Amp is alive ! 2 hours or so on nedium volume, heatsink seems to be colder then before, quiss current is same as before, more or less , 17mV based on service manual. DC offset is a few mV .
I feel like the bass isn't so great anymore maybe just in my head as I listened without the sub.
*Can changing the output transistors make it " have less" not so " full " bass ?
Also feel like I need more input volume
Maybe just in my head.
Overall is sounds great , the same .
I feel like the bass isn't so great anymore maybe just in my head as I listened without the sub.
*Can changing the output transistors make it " have less" not so " full " bass ?
Also feel like I need more input volume
Maybe just in my head.
Overall is sounds great , the same .
The transistors themselves won’t have anything to do with perception of bass, unless the stage runs out of gain. Not going to happen with an EF3. With EF2’s or any type of composite output stage with gain everything becomes more critical if original performance is to be maintained. Voltage rating needs to be high enough to work with the amp. If 200V types were originally used on +/-55V, you COULD use 160V parts. 120V not such a good idea. The second-breakdown breakpoint in the SOA curve is more important - where the slope changes needs to be at least as high as it was. Preferably above the rail voltage, but that’s often a tall order. The hFE at maximum normal operational current (Vcc divided by rated impedance) needs to be as high as the originals to maintain distortion performance. They fall at different rates, and it roughly follows the Ic rating.
fT and capacitance need to be in the ballpark. Lower fT or higher capacitance means harder to drive. Far more critical with EF2 than EF3. Most of the time the circuit will WORK even with gross changes like going to old epi-base 3MHz outputs. But stability and switch-off are affected. Stability changes may only manifest itself as added ringing on the square wave response. But that indicates a change. Switch off gets worse with low fT devices, and that causes an increase in high frequency crossover distortion. With EF3, you may very well get it low enough to be buried in the noise even with slow types. But not guaranteed. Worried about it? Put it on a distortion analyzer at one watt, and look at your small signal square wave response. If you see any red flags, the match to the original wasn’t good enough. What WILL happen is that given a sufficiently high frequency input, the output stage will go into cross conduction and burn up. The faster they switch off, the higher the frequency where this happens. With modern digital program source material there is almost no way for this to happen even with old hometaxial types. Feed it a 300 KHz full power square wave on the bench and you may end up with a ball of fire. But even a DJ won’t do that to it.
fT and capacitance need to be in the ballpark. Lower fT or higher capacitance means harder to drive. Far more critical with EF2 than EF3. Most of the time the circuit will WORK even with gross changes like going to old epi-base 3MHz outputs. But stability and switch-off are affected. Stability changes may only manifest itself as added ringing on the square wave response. But that indicates a change. Switch off gets worse with low fT devices, and that causes an increase in high frequency crossover distortion. With EF3, you may very well get it low enough to be buried in the noise even with slow types. But not guaranteed. Worried about it? Put it on a distortion analyzer at one watt, and look at your small signal square wave response. If you see any red flags, the match to the original wasn’t good enough. What WILL happen is that given a sufficiently high frequency input, the output stage will go into cross conduction and burn up. The faster they switch off, the higher the frequency where this happens. With modern digital program source material there is almost no way for this to happen even with old hometaxial types. Feed it a 300 KHz full power square wave on the bench and you may end up with a ball of fire. But even a DJ won’t do that to it.
I looked at the SOA of 2sc4468/2sa . , pretty similar to the originals , not the "The second-breakdown breakpoint in the SOA curve " need to read more about that and check.
from 30 to 20 mhz , 220p to 250p , not a big difference ( imo ). ?.
Top curve looks like 2SC3281, not 3182. It’s 15A/150W/200V, full power to 70 V. I think the 3182 did full power to 70V too, but it’s only 10A/100W/140V. That Sanken will handle just as much, at your 55V rail voltage. Sanken (and Toshiba) make cheap-o versions that start dropping off power handling around 30V Vce. Just don’t stick those in there.
30 vs. 20 MHz may mean nothing, especially if measured at 12V vs. 5 V. Cob drops with higher Vce, so the can play specsmanship games. Sanken’s top of the line is pretty comparable to Toshiba, but they are different processes. Sanken‘s top parts are epitaxial planar, Toshiba’s are triple diffused. Both are king of their respective hills. The old Motorola-type ON licensed Toshiba’s process, or at least parts of it. Fairchild always had an epitaxial planar process, and their copies of the C5200 and it’s brethren use it. You can basically end up with the same result, given optimized process. ANY of that is fundamentally different from the old-school epi-base process which is limited to about 7 MHz and usually gives about 3.
30 vs. 20 MHz may mean nothing, especially if measured at 12V vs. 5 V. Cob drops with higher Vce, so the can play specsmanship games. Sanken’s top of the line is pretty comparable to Toshiba, but they are different processes. Sanken‘s top parts are epitaxial planar, Toshiba’s are triple diffused. Both are king of their respective hills. The old Motorola-type ON licensed Toshiba’s process, or at least parts of it. Fairchild always had an epitaxial planar process, and their copies of the C5200 and it’s brethren use it. You can basically end up with the same result, given optimized process. ANY of that is fundamentally different from the old-school epi-base process which is limited to about 7 MHz and usually gives about 3.
2sc4468 ( triple diffused ) are rather cheap , not their top. But does that make them " bad " ?.
The original 2sc3281n are also triple diffused type I think .
Let's say we have two transistors , same specs , one is epitaxial planar and one is triple diffused , would you " hear " any difference, ?
From what I've read epitaxial are faster, more linear , triple diffused are more rugged , usually higher power , voltage .
I guess this depends from parts to parts.
Does one think to use epitaxial or triple diffused in their amplifier design, or they just pick the best candidate based on other factors like speed, linearity, SOA, low capacitance etc, regardless of the type ( epitaxial or triple diffused ) ?.
The C4468 isn’t “bad”. Just not as good as devices like the C3623 or C3519. Or the original C3281.
They generally pick the best candidate based on availability and price. If you look at the hFE curves, the Toshiba or ON versions generally win. The Sanken LAPTs (epitaxial planar) seem to be faster and don’t seem to be as linear, BUT I’m pretty sure differences are mostly in the test conditions.
Some of the old Sanyo ”complementary” pairs would mix and match the two processes. The NPN might be triple diffused, and the PNP epi planar. And theyd be about as complementary as anybody else’s. The most “complementary” of the available types (ie, NPN to PNP match) is actually the ON NJW0281/0302.
They generally pick the best candidate based on availability and price. If you look at the hFE curves, the Toshiba or ON versions generally win. The Sanken LAPTs (epitaxial planar) seem to be faster and don’t seem to be as linear, BUT I’m pretty sure differences are mostly in the test conditions.
Some of the old Sanyo ”complementary” pairs would mix and match the two processes. The NPN might be triple diffused, and the PNP epi planar. And theyd be about as complementary as anybody else’s. The most “complementary” of the available types (ie, NPN to PNP match) is actually the ON NJW0281/0302.