| stefanobilliani |
I haven't tried it . And most probably some values aren't right.
Mr. Pass , is it interesting or is best the way of a power front end like Stephano-x-soz file? |
|
|
| Ian Macmillan |
Stephano, your circuit will work just fine. In fact you have just (re)designed a pretty much standard X type amplifier such as that posted by Petter some while ago. However, this is missing the point somewhat in terms of an X-Zen. The latter is all about simplicity and using a single gain stage if at all possible. Remember the Koan: "the sound of one transistor clapping".
BTW, most would consider your XIdea a 2 stage amplifier.
Ian. |
|
|
| stefanobilliani |
| Then , other biasing ideas using it as a Power front end (whith folded cascode)? |
|
|
| stefanobilliani |
Of course , I feel ready to build my first Zen amp and ear how
one transistor clapping sounds .( instead of 2 )
As soon as possible:D |
|
|
| Nelson Pass |
If you want to go back to direct coupling of the
feedback, you can reference R10 and R11 to a
negative DC voltage. |
|
|
| Henrik |
Nelson, I dont have to think twice!
Thanks again. |
|
|
| Henrik |
I just tryied it out.
It works fine, besides from one thing, the dc-potential between R1 and R14 varies with the outputimpedance (simulated with R3) in the linestage. Is the solution to this problem a DCservo or could there be a more simple way to solve this? |
|
|
| Henrik |
| Sorry, I am too tierd I think, but here is the scematic to my former post. |
|
|
| Nelson Pass |
At this point I definitely would think about replacing the
1 and 8 ohm resistors on the negative half with an
active current source and a -10 volt rail. Less dissipation,
more gain, better input cmrr.
Also, put those zeners on the outside of the gate resistors
or I might be hearing you on my am radio :) |
|
|
| Henrik |
I gess you are wright on this one as well!
I will give it try with the active currentsouce of yours.
Are you listening to radio all day, checking zen-ers?
If you are, you wounīt hear mine, it was a sloopy scematic, sorry, the zeners is sitting on the outside of the gateresistors on my SOZ.
I will go on vacation with my wife for a month or so. You will hear from me again, hopefully not on the radio. |
|
|
| Henrik |
I have bin given these questions som thoughts.
Isnīt the active currentsource (like the Aleph-one) only useable at the drains?
I have tried simulating a normal currentsource at the tail, and i still have the same problem with DC offset at the input varying with the outputimpedance at the linestage when feedbackloop is DC-coupled. |
|
|
| stefanobilliani |
Well : I have tried the version posted by Ian.
Nelson said : The paradox is that feedback is best applied around circuits that need it the least.
I guess this is the case.
When I read the SOZ article for the first time I said : this is for me!
Now I try it with X and if it is possible , I am more impressed too.
What SOUND.
Ian : The arrangement I posted wasnt so good performing. Not like this One !
Nelson : thank you for the patience . I am anxious to try the current source .
What current value is preferable?
H Potter when finished your 10W Soz : add 2 resistor and 2 caps and try it .
Stefano. |
|
|
| Ian Macmillan |
Stephano, glad you like the sound of my simple X modified SOZ! Simple is often best from a sonic standpoint if not from an efficiency one. It is a fairly simple matter to drop in a current source for the diff pair 'tail' instead of the resistors. As for the current, it depends on the desired output power and associated power supply voltage. If you already have a working circuit, simply measure the voltage drop across the drain resistors and dive by the value of the resistor. Double this to give the required tail current.
Henrik, I am not sure that Nelson's reference to an active current source was meant to be interpreted as the Aleph current source, though of course I could be wrong. I suspect he simply meant a transistor (MOSFET) one rather than a resistor. It is not at all clear to me how one could take advantage of the Aleph current source in this application. Of course, Nelson could have something else up his sleeve...
Ian. |
|
|
| HarryHaller |
Uh Oh...... Poor Mr. Pass now has two unofficial spokesman. Can a career in politics be far behind?
H.H. |
|
|
| GRollins |
A normal (i.e. non-varying) current source is called for beneath the Sources of the output devices. Been there, done that...see SOZ w/current sources thread.
That's not to say that one couldn't use Aleph current sources <i>above</i> the output devices...
Grey |
|
|
| Henrik |
Ian
I have simulated your X-SOSAlt, and i found about -2,5v at the inputterminals with an outputimpedance at a linestage of 100K. This is because, R13, R11, R14 and R4 splits the dc-voltagediff ground/-Vcc/output and leave the -2,5v at the inputterminals. And since the outputimpedance at the linestage is a part of this voltagesplitting it will influence on the given DC at the inputterminals.
To get rid of this DC is my mainocupaton at the moment, you can get around this problem by adding a coupleing cap in the feedbackloop or at the input, but it would be prefereable to make it dc-coupled if possible.
Stefano
You did you use som caps, was it in the input or in the feedbackloop?
Grollins
Nelsons suggestion of using an active currentsource at the tail was (in my oppinion) aimd at solving the DC-problem i had, but i canīt see how it can. It should be possible to use alephcurrentsources at the drains, but that was not what Nelson was talking about. I will go back to read your post on SOZ and currentsources anyway. |
|
|
| gromanswe |
| quote: | Originally posted by Variac
So you agree that a true Zen doesn't have active current sources.
Despite Nelson steering us toward the active design, I am also really curious about the passive (?) approach. |
For ultimate teoretical performance
I would chose a passive.
I have tried it in one of my
most precious designs.
It uses passive bias in a way I have not seen yet.
You can not hear any difference,
but the performance is there.
Would probably not be accepted by "the market".
They are somewhat behind research and ideas
all the time
Thanks for reminding me, Variac |
|
|
| gromanswe |
If any man who treats others
with RESPECT
want to buy my idea
I am OPEN
gro |
|
|
| stefanobilliani |
Henrik , I put capacitors between R15/R5 and R14/R4
Stefano |
|
|
| Nelson Pass |
Sorry, I was in Bermuda.
The rationale for active current sources is mostly efficiency.
With active current sources such as the Penultimate Zen
and upcoming SOZ variations, we can start approaching 50%
efficiency instead of 5%.
That aside, I also prefer passives (even the occasional
light bulb). :) |
|
|
| JasonL |
| what is the power out put of this x-soz |
|
|
| gromanswe |
| quote: | Originally posted by Nelson Pass
Sorry, I was in Bermuda.
The rationale for active current sources is mostly efficiency.
With active current sources such as the Penultimate Zen
and upcoming SOZ variations, we can start approaching 50%
efficiency instead of 5%.
That aside, I also prefer passives (even the occasional
light bulb). :) |
But light bulbs can IandI use in HIFI
equipment
:confused:
fotenote:
only GO(O)D KNOWWWWWWSSSSSS |
|
|
| bob12345678 |
I am building a small SOZ with an active current source and 8 ohm
load resistors with 12 volt rails. Is there any way to calculate the
damping factor in paticular will the damping factor be different
with a active current sourse instead of using the source resistors.
Thanks Bob12345678 AKA Woody |
|
|
| stefanobilliani |
| The damping factor of the amplifier will change changing R1 and R2 values. Or adding feedback. |
|
|
| bob12345678 |
Thanks for the reply but the question still stands how can this
damping factor be calculated.
Regards, bob12345678 AKA Woody |
|
|
| Nelson Pass |
The output load divided by the sum of the Drain
resistors is the damping factor, which in this case
is 8 / (8+8) = .5 |
|
|
| bob12345678 |
Thanks for the responce a damping factor of .5 will work out
just about right for my plans. Building a 4 0r 5 watt SOZ in a
Dynaco st400 chassis I am almost tempted to use batteries!
Thanks Bob12345678 AKA Woody |
|
|
| FYC |
Based on Ian's circuit, I modify my SOZ into a x-SOZ today.
Power supply rail is +/-15V, I keep the 8 ohm drain resistors, removing all the 1 ohm power resistors, adding the 10k/1k feedback resistors, and the 15k bias resistor connecting to the Gate node.
a 2.2uF input coupling capacitor as the Gate is no longer at 0V.
all done within 2 hours.
result - hard to describe. definitely more details compared to SOZ. seems to sound sweeter, but apparently the dynamics is less than the SOZ. more audition is needed to draw a better conclusion. |
|
|
| gnomus |
I have some questions on what X circuit topology offers. From what I understand one benefit is to lower the distortion of the output mosfets. Does X also help clean up power supply noise? Or, improve low impedance speaker performance? Finally are there other benefits that I have not touched on?
Thanks in advance. |
|
|
| stefanobilliani |
X works to create a more perfect match between the 2 halves of the circuit. Due to cancellation the increased matching will reduce noise and distorsion . With active current sources I think you can go on with a higher damping factor.
Stefano |
|
|
| Nelson Pass |
You can get arbitrarily greater damping factors
with the X amps, particularly if you enclose the
output stage in the loop.
One point is worth making here, and that is that the
SuperSymmetric circuit is not very forgiving of excess
feedback, tending to form an "infinite hall of mirrors"
with large open loop gain.
Thus it is quite ideal for simple circuits - 2 gain stages
or less being preferred.
When I first developed the idea, I couldn't get it to
work properly for a long time, which is why the Alephs
were released first. Later, when the Alephs were
reduced to 2 gain stage designs, I went back at the
X circuit, and it began working beautifully. |
|
|
| Henrik |
I did it today, an acording to the attached scematic, but only as a test.
I will try to describe my first impression of this X-SOZ:
The bas is more tight, grainy and dynamic and still deep.
The trebele is even more transparet and detaild.
The midtone more detailed.
The allover dynamics is increesed.
This is a conclusion after listening to two of my favortes.
First, Leonard Choenīs "Ten New Songs" (CD). For instance on the number "Alexander leaving" there is a butiful welldefined space around the voices which gives a nice impression of their presence. These voises is very much alive in a space made soft and wide by bas and synthesizer-violin(s) with deep drums and some sharp highhat. Nice composition.
Last, one of my old favortis, Tom Verlaine, an especially "Flash Light" from 1987 (LP). This is Rocknīroll without any soft edges with two leadguitars, bas and drums. This music needs a lot of dynamics, and no matter how "noisy" and complex the music seems to bee, you can still hear evrey single instrument on the stage. This album include some of the best Rocknīroll guitars I have ever heard, there is a lot of energy and a lot of fantasy here.
The improvements is clear, but it is not a totally different amplifire, it has just become a little (how much is that?) better. May be it will be even better when I make these changes with parts of a quality as the rest of the amplifire. |
|
|
| Nelson Pass |
| Sweet. Looks real good. |
|
|
| stefanobilliani |
Mr pass , I notice your posts regarding SS circuits .
So , I have a question.
Does that circuit with 2 diff pair ( this thread ) suffer the effect of
" hall of mirrors " ?
Thanks |
|
|
| gnomus |
Henrik,
The resistor that ties the drains of the mosfets together is that 15k ohms? How was the value calculated?
The 47u capacitors what voltage and how did you arrive at the capacitance value of these?
Also, R12 and R13 how was the value calculated?
Finally, the round symbols labled with "V" these are test points? Where V4 is 19 volts?
I have a similar SOZ and was considering converting it. Now that you have listened longer to the the X-SOZ. Do you feel the change is for the better?
Thanks in advance. |
|
|
| Henrik |
Gnomus
R6, R7, R8, R9 are simple leftovers from the org. SOZ, and they now constitute a single resistor at 4E25. They are currentsources for the diff. pair.
C4, and R11+R13 forms a 1 Hz Highpass filter.
C=(1/(2*PI*R*f*1000000)) uF.
Frequences from the input under 1 Hz wil go to the output in openloopgain.
The calculation of R13 an R14 was made by Ian.
R11 and R13 divides the outputsignal, and R14 increases the source (preamp) impedance to allow the inputsignal to swing with the feedbacksignal.
This is my understanding, but may be there is more to this.
Other values could be used, but since these works fine, i will let them stay.
V2 an V4 is the powersupply at 19V each.
V1 and V3 is the inputsignal to simulate.
My X-SOZ still sounds good.
To do a test is very simple, so try it out and se if this is something for you.
Then let us hear how it suits you. |
|
|
| Ian Macmillan |
As Henrik says, this calculation is not difficult. Basically the ratio determines the gain, i.e. the gain of each half of the diff pair is R13/R14 although this will be reduced due the intrinsic gain being quite low. The absolute values of these resistors are not critical but R14 determines the input impedance (single ended) and hence a low value will make life hard for the preamp driving it. On the other hand, too high a value will significantly reduce the high frequency roll off. The values I chose were intended as a compromise between these conditions.
Ian. |
|
|
| Henrik |
Ian
With R13=10k and R14=1k (your choise) High roll off is 45kHz (-3db.
With R13=100k and R14=10k High roll off is 6kHz (-3db).
With R13=1k and R14=100E High roll off is 200kHz (-3db).
45kHz Seems to be a good choise, but woudīnt it be a little too hard job for some preamps to drive an impedance at 1k? |
|
|
| Henrik |
In my review of my new X_SOZ from a earleier post:
"The bas is more tight, grainy and dynamic and still deep."
I felt a littel hesitation wen i wrote "still deep".
Wen a bass seems more structurated or grainy it looses some softness wich can be interpretated as lack of deepness in the bass, but it isnīt rearley.
But I have a feeling, that it could be deeper without loosing the structure.
I still prefer my test X-SOZ, no doubt, even for the bass alone.
This lack of deepness could be caused by the lower inputimpedance in the X-SOZ, so by increacing the C1 and C2 in BOSOZ from 10 to 47uF it shoud go deeper.
I havenīt tryed it out yet, but i am pretty shure that tis is a problem, but how signifant it is i dont now before i have tried it.
The graph attached is from a simulation on BOSOZ an SOZ in combination. |
|
|
| Henrik |
Nice reading.
In my X-SOZ it is obvious that the impedance and the frequency respons is working at the edges, may be a problem, may be not.
I coud not resist to give it a sim, adding "the penultimate zen" buffer to this circiut.
Tis is a quick shut, but in the sim it is clear what it can do for X_SOZ. High input impedance and exelent low and high frequency.
And may be it also will benefit from other virtues from this buffer deskribed so well by Mr. Pass.
Any comments?
I must make two posts, each for one picture. |
|
|
| Henrik |
| Pushed the wrong button, so tree posts it will be. |
|
|
| Henrik |
| Graph for X-SOZ with buffer. |
|
|
| Henrik |
| Wrong picture, so here it is, the graph. |
|
|
| stefanobilliani |
:cool: It seem to be very interesting .I like to think the input buffer not as a second stage thus probably it will be a good
improvement.
These days I am enjoing the Zen -like Headphone amp (that with
an imput buffer too is eccelent) and I left soz a little aside .
Even becouse I am planning to buy more powered transformers.
By the way , what is the voltage at the output of the buffer?
Stefano |
|
|
| Henrik |
Hi Stefano
The voltage output from the buffer is the same as at the input since it is a follower.
The input signal presented at the node between R15 an C1 is 900mV (amplitude).
The bufferinput is 220 mA, that is at the node between the gate and R14. This lowering of the signal is caused by the NFB.
These values are just a first shut, for best performance they should be reconsiderd.
I am not shure, if the X-SOZ will benefit from these buffers, I havnīt actually builded them. I just felt, that the problems I had with the X-SOZ in a way was related to "The One And Onlyīs" considerations in Zen Var. 4.
I am looking foreward to rebuild my test-X-SOZ to the (final?) X-SOZ with better parts in the X-circiuit, and put some new blackgate caps into my BOSOZ.
And happy listening in your headbangers, i hope you will enjoy your X-SOZ too some day! |
|
|
| jh6you |
It was a very interesting read.
I have understood that, if the low input impedance is mostly resistive in character, it is not a problem to keep the wide bandwidth.
Meanwhile, Mr Nelson Pass comments on his article: "The Son of Zen can have any input resistor you care to use."
Why the low input impedance will make the life hard for preamps driving it...?
Look forward to hearing. |
|
|
| Henrik |
I wrote in my previus post: "The bufferinput is 220 mA".
I ment "The bufferinput is 220 mV". Sorry!
Low inputimpedance demands more current to keep the signal alive.
If you are using coupling caps in front of this input, then the value of the caps should increase with the lower inputimpedance.
This is my understanding. I am not a proffessional in these matters, so if i am wrong, i hope someone wil correct me. |
|
|
| stefanobilliani |
Hallo Henrik and all of you,
is interestin to know if is it possible obtain a zero volt potential at
the source pin of the imput buffer (dc) . To do this the gate of this buffer there must be -4 volt .Isn't it?
I tell you this 'couse in this way the output swing will be greater.
And the pass band too.
What do you think?
Stefano |
|
|
| Henrik |
Stefano
As I understand the "Penultimate inputbuffer", it is the low input capacitance at 50 pF on the gate of ZVP3310, where IRFP140 has a 1700 pF inputcapacitance.
This buffer makes it possible to keep a high input resistor (R14) in the front, and because of the low capacitance load (like a lowpassfilter) we get a significant higer highfrequense roloff.
Because of the higinputimpedance (mainly R14) we can use much smaller inputcaps, here 4,7uF, (like in a highpassfilter) and have a -3db rolloff a little under 1 Hz.
This is the main cause to the wider bandwidth as I see it, and again I am shure that tere is much more to this.
In the simulation, the sourcepin at the inputbuffer have a dcpotentiale at 0,08V and the gate have -3,79V, this could be adjusted through R10 to obtain 0V at the buffers sourcepin. But I donīt think that this 0 is important, as long as it stays within say +- 2V, but a difference in voltage between the two complementary buffres sourcepins ar important to keep dc at the output near 0.
I dont think that a Voltage change at the buffers sourcepins will alter the bandwidth, but it can have some major influence at the symmetry of the singlended voltage swing when clipping. |
|
|
| Ian Macmillan |
I'm glad to someone else taking up the x-soz challenge, Henrik. I hope you don't mind, but I've taken the liberty of modifying your circuit in an attempt to improve its performance further. The main change is to connect the input buffer to the diff pair 'tail' rather than to -ve. This operates it under conditions more like the Zen V4, reducing distortion and power dissipation. I've also substituted the ZVP3306 and I'm lead to beleive that the ZVP3310 may be even better due to its lower capacitance. I didn't have a model of the latter to hand which is why I didn't use it. I've changed the biasing a little to minimise distortion and reduced the current in the buffer. The latter change results in only a very small increase in distortion (approx 0.001%) and usefully reduces dissipation.
I hope this is useful and I would appreciate it very much if you would like to try it. I will do so too, but I don't have a SOZ available at the moment.
Ian. |
|
|
| Henrik |
Ian
I actually hoped You would come back, more than welcome!
Wen I was drawing the first scematic, I realised that the Vds at the buffer was too high, generating too mush unlinearity or distortion (outputcharacteristica from datasheet), so I would like to make a suitable negative voltage refrence somehow, but I did not find it at first, so therefore I quickly tied the drain to -ve. I was looking for the wider bandwidth, and that I got in this first shut.
I do have somequestions to You:
At the node for the two tied sourcepins of the IRFP140īs You will se som voltage swing from the signal, not much, but it is there. The amplitude of this swing is in bal. mode 15mV and in unbal. mode 300mV.
1) When You tie the buffer drains to this node, wounīt You then get som feedback this way and thus having two feedbackloops?
2) If You get this feedback, will it matter at all?
3) Is there other benefits tieing the bufferdrains to the sourcepins of the IRFP140īs besides from getting a suitable negative voltage refrence?
4) As You can see in my scematic, I have used some zeners to make this negative voltage refrence, it works fine in the simulator, but could there be some hidden problems in this?
5) You changed the bias a little, and as far as I can se, you have mediated the voltage betwin the gate and the sourcepin at at the inputbuffer. Why will this better the distortion figurs?
The current through the buffer in my first scematic was too high, In Zen Var4 Nelson used 7mA, therefore I have increased R3 from 680E to 2.7K.
6) Any particular reason You have chosen R3=1K or 19mA current?
Your aproach to this is very usefull to me, I am learning a lot from this too, and I hope You will answer at least some of my questions.
I need to get the ZVP3310 from Digikey first, so it will take some time to do this.
Thancks for Your post. |
|
|
| Ian Macmillan |
Hi Henrik, good to see that you are questioning my suggestions rather than blindly accepting them. I do have some answers for you, hopefully useful. First off, why did I tie the buffers to the diff pair tail? Your guess is at least partly correct: the diff pair tail is a useful source of negative voltage at about the right voltage. However, there is another and perhaps better reason. Connecting the buffer in this way effectively operates it at almost constant Vds (due to the Vgs of the diff pair MOSFET). This might be regarded as a sort of cascode arrangement and has the usual advantages, namely reduced distortion in the buffer stage. Of course you can use the zeners as per your schematic but then you will not get the cascode effect. However, Im sure they will work just fine.
Does tying the buffers to the tail result in another feedback loop? I dont think so, at least not in the conventional sense. Any voltage at the diff pair tail is effectively the result of the buffer output (which will be much the same as the input signal) and is in phase with it. Other than the cascode action already mentioned, I dont see any other form of feedback, undesirable or otherwise.
Regarding the biasing, this is the result of another almost inevitable compromise. The most important aspect here is arranging for the correct voltage at the drains of the output devices. Correct in this sense being a compromise between maximum output voltage swing and avoiding excessively low values of dynamic Vds due to signal excursion. The latter region of the output devices being less linear. I chose a value of bias that gave the least distortion at close to the maximum signal level.
Finally, on to the current through the input buffer device. In general, the higher the current, the lower the distortion. I started from your value and decreased the current until the distortion started to increase measurably. That said, this is a pretty low level effect and I am sure Nelsons figure of 7mA would do just fine.
I hope this helps. I too am keen to learn more as a result of our collective exploration.
Ian. |
|
|
| stefanobilliani |
Ian ,
for example I don't understand (or don't agree) that way to connect the buffer.
First I don't recognize a cascode operation,
second ,talkin' about phase ,the signal from the buffer goes to the source of the diff pair in phase, while if you connect a resistor
between the drain of the buffer and -V you will find at this point a
amplified signal wich will be inverted on the phase.
Notwithstanding this , I started to see clearly through the SuSy concept ,thanks even to our considerations and those of MR Pass.
Pretty soon I'll come back with a schematic for a comparison.
Stefano |
|
|
| Ian Macmillan |
You may be right but I don't think so. Perhaps I confused the issue with my reference to cascode operation. All I meant was that the buffer device is being operated at a near constant Vds (like, but not the same as, a cascode) since the Vgs of the output device acts like an almost constant voltage source. This will also effectively conduct any signal across the buffer device so I don't think there is any issue with an out of phase signal appearing at the buffer drain.
Ian. |
|
|
| stefanobilliani |
| I try to post the zip with schematic: |
|
|
| stefanobilliani |
Ok , seems allright ...
Here is a feedback arrangiament around the differential pair.
The loop is actually not including the output stage.
Ian , Henrik
I play around with some resistor values , but you with your
simulating sistems could help find more appropriate ones.
Any way , I started with :
R12-R15 1.5k
R10-R11 1k
R13-R14 200k
May be good put some imput resistor to ground.
Stefano |
|
|
| Henrik |
Ian
Thaks a lot for Your answers.
I try real hard to go deeper into the understanding of this buffer, it is hard to beleve, that scuch little peaces of plastic with only 3 legs can be so hard to understand, amaizing.
I got your point on tieing the the buffer sources to the tail, but I still canīt make room in my imagination to see if it works or not, and that is becauce I need som more bacic understanding of how little plastic-things works.
So I wil go on to get som literature on the subject. If You have some recomadations, please let me know.
Stefano
I know the benefits of using cascode, so it is obvius to me that this buffer is not, but there coud be some common operating conditions, as I read Nelsons article Zen Var4 I got that feeling, but since I dot understand his word fully, it is hard for me to get an opinion on this. You donīt beleve in this coupleing bufferscources to the diff. tail, and I am shure You have Your resons, but I cant at the moment see wether it works or not, sorry.
In Your scematic, You omited the the feedback at the output, I regard these as the fundamental diff. pair, around wich we shoud make the x-feedback, so taking them out is in my humble opinion to leave the X behind and make some local feedback at the inputs. Donīt the X-feedback need some "tails" to work through? In the Aleph-X it was the input diff. pair tail.
I am up to do this X-SOZ, but it might take some time.
But to day I order som parts at Digikey, but how long it takes before I get these parts is hard to say.
Keep up the good work!
I have some studing to do first, I will come back later. |
|
|
| Ian Macmillan |
I admit to being a little bewildered at what you are trying to achieve with your circuit suggestion. Why place feedback round the buffer stage, it is already quite linear and the circuit does not require any output from the drain. I agree with Henrik that the most non-linear devices are the diff pair and this is where the feedback really counts. Your alternative is not an X either, so any advantage from this mode of working will be missing from your suggestion.
Ian. |
|
|
| Henrik |
Ian
My frustration not understandinig it all yesterday was leading my post, an You did answer some of my questions to which I didnīt make any comments.
Your changes to my first circiut, except the main issue about tying the buffers drains to the tail at the output pair, I can agree with.
You changed the bias point for the diff. pair by changing R10 (in mys scematic) from 360K to 680K, thus giving a more symmetrical waveform when approching klipping. Nice, I have changed that too in my scemtic!
You also changed the current trough the buffer, and even if it only decreases the distortion very little, we take that too as long as we keep Vds around -4V, conciderd and done.
The size of the inputcap (22uF) in Your scematic gives som -3 db lowfrequince rolloff at 0,17 Hz, against my 0,79 Hz with 4.7uF.
I think this difference is more a private matter, I will prefer to use mor money on better quality caps at lower value than the other way around.
I canīt give you my opinion on where to put the drains of the buffer, You may get it if or when I get a better understanding of little peaces of plastic withe three legs. |
|
|
| Henrik |
Ian
You wrote in an earlier post:
"Connecting the buffer in this way effectively operates it at almost constant Vds (due to the Vgs of the diff pair MOSFET)."
I have simulated both connectings (tail/zener) and there is not any difference in the Vgs behaviour, at least not in the simulator.
But there is a difference in the Vds swing, not in the amplitude, but in the shape, wich is normal sinus when connectig is to the zeners, but som way irregular when connested to the tail.
But ther is no difference to se at the output to the spaeker. |
|
|
| Henrik |
The unregular shape of the Vds behavior at the buffer is caused by the Vgs of the diffpair, but it seems as if this irregularity dont show up at the gates of de diffpair. The buffer Vds is affecet by the tail, but this forced change in Vds is not to be seen at bufferoutput. It seems as if there is no feedback here at all.
So, for now I see the connecting of the buffer to the tail rather than to -Ve as somethig usable, but if there is some benefits to this I donīt know.
But again, can You trust simulation in this delicate matter, I donīt know, but I have an Idear of what to look for when making some experimens.
Ian, thanks for your answers. |
|
|
| stefanobilliani |
I'm not surprise that you don't see an X .
Or is it Just that you don't like a Bmp file?:zombie: |
|
|
| Henrik |
Stefano
I am confused over Your last post.
I am truley sorry if I have offended You in any way, that was not my intention, what so ever. |
|
|
| Ian Macmillan |
I have no problem with bmp files and I can view your circuit just fine. I pointed out the lack of the X factor since I thought that this thread was supposed to be about x-rated soz - it is after all called x-soz :)
Ian. |
|
|
| stefanobilliani |
Henrik , I'm truly not offended .
An Aleph circuit is a very good source for study ( even unbalanced)
say Aleph 4 .
The output on that circuits is not taken at the differential pair.
Neither feedback. |
|
|
| Henrik |
Stefano
I am rearley happy about that.
Aleph 4 is indeed a subjet to learn from.
So you was aming more to a bridged Zen-thing rather than an X? |
|
|
| Ian Macmillan |
Having read your comments and pondered the circuit some more, I find I am in broad agreement with you. The 4.7uF input capacitor will probably do just fine I only raised the value as the roll off is gentle and begins at around 10Hz with the 4.7uF value. In retrospect I agree that it is better to use a high quality part rather than a larger value.
Ive also taken a long hard look at my reasons for connecting the buffer drains to the diff pair tail and have concluded that it really doesnt offer any significant advantage over a fixed voltage reference, other than perhaps a reduction in parts. Distortion for a 10V peak to peak sine is about 0.39% either way.
There is of course another way of doing things entirely, i.e. use the buffer AC coupled in the same way as the Zen v4. The drain of the buffers can then be connected directly to ground. Of course some way must then be found to restore the correct bias for the output devices. I should say however, that when I tried this I could only see worse distortion no matter how I did the biasing (approx 0.6%). I dont much care for the additional capacitor either, so perhaps this is a dead end.
Finally, can (should?) we trust simulation in these delicate matters? Personally I dont think so, the final arbiter must always be ones own ears. Simulation is just a useful tool that can help in avoiding the pursuit of ill conceived ideas and sometimes save the odd blown device!
Ian. |
|
|
| stefanobilliani |
Sorry , on my Italian-English ditionary I don't find the verb aming.
But seems to me that the correction of error can be done adding feedback even were it seems not to be required. |
|
|
| Henrik |
Stefano
You know, itīs not always things comes out right, specially not when you pushes arond with a computer-keyboard. If You change aming to aiming, I am shure You will find it in Your dictionary.
Ian
I am too tired to give you some answers to nigt, but to morrow. |
|
|
| Henrik |
Ian
It seems as if we share the same basic poit of view on this X-SOZ, based on simulation.
4.5 uF at the input.
Buffer-drains to fixed voltge ref.
Since there is no advantaged in using the the diff. pair tail as voltage refrence, I suggests the use of Zener, all though it means a few parts more.
You tried to use ground as Vref., but it didnīt pay off, on the contrary, an also implied more parts in the signalpath.
I also like the ground to be used exclusively for the signal path, and nothing else, this also for aesthetic reasons.
I woud say we have tried to make it right and simple.
We also shares the same judgement on the simulator.
I have "trimmed" the peeks of a square wave signal with C3 and C4, like a nice haircut
So there is only one thing to do now, an that is to bring this X-SOZ out of the simulatoor and in to real life, music, air, ears and and all that.
I have named the attached file HenrikIan- X-SOZ with buffer Ver.01.pdf, so if You want to add something in a new ver. nr, You ar welcome. |
|
|
| Ian Macmillan |
Looks like you have done a nice job so I don't have much to add. The only change I would make is to set R16 and R10 at 470k to give around 5v at Q1/Q2 drains. This maximises headroom (symmetrical clipping) at +/- 12v peak. This also drops the voltage at the gates of Q3/Q4 and hence you may need to change the value of the zeners to compensate (I used 16v to set the drain at -3v).
Ian. |
|
|
| Henrik |
Ian
Glad that You like it.
I have redrawn the scematic, so most of the partnumbers have changed. I have done this to keep the good old SOZ numbers intact as this one is the mainissue.
My first values for R16 and R17 was 360k, a value that kept the original biaspoint at Q1 and Q2.
You suggested 680k, and I agreed with You.
Reconcidering these values, I found, that Yours 470k is the best solution for the headroom. I donīt know why I didnīt saw the reduced "headroom" at 680k, though it is verry obvious in the graph I have attached along with the new scematic.
So R16 and R17 = 470k and 16V zeners at the bufferdrains it will be for now.
C5 and C6 is there to remove the noise from the zeners, I donīt know if they really are needed or not, but if they shall remoove any noise, they have to be conciderable larger than the first 10nF, if they are 100uF, noise will have a -3db rolloff at 500Hz. Test in real life will tell if they are needed or not.
As You can see, I have added some back to back zeners and gateresistors at the input.
Is there any particular reason why NP didnīt put these in Zen Var. 4 ? I like the protection.
I was surfing for some information on the Millereffect, and I found som nice article by Kieth Herron at his homepage. http://www.herronaudio.com/tastubes.html
In this article he points out verry clearley the need for wider band amplification, the X-SOZ canīt deliver that bandwith he claims, but X-SOZ with a buffer can, when simulated.
But wee donīt know before wee have heard it. |
|
|
| Henrik |
| I forgot the attachment, here it is. |
|
|
| Nelson Pass |
Reading Herron's article, I observe that he regards limited
bandwidth as stuff that doesn't make it to 20 KHz, and
I'm certain most here would agree.
Why do you think any of the Zen amps don't make it
to 20 KHz, much less the 100 KHz numbers they measure? |
|
|
| Henrik |
Hi Nelson!
I am very sorry, that I have expressed my self in a way that could be interpreted the way You did.
So I will try to do better now.
I know that all the Zen amps makes it beyond 100KHz, and that is great.
When I wrote "the need for wider band amplification" it was not ment as to say that wee need more bandwidth than the Zen amps can deliver, I know that they do better than 100KHz.
Measurements by NP:
........................20KHz..........100KHz
BOSOZ................-0db...........-0.3db
Zen V4.............-0.1db...........-1.4db
SOZ..................-0db...............-0db (Flat from DC to several MHz)
Measurements in my simulator:
.......................20KHz...........100KHz
SOZ...................-0db.............-0.3db (I still beleeve in DC to XMHz)
X-SOZ................-1db...............-8db
X-SOZ+Buf......-0.05db............-1.4db
What is new to me, is this statement in Herrons article:
"We found that a 3-millibel variation in the frequency response curve can change a listener's perception of the unit. This could explain the audible improvements often associated with wide bandwidth in audio equipment. A loss of frequency response in terms of decibels at 100kHz will mean, in most cases, a drop in frequency response of several millibels at 10kHz - a perceptible change. Many of the colorations (sonic variations) heard in audio equipment will correlate with frequency response variations. It is somewhat surprising that one can consistently hear such small differences between electronic components when the listening is done with speakers that vary in response by several decibels."
Since amplifiredesign also is the art of compromises, all Herrons products keeps within 20 Hz to 20 kHz ą0.1 dB.
May be all this i ****, but there coud be somthing in this.
Said in an other way, all Your Zens do the same, and most of them even much better.
When I modifyed Your SOZ, wich by its nature have really nice frequencyrespoce, into an Henrik-X-SOZ, the cost of this was decreesed bandwidth. After reading Herrons article I got this overwhelming feeling of "Audiofile anxiety" (one of Your own expressions as far as I remember) and claustrophobia caused of the Henrik-X-SOZ limited bandwidth, and that, at least of all, we donīt want, donīt we?
Nelson, still friends!:angel: |
|
|
| yldouright |
Henrik
some people with good ears will hear 100mB (.01dB) differences in a fixed frequency controlled test. I have never heard of anyone being able to reliably discern more than 1dB differences of a complex waveform in phase. |
|
|
| Henrik |
yldouright
:ill: I allready feel much better, thanks!:)
Actually, I have never tested my hearing this way, it should be worth a try.
If these buffers produceses som sonical benefit, bandwidth or not, then they shall stay, else out.
I am looking forward to try them anyway. |
|
|
| Nelson Pass |
I would certainly agree that -1 dB at 20 KHz is going to
be audible to your more discerning listener.
There must be something going on with your simulation
to get so much drop off. Can you post a schematic? |
|
|
| Henrik |
Nelson!
Shure I can post a scematic, here it is, my X-SOZ without buffer.
I really woud like to get some feedback on the freequensy responce. Thanks!
C1 and C2 at the scematic is the outputcaps in BOSOZ (10uF at the moment), without these the low frequensy will become a "rollup" because of the cap in the feedbackloop have a roll off at around 0,7 Hz, and thus no NFB here.
But all this is simulations. I havenīt got a scpoe yet to do the measurement, but I will get one within a month or so.
Mesurements or not, I am still happy with the sound of my X-SOZ, but I have this feeling it coud do even better. |
|
|
| Nelson Pass |
Yeah, the 1 Kohm inputs will give you about that. You
can lower the values of both the 10K and 1K values and
pick up some more bandwidth or you can reduce the
gain from 20 dB to say 14 dB by replacing the 10K's with
4.7 Kohm, keeping the 1K inputs. |
|
|
| Henrik |
I will try to do that, in the sim and in real life.
Thanks a lot. |
|
|
| Henrik |
When gain changed to 14db then -0.3db at 20KHz, -3db at 74KHz.
I have only tried this in my sim.
To mee it still seems a little low allthough better.
Any Opinions on this one. |
|
|
| Nelson Pass |
The buffer in ZV4 does work very well.
;) |
|
|
| stefanobilliani |
Henrik,
I don't know if I was aming a bridged Zen rather than an X,
Actually I suppose that " only Freud knows it ":) ; by the way thank you for the input.
I 'm for the buffer. I have tried it - but with irf9610- and it has surprised me ( real life ).
I would like pospone any discussion for when you will try it too.:judge: |
|
|
| Henrik |
Hi Stefano!
May be You had a "Freudian slip".
Most certainly, you are quick at the "real life" trigger, nice.
You made the buffer with irf9610, and even with the higher inputcapacitanse You get -0,01db at 20kHz and -3db at 140 KHz. All seen in my "crystal ball" better known as a simulator.
So if this is real life to You, congatulation.
I do have a wordbook just like You, but I canīt find "pospone", but any way, you wil hear from me again when I have tried this buffer outside my "krystalball".
I am planning a test SOZ, with wich I can switch fast between SOZ, X-SOZ and X-SOZ with buffer, and I have all the difficult parts in house now. |
|
|
| AudioFreak |
| Postpone should give you the meaning you are seeking. |
|
|
| eLarson |
Hej, Henrik
It sounds like an interesting set of experiments you have planned. Are you going to use switches to bypass the buffer, connect/disconnect the "X" connection, and so on?
Just curious.
Erik |
|
|
| Henrik |
Erik
I plan to use some goldplated computerconnectors and some veroboards.
My setup looks like the attached pdf.
Curiousity is a good thing! |
|
|
| Henrik |
AudioFreak!
I just realised Your ansver to me.
Thanks for the interpetation of pospone.
Stefano
No discussion before entering real life, promise. |
|
|
| eLarson |
Henrik -
That looks fantastic! Not only that: it has made it considerably easier for me to get my head around the different variations that had been posted in the thread.
Erik |
|
|
| Henrik |
| When coupeling into X-SOZ with or without buffer, the terminals in "test SOZ.pdf" 13 and 14 is short. |
|
|
| Henrik |
To day I finished my Test SOZ, as described in "test soz.zip" in my post from 09-10-2002.
Stefano tried the Buffered X-SOZ the other day, and he was for it, as he said in his post from 09-10-2002.
I totaly agree with him.
It really was fun to listen to the different versions and to be able to compare them.
For this purpose I used Laurie Andersonīs LP "Strange Angels" for two reasons, first I like this album very much and second Laurie Anderson works with the sensation of space in her music as a part of then composition. I used the two first cuts, "Strange Angel" and "Monkeyīs Paw".
SOZ compared to Buf-X-SOZ.
I started listening to the Buf-X-SOZ, and afterwards the SOZ.
I coudnīt beleve the difference, the SOZ had only verry little depth in the soundstage, it was almost as if it wasnīt there at all when you just have heard the Buf-X-SOZ, unbelievable difference.
The SOZīs bas is deep, but it is pouring more or les out in the bottom of the soundstage like smoke on a Rock concert, but on the Buf-X-SOZ you can actually hear the different levels of deep and dynamic bass, wich also helps locating it in the room and makes it more tactile in contast to smoke. Amaizing.
All the midtone is far more dynamic and detaild on the Buf-X-SOZ, wish makes it possible to hear Lauriandersons voice tree diffrent places in the room, where she is triplicated, singing unison with her self. There is a lot of space around her voice made by overtones from accordion, guitars, tambourine, cymbals and echoes from dynamic drums.
So so so so wonderful.
My X-SOZ had all the treedimentional virtues besides the deepest bas, wich was caused by the low value of output caps (10uF) in my BOSOZ, and because of the lack of highfrequense response the sound was somehow harsh, and that destroyed the feeling of high space in the soundstage. So the easy coming sound from the SOZ came back because of the buffer, so nothing is lost, only much more had come to stage.
This is really some improvement worth talking about, not to say to listen to.
If this is how the magical X sounds, then Mr. Pass, congratulations with Your extreemly good looking XA200, the sound must be terific!!!!
Stefano.
I would appriciate some comments from You, we shoud be able to talk together since the "postpone" era is over now.
Many thanks for starting this thread and staying on board.
Ian
Thanks a lot for Your support, I hope You will try this Buf-X-SOZ Yourself.
Nelson "The One And Only"
Again, thanks a lot for Your support and will to share Your knowledge.
Cheers :drink: |
|
|
| Henrik |
| By the way, I used ZVP3310A in the buffer, not the ZVP3306A, just like "The One And Only" did in ZV4. |
|
|
| Peter Daniel |
| So in your opinion buffered X-SOZ sounds better than just X-SOZ? |
|
|
| eLarson |
Henrik,
Congratulations on completing your tests. What type of speakers do you run the buffered X-SoZ with?
Erik |
|
|
| Henrik |
Bas: Richard Allen CG 12 Super, 15 Ohm (sensitivity >94 db)
Midway: Richard Allen LP88, 8 Ohm
High: Dynaudio D28
Filter: Hybrid, 6db/oct at 340 Hz and 18db/oct at 4000 Hz.
These data are as I remember, could be a little wrong.
In my rooms I can hear some verry loud music with this SOZ! |
|
|
| stefanobilliani |
Henrik I'm very pleased with your posts.
I love Laurie Anderson too. Even Dynaudio!
Some listeneng tests: Archie Sheep-Dollar Brand , Moniebah;
Steely Dan, Your Gold Teeth 2
Dizzy Gillespie , Birk's Works
Tinsley Ellis , Pawnbroker
Billy Cobham, Two For Juan
Afterhours
Just some .
I feel like the musicians want to come out the speakers with their
instruments and voices and yes, they play fast and warmth and
the bass is round and deep. Touching .
Dinamics are very different from the xson of zen without buffer.
At the moment I put the feedback connections like the xson of zen without buffer and let the buffer out the loop.
But I will try your arrangement when Zvp's will be avaiable .Enjoy :Piano: |
|
|
| Henrik |
Stefano
When You have tried the buffer inside the loop, let us hear from You again.
And, I like Dollarbrand too.
;) |
|
|
|
