amp design

[demons_wing]

sorry if this is a silly question, but for a beginner, how do you start about designing an audio power amp?...i mean as in equations and generl rules of thumb

[Praise!]

Desired Power output -> circuit topology -> ohm's law and diode drop to deduce values -> choose part # -> PCB layout.

There is really no straight rule. In fact, no 1->2->3 kind of rules work perfectly.

[Solid Snake]

Decide the impedance, power handling and number of the speakers you want to drive and find build a power supply that will match. For example, if you want to build an amp to drive two 8-ohm 100w speakers, a +/-40v 300va transformer would work well. That same transformer would toast your output devices if you tried to drive a 2 ohm subwoofer (the voltage is too high). Also, your transformer sees no difference between driving four 8-ohm speakers and one 2-ohm speaker so take the number of channels into account.

Decide whether you want seperate power supplies for multiple channels. Seperate supplies eliminate crosstalk but if one channel is demanding a lot of power, the other supplies can't help out.

Make sure your output devices are rated to do their job. Low impedance loads require high current devices. High voltage rails require high voltage devices. It's a good idea to buy more than you need. For example, if you're building an amp to push 100 watts, you could use a transistor rated for 100 watts, but once that transistor warms up, it's safe power output falls due to a temperature derating curve. If you started with a device rated at 200 watts, it would be able to provide 100 watts safely while running warm.

Gain and negative feedback affect each other inversely. If I reduce the gain of the output stage on my amp by 3dB, the damping factor (a product of NFB) approximately doubles. Calculating how many dB of NFB you have is almost impossible and useless. Turning off the NFB of my amp to meaure open loop gain causes it to destabilize and is dangerous.

[traderbam]

how do you start about designing an audio power amp?

Nothing silly about this question. But it is rather broad as it stands. Could you bound the question a little?

[demons_wing]

thank you for the reply..i know it is a bit of a broad question, but i was wondering how you work out the tail currents for the input stage. is it related to the output power? and also how do you work out the lag compensation capacitors for the VAS..

i have noticed that some topogies tend to have more caps than others..does this not slow the amp down?..how do you also work out degenration resistors for the output?..

im not sure as to which topologies are the best..i was thinking of maybe 100Wpc, with separate external PSU...i hear the double differential input is good (like the opti-mos -> very similar to the leach!)..

[Praise!]

Originally posted by demons_wing
thank you for the reply..i know it is a bit of a broad question, but i was wondering how you work out the tail currents for the input stage.

Tail current is typ 2mA. If you are using lag comp, the comp cap will depend on the tail current. see below

is it related to the output power?

not at all!

and also how do you work out the lag compensation capacitors for the VAS.

go read the book by Randy Slone, High power audio amp construction manual.

i have noticed that some topogies tend to have more caps than others..

what caps? Power supply smoothing caps? you can use any value you want as long as it is big enough.

does this not slow the amp down?

power supply smoothing caps will not.

..how do you also work out degenration resistors for the output?..

depends on your taste, 0.1R to 0.33R all work.

im not sure as to which topologies are the best..i was thinking of maybe 100Wpc, with separate external PSU...i hear the double differential input is good (like the opti-mos -> very similar to the leach!)..

This is my opinion:
Fully symmetrical (mirror image), DC-coupled, Toshiba's JFETs 389/109 input stage driving cascoded Toshiba's 1815GR/1015GR driving 5 pairs of Toshiba's 5200/1943 emmiter folllowers with overall NFB. Use lead comp instead of lag comp. Input stage load use Toshiba's 3381/1349 as current mirrors. Use Toshiba's 5171/1930 b/w VA and OPS. Set VA bias @ 6mA.
VA cascode bjt use Toshiba's 3423/1360. Do not use caps in feedback path. Use nice quality resistors in feedback. Hope you can visualise the schematic and hope it helps!

Regards,
James
Praise! Audio
Australia

[lt cdr data]

Its an excellent question, one that I do not really know the answer to in depth, but I will say what I do know.

Your design almost ALWAYS starts from where you wish to end up, ie your output.

So you have your target speakers, you need a goal, commercially, it would be the price, the features of the amp, etc., try to get some concrete ideas, thats why sometimes commercially driven can be an aide.

so say 100 watts, 8 ohm speakers, from this you can work out your voltage rails, the rest of your psu inc resevoir cap sizes, diodes, transformer.

Your output device types...bipolar, fet, bias, then your heatsink size, this will all determine the required VA stage, which will go back again and determine the input stage topology.

Then your feedback amount, which determines closed loop gain and sensitivity, they are ALL interrelated to some extent. Its quite a logical and fascinating process.

You can start from the very simple and embellish. no current source in the input, jsut a resistor, replace this with active, then cascode load, current mirrors, mirror image all the way through, parallel up outupt devices for more power(current), as to determing caps, I think it is done by scopes, playing around until the thing doesn't oscillate. Maybe with spice.

The big problem is there is so little info., it has taken me ages to learn stuff, writing is by and large sparce and a lot is VERY poor, not logical, bits omitted, you end up wondering what is what.

Even Linsley hood, god save his soul, his books were not really lucid, a bit hotch potch. Still valuable though.

A lot of it stems from valves, contrary to opinion, they are very similar to transistors, the mechanism of conduction is different, but the design principles pretty much the same, they are simpler to understand, too, transistors build on this with diff pairs, and active loads, both of which came lateish in valves.

So understand valve cascodes and you will have more insight into transistor ones. the reasons behind the topology. Audio is a MASSIVE subject, but it all boils down to amplitude, frequency, waves, phase, resistance, caps, chokes, volts and current, and making things bigger and smaller. That in a nutshell is it.

As to setting currents, I am not sure how the exact value is determined, again valve design helps, its to do with the loading imposed by the next stage.

Its actually something I have thought much about myself, hence it must be a good one , a lot of designers are in the past in respect of many of the developments were done in the 1940s on, and there was loads of info. going around, this is now in old books which are getting dusty, maybe some of the designers grew up with these, now you just can't find the info., and I think some of the makers don't want you to either.

Ask the real experts, the borbelys, slone?, pass, etc. usually, these people can write well.

one thing is funny, like cars, they are all starting to appear similar, as we start to find the bits to optimise. Funny, too, there hasn't been much progress in topolgies since the late 60s.

Finally, what amazes me, though is the sheer variety of designs, even in the simple 3 stager, how many makers, how many different variations.

The last thing I have found is it may look good on paper, but sound quality IMHO is in inverse proportion to how elaborate the circuit is, ie measurement perfection kills sound.

[traderbam]

how you work out the tail currents for the input stage. is it related to the output power?

Typically about 1mA.
The bigger the better until: the power dissipation limit of the input transistors is reached OR the input transistors become too non-linear OR the input impedance becomes too low.
The current must also be adequate to drive the following stage during worst case voltage and current slewing.

and also how do you work out the lag compensation capacitors for the VAS..

Typically 20pF to 100pF.
The smaller the better until the closed-loop response of the amplifier becomes too unstable. This must be verified into worst-case loads.

i have noticed that some topogies tend to have more caps than others..does this not slow the amp down?

Depends where they are used. The voltage response of the amp needn't exceed 20kHz so slowness may be a red herring.

..how do you also work out degenration resistors for the output?..

Typically 0.1 to 0.5 ohms for bipolars.
Related to the thermal characteristics of the BE junction. Used primarily to prevent current drift or runaway as the temperature of the output transistors changes.

im not sure as to which topologies are the best..i was thinking of maybe 100Wpc, with separate external PSU...i hear the double differential input is good (like the opti-mos -> very similar to the leach!)..

You need to suck'em and see.

[millwood]

very good advice.

Originally posted by traderbam
Typically about 1mA.

I usually run mine 4ma for bjt input (driving mosfet vas). I have, however, ran the input stage at over 20ma (sometimes to 40ma) on mosfets. I don't detect any audible difference tho.

[johnnyx]

Douglas Self has published articles on this subject in Electronics World magazine, and published a book called "Audio Power Amplifier Design Handbook" which is very good. It explains pretty much everything. Well worth reading if you want to know about design. There's very little maths, and loads of explanations, illustrated by graphs of simulations and measurements. It's very good. His web site;-
www.dself.dsl.pipex.com
I repeat, It's very good.