The JLH simplicity makes it a slightly quirky affair... too much deviation from the original design and it gets upset, which is a shame as a chunk more feedback would make it fit much better in my system in terms of gain and also lower the distortion helpfully.
I also thought it worth to revisit the JLH to see what was possible with higher voltages and bias points... at this point, I'm just interested in the 1 watt into 8 ohms performance... here's the results I got from the experiments...
2H 3H 4H 5H
24V 560mA = 13.4W -59 -73 -89 -102
27V 667mA = 18.1W -61 -76 -95 -109
24V 835mA = 20.1W -67 -79 -104 -113
30V 774mA = 23.2W -63 -79 -99 -114
33V 922mA = 30.4W -65 -82 -104 -118
35V 973mA = 34.1W -66 -83 -105 -118
So substantial gains from higher voltages but also a lot more waste.
I have been looking at other simple designs too. A kind audio fellow in France sent me some original boards for the Hiraga 8W design. I've had my own boards based on this design sitting around for, ooh, probably a year now, including one based on PTFE (christened the "Slippy Amp" as ink would just slide off it) but never quite got around to finishing them. All that needed to be re-added was some 1 ohm power resistors, after I'd carelessly not mounted one of the power transistors very well.
The schematic can be found on the link above. It is a very simple amplifier with a lovely symmetry to the stages. It runs at fairly low voltage, but with very high bias... around 1.7A. You can run it off big lead acid batteries, if not for that long.
I haven't actually sat and listened to the Hiragas as yet simply because the measurements suggested a big difference in characteristics between the two boards, which is likely to lead to a flawed evaluation.
2H 3H 4H 5H 6H 7H
Hiraga Board 1 -42 -81 -76 -74 -83 -101
Hiraga Board 2 -45 -52 -86 -69 -81 -79
I'm not entirely sure why that is, whether it is careless abuse on my part or whether some of the parts are damaged, but in any case it will be interesting to compare to my own boards when complete. The basic distortion performance appears to be considerably worse than the even simpler JLH design for the same power consumption.
To throw something very different into the mix, I had decided to build a new amplifier for the bass drivers... rather than going for a Class D, I'd decided to try a Class AB which had the potential to run in Class A for a few watts. The design in question is the LME49830 reference design, originally from National Semiconductor, before being absorbed by TI.
The LM49830 is essentially a near-complete MOSFET driving solution for an amplifier design, containing all the front end and driving circuitry required for building an amplifier. It is very low distortion and results in a fairly simple PCB. Rather than go to the trouble of designing my own from scratch, I decided to use the reference design.
The boards are 3.2mm FR4, with 4 oz traces. Not cheap, but if you are building something to handle power and want something that will not flex, this makes a lot of sense. I decided to stay close to the original specifications of the parts - a lot of the parts are exactly as listed on the BOM. A few minor changes are the use of silver mica instead of polypropylene for the signal filtering, non-inductive wirewound power resistors, and slightly bigger local decoupling caps for the LME49830.
The power devices used in the design are the Toshiba 2SK1530 and 2SJ201, PDFs on Bob Cordell's website. These are beefy complementary MOSFETs with a lower than usual turn-on voltage - they should not be confused with lateral MOSFETs which have a very different structure.
When finally built up, they look something like this...
A parallel pair of N FETs and P FETs are used for high power handling and low output impedance. The pairs of FETs were as tightly matched as were possible from the ones I had.
As supplied, it appeared that the LME49830 reference design can be biased from approximately 200mA to 550mA. For normal use, that is plenty but for lower voltage operation I was interested in seeing what benefits there were from going that little bit further. Adding a resistor in parallel let me increase the maximum point to see what was possible. I decided that going beyond 700mA would probably be pointless so ended up setting that as the maximum. Here are the distortion results so far, again for 1W into 8 ohms, with +/- 23.5V rails...
2H 3H 4H 5H 6H 7H 8H 9H
201mA bias -98 -100 -111 -107 -124 -119 -124 -127
541mA bias -105 -107 -125 -118 -125 -124 -126 -127
700mA bias -115 -116 -133 -126 -133 -132 n/a -135
The distortion as can be seen is incredibly low. For the 700mA bias result, I suspect the D/A and A/D are actually the limiting factor in the results rather than the amplifier itself.
Listening to the amplifier is an unusual experience. I'm not sure what to make of it so far... it reminds me slightly of the ExtremeA amplifier, but will need a bit more time to make my mind up...
