I'm curious what exactly would make Maouna say the Krypton-C sounds more tube like than the CFA-X? Does this mean that the distortion profile has a higher 2nd harmonic distortion product, this is what I normally would expect from a vacuum tube design. So the question becomes does the CFA-X actually have lower harmonic distortion in general or just a very different harmonic distortion ratio of 2nd to higher 3rd and others multiples?
JBL UREI 6150 - TUBE LIKE
Hi Kindhornman,
he is not the first to say about it from a CFA.
" One could say the UREI is remarkably unremarkable - a supreme complement for any amplifier – no aspect of its sound jumps out to say, ‘Hey, look at me’, and nothing is missing. The sound is neutral and absolutely NATURAL. They say it’s solid-state (transistorized) but I'd swear there must be a tube or two in there somewhere as it's capable of sounding more tube-like than some tube-amps I’ve known. But the UREI’s greatest achievement is in the area where most other amps fall flat on their back-sides – in the lower mid-range. It facilitates the reproduction of all the detail in this area that those others don’t even acknowledge – the low-level sounds, the fade of notes, the weight and resonance of instruments – it allows a system to supply that elusive LOWER MID-RANGE BODY in spades. That, along with its other attributes, puts it in a class above the rest, as far as I’m concerned."
source : poweramp
Hi Kindhornman,
he is not the first to say about it from a CFA.
" One could say the UREI is remarkably unremarkable - a supreme complement for any amplifier – no aspect of its sound jumps out to say, ‘Hey, look at me’, and nothing is missing. The sound is neutral and absolutely NATURAL. They say it’s solid-state (transistorized) but I'd swear there must be a tube or two in there somewhere as it's capable of sounding more tube-like than some tube-amps I’ve known. But the UREI’s greatest achievement is in the area where most other amps fall flat on their back-sides – in the lower mid-range. It facilitates the reproduction of all the detail in this area that those others don’t even acknowledge – the low-level sounds, the fade of notes, the weight and resonance of instruments – it allows a system to supply that elusive LOWER MID-RANGE BODY in spades. That, along with its other attributes, puts it in a class above the rest, as far as I’m concerned."
source : poweramp
I find that comment about the aging UREI 6150 interesting, especially since that amplifier has a TL084 op amp at it's core. I am a big fan of the UREI 6250. I have a rack full of them in my home theater room!
The schematic for the UREI 6150 is here: http://www.jblproservice.com/pdf/Vintage%20JBL-UREI%20Electronics/UREI-6150.pdf
The schematic for the UREI 6150 is here: http://www.jblproservice.com/pdf/Vintage%20JBL-UREI%20Electronics/UREI-6150.pdf
I came across the strange interaction of the super-pair and a CFA stage by accident.
I was going to design a "APEX" like VFA with the baxandall.
I noticed the "current on demand" aspect of a CFA running the Baxandall gave a
MUCH more rounded overload plot - so I kept it.
Kypton-C has dominant second before clip and 2'nd/4'th even during heavy clip.
(much less 3rd/5th/7th)
A "normal" AB amp has 2-7th during overload. C actually simulates differently.
OS
So, you already forgot that I showed that soft clipping super pair behavior before.
Lazy Cat,
I take it from that response that this was also your own conclusion. Keeping things simple usually has the unintended result of a superior product to one with all the bells and whistles added to solve problems that may not really be a problem. KISS, the practice I use when doing my industrial designs, it is easy to overkill elements and then you have a mess.
Thanks OS,
When I can get my head above water, doing a new project with new people, I can finally get back to building a few of the input sections on some of Jason's boards. I can't wait to see if I can hear the differences between the CFA-X and the Krypton-C.
I take it from that response that this was also your own conclusion. Keeping things simple usually has the unintended result of a superior product to one with all the bells and whistles added to solve problems that may not really be a problem. KISS, the practice I use when doing my industrial designs, it is easy to overkill elements and then you have a mess.
Thanks OS,
When I can get my head above water, doing a new project with new people, I can finally get back to building a few of the input sections on some of Jason's boards. I can't wait to see if I can hear the differences between the CFA-X and the Krypton-C.
Yes , I know ... "you da man"
... I was always "scared" of the Baxandall.I was even more apprehensive of combining mirrors/servo to the "VSSA holy
grail" , but it worked out in the end.
Personally , the kypton-C will be the first CFA I will etch on my real amps.
It has "pretty " layout that will look good with my violet/UV (390nm) led's.
PS - I have not heard a CFA yet
.Edit - the real reason for the CFA-Baxandall "interaction" is the lack of saturation affecting
the collectors of the input pair. On Hawksford/ plain VAS's = IP Ic is affected on overload.
OS
Last edited:
The CFA secret is based, on my opinion, on a good balance between speed and stability. Any add of "complexity" *in the loop*, in order to reduce further distortion numbers will be paid cash by a reduction in gain/stability margin and speed. And, yes, lose of details in the musical landscape.
Only careful listening can tell if this reduction of distortion is worth the try and don't kill a major virtue else where.
I believe, on my side, that CFA need a very special approach, where simplicity and optimization are the key. Each stage has to be evaluated / modified / complicated only with careful listening. Step by step. No way to achieve the best results working only on paper or simulators with CFA, on my experience. Reason why i follow this thread with curiosity and a bit of skepticism on the approach.
Waiting for real and serious listening comparisons between those hundreds of proposals.
UREI PHILOSOPHY
Maybe it helps:
JBL Technical Notes
search for
"Vol.2, No.2 - JBL-UREI Power Amplifier Design Philosophy"
at the bottom of the list.
Maybe it helps:
JBL Technical Notes
search for
"Vol.2, No.2 - JBL-UREI Power Amplifier Design Philosophy"
at the bottom of the list.
Christophe,
I agree that you would have to fine tune any of these amplifiers that OS has designed. You can see by the fact that people have used different devices that each has had to tweak their particular combination to get it right. At the same time the basic designs that OS has come up with have all functioned correctly after what has seemed to be minor adjustments. I can't wait to hear one of these CFA's, I don't know if I have ever heard a current feedback amplifier, at least I am unaware of it.
I agree that you would have to fine tune any of these amplifiers that OS has designed. You can see by the fact that people have used different devices that each has had to tweak their particular combination to get it right. At the same time the basic designs that OS has come up with have all functioned correctly after what has seemed to be minor adjustments. I can't wait to hear one of these CFA's, I don't know if I have ever heard a current feedback amplifier, at least I am unaware of it.
BJT Output stage (schematic and layout files) is at posts #500 , 502 , 504
There is a lateral output stage but i dont know if its final ,tested and working.Some interesing comments noted about it :
'' P-channel laterals (2SJ162 in your case ~ 900pF) have got significantly higher input capacitance, than N-channel ones (2SK1058 ~ 600pF), resulting in rather unbalanced OPS at high frequencies.
Try adding a 330pF cap between the gate and the source of 2SK1058 as shown on the picture - this will make the whole thing much more balanced and stable.
Also, slight increased stoppers may help (I would use 470 ohm for R107, R108), however the cap is more important.
Another way to compensate for the different capacitances of the outputs is to use different gate stopper values between the N and P channel devices. Choose the values so the time constant if the same for both.
Let's say the 2SJ162 has a gate stopper of 100Ω and a gate capacitance of 900pF, the RC time constant is just RgatexCgate=RC time constant, so 100x900x10^-12=90nS. If we want 'balance' we need a resistance on the gate of the 2SK1058 that gives the same time constant. So, Rgatex600pF=90nS, solve for Rgate. In this case Rgate needs to be 150Ω.
This could be a valid experiment, remove the capacitor and increase R107 to 150Ω and see if the effect is the same. I'd even suggest since it appears sensitive to set R107 to 470Ω and R108 to 330Ω, same equalization of the time constant, just with higher values overall and see if you can get it to function stably without the added capacitor.
For the VBE, try R104 as 2K4 or even 2K2 if needed and see if you are closer to the sweet spot. ''
Attachments
That was not what i tried to say. I'm pretty sure that all the schematics of OS will work. Apart some minor adjustments of compensations, due to parasitic caps etc. of the builds.
It is a more general consideration. In his minimal topology, the CFA count 3 poles: Input stage, VAS, output stage. Every attempt to add a cascode, a driver, a super pair etc. inside the loop will add a pole.
You can reduce harmonic distortion, this way, but you have to pay-it cash with the stability margin, slew rate, linearity on square waves etc. And add a compensation to keep-it stable, reducing overall speed (open loop bandwidth).
Now, the question is: did this reduction of distortion was worth the try ? Or did the distortion level was low enough previously to be under the audibility threshold ? In other words, had-we improved something witch don't needed-it, reducing performance somewhere else where it is more noticeable, like dynamic behavior ?
If we compare the both topologies to cars, i should say a VFA is a limousine, and a CFA a race car. Is-it clever to tweak a race car in order to bring the same comfort than in a limousine, when we pay-it with performance's reduction ?
Both typologies need a very different approach in order to get the best of them. We cannot figure-it on paper or simulation.
CFA is not neutral. It is "expansive", while VFA is "compressive". The builders of the Dadod's gain wire had noticed a clear difference between the two topologies: No feedback and CFA that are switchable in his design.
Well. I pretend CFA over emphasis details and peaks, and that it compensates the losses that occurs in the rest of the recording/reproducing channel.
We have to chose our evils, and because all of this is a question of "balance", the main tool we have to refer during the design process is our ears. Simulation is just a calculator.
Last edited:
I tried simple CFA with single transistor VAS and Hawksford VAS. I like the sound of simple transistor VAS. May be THD of single transistor VAS is higher (I can not confirm, it only from simulator. I have no THD analyzer).
After I finished my Krypton-V variant, I will try Kripton-C variant.
Surprise...!
The attached FFT image could mean nothing to others because it has only subjective/comparative value. But I intentionally present all my simulation FFT in a manner where I can see consistent differences between circuits.
And what did I see in this FFT? Cool... I don't understand this kind of VAS. It's too good to be true, or it hides its "ugly face" somewhere. I will look deeper into it later.
It apparently is not the 2nd harmonic domination like commonly found with tube amps. So, this "tube sound" perception, does it sound "rounded"??
Attachments
SLEWMONSTER LF
While you all are waiting to hear the subtleties of the CFA "holy grail" , I want to see/hear
the "grunt" factor of these amps. Are they durable ?? we'll see .
At 60V , a slewmonster with 3 pair MT-200 (below 1) will have almost 25A SOA. And the hungry subwoofer is the true test (below 2)
I'll need to dab some silicone/or contact cement -on all screws/some caps , The LF
microphonics on this amp will be intense. Luckily , I'm ported ... (26hz Fc) with a
100cm port - air will blow on the amp .
I did not choose a CFA for this - just a non-cascoded "wolverine" with custom
comp. and NO cap multipliers. I want the WHOLE 56-0-56V rail. (EvanC's trafo).
OS
While you all are waiting to hear the subtleties of the CFA "holy grail" , I want to see/hear
the "grunt" factor of these amps. Are they durable ?? we'll see
.At 60V , a slewmonster with 3 pair MT-200 (below 1) will have almost 25A SOA. And the hungry subwoofer is the true test (below 2)
I'll need to dab some silicone/or contact cement -on all screws/some caps , The LF
microphonics on this amp will be intense. Luckily , I'm ported ... (26hz Fc) with a
100cm port - air will blow on the amp
.I did not choose a CFA for this - just a non-cascoded "wolverine" with custom
comp. and NO cap multipliers. I want the WHOLE 56-0-56V rail. (EvanC's trafo).
OS
Attachments
For the BJT Slewmaster, the VAS relies heavily on the 33p compensation cap. With mosfet output the VAS behaves much better. No compensation needed (I haven't checked PM/GM, which should be fine with this topology, but the pulse test was smooth) and 2 feedback paths can be separated. And this is by driving the latfet directly from VAS. So no need VBE and extra driver. Haven't checked THD20 but this could be a winner. So now I have 2 topologies competing for the best (or this one will be special for high power version)
It has second (and third) , of course. You just can't see it.
But I CAN (below) - there is the "ugly face" (at 150db down).
Drop that noise floor - there it is. But it is Loooooow ! "C" is the champ.
PS - it also keeps that same H2/3 ratio after it clips - sort of "tubey".
let's see the Valve do 1/2ppm. Ha !
OS
