A diy amp for looow impedance

This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Bottom line: I don't know much about amplifiers. I know just enough basic electronic theory to be dangerous. So I'm hoping some resident amp guru will answer the following naive questions for me:

I'm putting together a line array of around 30 drivers/channel (cheapo PE 269-469). This ought to be about 104dB/W efficient, so I figure 10 watts should drive me out of the room.

I'd also like to build a one-off amp to run these because... (Here's where I swerve into the weeds)... I'd like to wire all the drivers in parallel to eliminate back EMF interactions. In this case it would work out to a nominal load of just over a quarter of an ohm!

Before you laugh (or after you've finished), let me elaborate on the beauty of the concept: Such an absurdly low impedence certainly wouldn't need many volts to drive it. So I got wondering, could I construct a pure current amplifier that would simply use the line level voltage ( about 2-7 volts) from a CD player or DAC to directly drive a bank of MOSFET outputs?

I'm picturing a stepped attenuator on the line for volume control. This chain would eliminate the preamp and the voltage amplifier stages of the power amplifier. Less messing with the signal!

Now, I realize that there's probably a reason why I've never seen an amplifier capable of driving .25 ohm comfortably. What is that reason? Is it merely the grand ol' 4-16 ohm tradition?

Are there issues that arise when output voltage swing is so small compared to current? Are their weird, funky interactions with the inductive load of the speakers at these voltages/currents? Issues of cost/practicality/zodiacal alignment?

What would stand in the way of me doing this?

I'd deeply appreciate any/all input.

Increasing current increases distortion in the
output stage; can't get around that. You'd have
to parallel a bunch of MOSFETS and the power supply
would have to be rated for a lot more current--bigger
transformer with a hefty secondary winding. Even
with MOSFETs, the drive current might be significant.

Wiring resistance would start to become signficant, too.

Seems like a lot of trouble when existing designs are
already into vanishingly low distortion.
Yeah, the current output to the speakers at 10 watts would be the sq. root of 40A--a little more than 6.5A! Wiring would have to be heavy gauge.

While I recognize some of these practical issues, I'm still trying to figure out what is and is not possible--an exercise in "what if." (Also, I'd really like to wire my array in parallel!)

As for the attraction of the source directly driving the output stage, I was indulging in a little less-is-more philosophy. (However, in this case I guess less impedence equals more MOSFETs! What's that Enstein quote Nelson Pass cited? Make things as simple as possible, but no simpler?)

Anyway, you're saying that it is possible to drive .25 ohm with enough power supply, MOSFETs, and robust build? I understand (theoretically) that MOSFET gates draw almost no current. What about charging the gate capacitance of a large bank of MOSFETs? Would that put an undue load on a typical line-level source?

I'm also still curious if there are any interaction issues with driving speakers this way. Any increased chance of ocillation?
Bill, original thought scares people. Anxiety complex over, I don’t think that what you are suggesting is likely to be practical or an improvement over current designs, but I’ve been wrong before. There are lots of issues with your design, both with the speakers, cables as well as the amp, I think you’ve realised that already.

Just a few issues to add to those already mentioned.
I don’t think that the transformer size will be an issue, even though the amps efficiency is likely to be lower than normal. The power supply filtering will have to be massive, I can only imagine what the ripple current would be like. Also with so little voltage you may have to resort to current feedback. There are semiconductors around that will handle the current easily, check out some of the new IGBTs, they may be a better solution than FETs.
You may want to seek out designs for industrial servo amplifiers.

If you parallel drivers (speakers) they should be matched to have the same properties and enclosure loading/dampening, other wise each speaker can get driven by the other speakers as well as the amp. That is, at some frequencies they can start doing their own thing. Even the same model drivers from the same manufacturer can have different properties. Having said that it is often done

I can’t wait to hear all the other problems that will come up. If you really what do go ahead please let us all know how it turns out. He who is afraid of failure is unlikely to discovery anything new. Even if you don’t build it you’ve woken me up.

Regards WALKER

PS Check out Elliot Sound Products's site on speaker impedance.

[Edited by walker on 11-15-2001 at 03:15 PM]
Well, I guess that's the mixed blessing of being an amplifier neophyte--blissful ignorance! Until I accumulate a sufficient reservoir of knowlege/experience, I'll be both misguided and original.

As to the parallel wiring of speaker drivers, my array is open baffle, so cabinet loading/damping aren't invited to the party. Also, the drivers are not only identical models, but also have matching lot numbers. Short of hand matching, they're as close as I can get.

Correct me if I'm wrong, but doesn't parallel wiring minimize driver interaction, especially in a massivly parallel array? It seems to me that any series wiring would piggyback deviations in resonance and inductive reactance, forcing drivers to modulate each other.

My (possibly flawed) perception is that in a .25-ohm parallel array of 32 8-ohm drivers, the vast majority of back EMF will be dissapated in the amp output circuit (assuming reasonably low output impedence) rather than in any (comparatively high impedence) driver. That was the most attractive part of my .25-ohm amp, you see!

Is this misguided, original, both, or neither? :)

Bring me back if I'm in the weeds again.
Two points occur to me:
1) The current draw is going to be ferocious. Your main power transformers are likely to end up being filament transformers or the like. That current will need to be dumped into the circuit as bias, so depending on what rail voltage you end up with, you could very well end up generating a lot of heat, even if you go class AB or B.
2) You won't have much damping factor from the amp side of things. Something that frequently gets overlooked is that the lower the driver impedance, the lower the damping factor. If you're looking at .25 ohms for the speaker array, you're going to have to have a vanishingly small output impedance in the back end of this thing. That said, the drivers will to some extent serve to damp each other. I hope the cones/diaphragms/whatever are low mass (hence less needful of a damping factor to crack the whip should they get out of line). What kind of drivers are you planning on?

Bill, thankyou for a great thread mate, (look out he's from downunder). I would like to start by restating that I think you are barking up the wrong tree:) Having said that even if you go ahead and to doesn't achieve the desired results we will have learnt for your endeavours.

I think that paralleling drivers still allows rogues to miss behave but not as much as series connections. If one cone is doing it's own thing and say moving outward, another cone could be operating at the same resonant frequency but 180 deg out of phase. Nett result is that the amp's happy as are both drivers. But don't let this small issue worry you as many a fine speaker has paralleled drivers, (my beloved AR9s included).

Regarding the amp let's stray a little further into the weeds, it's a nice day for it. Have you considered individual, open ended, (without overall negative feedback, relying on the local device feedback) output devices connected to individual drivers. Bingo dampening issue over. I'm scaring myself now. Reduces some of the concerns of the output stage. Don't expect textbook distortion figures, but I rarely listen to textbooks on my system anyway.

Regards WALKER

Here's to the road less travelled!
Yikes! Yeah, you're really far off into the swamp on this one! Well, let me put on my hip-waders. Hmmm, lets see.....

Well, you're going to be battling the open-loop output impedance of the amp (read: inefficient output stage = lots of heat). Even after feedback, your damping factor will probably still be less than optimal, as Grey suggests. I would shy away from MOSFETs here, as they do have a high Vgs (potentially high power loss in the output stage, depending on your circuit), and less power handling capability than BJTs. A bipolar implementation might work better... but i'm waffling on this one. BJTs have their limitations too, such as poorer linearity at high current, and demanding more of the driver stage. hrm... this is a complicated equation to weigh. Not to stifle your creativity, as original thinking is always good, but I think this idea poses too many practical challenges... maybe a slight rethink is what's needed:

Have you considered multiple amplifiers to drive parallel arrays of fewer drivers? So instead of say 30 drivers, you might have six banks of 5 drivers, each bank with it's own amp, or at least portion of an amp. This might bring things down to a more manageable level, and each amp wouldn't have to put out 10W, maybe only 2W each. I think this would make the equation a whole lot more manageable...

as far as the current amplifier goes, I'll have to put on my thinking cap sometime... sometime when I can find it. Right now, I should get some sleep.
Ok, now for some synthesis:

It sounds like the recipe for a 10W amp at this spec would look like this: Take one half-ton toroid, two bushels of chips, and a substation, mix thoroughly, solder to taste, and serve with asbestos room treatment.

Maybe I'll shoot for just a couple watts pushing half an ohm--two amps and a volt at the output terminals. (Not much juice, but my speaker array models to be 104+ dB/W.) Could this work with plain vanilla MOSFETs? Is this moving in the direction of reasonability (especially considering from whence I came)? Should serve as a good disproof-of-concept, anyway.


I like your ideas of discrete drives for subsections of the array. How about one driver stage feeding 8 small complimentary pairs each driving 2-4 ohms? If I insert a continuously variable phase control in front of each output, I could dial in an electronically focused (virtually curved) array or maybe a polar response something like a point source...

...see how it is? just as I begin to find the path, I'm diving back into the weeds.

So much for synthesis.
Don't you want to go more than 10W? Sure that would be loud, but those 30 drivers must be able to handle some serious power. If I were you, I would put the drivers in series-parallel so they total about 2 or 4 ohms per channel stereo, and hook them up to an amp that can put out as much power as the drivers can handle in case you ever want to make LOTS of noise! Like others have said, there are many problems with building an amp for low such impedance, it would be much easier to do it for 2, 4, or even 8 ohms. There's nothing saying you have to put all those speakers in parallel, series-parallel should work much better. With all those drivers, you could achieve about any impedance you want.

Good Luck! :)

[Edited by Kilowatt on 11-16-2001 at 07:47 PM]
Yes, my arrays could easily handle 300W+ per channel. At that power, my smallish listening room would be serenaded with peak SPLs of nearly 130dB--and I'd be crawling down the hall in the opposite direction, blood trickling from my ears, asking myself if I was having fun yet.

I do like to turn it up till the sound becomes compelling, but quality is the goal.

That's some of the attraction of arrays. I'm planning on 32 full-range drivers per side (about 3.7" cone diameters, 90 dB/W 15W max pwr.) Working together, they roughly equate to a 20" driver of 105 dB/W while maintaining the mechanical characteristics of small drivers.

So, as I enjoy powerful tunes at peak SPLs near 110 dB, my amp is putting out 4 watts per channel and each driver is dissapating 1/8 of a watt. Cone excursions stay very small and IM distortion is a thing of the past. (Beware: IM distortion crops up way before X-max.)

Truth is, 10W is all kinds of headroom, as far as I'm concerned.
Does anyone here remember the discussion on the BASS List a
while back where Tom Danley and Nathan Sargeant were
speculating how to power the Servodrive? This was a clever
and original way to overcome distortion in sub-bass drivers
by driving the cone from a rotary electric motor with a
servo control to reduce the distortion. Only thing was, the
motor that was the most suitable in every other respect for
this application turned out to have a DC resistance of under
an ohm. I think Tom designed an amp with an exotic
transformer-coupled output, but I forget whether anyone
actually built the amp or they found another motor.

Details of the Servodrive for those interested are to be
found at http://www.servodrive.com/servo .

Dear Sir.....

I do believe it's a great idea... like the Startrek phrase..
... to (??) .. go where none have gone before.. well or something like that... anyways....

Look at the good side to it:

1] very high efficientcy: less power needed :)

2] using more drivers, every driver needs less excursion so will be more lineair, in you case : very lineair :)

3] current amplifier can be very very simple mosfet buffer, since 2 volt RMS will be just fine... .... :)

4] such a amp will have a distortion of less then 0.1% wich will be just fine since normal BASS drivers have distortion of 1-10% (!!!) .. your total system linearity will be very good indeed

4] using low voltage MOSFET's will keep input capacity low, but this doesn't matter since it's a BASS-amp, high Hz are not a issue ... :)

5] damping factor (not that important) .. parralel MOSFET have very low on-resistance.. no issue here

6] low voltage capacitor for the powersupply are cheap.. maybe 5 Volt Cap's will do just fine .. you may be able to get your self some whole Farads of capacitance :) .. you do need power rectifiers... hmmmmmmm

1] using a very simple 2 stage amplifier give feedback if you want to

2] use as many MOSFET as you have speaker ($$$...:-( )

3] use BIG cable, like car-battery cables, parralel

4] use heavy filament trannys, 5 or 6 of them will deliver 18 Amperes ... that not impossible to get relitive cheap.. or build yourself 10 regulated voltages supplies of 1 ampere.. also not a problem at 5-6 volt ..better yet

5] use ver very cheap voltage supplies from second-hand 10-15 hear old computer supplies... YES

6] you can try you concept first with just 4 or 6 speakers..

welllll....... just do it!!! and be the bravest DIY-man on this forum!!!.. and tell us about the performance...

A different way to go

I saw a project like this years ago - the idea was to limit distortion due to the Doppler effect in full range speakers.
Apparently it sounded superb.

Anyway, it seems to me like you have three choices:

1. Buy Krell or McIntosh monoblocs; they would have no problem driving the load, but you probably can't afford them.

2. Build you own; if you don't mind waiting 'til 2004 to listen to your creation - the learning curve on anything but the most mundane power amps is time-consuming and, again, very expensive.

3. Why not amplify each speaker individually? Use a chip like National's LM3876 (about $7cdn at Digi-Key) and make lots of small amps. Huge advantage - as soon as you make the first two, you'll have something to listen to.

Cheers, Peter
Petermac gave you a good tip, but, there are some problems in this arrange, specially to put all of these chips working at same way. But I think it will be easier than deal with the very low impedances needed to amplifier the current and to connect your array. I think, considering 2 or 3 volts RMS at output and 6 or 7 amperes, you won't need very much mosfets. Your cables, however, will need to be very huge.

By the way, in my opinion, this kind of question and thoughts are the best think in this Forum.

If you decide to go ahead with your idea, let us to know about its development.


I have built some years ago a dc-100 Hz 25 W rms amplifier for special purpose, with loading impedance of 1 ohm. It is difficult to maintain low level distortion at such low impedance, and the global efficiency of the amplifier is poor.

Another difficulty is to design dc power supply of low voltage, but high intensity, which requires big condensers and inductors. Efficiency is poor too, because of the voltage drop in the rectifiers.

In my opinion, series-parallel wiring of your drivers, with a resultant impedance of 4 ohms (8 parallel of 4 serial) is the most reasonable solution.

In another way, you can experiment with a cheap auto IC amplifier for *each* driver, paralleling inputs of this ICs, and a general stabilized power supply of 13.8 Vdc, 10 A, which is available at low cost for amateur-radio transmitters. Furthermore, you can modify level and/or frequency response of each driver, as necessary. Total power output would be 64 W RMS or so !

Regards, P.Lacombe
Square Wave

Be careful with your drivers. I don't know what the rms wattage of your drivers are, but watch for clipping in your amplifier. Clipping will burn the coils out of your drivers. At the output impedances your discussing, there will be a lot of demand on your amp.
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.