Antoinel, I would recommend to read this:
https://www.passdiy.com/project/amplifiers/zen-variations-2
https://www.passdiy.com/project/amplifiers/zen-variations-2
I choose Constant instead of Active current source
Thanks pr for the reference above which I will look at. I hope my following understanding is not controversial; because it is not my intent to stir controversy.
The stated objective for the power output stage of lhquam's model schematic is to have an asymmetric property/performance. The bottom two graphs of Fig. 3A [for CCS] and Fig. 3B [for ACS], and lines 42-50 in column 2 in the formal write up of US 5,710,522 help clarify my choice.
1. The two bottom graphs of Fig. 3B on the right are mirror images. A mirror is a symmetry operator. These two graphs say that current is pushed in/through Q1, and a simultaneous equal current is pulled in/through Q2, and so on back and forth. The SE amp of Fig. 3B operates in a mode like or is Push-Pull as also explained by the language in the paragraph of the patent. The operation of this amp is symmetric which is contra to the objective of lhquam; as one further sees its build of a balanced and complementary SE antiphase.... The schematic [e.g. diyF4 output stage] of his proposed model amp already has symmetric push-pull components.
2. The two bottom graphs of Fig. 3A on the left are not mirror images. Thus, the amp of Fig. 3A is relatively less symmetric than that of Fig. 3B. It follows that its balanced and complementary SE antiphase model amp of lhquam is more asymmetric than the aforementioned in point 1.
3. The choice was clear. Go with the asymmetric build utilizing CCS instead of that of the symmetric ACS.
4. Asymmetric operation of the balanced /complementary SE amp utilizing CCS generates inherent distortion which is then managed/controlled to good use.
5,Symmetric operation of the balanced/complementary SE amp utilizing ACS generates less inherent distortion as the patent teaches. I do not have enough distortion to work with [it is trivial], and so how am going to manage/control it towards good use?.
Best regards
Thanks pr for the reference above which I will look at. I hope my following understanding is not controversial; because it is not my intent to stir controversy.
The stated objective for the power output stage of lhquam's model schematic is to have an asymmetric property/performance. The bottom two graphs of Fig. 3A [for CCS] and Fig. 3B [for ACS], and lines 42-50 in column 2 in the formal write up of US 5,710,522 help clarify my choice.
1. The two bottom graphs of Fig. 3B on the right are mirror images. A mirror is a symmetry operator. These two graphs say that current is pushed in/through Q1, and a simultaneous equal current is pulled in/through Q2, and so on back and forth. The SE amp of Fig. 3B operates in a mode like or is Push-Pull as also explained by the language in the paragraph of the patent. The operation of this amp is symmetric which is contra to the objective of lhquam; as one further sees its build of a balanced and complementary SE antiphase.... The schematic [e.g. diyF4 output stage] of his proposed model amp already has symmetric push-pull components.
2. The two bottom graphs of Fig. 3A on the left are not mirror images. Thus, the amp of Fig. 3A is relatively less symmetric than that of Fig. 3B. It follows that its balanced and complementary SE antiphase model amp of lhquam is more asymmetric than the aforementioned in point 1.
3. The choice was clear. Go with the asymmetric build utilizing CCS instead of that of the symmetric ACS.
4. Asymmetric operation of the balanced /complementary SE amp utilizing CCS generates inherent distortion which is then managed/controlled to good use.
5,Symmetric operation of the balanced/complementary SE amp utilizing ACS generates less inherent distortion as the patent teaches. I do not have enough distortion to work with [it is trivial], and so how am going to manage/control it towards good use?.
Best regards
Last edited:
Please remember, that the ACS has an imbalance compared to the PP stage and hence creates second order harmonics. Reason for the imbalance: The little npn transistor that controls the MOSFET in the ACS controls to some degree the behaviour of the ACS. In contrast the PP MOSFETs act alone as source followers. As a result there is an imbalance in the transfer functions of the ACS and the PP stage.
So for me the question is still open:
Why no Aleph current source in the XA.8?
The original XA's (no . ) use the Aleph current sources, but ultimately we
decided that a push-pull output stage in parallel with a CCS was more
appropriate to customer's needs in terms of driving low impedances.
The configuration we chose was similar to a balanced Aleph 0. A push-pull
stage in parallel with an Aleph current source did not bring much to the
table.
😎
nelson, out of curiosity - did you end up with followers for the output stage in the XAs vs the common source config in the Alephs ?
Thank you very much Nelson for your kind reply.
Again I made some thoughts about it and collected them in a pdf file.
Some simulation files mentioned in the pdf file are also appended.
lhquam, please read the file. I found out some things that show, that
it is very unprobable that the former circuit idea will give the best
results. Instead, a different idea with solution is shown.
With this information I regard the topic "output stage" as solved.
I will now make my thoughts about the input stage and the power supply.
By the way: Is there a prize available for the person who
at first makes the correct guess about the cuircuitry of the
"local cascode feedback" LCF like many years ago for the
"Magic MacMillan Resistors"?😀😀😀
Again I made some thoughts about it and collected them in a pdf file.
Some simulation files mentioned in the pdf file are also appended.
lhquam, please read the file. I found out some things that show, that
it is very unprobable that the former circuit idea will give the best
results. Instead, a different idea with solution is shown.
With this information I regard the topic "output stage" as solved.
I will now make my thoughts about the input stage and the power supply.
By the way: Is there a prize available for the person who
at first makes the correct guess about the cuircuitry of the
"local cascode feedback" LCF like many years ago for the
"Magic MacMillan Resistors"?😀😀😀
The original XA's had the Aleph CS. Subsequent XA's used current sources
for SE bias, initially as simple as resistors off to the supplies, and later
constant current sources.
"Currently" X and XA and Xs product use constant current sources.
If you are referring to "Mu followers", no X amp has employed them,
although FW has on several pieces.
😎
Hello pr. I admire and commend you for the valuable, and generous content in your post above.
Best regards.
Best regards.
I have briefly examined the analysis in your pdf file and it looks interesting. I have done some similar simulations only to find that the harmonics change considerably due to many cancellations when the front end is added and feedback is introduced. The mathematical analysis appears to be intractable, but after simulations of many feedback schemes, I have a design for the XA30.8 that looks plausible. I am in the process to populating my custom PCBs, and expect to start bench testing very soon.
I will post photos and results in the near future.
Hello pr. Please show an enlarged schematic of the XAleph...MitGeggen...front end for further study by myself using LTspice.
Question: What is the harmonic structure and the resultant numerical values [if any] for the front end's outputs at max power output? My interest is to separate the contribution of the front end from that of the power output stages
Thank you and best regards
Question: What is the harmonic structure and the resultant numerical values [if any] for the front end's outputs at max power output? My interest is to separate the contribution of the front end from that of the power output stages
Thank you and best regards
Antoinel:
Once global feedback is introduced it is very difficult to separate the distortion spectrum contributions of the front-end from those of the output-stage. I have spent many days analyzing the problem both mathematically and via simulations. Perhaps someone has discovered a way to solve the problem, but so far it has eluded me. Roughly speaking, negative feedback compares the output signal with the input and adds the difference signal into the amplification chain in an attempt to force the output (divided by the closed-loop-gain factor) to agree with the input. Without trickery, this results in some degree of cancellation of both the even-order and odd-order harmonics. Thus, even if the output stage by itself generates a lot of K2, negative feedback through a "harmonically neutral" front-end will remove much of that K2.
The challenge is to create the desired end-to-end harmonic structure and also have some degree of negative feedback to increase the damping factor (lower the output impedance) in order to handle difficult speaker loads. Nelson has apparently found a good way to do this.
Once global feedback is introduced it is very difficult to separate the distortion spectrum contributions of the front-end from those of the output-stage. I have spent many days analyzing the problem both mathematically and via simulations. Perhaps someone has discovered a way to solve the problem, but so far it has eluded me. Roughly speaking, negative feedback compares the output signal with the input and adds the difference signal into the amplification chain in an attempt to force the output (divided by the closed-loop-gain factor) to agree with the input. Without trickery, this results in some degree of cancellation of both the even-order and odd-order harmonics. Thus, even if the output stage by itself generates a lot of K2, negative feedback through a "harmonically neutral" front-end will remove much of that K2.
The challenge is to create the desired end-to-end harmonic structure and also have some degree of negative feedback to increase the damping factor (lower the output impedance) in order to handle difficult speaker loads. Nelson has apparently found a good way to do this.
It is probably more complicated than you think.
When I look at the "simplified" conceptual schematic I have 4 complementary
Jfet input devices, 4 complementary bipolar cascodes, 4 complementary
voltage gain Mosfets and 4 complementary Mosfet output followers (in real
life as many as 80 followers) and these are just the actual signal path devices,
not counting current sources and voltage references.
There are potentially 37 resistors that affect the gain and feedback (not
counting output Source resistance, input resistors and Gate stoppers), 8 are
degenerative, 8 are Drain loading, 4 are main loops, 1 is both loop and
degeneration and 16 are cascode feedback.
This allows a lot of tweaking, best done over a long period of time, listening
with one eye on the scope.
😎
When I look at the "simplified" conceptual schematic I have 4 complementary
Jfet input devices, 4 complementary bipolar cascodes, 4 complementary
voltage gain Mosfets and 4 complementary Mosfet output followers (in real
life as many as 80 followers) and these are just the actual signal path devices,
not counting current sources and voltage references.
There are potentially 37 resistors that affect the gain and feedback (not
counting output Source resistance, input resistors and Gate stoppers), 8 are
degenerative, 8 are Drain loading, 4 are main loops, 1 is both loop and
degeneration and 16 are cascode feedback.
This allows a lot of tweaking, best done over a long period of time, listening
with one eye on the scope.
😎
I assume the resistor counts are for a single channel. Hmmm. 😱 The cascode feedback network is apparently much more complicated than I expected. It is a good thing that my PCB layout makes it easy to play around with that, but it might not have quite enough flexibility for what Nelson has come up with.
By "one eye of the scope" do you mean your AP distortion analyzer?
Last edited:
now I know better why it lasts about seven years to develop a new set of amps……
🙂
🙂
Last edited:
how can four points get so much connections….?
o.k. four resistors to every cascode point that is manageable……..
one to north-- one to south-- one to west-- and one to east--
🙂🙂🙂
o.k. four resistors to every cascode point that is manageable……..
one to north-- one to south-- one to west-- and one to east--
🙂🙂🙂
I am wondering about 4 resistors for the main feedback loops, not counting the "1 is for both loop and degeneration". Does the main feedback include both the output stage and the front-end outputs, is the topology of the feedback from the output stage rather different than "usual"?
Thank you lhquam for your post. It is clear, valuable in content, and well articulated. I was able to magnify the front end of pr's schematic in LT for further [curious] analysis. I did not find the node used by pr to inject overall feedback in schematic; but will look more closely. As always, I look forward to your additional input...and to your prototype.
- Status
- Not open for further replies.