Update: Still looking for default distortion profile when switching from lower powered transistor to higher power (with all other parameters equal).
Generally it's been brought to my attention that there are many variables that affect distortion, but the precise effect having a more powerful transistor in place (capable of higher max current) has on distortion - I'm still looking for
Because I changed the title, I need a preamble to explain a bit. Look at the chart below in OP.. you'll see that all 4 Denon AVRs ranging 105-150WPC and 9-11 AMPLIFIED channels, all begin distorting more starting at 12W. In my reading up to this point I feel comfortable stating all 4 receivers use 15A transistors. OP and post 9 explain the significance of this - I believe a big part of the issue this thread aims to tackle is that transistors require quite a bit of feedback to remain linear once pushed past 8-10% of their maximum rated current (for example, a 15A transistor at 50C behaves well (requires minimal feedback for output into resistive load to behave - up to 1.2-1.5A).
If you knew nothing of what I just said, you might think that if you had two very similarly designed AVRs, one rated for 105W RMS, and the other rated for 150W RMS, and you saw the 105W AVR's THD+n began increasing from 12W, you might hazard a guess that the 150W receiver might start distorting at 18W, because they're a similar designj, but the 150W receiver is capable of about 1.5X more power than the 105W. You'd also be wrong, because both receivers THD+n begin increasing at 12W.
Another interesting fact is that this distortion begins at 12W whether the load is 8 ohms or 4 ohms...
Yeah, that one's kind of a bombshell lol.
This evidence suggests the reason for the observed increased distortion after 12W is the transistor operating outside of its 8-10% zone previously mentioned. I have a class A amplifier I discuss later in OP/p9 which uses 10 transistors in parallel to accomplish operating with minimal feedback.
This thread is to brainstorm ways to reduce the amount of feedback required in push/pull designs. Obviously class A is easier to stick a bunch of transistors in parallel and call it a day.
Are we using greatly underpowered transistors in our class AB amplifiers? (the push-pull design biased for minimal crossover distortion)
I think we might be!
The reason is... actually, there are many. But to start, I'll begin with the thing most at the forefront of my thoughts - I have a Kinergetics KBA-280 - a stereo class A amplifier rated for 140W RMS into 8 ohms. It's a behemoth - it weighs more than any other amplifier I've lifted, and the transformer is a toroid. Inside there are... I want to say... 10 output transistors per channel, and they're not small. I forget what they are right now, but their package is like 1.8cm x 3.5cm x 0.4cm. They're MASSIVE transistors, and the amount of them in there? Is MASSIVE!
Stereophile reviewed it, and during their review, worked it HARD. How hard?
Single ohm hard.
Why? I don't know, probably had something to do with the weight of it and they're connoisseurs. Next is watts into 8, 4 and 2 ohms.
140, 270, 530, and it gets even better when the final power rating is taken while driving a single ohm...
A THOUSAND WATTS PER CHANNEL. But only one at a time, because the main fuse.
Right now we're talking about something else though: this thing's transistors. We can marvel later
For the sake of discussion, let's say behemoth's transistors are rated for 15A max. They're probably 15 amps (if they're not, they're more)
15A x 10 = 150
150 amps
That's a TONNE of amps! More than the service to my last place! (which I believe was 120A).
150A and an 8 ohm load? That's 1200 volts.
1200 x 150 = 180,000W
I looked at the datasheet and found out: where the transistor is most linear is in the first 7-10% of its "maximum" rating. Past that, it's pretty nonlinear. How does this nonlinearity increase? Nonlinearly! Funny, isn't it, that the increase in nonlinearity of a transistor is nonlinear itself! Lol. It kinda looks like y=a^x ("y" equals "a" to the power of "x").
Fortunately this unfortunate part of transistor existence doesn't affect its operation most of the time, because in addition to the transistor's input, feedback (how far off the transistor is from its intended path) is also applied, simultaneously. This ensures that the output becomes what's desired, and not just the transistor's nonlinear representation of its input. How is this feedback collected and how is it applied to linearize? It varies a lot, and can get complicated. Which is why, if you can, you minimize the amount of feedback required
And in the end there's less parts, too.
So moving on... let's use 8% of 150A
12A
with an 8 ohm load 1152 watts, and 96 volts
I think the transistors were 200V, which in class A gives a maximum of 70VAC. Since you can't hit the rails, call it 65VAC
528 watts, and linear behaviour to over 2x that amount of current.
From what I gather, these outputs would need minimal feedback because the transistors are operating well within their optimal operating range (that small fraction of their maximum rating).
There are 10 transistors being used to do the work that just one transistor can do.
The only real caveat to performance from running all 10 transistors in parallel in the way I've outlined (class A) is the noise floor would increase by 10dB. But even half decent transistors are pretty quiet, and 10dB is nothing compared to the benefits from running in linear territory.
Now the conversation is shifting though- background is done. The graphs below, though - they need to be considered as the first step to the next direction. Four of these charts are power vs. distortion of recent Denon home theater receivers. The fifth more bland looking one? Still from this site, just someone else measuring a Yamaha that has the same type of amplifier (AB), but it's rated for 60 watts RMS x2 instead of 105/125/140/150 watts (Denon X3700/4700/6700/8500H). Yamaha got a slot so that all the amps weren't Denons, and there was one significantly different. I believe all of the Denons were made with 15A transistors.
Aside that needs to be said that I wish I didn't because it's annoying: SO, the fourth Denon (X8500H) is represented by the 8 ohm chart instead of the 4 ohm chart like all the others, because the X8500H's 4 ohm chart is ugly as sin. One channel was very divergent, and ignoring it and putting a line through the other is stupid when the better looking 8 ohm chart is there to use. It shows the same result as the 4 ohm chart and does it lookin' good
'
3800 4800 6700 8500
Above you can see that the X3700H, X4700H, X6700H, and X8500H, are all capable of different maximum continuous stereo power outputs.
And you can see that they all four beasts are capable of the same amount of power before THD+N stops improving with increasing power.
WHY IS THIS?
8% of 15A is 1.2A
1.2A and 8 ohms is 11.52 watts.
The charts above, the vertical green lines I've marked between 10 and 20 watts, more accurately are between 12.2 and 14 watts.
Before accurately calling the change in a trend, a marked difference needs to happen. Those lines at 12.2 and 14 are after a noticeable change - a clear trend. If you look to precisely where the green line drawn from ~100mW to ~2000mW diverges.......
11.5W
11.5 watts x 2 channels is well below the maximum current able to be supplied by the power supply
11.5 watts x 2 channels into 8 ohms or 4 ohms requires much less than the amount of voltage supplied to the transistors
The new receivers for 2023: the X3800H, X4800H, X6700H, and X8500HA ALL have 15 amp transistors. I can't say for sure that the models above also have 15 amp transistors, but I know the X3700H does (I have one, wooo! partyy!), and I'm fairly sure the X4700H does. So if the X8500HA does too, I think it'd be stupid to think the 6700 didn't!
I think the reason distortion goes up at about 11.5 watts is because more feedback is needed.
Does anyone who knows more than me think what I described is plausible?
If so, what can we do moving forward designing amplifiers?
Putting a bunch in parallel works for class A amplifiers. For low distortion, class AB (working in B, push-pull) need to be biased to minimize crossover distortion as much as possible or none of the rest matters. If using multiple transistors in parallel for push pull, each transistor has a different voltage that it turns on. This voltage also changes with temperature. Question: if two turn on at the same voltage at 20C, will they both turn on at the same voltage at 50C? 90C?
Generally it's been brought to my attention that there are many variables that affect distortion, but the precise effect having a more powerful transistor in place (capable of higher max current) has on distortion - I'm still looking for
Because I changed the title, I need a preamble to explain a bit. Look at the chart below in OP.. you'll see that all 4 Denon AVRs ranging 105-150WPC and 9-11 AMPLIFIED channels, all begin distorting more starting at 12W. In my reading up to this point I feel comfortable stating all 4 receivers use 15A transistors. OP and post 9 explain the significance of this - I believe a big part of the issue this thread aims to tackle is that transistors require quite a bit of feedback to remain linear once pushed past 8-10% of their maximum rated current (for example, a 15A transistor at 50C behaves well (requires minimal feedback for output into resistive load to behave - up to 1.2-1.5A).
If you knew nothing of what I just said, you might think that if you had two very similarly designed AVRs, one rated for 105W RMS, and the other rated for 150W RMS, and you saw the 105W AVR's THD+n began increasing from 12W, you might hazard a guess that the 150W receiver might start distorting at 18W, because they're a similar designj, but the 150W receiver is capable of about 1.5X more power than the 105W. You'd also be wrong, because both receivers THD+n begin increasing at 12W.
Another interesting fact is that this distortion begins at 12W whether the load is 8 ohms or 4 ohms...
Yeah, that one's kind of a bombshell lol.
This evidence suggests the reason for the observed increased distortion after 12W is the transistor operating outside of its 8-10% zone previously mentioned. I have a class A amplifier I discuss later in OP/p9 which uses 10 transistors in parallel to accomplish operating with minimal feedback.
This thread is to brainstorm ways to reduce the amount of feedback required in push/pull designs. Obviously class A is easier to stick a bunch of transistors in parallel and call it a day.
Are we using greatly underpowered transistors in our class AB amplifiers? (the push-pull design biased for minimal crossover distortion)
I think we might be!
The reason is... actually, there are many. But to start, I'll begin with the thing most at the forefront of my thoughts - I have a Kinergetics KBA-280 - a stereo class A amplifier rated for 140W RMS into 8 ohms. It's a behemoth - it weighs more than any other amplifier I've lifted, and the transformer is a toroid. Inside there are... I want to say... 10 output transistors per channel, and they're not small. I forget what they are right now, but their package is like 1.8cm x 3.5cm x 0.4cm. They're MASSIVE transistors, and the amount of them in there? Is MASSIVE!
Stereophile reviewed it, and during their review, worked it HARD. How hard?
Single ohm hard.
Why? I don't know, probably had something to do with the weight of it and they're connoisseurs. Next is watts into 8, 4 and 2 ohms.
140, 270, 530, and it gets even better when the final power rating is taken while driving a single ohm...
A THOUSAND WATTS PER CHANNEL. But only one at a time, because the main fuse.
Right now we're talking about something else though: this thing's transistors. We can marvel later
For the sake of discussion, let's say behemoth's transistors are rated for 15A max. They're probably 15 amps (if they're not, they're more)
15A x 10 = 150
150 amps
That's a TONNE of amps! More than the service to my last place! (which I believe was 120A).
150A and an 8 ohm load? That's 1200 volts.
1200 x 150 = 180,000W
I looked at the datasheet and found out: where the transistor is most linear is in the first 7-10% of its "maximum" rating. Past that, it's pretty nonlinear. How does this nonlinearity increase? Nonlinearly! Funny, isn't it, that the increase in nonlinearity of a transistor is nonlinear itself! Lol. It kinda looks like y=a^x ("y" equals "a" to the power of "x").
Fortunately this unfortunate part of transistor existence doesn't affect its operation most of the time, because in addition to the transistor's input, feedback (how far off the transistor is from its intended path) is also applied, simultaneously. This ensures that the output becomes what's desired, and not just the transistor's nonlinear representation of its input. How is this feedback collected and how is it applied to linearize? It varies a lot, and can get complicated. Which is why, if you can, you minimize the amount of feedback required
And in the end there's less parts, too.
So moving on... let's use 8% of 150A
12A
with an 8 ohm load 1152 watts, and 96 volts
I think the transistors were 200V, which in class A gives a maximum of 70VAC. Since you can't hit the rails, call it 65VAC
528 watts, and linear behaviour to over 2x that amount of current.
From what I gather, these outputs would need minimal feedback because the transistors are operating well within their optimal operating range (that small fraction of their maximum rating).
There are 10 transistors being used to do the work that just one transistor can do.
The only real caveat to performance from running all 10 transistors in parallel in the way I've outlined (class A) is the noise floor would increase by 10dB. But even half decent transistors are pretty quiet, and 10dB is nothing compared to the benefits from running in linear territory.
Now the conversation is shifting though- background is done. The graphs below, though - they need to be considered as the first step to the next direction. Four of these charts are power vs. distortion of recent Denon home theater receivers. The fifth more bland looking one? Still from this site, just someone else measuring a Yamaha that has the same type of amplifier (AB), but it's rated for 60 watts RMS x2 instead of 105/125/140/150 watts (Denon X3700/4700/6700/8500H). Yamaha got a slot so that all the amps weren't Denons, and there was one significantly different. I believe all of the Denons were made with 15A transistors.
Aside that needs to be said that I wish I didn't because it's annoying: SO, the fourth Denon (X8500H) is represented by the 8 ohm chart instead of the 4 ohm chart like all the others, because the X8500H's 4 ohm chart is ugly as sin. One channel was very divergent, and ignoring it and putting a line through the other is stupid when the better looking 8 ohm chart is there to use. It shows the same result as the 4 ohm chart and does it lookin' good
'
3800 4800 6700 8500
Above you can see that the X3700H, X4700H, X6700H, and X8500H, are all capable of different maximum continuous stereo power outputs.
And you can see that they all four beasts are capable of the same amount of power before THD+N stops improving with increasing power.
WHY IS THIS?
8% of 15A is 1.2A
1.2A and 8 ohms is 11.52 watts.
The charts above, the vertical green lines I've marked between 10 and 20 watts, more accurately are between 12.2 and 14 watts.
Before accurately calling the change in a trend, a marked difference needs to happen. Those lines at 12.2 and 14 are after a noticeable change - a clear trend. If you look to precisely where the green line drawn from ~100mW to ~2000mW diverges.......
11.5W
11.5 watts x 2 channels is well below the maximum current able to be supplied by the power supply
11.5 watts x 2 channels into 8 ohms or 4 ohms requires much less than the amount of voltage supplied to the transistors
The new receivers for 2023: the X3800H, X4800H, X6700H, and X8500HA ALL have 15 amp transistors. I can't say for sure that the models above also have 15 amp transistors, but I know the X3700H does (I have one, wooo! partyy!), and I'm fairly sure the X4700H does. So if the X8500HA does too, I think it'd be stupid to think the 6700 didn't!
I think the reason distortion goes up at about 11.5 watts is because more feedback is needed.
Does anyone who knows more than me think what I described is plausible?
If so, what can we do moving forward designing amplifiers?
Putting a bunch in parallel works for class A amplifiers. For low distortion, class AB (working in B, push-pull) need to be biased to minimize crossover distortion as much as possible or none of the rest matters. If using multiple transistors in parallel for push pull, each transistor has a different voltage that it turns on. This voltage also changes with temperature. Question: if two turn on at the same voltage at 20C, will they both turn on at the same voltage at 50C? 90C?
Last edited: