Bias should be like 19.5 mA for best results.
Then the simulated THD is 0.00013% at 1 kHz 1 Watt into 8 Ohm.
The output uses 2SC5200/2SA1943 in Sziklai configuration.
Then the simulated THD is 0.00013% at 1 kHz 1 Watt into 8 Ohm.
The output uses 2SC5200/2SA1943 in Sziklai configuration.
Last edited:
That does only work in simulation. D44/5 unfortunately will not hold up to +/-25V operation. Too fragile - even 8 ohm is taking your life into your hands. On +/-15 it’s hard to beat them, but going higher is dangerous unless it’s a tweeter amp for 2KHz+. MJE15030/1 fare better, but unfortunately like most high voltage transistors, have quasi-saturation to deal with.
But datasheet tells D44/45 has max collector current 10 A.
And peak current 20 A.
These are high values.
30 Watt output is peak current 2.73 A.
So I have no fear D44/45 will be blown.
Or ... ?
And peak current 20 A.
These are high values.
30 Watt output is peak current 2.73 A.
So I have no fear D44/45 will be blown.
Or ... ?
Transistors designed for high speed switching operation often have poor SOA's. These appear to share that limitation.
I speak from experience. I designed a very nice-sounding amplifier that used fast switching transistors. It sounded very nice but I got tired of replacing blown-up output transistors.
I speak from experience. I designed a very nice-sounding amplifier that used fast switching transistors. It sounded very nice but I got tired of replacing blown-up output transistors.
Look at the SOA graph. Second breakdown limits start kicking in (with a vengence) at a whopping 10 V. Tried +/-27 once, never again. Worked beautifully at +/18 as an upgrade of TIP41/2.
Ya, 10 Amps at about 3.5V. I learned about SOA and secondary breakdown about 50 years ago.
At 25 Volts D45H11s are good for about 800mA, but capacitive loads will draw peak current at zero volts out, ie ~3A@25V. I would not run any TO-220 over 40V total (+/-20V). Also remember that your OPs are going to get hot, even with a good heat sink so derate the 25C power rating by about half. Also, no OP protection is dangerous, unless you like constantly rebuilding your amp. If you don't want to add protection, use very large transistors. When I was young, these mistakes were common, and amps blew up all the time. Today, Silicon is cheap so don't waste your time on < dollar power transistors.
BTW, I would put some gain in the CFP output so that you can use a regular op-amp (+/-15V), and you will not be limited by the op-amp output saturation. And you don't need the current sources because the op-amp will drive a couple bias resistors, no CCS or bootstraps required.
At 25 Volts D45H11s are good for about 800mA, but capacitive loads will draw peak current at zero volts out, ie ~3A@25V. I would not run any TO-220 over 40V total (+/-20V). Also remember that your OPs are going to get hot, even with a good heat sink so derate the 25C power rating by about half. Also, no OP protection is dangerous, unless you like constantly rebuilding your amp. If you don't want to add protection, use very large transistors. When I was young, these mistakes were common, and amps blew up all the time. Today, Silicon is cheap so don't waste your time on < dollar power transistors.
BTW, I would put some gain in the CFP output so that you can use a regular op-amp (+/-15V), and you will not be limited by the op-amp output saturation. And you don't need the current sources because the op-amp will drive a couple bias resistors, no CCS or bootstraps required.
I have changed output transistors to SC5200 and 2SA1943.
See new schematic in post number 1.
Now there is no fear of overloading.
See new schematic in post number 1.
Now there is no fear of overloading.
Like I said, tweeter amp. A couple of good kick drum hits into a 30 degree load and poof.
One can always use full size power transistors. When they do come out with a shrunk die 50-70 watt TO-220 type that is 50 MHz, handles full power to 30 or more volts and is free of quasi-saturation effects we’ll be all over it like flies on horse $#**. The way things are going I doubt it will ever be developed.
I used Irf530 and irf9530 for 15 years in a car amp driving 4 ohms with 35v rails .
(1 each per channel)
These transistors shouldn't have a problem.
(1 each per channel)
These transistors shouldn't have a problem.
Agree. The problematic amp I described was designed/built in the '80s. I don't see much point in revisiting it today.
IRF530/9530 are one of the old style hexfets which have good SOA. The 240/9240 are the good “big ones” which can take 200 wpc rail voltages. Some people don’t like using hexfets for audio, though.
Many reasons have been cited over the years.
Piispoor match between N and P
Higher optimum bias current (compared to bipolar)
Lower gm
Requirement to drive above the rails as followers (if excess losses not accepted)
Harder to drive by the dozen
No good p-channel for >200V operation
And of course, the biggie - when one’s objective was to build with bipolars in the first place. Some builders just stay away from them. Or insist on laterals if they use FETs at all.
Piispoor match between N and P
Higher optimum bias current (compared to bipolar)
Lower gm
Requirement to drive above the rails as followers (if excess losses not accepted)
Harder to drive by the dozen
No good p-channel for >200V operation
And of course, the biggie - when one’s objective was to build with bipolars in the first place. Some builders just stay away from them. Or insist on laterals if they use FETs at all.
The vertical MOSFETs all/mostly seem to be designed to deliberately use positive thermal feedback to increase switching speed. Even if the SOA allows linear operation without the device blowing itself up, the Vgs threshold voltage usually still has a huge gap between hot and cold. That one and only sole manufacturer of laterals is a saving grace, and I hope they come out with more of them.
Local NFB should be used, I think. Not just de-gen resistors but local loops. Relying solely on a global loop with maxed-out gain is not a good idea, IMO, because the N and P halves are never perfectly matched, so the individual harmonics are poorly defined, sometimes cancelling out, and sometime not, depending on whatever voltage and current offsets are caused by the signal and load.
What is interesting is the low bias level in the Sziklai output.
Only 27.2 mA was found to be optimal for lowest THD.
Only 27.2 mA was found to be optimal for lowest THD.
I have updated with a new version in first post.
The emitter resistors in the output is now 0.22 Ohm.
It is the amplifier I am most pleased with.
The emitter resistors in the output is now 0.22 Ohm.
It is the amplifier I am most pleased with.
