OK, I tend to agree - these outputs are too slow.
IPS is lightning-fast. Just terribly fast
Controlled CCSs for input buffer work perfectly.
Also, "20-th century" protection works perfectly, saving most of the circuit and PSU in case of "trouble" - I didn't have over-current sensor connected, but it shuts down on "AC Failure", catching the power supply drop when some output transistor shortens... great board Although, I'm going to connect over-current sensor now - it will shut down even faster then.
OK, ordering some NJW1302/3281 too.
MT200 Sankens are great, but a bit biggish for this PCB...
IPS is lightning-fast. Just terribly fast
Controlled CCSs for input buffer work perfectly.
Also, "20-th century" protection works perfectly, saving most of the circuit and PSU in case of "trouble" - I didn't have over-current sensor connected, but it shuts down on "AC Failure", catching the power supply drop when some output transistor shortens... great board
Although, I'm going to connect over-current sensor now - it will shut down even faster then. OK, ordering some NJW1302/3281 too.
MT200 Sankens are great, but a bit biggish for this PCB...
I have removed these caps... also slowed down the IPS a bit with more aggressive compensation...
I thought about CCSs, but closer research showed no real difference depending on frequency. By the way, I've tried to increase C35, C36 to 2.2uF - no difference in behavior noticed...
But Zobel! Of course! I think, that's the one, terrorizing the drivers now
Thanks LC
Tamed!
OK, I have tamed it. I would say, this is the first time I had to deal with IPS, which is that fast. MJL21193/21194 is not an option here. They work fine in audio range, but get killed easily as soon as 500KHz square wave appears at the input (closing too slowly).
I've got no MJL3281/1302 in hand at the moment. So what I did... you wouldn't believe it Or maybe you would, but it was a bit unexpected for myself - at least, I didn't plan it initially
"Composite" Sziklai with HEX FETs. Good old rugged IRFP240/9240. Their high non-linear input capacitance doesnt hurt in CFP arrangement with its 100% local NFB. Otherwise - they are fast and easily deal with high currents if necessary. Base stoppers and shunt capacitors at the drivers' bases play an important complex role here. Being a part of the global NFB loop they also prevent the "evil" CFP pairs from local oscillation. Optimal values are 68R / 47pF each pair. Miller caps around the super-pairs are 68pF each. No lead compensation caps at all.
Just in case, I keep the input HF filter cutoff frequency at 480KHz (lower than I usually do).
The screenshots show the square wave response at 1KHz, 20KHz and 50KHz as well as clipping at 20KHz (superpairs in action).
THD is virtually constant throughout the audio bandwidth - around 0.003% @ 2W, going up to 0.015% @ 20W and 0.03% @100W.
IMD @ 2W is less than 0.005%.
OPS quiescent current is 60mA per pair. Thermal stability is no problem (Vbe multiplier transistor senses the drivers).
BTW, if the quiescent current is set to 30mA, THD at high powers halves, but the one at the lower powers - roughly doubles, still being practically frequency-independent up to 20KHz.
It still may be tested with NJW3281/1302, as well as with EF2/EF3 OPS options.
It sounds... how to say it... sharp / precise / crisp / well-defined. A bit "harder" than my hybrid VFAs (Low TIM hybrid and TubSuMo).
I will post "as built" schematic and corrected PCB later today.
Cheers,
Valery
OK, I have tamed it. I would say, this is the first time I had to deal with IPS, which is that fast. MJL21193/21194 is not an option here. They work fine in audio range, but get killed easily as soon as 500KHz square wave appears at the input (closing too slowly).
I've got no MJL3281/1302 in hand at the moment. So what I did... you wouldn't believe it
Or maybe you would, but it was a bit unexpected for myself - at least, I didn't plan it initially "Composite" Sziklai with HEX FETs. Good old rugged IRFP240/9240. Their high non-linear input capacitance doesnt hurt in CFP arrangement with its 100% local NFB. Otherwise - they are fast and easily deal with high currents if necessary. Base stoppers and shunt capacitors at the drivers' bases play an important complex role here. Being a part of the global NFB loop they also prevent the "evil" CFP pairs from local oscillation. Optimal values are 68R / 47pF each pair. Miller caps around the super-pairs are 68pF each. No lead compensation caps at all.
Just in case, I keep the input HF filter cutoff frequency at 480KHz (lower than I usually do).
The screenshots show the square wave response at 1KHz, 20KHz and 50KHz as well as clipping at 20KHz (superpairs in action).
THD is virtually constant throughout the audio bandwidth - around 0.003% @ 2W, going up to 0.015% @ 20W and 0.03% @100W.
IMD @ 2W is less than 0.005%.
OPS quiescent current is 60mA per pair. Thermal stability is no problem (Vbe multiplier transistor senses the drivers).
BTW, if the quiescent current is set to 30mA, THD at high powers halves, but the one at the lower powers - roughly doubles, still being practically frequency-independent up to 20KHz.
It still may be tested with NJW3281/1302, as well as with EF2/EF3 OPS options.
It sounds... how to say it... sharp / precise / crisp / well-defined. A bit "harder" than my hybrid VFAs (Low TIM hybrid and TubSuMo).
I will post "as built" schematic and corrected PCB later today.
Cheers,
Valery
Attachments
Why not build a wee bit of asymmetry into the front end to increase H2 which will soften the sound, add euphony and perhaps, subjectively be more musical?
Hi Valery
Nice progress, measurements tends to be very promising, sound can still be shaped with fine tuning of OLG, HF compensation and bias currents of separate stages.
Can you please check R26 voltage drop delta from cold start to normal working conditions?
Any benefit if VAS voltage gain be increased and/or VAS bias would be higher, ie. 10-15 mA?
L.C.
Nice progress, measurements tends to be very promising, sound can still be shaped with fine tuning of OLG, HF compensation and bias currents of separate stages.
Can you please check R26 voltage drop delta from cold start to normal working conditions?
Any benefit if VAS voltage gain be increased and/or VAS bias would be higher, ie. 10-15 mA?
L.C.
Hi LC, R26 voltage drop goes like this:
10s - 0.79V
30s - 0.80V
1m - 0.82V
2m - 0.84V
3m - 0.85V
4m - 0.86V
6m - 0.87V
9m - 0.88V and then it stays at this level
I simulated higher VAS current (14mA with 3.3K input buffer CCS emitter resistors) - it doesn't show any advantages... almost 9mA seems to be enough for driving this OPS smoothly. I will probably try it live a bit later, just to make sure...
Attached is the schematic "as built" for now.
Updating the PCB
Cheers,
Valery
10s - 0.79V
30s - 0.80V
1m - 0.82V
2m - 0.84V
3m - 0.85V
4m - 0.86V
6m - 0.87V
9m - 0.88V and then it stays at this level
I simulated higher VAS current (14mA with 3.3K input buffer CCS emitter resistors) - it doesn't show any advantages... almost 9mA seems to be enough for driving this OPS smoothly. I will probably try it live a bit later, just to make sure...
Attached is the schematic "as built" for now.
Updating the PCB
Cheers,
Valery
Attachments
I don't see it as an issue - it's not dry or sterile, the character is there - just a bit tighter bass control, probably this is what gives a sort of "harder" impression. Highs are marvelous, very natural. Also, such OPS topology allows high currents with fast response - so, I think, what I hear so far, is just reflecting the specialties of this amp. Need more listening of different material though...
Hi Kimschips, I didn't measure overall slew rate yet, but from what I saw while measuring the front-end solely - rise/fall times barely differentiated from the ones at the input. OPS is the one, slowing it down to some extent, but it's still one of the fastest ones I tested live.
Overall phase shift at 20KHz is less than 3 degrees. IPS itself showed some 1.4 degrees
And yes, I plan to test it with Slewmaster-based MosFet OPS later on...
You may have a question about some "strange" 0.22R resistors at the output - the only reason for them is having additional holes in PCB for the vertical type of resistors and still maintain compatibility with schematic in DipTrace. I use 5W cement ones.
BTW, PCB is fully compatible, regardless of what kind of output devices you use (FETs or BJTs) - no changes required except some parts' values. BCE <-> GDS
BTW, PCB is fully compatible, regardless of what kind of output devices you use (FETs or BJTs) - no changes required except some parts' values. BCE <-> GDS
Hi Jeff, not sure about some audible difference - somebody has to compare
Dual-CFP is interesting in general, there are not too many amps using it, but the ones that do - worth listening to...
Hex FETs are really rugged ones, no problem running pretty hot at high power, also having high conductance when fully open. Great for the purpose. I like them a lot.
Dual-CFP is interesting in general, there are not too many amps using it, but the ones that do - worth listening to...
Hex FETs are really rugged ones, no problem running pretty hot at high power, also having high conductance when fully open. Great for the purpose. I like them a lot.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.