suck heat more effectively.

Convection is the main way to remove heat and the sink structure is usually different if the convection is natural or forced. Usually forced convection is dependant mainly on the conduction of the base plate so the fins can be placed further from each other and they can be thicker. Sinks based on natural convection will usually benefit from a large amount of thinner fins but a too close spacing of them can become detrimental. Usually the width of heatsink has more of an importance than its height. Convenction happens when cold air replaces the warm. Warm air rises upwards so with natural convection vertically placed fins are more effective. Also, the more turbulence there is the more effective the conduction since turbulence will cancel the "laminar effect" = air flow velocity is smaller close to surfaces.

With quite normal amount of convenction the radiation of heat can count as much as 25% of the efficiency to dissipate heat. Shiny surfaces radiate poorly so stay away from "sexy" looking shiny heatsinks, they are not as good as they look. A mat black is an ideal color... but don't paint a heatsink: The paint will create a large thermal impedance between air and the metal.

Which is the last factor: Between heat source and sink you will have a thermal resistance that defines how much heat is transfered to the sink from the device. This also allows you to count the required thermal resistance for heatsink. This is the resistance between air and the sink. It's approximately:

RSA = (TC – TA) / PD,

where RSA = heatsinks thermal resistance, TC = chip's maximum case temperature, TA = ambient air temperature and PD = dissipated (waste) power. RSA will be in a form of T/P (i.e. 2K/W), which means the sink gets 2 kelvins hotter than the ambient air per an increase of one watt of wasted power. The resistance of selected heatsink should always be smaller than the derived thermal resistance. So in our example, if ambient T is 25 degrees C and power dissipation is 20W the heatsink temperature is 65 degrees of C.

The equation is a "torso" but serves as an example. Basically you follow this equation:

RSA = (TC – TA) / PD - RI,

where RI = interface thermal resistance. This is the thermal resistance caused by interface between chip's case and heatsink. The ingredients for it are basically lubrication and insulation. If the heatsink surface is rough the lubrication can decrease the interface resistance... so can sanding the surface. The effect of interface can be quite large so it is a very important aspect in heatsinking design. If one can construct the interface properly he can choose to use a less effective heatsink.

[teemuk]

And the transformer question: Do not skimp in it! It is one of the most important components that define how much power your power amplifier can really put out. Basically the power amplifier is just a modulator for power that is delivered by the supply. Use a cheap supply and you quite likely wind up with a "cheap" amplifier as well. ...And by cheap I don't neccessarily mean the actual price. There is more to transformers than what they cost.

I wouldn't use anything lesser than an transformer with a VA rating of 150 and might even choose something closer to 250 VA. Above that it really starts to be overkill and can actually be quite quite tricky to use due to high inrush current. If you don't know what VA rating is I suggest you find out before buying the transformer.

Toroids are smaller than E-core transformers with equal VA rating. They have smaller external magnetic field and make less mechanical noise. They are also more expensive and have higher inrush current.

[joecool85]

Right on - but what about the transformer?  Do I need to buy a $50 toroidal, or can I get away with the 'el cheapo' one from Partsexpress?
Mike

On paper to get full power you should have about 150VA per LM3886, but in practice it has been proven that as low as 90VA or so has worked.  VA is volt x amps (aka watts), the one you posted is about 50VA...too small.  Also, ideally you need a tranny that puts out about 22v+22v.  But you don't have to go torroidal, if you notice on the pic of my LM1875 amp, it uses a center tapped tranny.

This one should do the trick for $23: http://www.partsexpress.com/pe/showdetl.cfm?&Partnumber=120-225

Or a torroidal for $41: http://www.partsexpress.com/pe/showdetl.cfm?&Partnumber=122-615

I did a toroidal for my LM3886, but in retrospect, I really should have saved my money and went with a CT tranny.  The amp I used one with it sounded fine and works great.

[teemuk]

Lower VA will work but it increases the amount of power supply sag under current draw and reduces the output power. Some manufacturers use a cheap transformer as a "limiter" technique to protect the output devices from excessive power. Not so good idea in my opininon. Unless you are short on money invest to a transformer with higher VA rating, it will provide steadier supply voltages under current draw, which leads to higher continuos power.

[joecool85]

If you go with either of the trannys I suggested, and run your amp at 8ohms (which gives you 50w RMS) it shouldn't be a problem.  There may be a tad bit of lag running at 4ohms (yielding 68watts), but I doubt it would be noticeable.

[mydementia]

Right - so I'm thinking I'll try the toroidal (http://www.partsexpress.com/pe/showdetl.cfm?&Partnumber=122-615).

If I would buy a power board from Brian over at chipamp.com like this one:


could I run a second lead from the +DC voltage (not sure what Brian's kit produces - no regulators are identified...but I assume it's around 20VDC to run the 3886) to a daughter board with a +9V regulator to give myself a 9V power source for external effects/preamp (and for potentially driving a fan)?
Mike

[joecool85]

Brian's kits are unregulated.  The output power will be roughly 1.4 x the amount you put in.  You could, in theory, tap the power to run a 9v regulator for effects.

[mydementia]

I must have mistaken the little giz-widgets behind the cap in the front of the picture for regulators... maybe I'll just build the GGG bipolar power supply...

[joecool85]

Unless there is a new one I don't know about, that is unregulated is well.  Brian's power supply board is perfect for LM3886 application as regulation is totally unnecessary.

[ApexJr.]

I have some toroidial transformers in stock

21- 0 - 21 Volts at 80Va using 2 would work   $12.50 each
25- 0 - 25 Volts at 100Va     $17.95 each
28.4- 0 -28.4 Volts at 160 Va at $22.00 each

I have been selling these for gaineclones

These are pictured on my website at   www.apexjr.com
Under misc...

Steve @ Apex Jr.

[mydementia]

Hey guys.
I just got my transformer from partsexpress and my boards from Brian at chipamp.com (prettiest boards I've ever seen!!).  All I'm lacking is some sort of chassis and a heat sink (the 'huge' box I wanted to use isn't quite 'huge' enough...).  I started really looking at the schematics and the transformer label and want to make sure I have it...  Does this look right?


Any recommendations on a place to find a cheap chassis and/or heatsink?

Thanks for looking...
Mike

[joecool85]

eBay for the heatsink, and I would go with an old VCR or stereo amp and gut it for the case.  You can find a busted one on eBay, or you can dumpster dive...or ask a friend who has lots of electronic junk laying around.  I got my case from an old CB radio that didn't work, $1 at a yard sale.