Little Dot MK8SE / MK6 Super Mods (All verified mods are on first page)
- Thread starter Redge78
- Start date
coinmaster
1000+ Head-Fier
Here's an interesting article by Nelson Pass, the jesus of audio.
http://www.6moons.com/industryfeatures/distortion/distortion.html
He explains why NFB is bad and why THD is a load of BS.
http://www.6moons.com/industryfeatures/distortion/distortion.html
He explains why NFB is bad and why THD is a load of BS.
claverocks
New Head-Fier
- Joined
- Mar 12, 2015
- Posts
- 6
- Likes
- 10
Look at these nice tubes!
The VT-99 look sweet and smaller..
Alot testing to commence.
The VT-99 look sweet and smaller..
Alot testing to commence.
coinmaster
1000+ Head-Fier
Wow nice, how much did those set you back? Are those the new PSvane treasure tubes?
I have news as well, I did a nice long simulation of the Mk6 with the impedance mod vs my new design I'm working on.
Here is the simulated distortion of the MK6 up to 20khz
This is the simulated distortion of my design, it took half the day to simulate due to the complexity of it and there is still room for plenty of optimizations I think.
Dayyyuuuum
It looks almost like the MK6 without all the distortion. That's with the triodes slightly out of balance too and very little NFB was used.
I'm confident I can improve it further without the use of NFB but the damn simulations for it take ages to load .
What type of distortion display was that?
And can you run a stock version as well?
It would have been better to test against stock as we don't know what parameter you chose to do impedance mod.
Only the bias mod was posted.
The impedance mod may vary upon setting.
And can you run a stock version as well?
It would have been better to test against stock as we don't know what parameter you chose to do impedance mod.
Only the bias mod was posted.
The impedance mod may vary upon setting.
coinmaster
1000+ Head-Fier
FFT analysis
Yeah, I'll load it up and post it when it's done.
I used 800 ohm resistor for 400 ohm load.
The simulation is showing perfect balance between triodes so it is working as intended.
Interestingly I can't remember how I originally came up with the value of 800 but it does seem to be correct.
I went back to the cavalli papers and redid the calculation but I can't seem to come to the same conclusion, hmmm. Maybe I just suck at math
I did another sim of my amp, strangely it loaded quickly this time, only about 30 minutes to load the sim and about 1 second to actually do the simulation, it's almost comedic.
I removed all NFB and this is what I got
Interestingly a lot of the slant is gone but the the harmonics are back. Still a great improvement over the MK6 though since the MK6 has a lot of NFB and this has none and it's still better.
The design still has some kinks to be fixed so I suspect it will get reasonably better, hopefully I can get the .0001% open loop distortion MrCurwen spoke of
Unfortunately as soon as I removed the NFB it started clipping balls, it seems I inadvertently created a cascode between stages, I'm going to have to reduce gain majorly or end up using autoformer pots for gain reduction.
MrCurwen
100+ Head-Fier
- Joined
- Dec 26, 2015
- Posts
- 337
- Likes
- 133
NFB as a theoretical concept is not bad for audio signal reproduction, not bad at all.
For example, the reason a triode with good curves is linear, is because of NFB inside the tube.
The electron stream traveling in vacuum and different electric fields (such as the field around the control grid) has a feedback loop in it. This is what linearizes a triode.
The strenght of this NFB inside a triode is dependent on the geometry of the tube; how the electrodes (cathode/filament, control grid, anode) are placed in relation to each other.
There are triodes with "tetrode like curves" that don't have the screen grid. This is because the distance between the cathode and anode is so great that the feedback effect is very weak. (Very very high rp)
There are even triodes with pentode like curves (although a weak pentode); they even have beam plates! They are called beam triodes. They are usually for some extremely high voltage duties. The distance between the electrodes is so great, the tube can withstand something like 10kV of voltage. Good for some radar and large screen CRT applications.
For audio use triodes have a lot of NFB inside them. This is not bad, because what makes NFB bad is WHAT IS INSIDE THE LOOP. In a triode, there is vacuum.
Put a resistor inside a NFB loop. Not bad. (For example a CCS)
Put a capacitor inside a NFB loop. Not bad (if decent film cap).
Put a resistor and a capacitor inside NFB loop; not bad at all. (Say, the gyrator.)
Put a complex system of voltage and/or current gain stages inside a NFB loop, and you've got big big problems. This is the SS SOUND.
A regulator is a 100% degenerated NFB loop. The loop is local, very short, and doesn't contain any significant amount of reactive elements. No SS nastiness or problems.
The Little Dot's 'regulator' in the PSU (it's really a voltage setter, since the feedback loop is only (mainly) affecting the PSU ripple, not the audio side of the circuit; it sets the voltage, it doesn't REGULATE the B+) has a 100% degenerated NFB loop. Does it bring SS nastiness into the sound?
SS is not bad inherently. If you put SS inside a local, small and simple NFB loop, it's just as un-nasty as a triode with it's local, small and simple NFB loop.
This is a rudimentary explanation of the subject, but just to illuminate a bit about these things.
A tetrode or a pentode don't have triode curves because the screen grid breaks the NFB loop between the plate and the cathode. It breaks it by having an independent electric field in the path of the electron stream. If you want to have your pentode behave more like an ideal pentode (so, more like a MOSFET in essence), you regulate the screen supply, making it very sturdy. This makes the screen grid electric field more independent of the anode conditions, and breaks the feedback loop in the electron stream more, thereby making the curves more horizontal and flat.
If you connect the screen to the plate, you no longer break the NFB loop, therefore you have more linear curves. After you do that, the linearity of the system is again dependent on the inner geometrics of the tube. Most tetrodes, beam tetrodes or pentodes are not designed with triode operation in mind, so they have rather poor geometry for triode operation linearity. There are a notable number of exceptions, such as 4P1L.
4P1L was actually designed by nazi scientists. It was probably meant for a very wide variety of applications, and that is why it was designed with geometry suitable for triode operation as well.
Most pentodes were not. Linear triodes were around, so why bother? The nazis were operating under very severe shortages, and couldn't afford to make that many different types (each needing their own factories etc.) Same goes for the soviets.
People think because SS doing voltage amplification necessitates a LOT of NFB to make it linear, thereby causing A LOT of SS nastiness, that all SS components by nature produce this nastiness in all circuits, no matter what they are used for.
A general rule; if the SS component has 100% degeneration, meaning it exhibits no voltage gain, it WILL NOT PRODUCE ANY NASTINESS. At all. No more than your 1930's triodes. Actually, less than that if the 1930's triode is exhibiting voltage gain, it will produce some harmonics. The 100% degenerated SS component will not produce any harmonics, because 100% degeneration means just that.
If you have 97% degeneration, you will have harmonics.
In short, it's simply idiotic to compare SS with voltage gain to SS with no voltage gain. Any articles discussing the first and reaching a conclusion IT'S BAD FOLKS is not applicable to source followers or 100% degenerated regulators and such. They are different things.
A CCS is not exhibiting any voltage gain to the circuit. There is gain inside the loop, but it's 100% degenerated. This means 100% of the gain is dissipated in the loop, in audio frequencies. Sometimes there is gain with radio frequencies (oscillation) and that needs to be dealt with.
A CCS will not contribute any signature to the audio, other than by correctly organizing the circuit for example balancing a triode pair in a LTP, or sinking grid current from a power tube grid.
For example, the reason a triode with good curves is linear, is because of NFB inside the tube.
The electron stream traveling in vacuum and different electric fields (such as the field around the control grid) has a feedback loop in it. This is what linearizes a triode.
The strenght of this NFB inside a triode is dependent on the geometry of the tube; how the electrodes (cathode/filament, control grid, anode) are placed in relation to each other.
There are triodes with "tetrode like curves" that don't have the screen grid. This is because the distance between the cathode and anode is so great that the feedback effect is very weak. (Very very high rp)
There are even triodes with pentode like curves (although a weak pentode); they even have beam plates! They are called beam triodes. They are usually for some extremely high voltage duties. The distance between the electrodes is so great, the tube can withstand something like 10kV of voltage. Good for some radar and large screen CRT applications.
For audio use triodes have a lot of NFB inside them. This is not bad, because what makes NFB bad is WHAT IS INSIDE THE LOOP. In a triode, there is vacuum.
Put a resistor inside a NFB loop. Not bad. (For example a CCS)
Put a capacitor inside a NFB loop. Not bad (if decent film cap).
Put a resistor and a capacitor inside NFB loop; not bad at all. (Say, the gyrator.)
Put a complex system of voltage and/or current gain stages inside a NFB loop, and you've got big big problems. This is the SS SOUND.
A regulator is a 100% degenerated NFB loop. The loop is local, very short, and doesn't contain any significant amount of reactive elements. No SS nastiness or problems.
The Little Dot's 'regulator' in the PSU (it's really a voltage setter, since the feedback loop is only (mainly) affecting the PSU ripple, not the audio side of the circuit; it sets the voltage, it doesn't REGULATE the B+) has a 100% degenerated NFB loop. Does it bring SS nastiness into the sound?
SS is not bad inherently. If you put SS inside a local, small and simple NFB loop, it's just as un-nasty as a triode with it's local, small and simple NFB loop.
This is a rudimentary explanation of the subject, but just to illuminate a bit about these things.
A tetrode or a pentode don't have triode curves because the screen grid breaks the NFB loop between the plate and the cathode. It breaks it by having an independent electric field in the path of the electron stream. If you want to have your pentode behave more like an ideal pentode (so, more like a MOSFET in essence), you regulate the screen supply, making it very sturdy. This makes the screen grid electric field more independent of the anode conditions, and breaks the feedback loop in the electron stream more, thereby making the curves more horizontal and flat.
If you connect the screen to the plate, you no longer break the NFB loop, therefore you have more linear curves. After you do that, the linearity of the system is again dependent on the inner geometrics of the tube. Most tetrodes, beam tetrodes or pentodes are not designed with triode operation in mind, so they have rather poor geometry for triode operation linearity. There are a notable number of exceptions, such as 4P1L.
4P1L was actually designed by nazi scientists. It was probably meant for a very wide variety of applications, and that is why it was designed with geometry suitable for triode operation as well.
Most pentodes were not. Linear triodes were around, so why bother? The nazis were operating under very severe shortages, and couldn't afford to make that many different types (each needing their own factories etc.) Same goes for the soviets.
People think because SS doing voltage amplification necessitates a LOT of NFB to make it linear, thereby causing A LOT of SS nastiness, that all SS components by nature produce this nastiness in all circuits, no matter what they are used for.
A general rule; if the SS component has 100% degeneration, meaning it exhibits no voltage gain, it WILL NOT PRODUCE ANY NASTINESS. At all. No more than your 1930's triodes. Actually, less than that if the 1930's triode is exhibiting voltage gain, it will produce some harmonics. The 100% degenerated SS component will not produce any harmonics, because 100% degeneration means just that.
If you have 97% degeneration, you will have harmonics.
In short, it's simply idiotic to compare SS with voltage gain to SS with no voltage gain. Any articles discussing the first and reaching a conclusion IT'S BAD FOLKS is not applicable to source followers or 100% degenerated regulators and such. They are different things.
A CCS is not exhibiting any voltage gain to the circuit. There is gain inside the loop, but it's 100% degenerated. This means 100% of the gain is dissipated in the loop, in audio frequencies. Sometimes there is gain with radio frequencies (oscillation) and that needs to be dealt with.
A CCS will not contribute any signature to the audio, other than by correctly organizing the circuit for example balancing a triode pair in a LTP, or sinking grid current from a power tube grid.
MrCurwen
100+ Head-Fier
- Joined
- Dec 26, 2015
- Posts
- 337
- Likes
- 133
So in short: NFB is totally ok or completely crap depending on what is inside the loop.
SS is totally ok or completely crap depending on if it has voltage gain or not.
There is a feedback loop inside the voice coil of your headphones. The current traveling inside the coil is producing an electric field that is affecting the other parts of the coil, giving the coil IMPEDANCE.
Partial NFB (not 100%) over gain stages, worst case over multiple gain stages or worst possible case over multiple gain stages, multiple electrolytic caps, and a transformer ... You've got problems.
NFB from FET source via a resistor and a film cap to it's gate; no problems, no sonic impact. Unless cap is electrolytic, cap quality is irrelevant.
A NFB loop exchanges gain / transconductance to linearity.
Increase the gain and/or transconductance inside the loop, and you've got a better loop, assuming you're operating under 100% degeneration.
For example; add a darlington pair of 2N2222 inside the CCS feedback loop, thus introducing voltage gain of like 15 000, thus increasing the FET transconductance to nearly infinite. Now you have a CCS that is behaving more like an ideal CCS.
Replace the gyrator plate load FET with a weak triode (weak in transconductance) like a 6SN7. You've got less transconductance inside the loop, you've got a weaker loop.
It behaves much much less like an ideal gyrator. Impedance provided to the tube underneath it much much less. More distortion.
Replace the 6SN7 with a strong triode, like a 6S45P with very high transconductance. Now you have more transconductance inside the loop, stronger loop. More impedance, less distortion. Better plate load.
Ok, so now replace the 6S45P with a pentode. More transconductance, more impedance. Less distortion.
Replace the poor FET-lookalike pentode with a real FET; you've got more transconductance and you get the rest.
CCS's can be made with triodes or pentodes. There's no point really, outside of for giggles.
The progression I described is what I personally went thru. Although I skipped the pentode and went straight for the FET. And I did try more different types of tubes. More transconductance inside the gyrator 100% NFB loop --> better sound.
SS is totally ok or completely crap depending on if it has voltage gain or not.
There is a feedback loop inside the voice coil of your headphones. The current traveling inside the coil is producing an electric field that is affecting the other parts of the coil, giving the coil IMPEDANCE.
Partial NFB (not 100%) over gain stages, worst case over multiple gain stages or worst possible case over multiple gain stages, multiple electrolytic caps, and a transformer ... You've got problems.
NFB from FET source via a resistor and a film cap to it's gate; no problems, no sonic impact. Unless cap is electrolytic, cap quality is irrelevant.
A NFB loop exchanges gain / transconductance to linearity.
Increase the gain and/or transconductance inside the loop, and you've got a better loop, assuming you're operating under 100% degeneration.
For example; add a darlington pair of 2N2222 inside the CCS feedback loop, thus introducing voltage gain of like 15 000, thus increasing the FET transconductance to nearly infinite. Now you have a CCS that is behaving more like an ideal CCS.
Replace the gyrator plate load FET with a weak triode (weak in transconductance) like a 6SN7. You've got less transconductance inside the loop, you've got a weaker loop.
It behaves much much less like an ideal gyrator. Impedance provided to the tube underneath it much much less. More distortion.
Replace the 6SN7 with a strong triode, like a 6S45P with very high transconductance. Now you have more transconductance inside the loop, stronger loop. More impedance, less distortion. Better plate load.
Ok, so now replace the 6S45P with a pentode. More transconductance, more impedance. Less distortion.
Replace the poor FET-lookalike pentode with a real FET; you've got more transconductance and you get the rest.
CCS's can be made with triodes or pentodes. There's no point really, outside of for giggles.
The progression I described is what I personally went thru. Although I skipped the pentode and went straight for the FET. And I did try more different types of tubes. More transconductance inside the gyrator 100% NFB loop --> better sound.
coinmaster
1000+ Head-Fier
Yeah, I'm beginning to understand that now between nelson pass's article and yours.
What about source followers? They have no gain but they do produce harmonics, plenty of data around to suggest that.
Nelson pass said the NFB decreases total harmonic distortion but increase high order harmonics which are the non-musical kind. Does that not apply to 100% degenerated local NFB?
Also it seems the above FFT analysis of my design was of an older version with a major imbalance in the output tubes. Spice refuses to load when I apply a solution.
Plus I ran the MK6 stock in simulation at a 1 ns timestep all night and it produced a way too perfect distortion reading.
I'm thinking there is some sort of optimization I need to make on both ends, need to look into it before I post MK6 stock vs impedance mod vs my design.
You may need to break down / interpret info to mainly worry about odd order harmonics ,
and also the level I see is already over minus 120db down so it is of questionable importance, unless I am not interpreting correctly
and also the level I see is already over minus 120db down so it is of questionable importance, unless I am not interpreting correctly
coinmaster
1000+ Head-Fier
Reading from pass's article 2nd and 3rd harmonics are plentiful in real life music which is why they are more acceptable and considered musical, even if they do have downsides.
The higher harmonics sound bad.
I'm not sure what the deal is with the -120db on the FTT response.
I'm trying to set up some THD addon to LTspice that does all the complicated work of THD analysis for you since apparently it's not as simple as clicking on FFT after simulation but it wants to run the simulation 3001 times for different frequencies. We'll be in the space age by then
Where's a super computer when you need it.
The higher harmonics sound bad.
I'm not sure what the deal is with the -120db on the FTT response.
I'm trying to set up some THD addon to LTspice that does all the complicated work of THD analysis for you since apparently it's not as simple as clicking on FFT after simulation but it wants to run the simulation 3001 times for different frequencies. We'll be in the space age by then
MrCurwen
100+ Head-Fier
- Joined
- Dec 26, 2015
- Posts
- 337
- Likes
- 133
It would seem he doesn't know how to use the program. 1nS timestep is useless, info provided on low levels not realistic. FFT is of current thru load, the scale is wonky and unreadable... I'm guessing the improved accuracy math setting is not on (it's off by default) as well.
Proper way to measure distortion is voltage over load. For 1kHz signal proper scale is 800 Hz to 10 kHz since 9 first harmonics are the ones of interest. At least in open loop circuits. The further you go up, the less realistic the simulation is anyway.
Proper Y axis scale is about top of 1kHz signal minus 100 dB, since that's the absolute bottom of any importance. I've found that as far as distortion goes, my ears (and my friend the high end hobbyist with golden ears and Sennheiser HD 800) tell me -80dB from signal peak is "not there anymore", and -70dB is "excellent".
Spice overestimates 2nd and 3rd by a varying degree. With tube models, usually a great degree. Some things that model as "distortion free" are not distortion free in real life. Simulation has it's uses and also it's problems.
With 1µS timestep and improved accuracy math on it takes a couple of minutes to do a 100mS sim of a complex circuit. And my computer is made of potatoes from 2004.
Spice has inbuilt THD measument, usable by one line;
.four 1000 9 V(n002)
Results in Spice Error Log Ctrl + L
I just simmed a source follower with unoptimized FETs (general purpose); CCS load under it, set for 15mA, +310V B+, a very very demanding 5k load and 100VPP signal.
Total Harmonic Distortion: 0.000389%
These numbers are for entertainment purposes really, but a source follower really has no THD. I've never heard it myself. If there is any, it is way down on list of distortion contributants.
MrCurwen
100+ Head-Fier
- Joined
- Dec 26, 2015
- Posts
- 337
- Likes
- 133
If somebody is still scratching their head thinking "an SS circuit module with no real THD and no high or odd order harmonics? How could that be, every single SS amp I've heard have had that tiring, shrill sound to them!"
There is not one single SS amp on the market that doesn't have global NFB (NFB over the whole amp, multiple gain stages). It is the gNFB that produces the "SS sound". There are some tube amp designs that have more than a little bit of the "SS sound"; they utilize very high gNFB.
It's gNFB sound, not SS sound.
Even if you have an SS amp with source followers or other followers as the output, the low / no distortion output doesn't negate the nastiness made in the preceding sections. They will have gNFB if they are SS based.
So once again, it's about topology and circuit design. It pretty much always is.
There is not one single SS amp on the market that doesn't have global NFB (NFB over the whole amp, multiple gain stages). It is the gNFB that produces the "SS sound". There are some tube amp designs that have more than a little bit of the "SS sound"; they utilize very high gNFB.
It's gNFB sound, not SS sound.
Even if you have an SS amp with source followers or other followers as the output, the low / no distortion output doesn't negate the nastiness made in the preceding sections. They will have gNFB if they are SS based.
So once again, it's about topology and circuit design. It pretty much always is.
Users who are viewing this thread
