I recently changed my input resistor from 1K to 22R.
This radically improved the sound - much cleaner & and clearer.
The difference can be seen very clearly in spice AC analysis of CLG - big improvement in stabilty but with still quite small ( 22p ) miller cap
Based on this I conclude that ultimately this amp will sound better with an active preamp.
I might even build one on the front end to include an i/p filter which cannot really be done with 22R.
Yet again this amp exceeds my expectations and anything I've heard before
This radically improved the sound - much cleaner & and clearer.
The difference can be seen very clearly in spice AC analysis of CLG - big improvement in stabilty but with still quite small ( 22p ) miller cap
Based on this I conclude that ultimately this amp will sound better with an active preamp.
I might even build one on the front end to include an i/p filter which cannot really be done with 22R.
Yet again this amp exceeds my expectations and anything I've heard before
Last edited:
Pawel,
My sim settings can be seen in the bottom left corner of the schematic image. Please do tell if I have something wrong!
In that case after some more final touches I will move on . . . .
or possibly develop it some more.
It's a bit disappointing to hear this damming verdict cos as JLH said what's the point of building yet another amp if it is not at least as good if not better than anything out there . . .
I wish I would!
but I am not the expert in LTSpice
Using b2spice I got very different distribution of the harmonics
confused,maybe you should rise the resolution, or use shorter timestep?
I wish I would!
but I am not the expert in LTSpice
Using b2spice I got very different distribution of the harmonics
maybe you should rise the resolution, or use shorter timestep?
Use something like this:
.options plotwinsize=0
.options method=gear
.options numdgt=7
.param Freq=20k
.param numcyc=10
.param dlycyc=5
.param FFT=2**16
.param simtime=numcyc/Freq+dlytime
.param dlytime=dlycyc/Freq
.param timestep=(simtime-dlytime)/FFT
.four {Freq} V(Vin) V(Vout)
.tran 0 {simtime} {dlytime} {timestep}
.option numdgt=15
.option reltol=1e-6
.option ptrantau=0
;ac dec 1k 1 1G
and for generator:
SINE(0 {sqrt(.9)} {freq})
dado
Sam
This is my favourite version so far. It sounds airy & relaxed but also very detailed and refined with nice sound textures.
The i/p stage ccs adds a lot more gain in the audio range but hardly any in the very high frequencies so stability is still good with 22p but I will also try 33p. Perhaps it also further reduces noise in the i/p stage.
I think we all have different tastes in amps but this one is now giving reproduction that I like very much - but still work in progress . . .
cheers
mike
p.s. I also have filters between o/p & other stages - 20R + 470uF
This is my favourite version so far. It sounds airy & relaxed but also very detailed and refined with nice sound textures.
The i/p stage ccs adds a lot more gain in the audio range but hardly any in the very high frequencies so stability is still good with 22p but I will also try 33p. Perhaps it also further reduces noise in the i/p stage.
I think we all have different tastes in amps but this one is now giving reproduction that I like very much - but still work in progress . . .
cheers
mike
p.s. I also have filters between o/p & other stages - 20R + 470uF
Attachments
Last edited:
Dadod, that gets us part of the way there, but it is still not enough. The slanted "noise floor" error in LTSpice is caused by DC drift of the DC blocking caps, most often the input and feedback caps. Make these 1 farad and the noise floor should fall greatly. You should also consider power supply DC drift, bootstrap drift, etc if this doesn't work.
After this is done, I find that I don't really need high-resolution simulator settings most of the time, except for perhaps a small timestep so that oscillation/stability is simulated accuracy. A large timestep will cause emulated RF loss at oscillation frequencies and so if it is too large, your circuit's simulated stability will not be anything like real life.
SWF, try putting R3 in series with C3 and see if this improves stability. My studies have suggested that this is a better way to do it, giving better stability and not limiting performance. If deep RF filtering is desired, I think it should be placed before R22, perhaps before the volume POT. I am considering soldering mine onto my volume POT so it'll be handy for any amplifier I experiment with.
- keantoken
After this is done, I find that I don't really need high-resolution simulator settings most of the time, except for perhaps a small timestep so that oscillation/stability is simulated accuracy. A large timestep will cause emulated RF loss at oscillation frequencies and so if it is too large, your circuit's simulated stability will not be anything like real life.
SWF, try putting R3 in series with C3 and see if this improves stability. My studies have suggested that this is a better way to do it, giving better stability and not limiting performance. If deep RF filtering is desired, I think it should be placed before R22, perhaps before the volume POT. I am considering soldering mine onto my volume POT so it'll be handy for any amplifier I experiment with.
- keantoken
More about amp diagram at post 2034
I listened more extensively last night to this version and it really is quite remarkable - It's extremely revealing without being harsh in any way. If you have a good source and you prefer to listen to good recordings this may be the amp for you - good recordings sound truly wonderful.
But . . . everything is revealed, which means that the shortcomings of lesser recording are quite obvious. However, I wouldn't say that it's ruthlessly revealing because the benign character of the amp seems to remain unchanged.
For instance "Ricky don't loose that number" just sounded like a great band standing around playing very well recorded great music, whereas Richard Bono's Album "Reverence", although the recording quality was truly excellent, sounded a bit synthetic because of the extensive compression that had been applied. Hadn't noticed it before, now couldn't miss it.
In spice it still complies with descending the H2345 goal but this series now begins at -117 & ends at -133db ! The question was raised earlier in the thread "Although using a ccs as i/p stage load resistor will increase OLG and drastically reduce distortion would it sound good ?" For me the answer is very clearly yes - providing you feed in a good signal
If I ever manage to build an amp with less / no feedback that still sounds this clean I might prefer it, but for now it's gonna be hard for me to go back previous versions.
Provisos & Cautions
This amp is now no longer unconditionally stable - if you would intend to use it with challenging loads you will need to take the appropriate measures.
Also this amp is now heavily dependent on FB for sound quality - even if you do not use them elsewhere I would recommend using high quality components for the feedback & gain resistors. I'm using Caddock TF020's from parts connextion.
The comments above refer to a sound system with choke regulation throughout which to a large extent prevents RF noise getting in anywhere and messing up the sound. This particularly applies to the earth.
cheers mike
I listened more extensively last night to this version and it really is quite remarkable - It's extremely revealing without being harsh in any way. If you have a good source and you prefer to listen to good recordings this may be the amp for you - good recordings sound truly wonderful.
But . . . everything is revealed, which means that the shortcomings of lesser recording are quite obvious. However, I wouldn't say that it's ruthlessly revealing because the benign character of the amp seems to remain unchanged.
For instance "Ricky don't loose that number" just sounded like a great band standing around playing very well recorded great music, whereas Richard Bono's Album "Reverence", although the recording quality was truly excellent, sounded a bit synthetic because of the extensive compression that had been applied. Hadn't noticed it before, now couldn't miss it.
In spice it still complies with descending the H2345 goal but this series now begins at -117 & ends at -133db ! The question was raised earlier in the thread "Although using a ccs as i/p stage load resistor will increase OLG and drastically reduce distortion would it sound good ?" For me the answer is very clearly yes - providing you feed in a good signal
If I ever manage to build an amp with less / no feedback that still sounds this clean I might prefer it, but for now it's gonna be hard for me to go back previous versions.
Provisos & Cautions
This amp is now no longer unconditionally stable - if you would intend to use it with challenging loads you will need to take the appropriate measures.
Also this amp is now heavily dependent on FB for sound quality - even if you do not use them elsewhere I would recommend using high quality components for the feedback & gain resistors. I'm using Caddock TF020's from parts connextion.
The comments above refer to a sound system with choke regulation throughout which to a large extent prevents RF noise getting in anywhere and messing up the sound. This particularly applies to the earth.
cheers mike
Last edited:
Mike,
I definitely agree with You about quality of recordings as I was make a lot of them (classical). Now You are able with very high quality playback chain to hear each "layer" of "messing" within the making of each track and everything is audible every single master fader when punch in acquired and not to mention the mental condition of the mixing and mastering person . . . Yes You enter now in this secret, mostly masked world of todays making music production . . . and If for instance You'll ever try the O/P of Yours amp in series to a 670R best Caddock 3W resistor following to a Sennheisser HD800 headphones, then You'll be surprised a lot more, what You'll hear from even the best current SACD music production money can buy.
I wish You a lot of Joy with Yours amp.
Andrew
I definitely agree with You about quality of recordings as I was make a lot of them (classical). Now You are able with very high quality playback chain to hear each "layer" of "messing" within the making of each track and everything is audible every single master fader when punch in acquired and not to mention the mental condition of the mixing and mastering person . . . Yes You enter now in this secret, mostly masked world of todays making music production . . . and If for instance You'll ever try the O/P of Yours amp in series to a 670R best Caddock 3W resistor following to a Sennheisser HD800 headphones, then You'll be surprised a lot more, what You'll hear from even the best current SACD music production money can buy.
I wish You a lot of Joy with Yours amp.
Andrew
Just some remark.
Using a MOSFET as a VAS device (and trying to use a degeneration resistor) just won't work (well). It's because a MOSFET is a voltage-to-current device and not a current-to-current device is. You can attempt to use HEXfets for this as they generally have a huge Vgs(d)/Ids(d) slope, but consequently this needs a current source for the VAS rail since getting a highly accurate/stable Vgs for your intended driver current is nigh impossible. BJTs are unbeatable for this particular amp stage.
Using a MOSFET as a VAS device (and trying to use a degeneration resistor) just won't work (well). It's because a MOSFET is a voltage-to-current device and not a current-to-current device is. You can attempt to use HEXfets for this as they generally have a huge Vgs(d)/Ids(d) slope, but consequently this needs a current source for the VAS rail since getting a highly accurate/stable Vgs for your intended driver current is nigh impossible. BJTs are unbeatable for this particular amp stage.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.