A cool 14K/W
Fischer Elektronik SK104
they Don't fit the 303.
I remember first seeing them used in JLH PSUs and Krell REGs.
Fischer Elektronik SK104
they Don't fit the 303.
I remember first seeing them used in JLH PSUs and Krell REGs.
I was only querying whether the original heatsinks could have been re-used. The reason I ask is that they appear to have been constructed from two parts - probably aluminium- with the transistor clamped onto the base plate by the upper plate. Bolting a TO-5/TO-39 onto a metal plate is the best method of removing heat from those devices.
The original RCA heatsink which was attached to TO-5/39 devices was a rectangular folded sheet of steel, welded to the TO-can perimeter. That would have taken a bigger footprint than the sinks Quad used with fins over the top of other components.
I'm not sure what you are implying by the 0.68V of the power transistor- across 68 ohms that is 10mA to be added to the base current of the output devices. Base-emitter voltages vary considerably with current in power devices (greater than the theoretical 60mV/decade), so will the current in the 68 ohm resistor.
The driver power has to take into account the gain of the output transistor as that determines the demanded base current. The gain of the MJ15003 is higher than the old 2N3055 (25 at 5A compared with 20 at 4A) and taking 3.5A as the peak for 50W into 8 ohms that makes the base current around 140mA+the current in the 68 ohms, peak, in the worst case.
Those old RCA heatsinks were rated at 50C/W. The ones you mention are like the old Fischer SK510 at 44C/W. I tended to use the Fischer KK510 which were larger in diameter (20mm) star design which came in two parts held together by a spring clip. They were rated 38C/W, but the only To-5/39 heat sinks available now are typically those you mention or similar at up to 60C/W. Since device reliability is improved with lower temperature excursions, I recommend using lower thermal resistances where possible, thus the taller version is preferable not the shorter one, merely being one that "works perfectly".
And to be pedantic, silicon junctions of devices packaged in metal cans operated up to 200C. It was when most devices were shipped in plastic that the 150C limit became a standard.
Oops - you're right. The current is 10mA not 100mA. and you're also right have to take into account the base current of the output devices.
Exactly what the practical dynamic power dissipation of the driver and output devices is depends on the dynamics of music. Self tried to get under the skin of this in Chapter 16 of his book Audio Power Amplifier Design. Chapter 16 is "Power Dissipation in Amplifiers" using estimated peak to mean ratio of different types of music. A reference level is the PMR of a sine wave is 4dB, with rock around 10-20dB and classical 20-30dB. Really compressed rock, or musak, approaches the PMR of a sine wave. PMR determines how power is partitioned between power amp dissipation and loudspeaker power for various amplifier classes.
But the 2N3055 is not a very good transistor, because beta is strongly dependent on collector current https://www.onsemi.com/pdf/datasheet/2n3055-d.pdf.
There are far better much more recent modern power devices which have constant beta up to many amps such as https://www.onsemi.com/pdf/datasheet/mjl3281a-d.pdf , alas unsuitable for the 303 because of their fT in the tens of MHz. They are however the preferred power devices for Douglas Self Blameless power amps.
"The 2N3055 is not a very good transistor". Well, there are indeed much better devices today. But again let's be clear about which 2N3055 is being discussed. The old RCA hometaxial types were slow. Epi devices were available by the end of the 60's, and Bailey's 30W design was well respected. The modern 2N3055 is an epi based device with an fT of several MHz (2.5MHz min). It and the MJ2955 can be used in Bailey-like designs with similar or better performance despite the gain droop as long as that is designed for. Compared with Self's Blameless, he used MJE340/350 drivers which drooped at 50mA but the output transistors were MJ802/4502 which only had fT's of 2MHz, but maintained gain at higher currents. His original Blameless swapped droop in the output for droop in the drivers, it appeared to me.
Nevertheless, the modern high gain devices offer better performance all round.
And at risk of repetition, it is possible to use the modern epi 2N3055 in the Quad 303 (essentially the circuit- not as an original box refit). I've managed to halve the distortion and double the bandwidth at the same time, with some optimisation of the frequency compensation. I suspect it would be possible to use 3281's or 5200's in the output with similar considerations, but whether they can be used with existing wiring between the PCB and heatsink in a refit is another question.
Nevertheless, the modern high gain devices offer better performance all round.
And at risk of repetition, it is possible to use the modern epi 2N3055 in the Quad 303 (essentially the circuit- not as an original box refit). I've managed to halve the distortion and double the bandwidth at the same time, with some optimisation of the frequency compensation. I suspect it would be possible to use 3281's or 5200's in the output with similar considerations, but whether they can be used with existing wiring between the PCB and heatsink in a refit is another question.
Last edited:
You are right about the original Blameless in Self's articles in '93 and '94 Electronics World. The design has evolved quite a lot since, although still retains the MJE340/350 drivers.
But most of Self's work on blameless power amps was done by the late '90s, see for example https://www.worldradiohistory.com/UK/Wireless-World/90s/Electronics-World-1997-01.pdf which is a "load invariant" design which doubles the output devices.
FWIW I use 10 blameless power amps driving my Linkwitz LX521.4 speakers. Four load invariant ones for the 4 open baffle bass units, and regular single power transistor ones for the lower mid, upper mid and tweeters,
Would I do it that way now? Probably not. I'd either use Class D, or power amps from https://neurochrome.com/
But after lusting over the 33 and 303 so badly that it almost hurt when I was a kid, but could never remotely afford them, I now am the second owner of a cosmetically faultless pair that (shock horror) I use via ceiling speakers in the kitchen. The only things I've changed are the tantalum input caps, and the reservoir and output coupling caps, supplied by Quad Huntingdon's service centre, along with a kit of bits (including the correct colour coded lengths of wire) to put them the other way up - with the terminals and vent uppermost, and not pointing downwards at the circuit boards.
But most of Self's work on blameless power amps was done by the late '90s, see for example https://www.worldradiohistory.com/UK/Wireless-World/90s/Electronics-World-1997-01.pdf which is a "load invariant" design which doubles the output devices.
FWIW I use 10 blameless power amps driving my Linkwitz LX521.4 speakers. Four load invariant ones for the 4 open baffle bass units, and regular single power transistor ones for the lower mid, upper mid and tweeters,
Would I do it that way now? Probably not. I'd either use Class D, or power amps from https://neurochrome.com/
But after lusting over the 33 and 303 so badly that it almost hurt when I was a kid, but could never remotely afford them, I now am the second owner of a cosmetically faultless pair that (shock horror) I use via ceiling speakers in the kitchen. The only things I've changed are the tantalum input caps, and the reservoir and output coupling caps, supplied by Quad Huntingdon's service centre, along with a kit of bits (including the correct colour coded lengths of wire) to put them the other way up - with the terminals and vent uppermost, and not pointing downwards at the circuit boards.
Interesting article.
I see the mention of some related transistors:
Looks like the MJ21194G or if you cant find one, the MJ15022/24 could be used for TR3 'duty'.
This explains why the over rated MJ150003G works so well in that role for the main power amp transistors, is unlikely to ever reach max gain.
Tiny ferrite beads in the right place would mitigate RF... perhaps some one would like to comment where would be best.
I see the mention of some related transistors:
Looks like the MJ21194G or if you cant find one, the MJ15022/24 could be used for TR3 'duty'.
This explains why the over rated MJ150003G works so well in that role for the main power amp transistors, is unlikely to ever reach max gain.
Tiny ferrite beads in the right place would mitigate RF... perhaps some one would like to comment where would be best.