I have not read all 700 pages, but took a good look at the schematic on post 13981. Looks very nice, Damn good maybe, but not quite ultimate.
A few thoughts from my experience. Take with a grain of salt.
Giving up a a lot of swing with the size of the rail resistors. Maybe a "ripple eater" or as I used, a simple pi filter. ( 8X improvement in ripple)
Ultimate would be an independent regulated supply a few volts higher than the output rails.
IPS CCS is not as stiff as could pretty easily be achieved. An led is a bit more thermally stable than the two diodes. BJT/FET cascode is stiff as all get out.
I stumbled on playing with the amount of LTP degen vs global feedback and find there seems to be a sweet spot for every case. It seems to trade off even vs odd harmonics. One amp came out as the best compromise with 10 Ohms, another with 43. Self likes 63 but does not say why.
No output filter network? Fine if driving a resistor, but if a speaker, you might run into stability trouble.
I would have put the rail fuses outside the bypass caps. Electrolytics can short. But if one blew, it would not start a fire or blow up the amp.
No output protection? I understand even a fuse can cause distortion, but real world, DC sense, thermal and an a RF detector to detect instability? Or a fuse.
I can tell you, MOSFETS blow faster than fuses. Personal experience.
Some consider it evil, but I hate turn on and off thumps, so I sacrifice ultimate to use a muting relay. You can use the same sensing and relay driver to disconnect the rails as well as the speaker. Rail relay does not have to be good. Cheap will do.
One can then argue if the feedback should be in or outside the fuse.
The current fad is for differential input. I agree with your design it is not where to spend your design time for a home amplifier. If you already had a DAC and preamp that was diff, then I guess why not.
A feature I have added to amps is a low frequency HP filter. If driving full range speakers, it is wise to have an out- 20-30-40 Hz or so HP to limit cone excursion of your woofers below their F3 tuning. I think it would be smart to have an option of a 80Hz HP built in for running a sub. When I think ultimate, I think the entire system, not just the amp.
I was also looking at 139282. How much do you gain with the FETs? I have thought about them, but I can achieve simulations in the .00003% range ( on the bench, .003%) "FFT 10" levels of distortion without them.
Curious. I'm not energetic enough to build another amp. If mine ever failed, I would just buy a Benchmark. Just did a fresh re-cap, so it should last me until I just need a table radio in the nursing home.
A few thoughts from my experience. Take with a grain of salt.
Giving up a a lot of swing with the size of the rail resistors. Maybe a "ripple eater" or as I used, a simple pi filter. ( 8X improvement in ripple)
Ultimate would be an independent regulated supply a few volts higher than the output rails.
IPS CCS is not as stiff as could pretty easily be achieved. An led is a bit more thermally stable than the two diodes. BJT/FET cascode is stiff as all get out.
I stumbled on playing with the amount of LTP degen vs global feedback and find there seems to be a sweet spot for every case. It seems to trade off even vs odd harmonics. One amp came out as the best compromise with 10 Ohms, another with 43. Self likes 63 but does not say why.
No output filter network? Fine if driving a resistor, but if a speaker, you might run into stability trouble.
I would have put the rail fuses outside the bypass caps. Electrolytics can short. But if one blew, it would not start a fire or blow up the amp.
No output protection? I understand even a fuse can cause distortion, but real world, DC sense, thermal and an a RF detector to detect instability? Or a fuse.
I can tell you, MOSFETS blow faster than fuses. Personal experience.
Some consider it evil, but I hate turn on and off thumps, so I sacrifice ultimate to use a muting relay. You can use the same sensing and relay driver to disconnect the rails as well as the speaker. Rail relay does not have to be good. Cheap will do.
One can then argue if the feedback should be in or outside the fuse.
The current fad is for differential input. I agree with your design it is not where to spend your design time for a home amplifier. If you already had a DAC and preamp that was diff, then I guess why not.
A feature I have added to amps is a low frequency HP filter. If driving full range speakers, it is wise to have an out- 20-30-40 Hz or so HP to limit cone excursion of your woofers below their F3 tuning. I think it would be smart to have an option of a 80Hz HP built in for running a sub. When I think ultimate, I think the entire system, not just the amp.
I was also looking at 139282. How much do you gain with the FETs? I have thought about them, but I can achieve simulations in the .00003% range ( on the bench, .003%) "FFT 10" levels of distortion without them.
Curious. I'm not energetic enough to build another amp. If mine ever failed, I would just buy a Benchmark. Just did a fresh re-cap, so it should last me until I just need a table radio in the nursing home.
Some members here design an amplifier which is the measurement very close to the simulation. If you get measurement very far from simulation, there are some possibility:
1. the components model in simulation is not accurate
2. the implementation is not good, like pcb layout design and components choice.
3. the measurement is not accurate or wrong.
My simulation does not include the parasitics. Generic resistor models are perfect resistors. Transistor models questionable, though most of mine are Cordell.
A power amplifier that measures .003 is rather outstanding. I am talking 20-20 driving an actual speaker, Not 1K best case driving a resistor. And yea, my measurement capability is not big bucks. The amp barely shows up above my residual. Quite pleasing to listen to.
I would love to have it benched on some mega dollar professional system as it might be informative to compare to others that measure well and sound poor, and some that measure well and sound great. At what point are the measurements we can make separating the amps with a sonic signature from ones without? Is there something else to measure? Of course if we knew that, a lot of magazine reviewers would be out of business along with a vast majority of "prestige" brands.
A power amplifier that measures .003 is rather outstanding. I am talking 20-20 driving an actual speaker, Not 1K best case driving a resistor. And yea, my measurement capability is not big bucks. The amp barely shows up above my residual. Quite pleasing to listen to.
I would love to have it benched on some mega dollar professional system as it might be informative to compare to others that measure well and sound poor, and some that measure well and sound great. At what point are the measurements we can make separating the amps with a sonic signature from ones without? Is there something else to measure? Of course if we knew that, a lot of magazine reviewers would be out of business along with a vast majority of "prestige" brands.
then TPA3255 is your choice, I think it will show itself well on a real load. But, I have not seen such comparative measurements.
I don't understand enough about specific chips, but it seems several newer amps using custom, Purifi, and Hypex have fantastic specs in SNAID measurement. It is likely the results have most to do with implementation, just as DACs seem to be more implementation than chip. Same even with the old Tripath. Some were made into very nice amps. Most were not as they were cheap.
I am considering ordering one to see if the numbers tell enough of the full story. I wish I could borrow one just to see as I really like my own amp, but that is not to say another is not better. Don't want to jerk around Amazon or Crutchfield, let alone March clear on the other side of the world for curiosity. I have no ego problem if my amp gets shown up. It is only about the sound. Actually, if a class D does show up my amp, I would gladly say they have arrived.
In my dreams: A March, Atoll, Vidar, and Benchmark for a month. Gad, if they actually were all good enough to sound the same! The Topping PA5 is maybe not quite enough power, but my Creek did pretty well while I was refurbing my MOSFET.
yes, for 99% TPA3255 will be ok. I prefer with FB. without PFFB, sibilants on HF more able to be heard ...
Certainly worth a try. But maybe AX-20 right away?
and what amplifiers do you have now, what do you consider so far unsurpassed?
Hello everyone... just out of curiosity, I simulated the FH11 amplifier circuit and added a current mirror in the differential pair, which drastically reduced the distortion. Layman's question: is there any obstacle in including a current mirror in the differential pair circuit in FH11? Thank you in advance for your patience and responses.
It’s DIY so you are free to try out anything you want. Sort of like how I tried to make the FX8 a vertical MOSFET from a latFET. Adding a current mirror will change the amp’s signature sound for sure. Have you listened to a base FH9 or FH11? Try one first and have a reference to compare after you make a change that is supposed to make it better. The FH9/11 is already one of the best sounding amps there is out there IMO. It’s got a lot of solid bass and the mids and highs are all clear. The engagement factor is very very nice.
Can you please give the best way to assemble PCBs of these type of layouts on heat-sink by maintaining the flexibility to access the soldering side for measurements..
Last edited:
