Mismatched hFE... important?

[bleachershane]

I've had my Rotel RB-870BX for a few years now and after getting a Cyrus Two and restoring it to working order I wondered "Why doesn't the Rotel sound like THAT?"

Anyway, I'd long known there was some DC offset unbalance on the Rotel, one channel coming up about 20mV, one channel coming up at 70mv.

Eventually I went and bought a Peak DCA55 semiconductor tester (for use with future projects and jobs) and myself and my electronics engineering mate took out all the power transistors (all 12 of them, 6 per channel) 2SB817/2SD1047 PNP/NPN, and measured them with this new fangled toy.

Noting the results we were a bit baffled by the transistors measured hFEs...

LCH:
2SB817E (NPN)
104
97
96
2SD1047E (PNP)
60
59
49

RCH:
2SB817E (NPN)
111
108
108
2SD1047E (PNP)
56
56
63

So some questions, should these transistors with a rated hFE of 100-200 for both the PNP and NPN be measuring so differently? Why are the PNPs showing such a consistent hFE drop/difference to the NPNs?

Taking the power transistors down to 4 per channel (8 in total) by removing the lowest hFE specced, and finding the best matching transistors (2SA1016K) for the differential pairs Rotel from various postions on the boards, the amp does indeed seem to sound a bit more open, a bit less muddy and with a better soundstage. And the DC offset seems to measure a bit more favourable 15mV across both channels.

Is it worth completely changing the mismatched power drivers for newer, hopefully better matched complementary pairs? I've been looking at what's available from Farnell (my suppliers of choice) at things like the 2STC5242/2STA1962 or the 2STC5200/2STA1943... how will these newer transistors fall into this 30+ year old design?

Any help really appreciated, especially on the mismatched pairs!!!

RB-870BX Service Manual
2SD1047/2SB817 Datasheet
2STA1962 Datasheet
2STC5200 Datasheet

[Conrad Hoffman]

My experience, and I'm speaking strictly as some joker on the Internet here, is that the devices in the same group need to be matched well enough to share current reasonably closely for reliability. Emitter resistors reduce the need for matching. As for the NPN vs PNP, it's almost impossible to find complimentary devices with near the same Hfe. When I've been able to do it, the THD comes out lower, maybe with more favorable harmonics, and I wanted to believe the amp sounded a bit better. Could be all in my head. FWIW, I also believe that no decent design requires specific or very closely matched Hfe. If it does, it should go back to the drawing board, because it's just a laboratory curiosity, not a production ready circuit.

[schubert]

I'll second Mr. Hoffman - usually when I recap an amp I replace all the transistors as well, and I always order about twice as many as I need and then try to match everything for hFe as closely as I can. The amps always sound much better, though how much of the improvement is due to the transistors I can't say.

[AndrewT]

the hFE read oddly, because they have not been measured to comply with the manufacturer's specification.
That does not give any assurance that measured properly you will find a closer matching of hFE.

BTW,
I consider Vbe matching of paralleled output devices to be far more important than hFE matching.

[ingenieus]

I looked up the Peak DCA55 semiconductor tester and I want one! That said - it makes its measurements at fairly low voltages and currents. Are these the right measurements to base the matching of power transistors on? Under full load, the value of hFE will be very different from that measured by the DCA55.

[Salas]

Vbe, hfe, are curves. The DCA has a CCS so it measures all at 2.5mA Ic for Hfe at least and its comparable spot check for same type. Full display takes a curve tracer. For Vbe it forces high Ib current so its more consistent to match but don't expect to see the same value when biasing small transistors in circuit.

[jez]

Those Hfe results are pretty normal. Nothing to worry about and nothing to do with the better sound of the Cyrus 2.

[ingenieus]

From the datasheet for the 2SB817/2SD1047 it appears that hFE is (nearly) constant for collector currents up to 1,5A or so.The measurement of hFE taken at 2,5mA should therefore apply to currents up to this value. Not perfect perhaps, but a whole lot better than no measurement.

[kenpeter]

If offset is your worry, barking up the wrong tree looking at output mismatch.
The VAS is single ended and couldn't care less. Except in terms of sharing the
burden between parallel devices, 2H, and crossing distortions. Mismatch here
has nothing to do with offset...

Your offset is mismatched VBE at Q601 Q602, input pair.
The devices could be perfect match, but the currents flowing are probably not.
Q601's current is dictated by the VAS, and Q602 must conduct the remainder.
Correct proportion of "remainder" must be provided the tail. Just make your
tail adjustible, and tweak out the offset....

[bleachershane]

Right, my techy friend went and read this and kindly pointed out to me that the way I worded it made it sound like I was linking the DC offset imbalance with the unbalanced hFE in the power transistors... sorry, I wrote that just before heading to bed!

The DC offset issue was helped by better matching the input pair, an imbalance found between the pair on the channel with the higher DC offset was remedied without having to buy any parts whatsoever! With both DC offsets now at 15mV I consider that 'sorted'...

The Cyrus Two is a brilliant sounding amplifier, I didn't think I could make the Rotel sound like that, totally different designs but I wondered why it just sounded, well, lifeless with a huge lack of soundstage... surely it's not right...

So, the unbalanced power transistor hFE readings. I guess this is a case of OCD tweaking seeing as the design compensates for this through the feedback loop, etc...
Just how much better would the amp sound if I were to go for new complimentary pairs (and chancing that they probably will not 'match' that well anyway)?

It makes me wonder, is this why Cyrus went for the quasi-complimentary design which takes out the usual issues of NPN/PNP matching (or lack of...)? Another point being that the Cyrus used some well specced transistors elsewhere in the circuit that to this day are hard to substitute for...! It was a nightmare trying to track down 2SA1775As *argh!*