A Simple 10W Direct-Coupled Class-A Power Amplifier

[JR.]

I am not a serious student of class A amp design but apparently some of them did use regulated supplies to reduce PS ripple (like old Mark Levinson class A).

In fact that same PS ripple will be there for class A/B but only during loud parts, more of a modulation noise under and mostly concealed by the loud music. Having worst case PS ripple while the amp is idling makes it easier to hear.

I have been thinking about this some, after beer o'clock and lean toward local independent output stage class A bias. One for each output polarity where both output stages have a minimum current output (say 50 mA). So they can make more output current but never less than minimum. In use this would look like a class A/B with slightly higher class A bias, but output devices never turn off for low/no crossover distortion.

The output stage bias would use local feedback, while global overall negative feedback would control both outputs.

Sorry for the veer, just thinking out loud. This is not practical for your design. For your design i still lean toward experimenting with modulating the class A bias, knowing the anticipated load impedance. So you could dial in less class A bias for low level signals and ramp it up for louder music.

Sorry, now I'm worrying about efficiency in class A... color me green.

JR

[mediatechnology]

I have been thinking about this some, after beer o'clock and lean toward local independent output stage class A bias.

That's exactly what we find here:


Image from: http://www.mimave.net/chibidac/mdd2955.html

ON Semi make the NJL3281/NJL1302 TO-264 "Thermal Track" devices that have sense diodes in the package: http://www.onsemi.com/pub_link/Collater ... 281D-D.PDF
There's a whole series of these parts.

With regard to 120 Hz ingress into clipping I don't seriously think I'll be driving it into the walls that often.

With an unregulated bulk Vcc/Vee supply and a regulated sub-supply for the pre-driver I can fiddle the pre-driver supply to pre-clip the output base drive.
I realize that doesn't give me that "class-A" clipping warmth but what I'm really looking for is just good low-level performance.
At right up to the clipping point, wherever that turns out to be, the THD should be around the measured 0.04%. (1 kHz)
And if I do want a lot of warmth (i.e. color) I can just crank down the idle current.
I'm giving myself a lot of room to experiment: I suspect that ultimately the heatsink will limit the overall output level and THD performance.
As far as idle current goes there does appear to be a sweet spot. Too much at some levels and THD increases.

I do like ricardo's switcher idea.

I found some HP 65W 18V on eBay for $8.95.
Mouser have them for around $31 each.
(Mouser also has a 48V medical-grade switcher that I found interesting.)
Two table-top switcher bricks are required.

I'd like to be able to show both power supply options, linear and switching, to someone interested in building one.

Based on the amount of discrete devices used in the JLH, Zen and DoZen (and the issues with PSU hum rejection) the DCAO board or the schematic above seem to have the "Simple" factor in their favor.

[JR.]

That's exactly what we find here:


Image from: http://www.mimave.net/chibidac/mdd2955.html

Not exactly but close (there are probably more). That schematic delivers class A as low as about 4 ohm load before the opposite driver cuts off. But... I was talking about local bias control with a global NF loop around the whole thing. That amp is open loop a little like yours with the voltage feedback taken from before the emitter resistors.

ON Semi make the NJL3281/NJL1302 TO-264 "Thermal Track" devices that have sense diodes in the package: http://www.onsemi.com/pub_link/Collater ... 281D-D.PDF
There's a whole series of these parts.

Nice, but I did not see a mention of diode voltage tolerance. Back at Peavey we made amps without bias trims so purchased special diodes from Motorola that were guaranteed to hit predictable voltage vs current. These diodes used in combination with known motorola power transistors delivers acceptable trim free class AB amps. Those parts sure look like the Vbe and diode voltage should be similar and repeatable from identical process.

I mentioned in passing, and may have even posted a schematic somewhere around here of my auto bias scheme that did not rely upon power transistor Vbe thermal tracking. IIRC I used a transistor pair sensing across the emitter resistors in a low gain amp inside the bias network.

With regard to 120 Hz ingress into clipping I don't seriously think I'll be driving it into the walls that often.

With an unregulated bulk Vcc/Vee supply and a regulated sub-supply for the pre-driver I can fiddle the pre-driver supply to pre-clip the output base drive.
I realize that doesn't give me that "class-A" clipping warmth but what I'm really looking for is just good low-level performance.

I am not aware of a class A clipping warmth.

At right up to the clipping point, wherever that turns out to be, the THD should be around the measured 0.04%. (1 kHz)
And if I do want a lot of warmth (i.e. color) I can just crank down the idle current.
I'm giving myself a lot of room to experiment: I suspect that ultimately the heatsink will limit the overall output level and THD performance.
As far as idle current goes there does appear to be a sweet spot. Too much at some levels and THD increases.

I do like ricardo's switcher idea.

I found some HP 65W 18V on eBay for $8.95.
Mouser have them for around $31 each.
(Mouser also has a 48V medical-grade switcher that I found interesting.)
Two table-top switcher bricks are required.

I'd like to be able to show both power supply options, linear and switching, to someone interested in building one.

Based on the amount of discrete devices used in the JLH, Zen and DoZen (and the issues with PSU hum rejection) the DCAO board or the schematic above seem to have the "Simple" factor in their favor.

KISS is good...

JR

[edit] re switchers +1 I purchased a few relatively inexpensive switchers for my Peltier heater project. Seem like good value [/edit]

[ricardo]

I even had one disappointing listening trial where a 1500W amp without DDT was considered better/louder than a 1800W amp with DDT.

First you have to 'calibrate' your Blind Listening Panel.

Once you weed out the wannabe Golden Pinnae and are confident you have true golden pinnae who can return repeatable & reliable results, you learn to take their comments very seriously.

My most successful compressor/limiter/clipper application, I dubbed Powered Integrated Super Sub technology. It limited & compressed active EQ'd speakers & subs so the bass unit never exceeded its limits. The limiter/compressor had a tailored THD profile so the resulting speaker sounded like a much larger, more powerful one. If you turned the limiter off, you didn't actually gain any loudness or bass (cos that was determined by the physical drive unit) .. but it sounded worse.

For some reason, Marketing never liked the name.

IMHO, for designing stuff like limiters & compressors, Listening Tests are essential.

FWIW these tests were performed decades ago and the amount of clipping tolerated in recorded media has increased dramatically since then.

It's a shame when we finally have inexpensive gear that will record stuff to quality well beyond, the limits of human perception, the industry is obsessed with destroying Dynamic Range. Dynamic Range .. you know, Loud bits and Soft bits

One of Bob Cordell's favourite tracks for demonstrating Dynamic Range (by Riki Lee Jones IIRC) is apparently compressed by more than an extra 10dB on the latest re-issue.

[JR.]

IMHO, for designing stuff like limiters & compressors, Listening Tests are essential.

I spent way too many solitary hours on my design/test bench designing dynamics by looking and listening.

FWIW there are objective aspects and subjective aspects to dynamics design. Objective stuff like control voltage feedthrough generally sound bad to everybody, some people may like distortion but VCA designers (I have one patent for a VCA design) prefer low distortion. Time constants are more subjective for single ended dynamics... I killed a bunch of brain cells designing encode/decode NR where the overall goal was to sound transparent (just quieter). Actually more complicated that it first appears due to cassette tape as the medium between compression and expansion.

To bring this back on topic of clipping power amps, I would characterize the Peavey DDT as a fast/clean limiter. I won't name names, but I am aware of competitor power amp clip limiters that actually under-protected with a slower than most effective attack time, that sounded "louder" because it was. , but the customers were happy. All my customers heard from clean clip limiting was less loudness than just letting the amp clip.

It's a shame when we finally have inexpensive gear that will record stuff to quality well beyond, the limits of human perception, the industry is obsessed with destroying Dynamic Range. Dynamic Range .. you know, Loud bits and Soft bits

One of Bob Cordell's favourite tracks for demonstrating Dynamic Range (by Riki Lee Jones IIRC) is apparently compressed by more than an extra 10dB on the latest re-issue.

I was not satisfied with just using a dynamic recording, I designed a special piece of test equipment that could make sound sources more dynamic. Basically it was a tone burst generator when I fed it a sine wave source, but when I fed it music it would gate that signal on/off synchronized to zero crossings so it didn't sound clicky. The amount of off attenuation was adjustable, so i could literally use a music source and add bursts of 3-10 dB more gain for short periodic on/off intervals. While never very natural sounding, I could make any source dynamic enough the stress the most well behaved compressor/limiter. I was obviously parsing out artifacts that may be hard to replicate using even Riki Lee Jones. It is frustrating trying to replicate a random beep or fart, so I decided to make them on demand with hyper dynamic sound sources.

JR

PS sorry about the veer...

[mediatechnology]

That amp is open loop a little like yours with the voltage feedback taken from before the emitter resistors.

I think where they differ is that mine takes feedback from the outputs of the DRV134 which is the base drive.
The one I cited from the emitters.
Neither take feedback from the final output however so I guess its fair to call both open loop outputs.

I bet his circuit performs pretty well.
I'm not complaining about mine however...
I think we both use about the same idle current.

I am not aware of a class A clipping warmth.

I think that's the sound Doug and I were discussing earlier when class-A circuits without feedback compressively soft clip.

Got my big transformer today but haven't unpacked it yet.

[mediatechnology]

Power Supply Tests

Got the Hammond 1182H15 30VCT/7.5A transformer unpacked and hooked up.
Its perfect for the linear supply.

With 120V line and about 2.6A idle current (1.3A/Ch) the bulk supply is +/-18.8V.
This puts the idle power at about 49W/Ch.
The loaded ripple voltage using the 10,000 uF caps that I have is about 600 mV p-p.
The secondary rms current is 3.9A.

What I do need is a sub-regulated supply for the driver board.
600 mV is too much ripple.
I can do that two ways.

One would be a zener regulator using component locations already on the board.
To have enough regulation that solution would dictate +/-15V to +/-16V Zeners and limit the output to just under 10W.

A second option, one I prefer, would be to use one of the vintage Pico Compressor Power supply boards and use it as a sub-supply.
A second low-current transformer, 36VCT, might be needed.
Using a sub-supply board allows me separate filtering and the ability to dial the driver supply voltage in to +/-17V to provide 10W and a little change.
(The +/-18V limit for the DRV134 establishes the maximum output power.)

I can also use the sub-supply for the headphone amp.

[ricardo]

To bring this back on topic of clipping power amps, I would characterize the Peavey DDT as a fast/clean limiter. I won't name names, but I am aware of competitor power amp clip limiters that actually under-protected with a slower than most effective attack time, that sounded "louder" because it was. , but the customers were happy. All my customers heard from clean clip limiting was less loudness than just letting the amp clip.

That's actually my point.

If you can arrange your 'clipper' to introduce mainly 3rd harmonic (on a sine wave), it can sound both louder AND cleaner than stuff which either reduced the level such that no clipping occurred (Peavey DDT) or that allowed squaring of the wave forms. One could liken this to 'compression' introduced by mag. tape saturation.

But I agree with JR about control voltage feedthrough and other evil stuff.

There are all sorts of subtleties which can aid or hinder good sound. eg the Powered Integrated Super Stuff was sorta 'band-limited' so I could do some really dubious stuff and get away with it (have it sound better bla bla).

Again I hasten to add, the serious work was on the Powered Integrated Super Stuff ... where speaker 'clipping' & protection was paramount and amp clipping an important but secondary concern. The simple amp diode clipper was only a short exploration.
__________________________________

On an even more OT vein, but perhaps relevant to 'Class A' amp sound, you can make a clean digital source sound very much like vinyl by simming the huge increase in HF THD with level. I chanced upon this in the early days when trying to soup up my early Sony PCM F1 recordings.

You use a crude single BJT Baxandall tone control to apply max HF boost, followed by another applying max cut. All you need then is something to do the clicks & pops The 2 x crude single BJT tone controls introduced sufficient extra hiss.
__________________________________

Wayne, you gonna try the HP laptop brick PSUs?

[mediatechnology]

Wayne, you gonna try the HP laptop brick PSUs?

Yep, either the HP or ones like them from Mouser.
I think the HP at the too-good-to-be-true price of $8.95 may be the recalled ones that start fires.
Mouser's price on 2X 65W 18V are about the same as the power transformer alone around $62.

But wow, that 7.5A toroid is a thing of beauty.
Feels good in your hand. Like a bowling ball.

I bought several of the Meanwell 25W bipolar supplies last year and may try them on the headphone amp first.

[JR.]

There are all sorts of subtleties which can aid or hinder good sound. eg the Powered Integrated Super Stuff was sorta 'band-limited' so I could do some really dubious stuff and get away with it (have it sound better bla bla).

__________________________________

Yes, you can do things to band passed signals you can't get away with on wide-band signals. One of my patents was for a (very) inexpensive bass clipper used in fixed install products. A common problem for 70-100V background music systems is that the outputs go through step up and step down transformers, so too much bass boost can cause the marginal magnetics to saturate which can then damage the amplifiers driving them with low impedance loads. Customers want their loudness and bass boost capability but can't be trusted to not over do it.

My patented circuit that sounded surprisingly good for a simple diode clipper was to add a pair of anti-parallel diodes across just the boost leg of the bass section of a Baxandal tone control. The diodes were out of the circuit for most normal signal levels, but if/when the operator was commanding a lot of bass boost "and" there was significant bass signal present, the diodes would clamp just that bass part of the signal at +/- a diode drop. Since this bass leg of the Baxandal tone control is naturally LPF by the circuit the HF distortion from diode clipping is reduced and then masked by the normal HF audio signal being passed cleanly by the treble side. Since the fixed instal products were combination mixer-amps the gain structure was fixed and I could index the bass diode clamping to specific output power levels.

While I would not call this a hifi circuit it was surprisingly good sounding and very effective, for that application all for the cost of a couple diodes.

JR