Andrew: the kit was pre-assembled. Apparently RTR. Looks like I need to uninstall 2 more devices... Issue is, I tested 7 other njw0302 from digikey, all were within 1% match in hfe/Vbe of Dick. What if Tom , Harry and Lucy are the problem? I have 4 boards and 8 channels, and the devices cannot be tested installed, each one needs to be unsoldered ... 
mchambin: Yes... also the value of its base resistor.
mchambin: Yes... also the value of its base resistor.
Last edited:
Shotgun...
Also consider the effect that the differences between the emitter resistors will have also.
For example if one is 5% lower than the rest, and that one is attached to an output that is prone to hogging then the bad thing will happen sooner.
I will try and get a couple pots to attach as you have and see if I can generate a small amount over the emitter resistors. Might be a while, however, since electronics supply shops are nearly gone, have to wait for the mail.
I could use resistors if I knew what a good value would be considering my 82v supply...
Also consider the effect that the differences between the emitter resistors will have also.
For example if one is 5% lower than the rest, and that one is attached to an output that is prone to hogging then the bad thing will happen sooner.
I will try and get a couple pots to attach as you have and see if I can generate a small amount over the emitter resistors. Might be a while, however, since electronics supply shops are nearly gone, have to wait for the mail.
I could use resistors if I knew what a good value would be considering my 82v supply...
2.7k will give you about 23mV per device which is on the higher side but what Cordell suggests. 2.4-2.5k will be closer to 10mv which is safer and what the gfa -555ii manual suggests.
These value resistors will replace not add to the current 2.2k or you can add 200-500ohm in series to the current 2.2k to get 2.4-2.7k
Use a 300 watt incandescant bulb in series with the mains supply. Anything else and the voltage will drop like a brick at those bias levels
These value resistors will replace not add to the current 2.2k or you can add 200-500ohm in series to the current 2.2k to get 2.4-2.7k
Use a 300 watt incandescant bulb in series with the mains supply. Anything else and the voltage will drop like a brick at those bias levels
I fired it up a couple hours ago with a 150R tacked on that I found in the parts can, allows for 5-7mv on the er. Minimal heat sinking has it running at 85f.
I may try and find either a pair of 2.4k, or just some 200ohm parts to tack on and see if that gets closer to 10 - 12mv like I was hoping for.
Sounds much better, is odd to listen while hanging out near the good speaker, need to get both going here soon.
I did actually try and use my bulb tester, but it seems that the filament must have gotten broken, so I did the holding my breath thing instead.
I may try and find either a pair of 2.4k, or just some 200ohm parts to tack on and see if that gets closer to 10 - 12mv like I was hoping for.
Sounds much better, is odd to listen while hanging out near the good speaker, need to get both going here soon.
I did actually try and use my bulb tester, but it seems that the filament must have gotten broken, so I did the holding my breath thing instead.
Lol thats funny, I can picture you over the pcb with breath held,,, :-D
Great that its sounding much better. I know that over-biasing adds even order harmonics, making the sound fuller/ warmer. So I will keep trying to push this guy to the limits. I have a large enough heatsink. 15" x8" x4" about 20lbs.
With LED's swamping the market, we may soon not be able to source filament bulbs....
Will replace Dick with new part and see... although I dont see it improving current sharing since it measures the same as Digikey part... but lets try and see.
Oh... unless you have an IR thermometer etc... I would touch the tops of the OP devices to check if anyone is getting warmer than the other in addition to measuring the ER resistances and voltages.
Great that its sounding much better. I know that over-biasing adds even order harmonics, making the sound fuller/ warmer. So I will keep trying to push this guy to the limits. I have a large enough heatsink. 15" x8" x4" about 20lbs.
With LED's swamping the market, we may soon not be able to source filament bulbs....
Will replace Dick with new part and see... although I dont see it improving current sharing since it measures the same as Digikey part... but lets try and see.
Oh... unless you have an IR thermometer etc... I would touch the tops of the OP devices to check if anyone is getting warmer than the other in addition to measuring the ER resistances and voltages.
You have to squint too, for that extra margin of safety.
I actually feel more comfortable with a resistor of low value, soldered in than a pot, so wasn't too worried.
We may have to come up with alternative schemes for the almighty bulb tester, maybe some large resistors along with some leds or an analog meter shunt somehow to give a similar result.
I would strongly recommend replacing the outputs on yours, now that you have discovered the differences. The one here is showing about a 4% difference at low levels, just going by the readings off of the big resistors.
I will try and find some more resistors here soon and try it again.
I actually feel more comfortable with a resistor of low value, soldered in than a pot, so wasn't too worried.
We may have to come up with alternative schemes for the almighty bulb tester, maybe some large resistors along with some leds or an analog meter shunt somehow to give a similar result.
I would strongly recommend replacing the outputs on yours, now that you have discovered the differences. The one here is showing about a 4% difference at low levels, just going by the readings off of the big resistors.
I will try and find some more resistors here soon and try it again.
the tungsten filament bulbs for the Mains Bulb Tester are available.
All the small specialist types are still being made and sold.
The bigger ordinary bulbs are available as halogen versions, that work just as well as the ordinary versions.
In the UK these are likely to stay on the market for many years.
Stock up now with values form <20W in increments to 150W.
multiples of ~40W and ~60W may be needed.
All the small specialist types are still being made and sold.
The bigger ordinary bulbs are available as halogen versions, that work just as well as the ordinary versions.
In the UK these are likely to stay on the market for many years.
Stock up now with values form <20W in increments to 150W.
multiples of ~40W and ~60W may be needed.
Over-biasing a ClassAB BJT output stage increases the crossover distortion.
This crossover occupies only a small portion of the whole 360 degree cycle and thus is dominated by high harmonics. These high harmonics are not necessarily even order and are more likely to be odd order.
If you are experiencing an increase in 2nd harmonic then it is probably not due to over-biasing the output stage. Look for another cause.
Maybe an early stage is being loaded more heavily by the increased output driver current and that ClassA stage is being forced into a higher distortion regime.
If this were the case you could create that extra loading without incurring the crossover distortion that sounds quite bad.
Last edited:
Thanks Andrew for that, explains what I have experienced with different bias settings on consumer products.
I almost always wind up with a bit more bias than has been recommended by a mfr. but not by a whole lot. This is just my ears, and usually on the test bench speakers with horrible plastic tweeters that accentuate anything bad.
I almost always wind up with a bit more bias than has been recommended by a mfr. but not by a whole lot. This is just my ears, and usually on the test bench speakers with horrible plastic tweeters that accentuate anything bad.
Not what I heard... I guess then its still misunderstood enough for people to have differences in opinion including gurus... Cordell refers to overbiasing issues as GM doubling, not increase in crossover distortion which as I understand manifests in dead zones on the sinewave, when the transistor switches off due to no turn on signal. The more you bias, the more it conducts and eliminates the dead zone... some people say GM doubling is very minor and inaudible...
Ok guys, update. I put back Dick and replaced other three. Now I am within 10% variation.
the ones I took out all had 110 hfe (compared to 116 on replacements) but more importantly, had 556mV vbe's vs 530-535mV on the others... does the vbe parameter dominate current sharing more than hfe?
As a first test I used low bias of 7mV. And got between 6.8mv to 7.4mv on all 4. Will push it up some more later in the week.
the ones I took out all had 110 hfe (compared to 116 on replacements) but more importantly, had 556mV vbe's vs 530-535mV on the others... does the vbe parameter dominate current sharing more than hfe?
As a first test I used low bias of 7mV. And got between 6.8mv to 7.4mv on all 4. Will push it up some more later in the week.
Read R.Cordell again.
Increasing the bias above the optimal ClassAB level increases the crossover distortion.
Deceasing the bias below the optimal ClassAB level increases the crossover distortion.
R.Cordell spends a lot of pages explaining this and showing what to avoid in attempting to minimise crossover distortion.
Yes
draw an LTP (BJT pair) with the Power supply applied to the collectors, the Zero volts to the two emitters resistors. Add a variable voltage feeding the two bases.
The variable voltage is @ Vbe + Vrs
If Rs are equal and Ie are equal then Vbe must be the same. That's our ideal case that we are aiming for when matching devices.
Now consider what happens when Vbe is slightly different between the LTPair
A small difference in Vbe means that Vrs must be unequal. That means Ie must be unequal.
Vbe is the DOMINANT requirement for matching paralleled output devices.
BUT !!! it must be measured when Tj is the same and when Ie is the same. This is the difficult bit.
Yes, but sonically, a high bias distortions sound better/ warmer. You narrative seems to relegate the two in the same bucket as needs to be eliminated. This may not always be the case even if textbook proper.
Glad to hear that things are improving.
A single ended circuit will respond differently than an a/b with bias I would think, maybe that's the difference between what you are thinking of?
I'm at 7ma now, and it sounds pretty good, considering that there is minimal local decoupling.
I had before tried adding 100uf to the smaller decoupling on the board, but oscillation followed...
A single ended circuit will respond differently than an a/b with bias I would think, maybe that's the difference between what you are thinking of?
I'm at 7ma now, and it sounds pretty good, considering that there is minimal local decoupling.
I had before tried adding 100uf to the smaller decoupling on the board, but oscillation followed...
Great info on the decoupling... I was going to try it but now will be careful doing it. 0.56v across the B and E junctions should wake up the device... the rest is a personal preference.
No I was thinking a push pull class like the KSA-50 many of us built... its class of operation was purely a function of bias current...
No I was thinking a push pull class like the KSA-50 many of us built... its class of operation was purely a function of bias current...
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.