best topo for 25W class A into 1 ohm resistive load?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
LineSource said:
What is the best sounding circuit topology to drive a 1 ohm pure resistive load with 25-35 watts of class A power?

The "first watt" is the most important.

This shouldn't be too bad of a problem, and just the sort of thing for a DIY project. (There aren't really any commercial products that will do what you want.)

Let's start with 25 watts, and assume a complementary output stage. This will require 5 A rms, or 7.1 amps peak. Therefore the output stage will need to idle at 3.55 amps in order to maintain class A to full output.

Now this is a fair amount of current, but on the other hand you will only need to swing +/- 7.1 volts peak-to-peak. So let's be generous and use +/-10 volt rails in the output stage. That means the total dissipation in the output stage will only be 71 watts. This is easily handled with a moderately big heatsink.

The latest edition of Self's book will have a lot of useful information for you. I would suggest at least 4 pairs of what he calls "sustained beta" devices such as the MJL1302/MJL3281. You'll want to use very low values of emitter resistors when driving this low impedance load. I would suggest 0.1 ohms if the devices are unmatched but come from the same date code, or else 0.05 ohms if you can match them carefully. They are pretty cheap to purchase ($2 or $3 each), so it's not too painful to buy extras for matching.

Good luck!
 
Mosfets are poor devices for driving low impedance
loads, unless you have loads of them in parallel.

Gate voltage requirements in this case would change
the needed voltage rails from 10-0-10V to 14-0-14V
just for the same output voltage swing into a normal
load, adding the fact you need more gate voltage for
high current your heading towards 18-0-18v or more.

BJT's are the only sensible option here, a Mosfet output
stage would have at least 3 x the static dissapation,
i.e. > 200w for 25W output.

:) sreten.
 
Hi Linesource,

A 1 ohm load ? and class-A ?

First thought is a matching transformer.

Second thought, I agree with Millwood, the heat in the output devices will be high, even with paralleled devices.

Bipolar transistors need a headroom voltage to operate correctly at high audio frequencies, generally 5V minimum across each half, so you are immediately thinking about Vp-p plus 10V at whatever quiescent current is necessary.

The quiescent current also needs to be more than half of the peak current for proper high audio frequency linearity, this is because of unavoidable gain droop at the high class-A operating temperature with low Vce at maximum output amplitude.
I-q often needs to equal I-p !

So for 25W I am suggesting you'll need 25V @ 7A = 175W, which means a double parallel push-pull output stage, using the devices that Charles suggests and massive heatsinks. The voltage is low so SOAR limitations would not be a problem.

Topology, take your pick.

If your design outputs more than 25W then you won't grumble, and 40W might be possible at low frequencies.

What about the never-off class-B biasing suggested elsewhere, in diy.audio. Maybe by John. Basically this is class-B operation but no crossover switching distortion because neither device actually turns fully off.

Maybe Mosfets would do better, I do not have experience here.

Cheers ........ Graham.
 
Graham Maynard said:
Bipolar transistors need a headroom voltage to operate correctly at high audio frequencies, generally 5V minimum across each half, so you are immediately thinking about Vp-p plus 10V at whatever quiescent current is necessary.


So for 25W I am suggesting you'll need 25V @ 7A = 175W, which means a double parallel push-pull output stage, using the devices that Charles suggests and massive heatsinks. The voltage is low so SOAR limitations would not be a problem.


Maybe Mosfets would do better, I do not have experience here.

Cheers ........ Graham.

Well I'd suggest your 5V requirement per BJT is garbage,
and contrary to the laws of semiconductor physics,

:) sreten.
 
INPUT and DRIVE STAGEs

Any better output solution than this?
Output Power Supply +/- 12V - 15V transformers that can deliver at least 15 amps on each leg, 750-1,000VA.
Outputs 6 pairs of the MJL1302/MJL3281 with about 0.2 ohm emitter resistors...Buy from ON semi...(Motorola)
Large 0.35 C/watt heatsink for output transistors and thermal compensation components.

Now...any suggestions or pointers to schematics for a very low noise input and voltage drive stage? Inputs JFETs? Low noise bipolar? The driver power supply will be regulated, but I think current sources sound better than resistor loads. I'll add an opamp to null the DC offset.


Graham...I have found that step-up transformers shift the phase and move the soundstage around. I currently have a big Krell, but I can hear noise in the first 1-2 watts, the most important region. I can hear the ocean waves breaking..........unless I pad it up to 2 ohms.
 
I see no reason to go past +/- 10 volt for the output stage rails. Of course, this means that the transformer only needs to be around 14 VCT. This is kind of an odd value, it will be much easier to find 12.6 VCT, which will give rails around +/- 8 volts. This will still give the 25 watts you are looking for.

As far as current rating on the transformer, you only need 7 amps per channel as a minimum. Remember the circuit idles at 3.5 amps per channel, the only reason you need 7 amp transformers is because a capacitor input power supply is only about 50% efficient in terms of transformer utilization. But ths means that for a 12.6 volt center-tap transformer you only need 12.6 x 7 = 88 VA, which is 10x less than you were speculating! Going with more current won't hurt, but it will be pretty hard to find something much beyond 10 - 15 amps. The part you were suggesting would have a secondary rating somewhere between 50 - 100 amps. You might find something like that in an arc-welder!

Six pairs of outputs should be enough. I would use a smaller emitter resistor than 0.2, but that's a relatively small detail.

As far as schematics, I would suggest grafting the output stage of the Leach amp (triple emitter follower):

http://users.ece.gatech.edu/~mleach/lowtim/

on to the front end that Jam has successfully built:

http://www.diyaudio.com/forums/showthread.php?s=&threadid=30439

In fact for that application, you won't even need that much gain. Instead of the normal 20x used for power amps, you probably only need 3x or so. That way it will clip with around 1.5 Vrms input.

Good luck!
 
Hi sreten and Millwood.

I'm writing about class-A audio reproduction, not just the barest requirements for steady state bench testing into a passive resistor load; been there - 35yrs ago - and won't do it again.
You end up with a chassis that will heat a resistor to 25W but sound weak beside other 25W designs.

Cbc and pull-down requirements increase with low Vce, also the gain bandwidth product additionally drops with junction temp; both impairments compromise drive linearity and NFB capabilities when the load is a dynamic loudspeaker.

sreten - 'garbage' is not a technical term, nor criticism, nor proof !!! Are we talking about a consummate 'audio' design or a simulation / workbench project ?

One ohm is not a solid state friendly load, the junctions will dissipate so much more than the load during bass waveforms, especially at frequencies just above the main loudseaker resonance.

Of course the nature of the drive and bias has relevence here, but I have not made any assumptions in this regard. Maybe you have something specific in mind ?

Also 25W to 35W was requested. It is much more difficult to retrospectively upgrade a completed amplifier, thus better to play safe during construction.


Hi Linesource,

Six pairs is way more than OnSemi-Motorola designers would use, and now I realise that you are thinking about a significant construction. They suggest 3 pairs for 100W into 4 to 8 ohms, which is the same order of current, but with much greater SOAR limit incursion. Two pairs would work with 5V headroom, so more can be more efficient, but as you increase the parallels you must consider the capacitive loading that is presented to output stage drivers.


I quite understand the soundstage effects you mention due to very low levels of series inductance and resistance altering loudspeaker waveform. The waveforms, and hence the image, shift within minute fractions of time and relative position with amplitude and frequency, effectively shimmying about their nominally intended position, thus losing positional focus due to some dominant note inducing a loudspeaker generated back emf. (Change the sound level, or the loudspeakers or the amplifier and you change the effect, or make it disappear from that same moment on a repeat playback.)

You most certainly would not want this to happen due to dynamic audio waveforms causing loudspeaker generated back emfs to drive a simplistically biased class-A beyond its nominal steady state range.

If you can match/select output devices then 0.1 ohm emitter resistors should be fine...... non-inductive though !


For long term class-A usage I triple the DC power consumption in Watts when choosing a suitable VA transformer rating. It is possible to get away with VA = 2xWdc if the transformer has good convection cooling, otherwise temperature will steadily rise with long term playback.

My choice is for seperate channel monoblocks, each directly behind the loudspeaker with short cables, and each with its own mains/line transformer. Also gives you the chance to test a single prototype before doubling up on parts cost.



Cheers for now ............... Graham.
 
seems like fets and op amps could be made to work

i particularly like the "monoblock" idea, try floating high current supplies (use split bobbin EI xformer for low pri-sec C)

this is just a suggestion, needs some engineering to be a amp...

(D1 just helps spice find the bias point)
 

Attachments

  • 1ohm.png
    1ohm.png
    43.6 KB · Views: 3,289
Hi jcx,

I've had under-floor and isolated PSUs connected via thick cables plus surface mounted wall boxes beside each amp. Transformers and especially electrolytics were much larger in the seventies.

Anyone know if Bengt Ollsen is still alive ?

He designed a fully working amp in the 90's using a mirrored differential input, fet VAS, and Mosfet non-complementary amp that was excellent with awkward low impedance loads.

jcx, where you have floating power supplies, Bengt's were conventional, however he had a floating mirror where you have the ground referenced 'I1' His circuit was published in Electronics World in 1994, but would not be quickly accessible in my reference (piles) files. Give me a day and will see if I can post it.

Cheers ............. Graham.
 
Graham Maynard said:
Sreten,

I think you are thinking about bipolar class-B, where there is more drive to pull down the base charges in a low Vce transistor, which due to semiconductor fundamenals stores much more (NFB driven) charge and then cannot turn off again quickly.

Cheers ........ Graham.

????????

Its not a question of what I may be thinking about,
and implying I'm confused is smoke and mirrors.

You've made two statements about BJT Class A operation
which have no theorectical basis and you cannot justify.

I cannot disprove the first it as you've given no
sensible basis for the alledged phenomenon.

The 2nd is simply untrue, gain droop in the output transistors
does not affect the class A biasing current required at all.

:) sreten.
 
Hi Sreten,


I repeat that I am not writing about steady state bench testing, but about the real-world loudspeaker reproduction requirements that are dynamically imposed upon an amplifier, and which can cause premature stereo image degradation.

The nature of output stage drive is important, but a circuit is not class-A if it runs the devices close to transistor action degeneration due to falling operating voltage, or hf slew is limited because one half is not quickly enough turned off by class-A drive.

Unless you ensure about +/- 5V headroom and make quiescent way above 1/2 peak the amplifier will distort prematurely due to loudspeaker induced back emfs. Unless you listen to and study your class-A circuits with the real world loads you expect them to drive, then you will not know of the out of phase demands that are placed upon output devices and their drive circuitry.

I am attaching here the current and voltage plots of a 200Hz sinewave into the Ariel loudspeaker. Information is freely available on the web. Note they say that most transistor amplifiers do not drive it as well as valve amplifiers normally do.

First trace is amplifier output. Second trace bass driver output which is delayed due to crossover circuitry energy storage. Third trace is amplifier current. Note how the current increases after the first cycle.

A class-A design that is biased to output only the theoretical resistive maximum, will distort the second cycle and oblige you turn the volume down. Result - it won't sound like the amplifier you thought you had, in spite of what you saw with your own eyes on the test bench. It is when these unexpected current demands arise that the simplistically biased output transistors cannot behave as expected.

Similar affectations arise with hf second cycles where base junction carrier retention can prematurely cause slew rate limiting unless the charges are pulled out by a powerful driver stage, much as with Mosfets. The MJL3281A is a brilliant transistor but has appreciable junction capacitance if run to a low collector minus base voltage.

I have NEVER seen a published loudspeaker tested class-A bipolar design where the quiescent was half of the resistive peak. There are however some class-B designs that are class-A biased, but which then always run into class-AB through loudspeaker back emf induced impedance dips.


I cannot advise linesource to go below +/- 5 headroom and quiescent less than peak with shared devices for bipolar circuitry.

Mosfets might work better with at 1 ohm. I believe there are more 1 ohm capable Mosfet amps in the car audio scene than bipolar.


Cheers for now.............. Graham.
 
Transistor turn-off is not an issue for Class A - thats one its points.

The question is 25W class A into 1R.

Its not for me to decide if this wrong, intended load is not given.

Certainly for 8 ohm nominal loading you shouldn't design
class A for 8R, you should design for 4R to 6R depending
on the impedance characteristic of the 8 ohm speakers.
For 4ohm speakers design for 2R to 3R.

Your 5V headroom for each BJT's I've never seen described,
or even alluded to in all the technical literature I've read.

Its Mosfets that need at least 5V each, at least in
voltage swing and thus rails of the driver stage.

:) sreten.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.