Measuring the FPB 300 damping factor [Updated Jan 13 2019]

With sadness but hope, I've decided to mothball the Krell FPB 300.  Though it has been very entertaining to try to work out the bugs, after 3 perfect repairs, and 3 years of experiments to get it to work properly, I've decided to hang it up until I have the spare time to disassemble and repair, rebuild, or re-engineer myself.  Which may never happen, but at least I'll have time someday to take it apart which would be worth something to me.  I can't bring myself to sell it after all this time and money invested.  As they say, as long as there's money, there's hope.  I also have hope that eventually suitable replacement transistors will be available.  I remember the way this went with tubes, which have never been more available than today.

Meanwhile, I have measured the distortion performance in both channels at 4 and 8 ohms, and the numbers are excellent in the good channel, at all power levels, basically at spect at 8 ohms and slightly worse than 8 ohm spec at 4 (which is not specified).  The numbers are not quite as good in the "bad" channel, but 0.04% THD at 8 ohms taint that bad.  I actually measured 0.09% THD at full power at 10kHz after a full warmup in that "bad" channel, beating the spec IIRC.  (Now I can't seem to warmup anymore...)

The Hafler might measure slightly better than the Krell good channel into 8 ohms, I hadn't measured 4, up to significantly above rated power.  It seemed to have peak output of about 60V, which surprised me.

The Aragon hasn't measured quite as well, however, I haven't done a full set of measurments only a quick check while tuning the bias, and distortion at final bias was never checked, and actually the final bias may not be the final final bias either.  Still, I got the Aragon down to 0.07% at 8 ohms at modest levels at 1khz, in bias fiddling a couple of years ago.

Tonight I was able to measure the Krell at 220 Hz at 8, 4, and 2 ohms, and unloaded.  (This is the "bad" right channel.)  I put the load on one set of outputs and a 6 foot cable to DVM on the other output.

1 watt nominal level into 8 ohms, about 2.88V

220 Hz

8 ohms  2.797V (warmer)

4 ohms  2.785V (coldest)

2 ohms  2.762V (cold)

noload   2.812-2.810(warmer)

DF(220hz,8ohms,1w) = Eunl / (Eunl-Eload) = 2.810 / (2.810-2.797) = 216
DF(220hz,4ohms,1w) = 2.812/(2.812 - 2.785) = 104
DF(220hz,2ohms,1w) = 2.811/(2.811 - 2.762) = 57.4

(Note: the DF should scale as half for each halving of load.  8 ohm damping factor is the one normally specified.)

The amp did not run long enough to actually get very warm, but even during the 10 minutes or so of operation there was improvement from 28.12 to 28.10 in the unloaded output voltage.  I could not go back and remeasure 2 and 4 ohms because the amp shut down twice when doing these measurements, and I didn't want to push it any more than that.

I was also able to measure 10kHz at 8 ohms:

10kHz

8 ohms 2.201

noload  2.213

DF(10kHz,8ohms,1w) = 2.213/(2.213/2.201) = 184

Very little change in damping factor from 220Hz to 10khz.

The apparent HF loss of -4.5dB compared with 220 Hz is probably mostly being caused by my crossover and EQ....more than I expected but it achieves the desired result here (flat response).  Also a dB or so HF loss in the Fluke 8060 meter I am using at 10khz.  I know from previous measurements, specs, and reports, that the Krell is virtually flat well beyond 10kHz, but that is not the question here, so I'm not going to turn off the crossover, bring out a better meter, etc., to measure this better. Turning off the crossover requires a reboot of the DEQ's and has messed me up in the past (when I forgot to turn it back on, etc).

****

In the "good" channel, that runs cooler with no suspicious hotspot and seemingly does not trigger shutdowns, I measured slightly better low frequency damping factor, but more than just slight improvement in the change in damping factor at 10kHz.

DF 8 ohms, 1w, 220 Hz, 222 (equivalent to 0.036 ohms)
DF 8 ohms, 1w, 880 Hz, 207 (equivalent to 0.039 ohms)
DF 8 ohms, 1w, 10kHz, 204 (equivalent to 0.039 ohms)

This time I normalized all measurements to 1 watt by adjusting the volume to 2.88v in all cases, overridding what are mainly affects from my crossover on the frequency response.

The improvement is consistent  with a theory that that feedback and servo are keeping the bad channel in line, but also reducing the ability to cope with changes in load at high frequencies, if only by a small degree, since it's "resources" are already being used to fight the faulty transistor(s).

Though perhaps there was some improvement due to change in measurement strategy.

And you might imagine if there really were a transistor in very bad shape, the result would be much worse.  Differences like this might occur from differential aging as well as manufacturing and repair histories and usage histories.

I was hoping to see a bigger difference, so that the good channel might have damping factor like 500 or something, and there being something "obviously" wrong with the bad channel, more than just a slight variation, even if that variation is consistent with the theory that the bad channel is slightly faulty.  The distortion measurements showed the same thing, with the good channel meeting spec and the bad channel being slightly off spec.

The damping factor for the FPB 300 is not specified, actually.