If you listen in softer level, I am sure it works. I bias my Acurus to 100mA each pair and it has 4 pairs. Big improvement, but if you crank it beyond Class A, I think it will make a difference. 200mA will get 8W or class A for 5 pairs.
Basically, if you want to satisfy Oliver's criterion, for a given value of RE you are limited to a given amount of bias current. So, with RE=0.22, you cannot go above about 120mA, corresponding to 26mV across RE. As mentioned before, the actual amount of current to satisfy the "real" optimum will be somewhat less as a result of the ohmic contributions from the power transistor and base stopper resistor to effective emitter resistance "internal" to the power transistor.
Cheers,
Bob
1) I'm afraid there's no recipe. You have to determine this yourself, best by measurements. Too many variables to make theoretical predictions. As I have already mentioned, and somehow funny, matching 5 pairs of output devices is not that good for distortions; having the idle currents spreaded over a reasonable range (say, between 120mA and 160mA) helps spreading the transconductance and therefore lowering the distortion hump beyond the Oliver point.
2) There are to many variables to account for. Measure your amp distortions vs. the bias current and push the bias as high as you can, given the heat sinks capability to keep the devices thermally stable, under the worst case conditions (that is, at about 1/3 the maximum output power).
As others have mentioned, keeping the "first watts" in class A has virtually zero benefits in terms of sound quality. You'd be much better and safer keeping the bias around the Oliver point. For 5 pairs, I would use 0.33ohm and around 60-80mA per device, no need to hand match anything.
Attachments
Last edited:
Are all output BJT's the same ?
I've just noticed , from OEM recommendations and using ON/Sanken.
What is optimum for one , is not for the other.
Might it be the "ring emitter" ? Xover glitch disappears way down at 20-30ma
on a 2sa1216/2sc2922 pair.
From what I have read , these are like many paralleled power BJT's in one package.
From what I have experienced , you also might be able to "bend the rules"
with these devices. (high bias class AB).
Personally , I just bias either the ON/Sanken to 16mv/.22R . All this
fuss over a 3-8W class A region .... any modern digital source will easily peak above this at <1W.
You will always be reaching AB , why not run cool ?
OS
I've just noticed , from OEM recommendations and using ON/Sanken.
What is optimum for one , is not for the other.
Might it be the "ring emitter" ? Xover glitch disappears way down at 20-30ma
on a 2sa1216/2sc2922 pair.
From what I have read , these are like many paralleled power BJT's in one package.
From what I have experienced , you also might be able to "bend the rules"
with these devices. (high bias class AB).
Personally , I just bias either the ON/Sanken to 16mv/.22R . All this
fuss over a 3-8W class A region .... any modern digital source will easily peak above this at <1W.
You will always be reaching AB , why not run cool ?
OS
I use RE of 0.16R in my TT amp with no problem what so ever already for more than three years. http://www.diyaudio.com/forums/solid-state/182554-thermaltrak-tmc-amp-10.html#post3031504. Total emitter resistance is RE +Re +base stopper/beta + drivers output impedance/beta. In my case I run it at about 125 mA and I think it's good approximation of the Oliver's criterion. Advantage to use lower value for RE is lower OPS distortion and lower resistor wattage and then is possible to use precise metal film resistors in parallel instead of 5% wire resistor (probably inductive).
Damir
Damir
Well, I just bought 200 of the 0.22ohm 3W metal film on ebay. I am going to do as before, string them together and burn them in for two hours at over 2W to screen out the bad ones as I don't trust China. So far, I have good luck, I burn in all the 0.12 and 0.15 that I have. It got so hot that some got darken. But none failed.
I use 2.2ohm base stop resistors.
I have enough range to adjust from 120mA to over 200mA per stage, we'll see whether it makes any difference in the sound.
I use 2.2ohm base stop resistors.
I have enough range to adjust from 120mA to over 200mA per stage, we'll see whether it makes any difference in the sound.
I know it`s late but IMO you should bought 0.47ohm 3W metal films and run them paralleled so you have 0.235ohm but 6W resistor per output bjt. Just to be on the safe side
But in my case, I am trying to get the first 10W+ in Class A!!! I really consider my amp is more Class A with a flare!!! Or Class A with reserve. I am just trying to optimize the next 20 or 30W of Class AB as good as possible.
I am using very low voltage rails. In another thread, I was talking between using 18V-18V, 20V-20V or 22V=22V toroidal transformer so the rail is +/-25V, +/-28V or +/-31V respectively. That gives 12.5W, 10W and 8W of pure Class A respectively.
I even consider using 15V-15V to get 18W class A for both 4ohm or 8ohm load. with 4ohm, I get total of 36W class AB power.
I think all these will cover loud home use within Class A 99% of the time. then the 1% of the volume spike in Class AB. This is going to be louder than any single end tube power amps.
Last edited:
You'll get virtually nothing in terms of sound quality. Why don't you build one of Mr. Pass low-ish power class A designs? A class A push pull amplifier has a yield of about 40% (that is, 100W dissipation for 40W in the load), not that bad, you need to design the thermal part accordingly, though. You'll be much happy with that than with a half baked compromise.
Your call, I'm out.
Can you post the schematic? I want to see how he get 40W of class A into 4ohm.
Last edited:
Do you have the schematic?
I am using a quite expensive chassis, the heat sink is 12" X 5.75" X 2" already, you saw that in the other thread. It can only handle less than 70W of heat dissipation safely. How do you do a 200W Class A? Liquid cooled?
I don't care what transistors, you need rail voltage to get the power, then you need idle current to get Class A. It's the heatsink that's the limit.
Last edited:
I seen a member that used the "extreme A" OPS with 2 pair sanken BJT
(2sa1216/sc2922) and used a PC CPU liquid cooled setup for a 100W
monoblock.
The basic "extreme A" circuit -http://www.diyaudio.com/forums/solid-state/96853-extrema-class-strikes-back.html
I don't like the rest of his circuit , that OPS runs well with all my IPS's.
As it is (BJT) , it works. I just think a VFET version would be "hardier".
But , At <60V rails the Sankens are impressive.
Search the forum for that CPU cooler class A , it was sweet !
PS - Pass uses a much larger EF2 IRFP setup in the X60.8 and up.
Also uses the IRFP's heavily biased in the X250.5 (50W class A/250W AB)
OS
Ha ha, I was joking about liquid cooled. Did not expect that's for real!!! I think that's where I draw the line.
So is it true from all the comments that it's better to not worry about the large Class A using high current, instead concentrate on running Oliver's optimization. Using Re=0.22ohm for safety. Using 120mA bias for Oliver's condition and use the highest voltage rail to keep the heat to around 60W per side and call it a day?
This is getting into the grey area where you guys that have experience with real amp know better which way to go. I have no idea.