Linear Audio had an article in volume 4, "The Terrible Twins", which i found interesting mostly becuse he tested many different devices for performance. THe ones that consistently seemed to perform best where those that had larger current ratings, large die sizes, and higher Yfs. It was about the SJEPr100 from SS, but oddly enough, he did not provide the numbers for it. I also recently saw another interesteing post by RNMarsh, showing results concerning matching and what was the important factor for THD reduction.
http://www.diyaudio.com/forums/solid-state/240712-cfa-topology-audio-amplifiers-13.html#post3596502
http://www.diyaudio.com/forums/solid-state/240712-cfa-topology-audio-amplifiers-13.html#post3596502
i have this one at home actually but forgot about that article will look it up thanks.
The point about matching is a good one and doesn't just stop at the parts - the whole topology contributes. I measured the distortion at one end of the op of this amp against ground and it looked like a dogs dinner. But it was the same at the other end vs ground too!
So when measured across the load, all of the crud dissapears leaving just a clean signal and very little of anything else!
Call me naive certainly but looks like magic 😀
makes sense I guess - both halves are absolutely identical and being driven exactly identically (out of phase but same) so we don't need to go to superhuman efforts to linearize each half - its enough to make sure that they're as identical as we can make it. Its magic man i tell ya
The point about matching is a good one and doesn't just stop at the parts - the whole topology contributes. I measured the distortion at one end of the op of this amp against ground and it looked like a dogs dinner. But it was the same at the other end vs ground too!
So when measured across the load, all of the crud dissapears leaving just a clean signal and very little of anything else!
Call me naive certainly but looks like magic 😀
makes sense I guess - both halves are absolutely identical and being driven exactly identically (out of phase but same) so we don't need to go to superhuman efforts to linearize each half - its enough to make sure that they're as identical as we can make it. Its magic man i tell ya
Last edited:
I played around with 15000uF caps across Rs on an F5. My objective wasn't so much to increase power (transconductance) but to minimize degeneration which greatly reduces H3 (3rd harmonic).
What does CFA topology mean? I looked at the thread, but it is not spelled out enough for me to understand.
Rush
Current feedback amps
Nelson's F5, bonsai's ovation nx, lazycats vssa amps and many others I have missed.
Feedback goes to source or emitter instead of gate/base. Don't know much more about them but reading the posts seems clear that there isn't a simple definition.
I should take this to the other thread, but in the F5 Manual, Nelson Pass says, at least in the case of the F5, "Low impedance feedback has been (incorrectly) referred to as "current feedback...."
I don't know enough to argue with anyone, so I will continue to be ignorant.
Rush
Buzz: Some comments on the PCB.
Crank up the trace widths of the high current paths (power and output). Don't forget that you have a 2-layer PCB. You can route high currents on either or both layers.
Are the output FETs were positioned compatible with the F4 and BA PCB (ie. the UMS)?
Now about pads on power, gnd, outputs that have holes large enough for a #4 screw? This allows connections without soldering.
What is the connector used for inputs? Does it have .1" pin spacing?
The main traces are double sided. I cannot go any larger with the pre-defined trace widhts. I have consider the screw option for the power holes, but i like the tab connectors. If Kasey wants the holes. ill happily change. Very strong and easy to disconnect. The fets are according to universal mounting spec. I do believe it is .1" headers, but illl have to check.
Can't you have both #4 screw holes and tab connectors?
BTW: I think I could do a layout for 3 output MOSFETs per rail and UMS. There is also a possibility for 2 drive FETs and one cascode FET per rail. EUVL did a cascoded circlotron design. I played around with Spice sims some time ago of a different cascoded circlotron which got only a factor of about 2 reduction in harmonics. I need to revisit that simulation with EUVLs design.
In my F6CC design, I was seeing much larger improvement with cascoding. I am not sure if the difference is due to push-pull vs circlotron topology or something else.
I chose the spacing for better heat distribution. IT will get hot if biased at all like other Pass based amps. The dual fet AJ board REALLY pushes the big 5U chassis based on another members experience. The boards can be as big as needed, but Kasey only suggested 2 pairs. If the amp performs at all as expected, the cascode may be unecessary. Its up to Kasey. I was going to send the boards to the maker tonight, so if changes are wanted, I need to know.
THey can be had on Ebay from Alweit. IN truth, I would think part choice is not a huge thing in this instance. THe IRF9610would probably perform very similarly. I have built a lot of amps and have only noticed a significant difference in sound from two part subsititutions, the SS fets and laterals from Renasas. I am not saying I have heard them all and there is no difference, but I am talking about changing the sound difference. The topology usually defines the sound to the greatest degree.
Does the IRF9610 have the frequency dependent transconductance like the IRF9640? Is the FQP3P20 a reasonable alternative?
Last edited:
- For the input fets: IRF9Z24 did well as a substitute for the 2SJ313, but as buzz points out pretty much anything To-220 will work to some extent. I liked these two compared to others, but YMMV. These seemed very linear at the operating points i needed for this amp.
- Layout: I think lets just nail this down as is - buzz can add some holes next to the Tab connectors if theres space/time. Even if not, it won't be too hard to just drill it myself if i wanted to go down that route.
- Cascoding - lhquam makes some good points - it may be worthwhile trying if you want to improve distortion even more. I'm focussed on the practical side of building, understanding PSU sentsitivities (should be none), heat and listening but if you're minded to try the cascode design - why not 😀
My initial impressions are that the bass 'slam' and realistic thwacks (as opposed to a thud with treble) gained from running the amp at higher voltages (36 volts up) are pretty nice.
Buzz all your's man....
- Layout: I think lets just nail this down as is - buzz can add some holes next to the Tab connectors if theres space/time. Even if not, it won't be too hard to just drill it myself if i wanted to go down that route.
- Cascoding - lhquam makes some good points - it may be worthwhile trying if you want to improve distortion even more. I'm focussed on the practical side of building, understanding PSU sentsitivities (should be none), heat and listening but if you're minded to try the cascode design - why not 😀
My initial impressions are that the bass 'slam' and realistic thwacks (as opposed to a thud with treble) gained from running the amp at higher voltages (36 volts up) are pretty nice.
Buzz all your's man....
I might have missed a discussion about this question. In a circlotron amplifier, how significant are the power transformer interwinding capacitances and possible 60Hz (and 120Hz ...) leakage currents to the resulting amplifier output, as well as the capacitive load on the output.
capacitive load on output was tested in post # 30 http://www.diyaudio.com/forums/pass...regation-pass-inspired-ideas.html#post3589317
given the above, i expect that the transformer interwinding capacitances effects (to the extent that they exist) will not be discernible. But we'll verify when we test the sucker.
I will be subjecting the first build to a 'worst-case' PSU with ALL the windings for both channels on one big fat toroid 😀
So no prettied-up results from this one .... honest research is the best kind there is seyz feynman. I'll do my best to present everything i find - warts n all.
given the above, i expect that the transformer interwinding capacitances effects (to the extent that they exist) will not be discernible. But we'll verify when we test the sucker.
I will be subjecting the first build to a 'worst-case' PSU with ALL the windings for both channels on one big fat toroid 😀
So no prettied-up results from this one .... honest research is the best kind there is seyz feynman. I'll do my best to present everything i find - warts n all.
- Status
- Not open for further replies.