| Nelson Pass |
I prefer without the buffer if your source can drive a few
hundred ohms without distortion, but often they can't,
and this often makes the buffered piece sound better.
:Pirate: |
|
|
| stefanobilliani |
| Thinkin' at the I/V converter (balanced) that I 've build (with reference D 1 schematics) out of the dacs , probably if I need that lower output impedances is better put a follower at the output of the preamp(!?) or paralleling devices and circuits...:radar: |
|
|
| Henrik |
Nelson
I beleve You.
The only problem with the X-SOZ (without buffer) is the limited high roll off at inputimpedances > 200-300 Ohm.
In my listening test, the unbuffered X-SOZ with an inputimpedance at 1K showed some really nice manners, execpt for the highest frequences.
But BOSOZ needs outputcaps at 100uF to keep a deasent low frequence roll off when driving an impedance at 200 Ohm, and it seems to loose haf the gain when it does, so therefore I went for the buffer.
Do You think BOSOZ with output caps at around 100uF will do a good job as source for an unbuffered X-SOZ with an inputimpedance at say 220 Ohm?
By the way, the buffered X-SOZ still sounds verry nice. |
|
|
| Peter Daniel |
Henrik,
What happens if you increase input impedance on X-SOZ, lets say 10k or more? |
|
|
| Henrik |
Peter
I have some figurs to You, which illustrates the tendencies from one of my earlier posts:
With R13=10k and R14=1k 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).
You asked me in an earlier post:
So in your opinion buffered X-SOZ sounds better than just X-SOZ?
And I answerd: Far better!!!!
Nelson just pointed out, that if You have a source that can drive 200 Ohm load, this might not be the case.
Nelson keeps us on the track, realy nice. |
|
|
| Henrik |
Peter
In other words:
100 Ohm impedance High roll off is 200kHz (-3db).
1k impedance High roll off is 45kHz (-3db.
10k impedance High roll off is 6kHz (-3db).
This is du to the input capacitance of the mosfet, say 1700pF. |
|
|
| jh6you |
Should I alter my perception...?
Perception so far:
The input buffer is “mainly” to improve the distortion and corresponding sound.
Perception in the future:
The input buffer is “mainly” to raise the critical highest frequency.
Should I...?
:spin: |
|
|
| Henrik |
| To keep a reasonable high frequence roll off, the X-SOZ needs a very low inputimpedance around 200 Ohm. To raise the inputimpedance to a more normal level say more than 10K, You need the buffer. |
|
|
| jh6you |
Thanks Henrik
So, the input buffer works for three:
(a) It increases the input impedance.
(b) Accordingly it improves the high frequency distortion.
(c) And it raises the upper critical frequency (-3dB).
One more question, please.
Could I expect the sonic improvement mainly from (b)
or mainly from (c)
or equally from (b) and (c)?
JH |
|
|
| till |
| may this depend on the characteristics of your source? if the preamp isn´t able to handle a low impedance as a load sonical improvements result out of a). THD of preamp will be lower. |
|
|
| Henrik |
As Nelson pointed out, the buffer itself dosn´t improove the sound, on the contrary, but it helps the less ideal source to drive the input load. The ideal source is not at all affected by the character of the load.
a) Therefore, impedance itself has very little to do with soundquality, it depends on the environment.
b) Distortion (high or low frequency) is mainly produced in the mosfets, the buffer can´t remoove that, acctually it adds some.
The X-feedback lower the distortion both in buffer (if it is inside the loop) and output stage.
c) The buffers higher input impedance helps the X-SOZ keeping the high and low frequency roll off at a more reasonable level when the source is not ideal.
To me this is much more a question of compromises.
This is my understanding, I am relatively new in this, so I coud be wrong or not too precise. |
|
|
| Nelson Pass |
Sounds like a concise summary to me.
:) |
|
|
| Henrik |
Thanks :wiz:, :captain:,:Pirate: or whatever Nelson, nice to know.
jh6you
:spin: Selv tak.
In Korea they obvious also speaks Danish.
Mange tak means: Thank you very much.
Selv tak means: You are welcome or don´t mention it. |
|
|
| Variac |
So Nelson, I'll bet a large percentage of the x zens are used with the Balanced line stage, BOSOZ. Do you feel that the X SOZ used with the BOSOZ benefits from the buffer or not? Are there suggestions you would make to mod the BOSOZ to make the fit the best?
This thread is an example of the best of DIY Audio. Lots of brainpower, thoughtfulness, great consideration for the feelings of others and communicating well even in a language not totally familiar to all (which has worked because of hte previouly mentioned attributes)
Stegano's cry in the wilderness about the possibilities of the XSOZ and sticking with the designing and real life making. Ian's great inputs, Henrics also and his beautiful test schematic and more beautiful real life tests! Of course The One and Only! and lots of contributions by others! :wave: WoW ;) |
|
|
| Nelson Pass |
I was considering this issue the other night, since it is
my observation that a properly set up BSOZ or even
BOZ can be perfectly happy driving the virtual ground
of the Zen amps and their variations. Their output is
simply a current source from the Drain of the Mosfet,
and their output impedance is whatever you load the
Drain with.
Actually, they are even happier driving a short with current,
as the Miller effect and the distortion associated with varying
voltage across the Mosfet disappear.
If I had been designing these preamps for Zens only, I
would have used this approach.
It's a simple matter to optimize either preamp circuit
for this, and short the input resistor of the Zen amps, and
I'll see what I can work up in the next few days, unless one
of you guys would care to beat me to it.
:yes: |
|
|
| Henrik |
Variac
Glad you like this thread, me too.
It has ben extremly educational to me, allmost everything i know about electronics i have learnd on this forum.
And thanks for asking Zenmaster Pass about the BSOZ and X-SOZ.
Nelson
Thanks a lot for your answer.
I believe i got your point.
Quote: "Their output is simply a current source from the Drain of the Mosfet, and their output impedance is whatever you load the
Drain with. "
Ok, I have scaled the load (R29/30) to 200 Ohm.
Quote: "It's a simple matter to optimize either preamp circuit for this, and short the input resistor of the Zen amps"
Ok, I have short (remooved) the inpuresistor at the X-SOZ. If I have understood You correct, the loadresistors R29 and R30 has replaced the former input resistors at the X-SOZ. The outputimpedance (R29/30) of the BSOZ forms the gainratio with R13/36 in the X-SOZ.
I have made this experiment as an integrated amp. in order to get down to basics, when this works we can separate them afterwards.
If my suggestions is correct, it is almost too simple, and it seems to work perfectly well, the only thing is the high value for C1 and C2. I have attached the file "Integrated BSOZ and X-SOZ.pdf" and a graph showing the frequensy responce. All made in my simulator.
Any comments on this one? |
|
|
| Henrik |
| Some resistor values shoud be finetuned, but i will wait until i have got some feedback on the more fundamental questions of this design. |
|
|
| Nelson Pass |
Looks really good to me!
:) |
|
|
| Henrik |
Thanks Nelson!
Now that I have the basic idear behid this design, I will try to take it further. |
|
|
| Joe Berry |
If you were to convert the output stage to a follower, I think you could increase the values of R10 and R11 to 100K or so with acceptable high frequency performance. This would let you reduce C1 & C2 to 2.2uF for about the same low frequency cutoff.
You could try this in the simulator to see how it works. Just move the speaker between the two output stage sources, reference R10 and R11 to ground instead of to V-, short out R4, R5, R6, and R7, and omit R13 and R36.
Follower operation of the output stage will require more voltage swing from the BOSOZ input section, but it can handle this easily if you increase R29 & R30 as needed.
I realize that this takes the idea in a different direction from the X-SOZ concept, but it might still be an interesting thing to try. |
|
|
| eLarson |
Hej, Henrik
I like the look of the circuit and the accompanying graph!
BTW: Where did you collect the MOSFET models for the IRF610 and IRF140?
Best,
Erik |
|
|
| Henrik |
Joe Berry
Thanks for your suggestions.
You can´t X the topology you mention, as you yourself said.
It certainly coud be an interesting construction, but my head is X´ed these days. Thanks anyway!
It seems as I can reduce the C1 and C2 to 10-100uF by a special trick without more limited lowfrequensy roll off. I have to try it.
Erik
You can get them from IRF´s homepage: http://www.irf.com/product-info/models/
By the way, i am using SIMetrix intro v. 4.1, wich is very nice and free. |
|
|
| stefanobilliani |
Hallo!
Just today I try to modify substantially the original circuit.
The results are on the attached schematic.
The idea comes from the Bride Of Zen.
(Be ready to build an x Boz...)
For real life test (sigh ... I haven't got a simulator) I use instead of R1 R2 two ligh bulbs that has a value of resistance around 5 ohm or so. That seems to me right since the gain decrease with the resistors.
I trimmed R5 and R6 to obtain across R3 and R4 1.5 volts that let flow current for 1.5 Ampere par halve.But I would like to run each halve at a current of 2 amps ( when I'll buy the right transformers)
And rise a bit the imput impedance.
How it sounds? Incredible ! |
|
|
| stefanobilliani |
I forget...
:drink: |
|
|
| Henrik |
I have worked a little with the basic idea and included some X to the BSOZ too.
I have reduced C1 and C2 from 560uF to 100uF, and the reason this coud be done is, that I have included C3 and C4 in the feedback loop at the XSOZ.
C3/4 makes the feedback roll off at around 4Hz and is almost gone at 200mHz. This roll off makes the signal at the input pass through in total open loop under 200mHz. Since C1/2 starts rolling off at 4Hz, this roll off takes the amount of open loop down to equal the flat toplevel until you gets under 200mHz. By ajusting C1/2 and C3/4 you can modulate the frequency response, but the X will start roll of at 4Hz and be gone at 200mHz. This partnership between C1/2 and C3/4 makes it difficult for the X-SOZ to be coupled to other preams, it really needs thos caps in (X)BSOZ. Some test is needed to hear if this is a good idea. May be you coud reduce C1 and C2 even more.
The implementation of the buffer was easy, an helpd the X-BSOZ to do better than 82kHz high roll off at -3db. If the X-BSOZ benefits from the buffer, I don´t know, but it is worth a try.
Some resistorvalues shoud be adjusted, this is a fairly raw sketch.
These are the main reasons why the scamatics loks like the attached file, wich alo contanis some graphs.
I would very much apreaciate some comments. |
|
|
| Henrik |
I didn´t realise you have posted, you must have done it while i was writing mine.
I will take a look at your scematic!
:drink: |
|
|
| Henrik |
I have just tried your scematic in the simulator, it could do much better.
You have biased (in your scematic) the fets with only 3.6V, wich produces only 300mA draw in both mosfets, you wanted a reasonable draw at 1.5 A.
You are having a bias voltage at the drains at about 22.5V, wich leaves only a 1.5 voltage swing with 8 Ohm speaker, and no headroom at all.
By the way, I think your scematic looks more like Nelsons Zenlite.
You shoud get your self a simulator :Present: , it is a great tool in the designing process :trapper: . SIMetrix free simulator is so easy to use :o .
You can downlod it here: http://www.newburytech.co.uk/Pages/download.html
I have attached a pdf and a simetrix-file of your scematic.
Hope you can use my feedback.
:spin: |
|
|
| eLarson |
I've downloaded the Newbury Tech simulator, and I can see why it seems to be so popular around these parts. :)
Has anyone tried out any Linux schematic capture/Spice simulators? There seems to be quite an array, each in a varying state of doneness according to the project pages. Rather than download each one, I figured I'd check here to see if that trail has already been blazed.
Erik |
|
|
| stefanobilliani |
Hallo!
Henrik,
please take a look at the schematic of Bride Of Zen: I use the same sistem to fine biasing the circuit except for R105 10k instead of 100k.My error , the results are not 10k to ground but at least 20k(or more).
But my circuit effectively runs 1.5A par halve.The drains are at about half the PS or to say 10-12volts.
Probably with light bulbs(24v 60W each) things are different (?).
I'm not Joking , the performance has nothing to do with the previous Xsons.
No doubt , this is X.
Thank you for the simulator: I will download it sooner.
:bigeyes: ;) |
|
|
| Henrik |
Erik
I have only tried simulators to windows, so I don´t know those for Linux.
Stefano
I am shure, that your bridged or balanced Zenversion souns good.
But in my opinion it contains no X.
You have build a diffrential stage without a tail, both mosfet-sources are tied to ground through its own resistor, and is hereby separated from each other.
In my understanding they need to share one and the same currentsource, thus the feedback can be comunicated from one mosfet to the other through their sources.
I am pretty shure You need a diff. pair with a tail.
I have bin dooing some sims since my last post.
I am still not satisfied with them, so no tests has ben done.
My goal is to make a seperate preamp and outputstage, but in order to understand the scirciuts interplay i have made the sims as integrated amps. I found one solution for an integrated amp wich looks very promising, but it can´t be separated besause af the global feedbackloop. |
|
|
| stefanobilliani |
Please note the original post by ERGO at page 4 and relative comment by Nelson Pass.
Thank You. |
|
|
| stefanobilliani |
As far as the current source concerned , the US Patent number
5,376,899 shows a current source for each halve.
ciao ;) |
|
|
| Henrik |
Hi Stefano
Shure, but even more important, they are tied together with item 40 (low value resistor), otherwise there would not be any diffrential behaviour, wich is the path for the X-feedback.
By item 40, 42, 43 (resistors) and the two currentsources 26 an 27 you can controle the degree of diffrential behaviour to introduce some coupelig loss in the recoursive feedbackprocess, and thus reduce "the hall of mirrors" effect in ceartin circuits.
Doesn´t seems to be a problem for the Aleph-X or Grey have fixed this in an other way, since it only have one currentsource at the diff. pair.
This is how I understand the patent and the X, and I am shure there is much more to this, and also, I could be wrong.
:) |
|
|
| Ian Macmillan |
Henrik, You are absolutely right regarding your comments about what it takes to have the "X Factor". Whether it is two current sources separated by a resistor (PI network) or one current source and two resistors (T network) is pretty much irrelevant (except perhaps for power dissipation).
The coupling loss to reduce the hall of mirrors effect is really only an issue for preamps where a significant amount of common mode noise might be a problem for whatever it is connected to - particularly if used single-ended.
For a power amp, a relatively large amount of common mode noise is less of an issue as it is effectively rejected by the speaker, i.e. it does not see common mode voltage as a signal at all.
Ian. |
|
|
| Nelson Pass |
In simple circuits I find that the transconductance of the
Jfets or Mosfets on the input stage limits the open loop
gain enough to prevent the "hall of mirrors", and so we
typically don't bother to bleed this junction off to ground. |
|
|
| Henrik |
Ian
Thanks for pointing out the terms "PI-network" and "T-network" and for confirming my thoughts abot how the "X" works.
I got your point on the preamp consideration when driving a singleended poweramp.
Nelson
I remeber an earlier post of Yours, where You explaind the reason why it took you so long to get the X to work, You then mentiond the need for a simple circiut (2 gainstages or less).
And that is why Grays Aleph-X don´t have this problem with "hall of mirrors" effect when using one currentsource at the input diff. pair.
So thanks, I wont bother to do this with a T-network with one resistorbased currentsource, but witout additional resistors to reduce the differntial gain, since this gain is needed in the feedbackloop.
I try to make X-BSOZ and X-SOZ witout buffers, I think I can do it, but the problem is to keep the high roll off at a reasonable high level.
:) |
|
|
| Nelson Pass |
I don't find I need buffers unless the source starts
sounding crappy into 500 ohms or so. Unfortunately,
this seems to be very common.
In addition, I talk to lots of people who get nervous with
loads less than 47K. Myself, I say it's a pitiful preamp
that can't source the 200 microamps that it takes to clip
a 10K input.
On the other hand, many of the complaints about the
sound of the Zen amps are traced back to input impedance. |
|
|
| sonnya |
It is funny because i was thinking about why you raised the input impedance of ZEN V4 to 47K... Even when you run the design inverted to minimize capacitance modulation it will not go completly away....
But i can see know when you drive it from a passive preamp.
The most opamps and discrete should have no problem driving into 500Ohm or less. i guess this is a bigger problem as mentioned?
Sonny |
|
|
| Henrik |
Nelson and Sonny
I wondered too why you Nelson raised the inputimpedance to 47k at Zen V4, working with tis XSOZ stuf it became very clear to me. And I understand this compromise, it is my impression, that the buffer only degrees the soundquality very little, specially compared to the sins comitted by an preamp not able to drive the presented load, that is some degradation worth talking about.
It is easy to make the BSOZ drive even 100 Ohms load.
I have made some simulations of a X-BSOZ and a X-SOZ connected to each other, where I got a frequensey respons like this:
-0 db @ 1Hz.
- 0,15db @ 20 kHz
-3db @ 100kHz
This is the best I can do for now, with a little less gain It can go -0.1db at 20kHz, wich was my goal, tests will show how much gain I need.
I tend to do some comparatative tests between the different types, with and without buffers.
I will come back when this is done, though I think it will take quite some time to do this, as I also have to work for living. |
|
|
| Henrik |
Yesterday evening I finished my test setup.
"The one and only" is so right about the use of buffers.
This time the buffers didn´t add anything, not even in the high frequense area as it did in my first testsetup, on the contrary, a significant part of the directness of the sound and soundstage dissapeard.
The addition of X to the BSOZ was quit the same as what happend when adding X to the SOZ.
I really am happy with this result.:bawling:
I will be back with some conclusions and more descriptions on the change of sound and some conciderations on the scematic.
Nelson, thanks for driving me to these tests by your short answers dropped at the right spots. |
|
|
| Henrik |
I have finished my description of the X-BSOZ and X-SOZ for now, so this is my point of view.
Gain
The gain at the original BSOZ an SOZ was controlled by R115 and R207, now it is by R120, R121, R122, R123, and R212, R213, R214, R215, all resistors in the X-feedbackloop.
I have kept the original gainlevel, and also this gave the best frequency responce.
Impedance
I have mesured the impedance by placing a resistor in series with the impedance to measure, and then scaled this resistor until it lowered the signal to the half. Thus the scaled resistor shoud be equal to the imedance.
If this is correct, then my measurements are reliable with recpect to the simulator.
I wanted drastically to lower the output impedance when changing BSOZ to the X-BSOZ, actually this was an easy task, because I got that as a free sideeffect when adding the X-feedback, it was lowered from 1.5kOhm to 20 Ohm, and without changing the values of R101 and R102, so I kept the original values of 750 Ohm each.
The X-SOZ output impedance was lowered by the X-feedback, just like it happend for the X-BSOZ, only not so dramatically, it went down to 4.7 Ohm from 15 Ohm.
The dampingfactor at the X-SOZ is the load divided with the outputimpedance, and at a 15-Ohm load this factor will be about 3. If this is the main reason to the better bas control, I don´t know, but i theory it should, and ceartenly the bas is far more controled with this X-feedback.
Capacitors in the signalpath and X-feedback loop.
A great deal of concearn (read audiophile anxiety) is caused by the use of capacitors in the signalpath or in feedback loops.
When I added the ZV4 buffer to the X-SOZ I also added some highquality filmcaps in the signalpath, and despite of those, the buffer turned out to degrade the sound compared to the X-SOZ without buffer. But more important, this X-SOZ had some absolutely nonexotic 220uF electrolytics in the X-feedback loop. So the nonexotic high value electrolytics dosn´t seem to degrade the sound as much as even the best constructed buffer, at least not when placed in the X-feedback loop. The X-feedback wil minimise the noise and distortion from thise electrolytic capacitors since they are inside the loop, but if this alone can explain the sonic abcence of these caps, I don´t know.
Another thing is that the outputcaps in the X-BSOZ are inside the X-feedback wich will reduce theier sonic precense, in BSOZ they were directly in the signal path.
Nelson emphasises the use of simple circiuts with few gainstages, some times also if the price paid is electrolytic capasitors in the signalpath or in the feedbackloops, and he also points out, that his designs are much more a question of construction rather than exotic parts.
I really agree with him.
Critical values
C101 and C102 determine the low-end frequency roll off, both in the X-loop and at the output at the X-BSOZ.
At the X-SOZ it is a little different, C201 and C202 determines the lowend roll off only in the X-loop and not at the output, since the output is before the caps. The input signal at the X-SOZ will be lowered by the negative X-feedback, but as the lower end is “rolled off” at frequencies lower than that. So these low frequencies will pass the gainstage in open loop gain, and since there is no cap at the output, it will go directly to the speakers unaffected by the X-feedback. Avoiding any raise in the low-end response under 4 Hz, the preamp must start rolling off at about 2-4 Hz. The low-end response is formed by the relationship between C101/102 and C201/202.
R208 and R209 must be as small as possible, but still able to reject parasitic oscillation, a relatievly small change in value do have a significant influence on the high frequency roll off, if you raise the value, the point of roll off will go down.
Tests
Because I have to work for living (sort of freelancer), I haven´t had so much time lately to describe my experiences from the last testsetup. But it is verry clear to me that the addition of the X to both the BSOZ and the SOZ is remarkable positive in any acpects of the sonic performance, and the best result is most certainly without the buffer.
The sound is more controlled than ever, and it still have this easy and almost scilent flow, and when nessesary the dynamicks taks over. The soundstage is deeper, higher and wider.
I am shure that this X coud be applied in different ways, but this is how I did it until now, and Nelson and others are more than welcome to make their comments and suggestions.
I have attached the testfiles in pdf and scematics in pdf / SIMetrix simulator formats.
Thanks everyone for this nice thread on the X-SOZ.
:) |
|
|
| Nelson Pass |
Nice piece of work, Henrik.
:cool: |
|
|
| eLarson |
Isn't the original Zen Daughter-in-Law linestage pretty "X-ish" already?
It doesn't feature a folded cascode (mentioned in the patent), but it already has coupling between the noninverting ports of two identical amplifiers.
Maybe the question should be: How important is the folded cascode idea to the Super Symmetry concept?
Erik |
|
|
| Nelson Pass |
| Not important, but X does require the symmetric feedback. |
|
|
| eLarson |
I was under the impression that the feedback from the drain of the output section back to the gate of the input section was just to allow the distortion products of the output section to be communicated through to the other side. (I'm thinking of the drawing from the cover sheet of the Super Symmetry patent.)
That is why I was left scratching my head about the need for a drain to gate feedback path when all there is in a Bride-of-Son-of-Zen are two common-source amps tied together: there is no second transistor to worry about distortion products from.
See what I mean?
Erik |
|
|
| Henrik |
Thanks Nelson, this means a lot to me.
Erik
The BSOZ was born diffrential, and thus both mosfets reflects each others work. But the X is as Nelson said a symmetricl local feedback, wich more or less cancels the noice and distortion from the mosfets. The center for this communication is the diffrential behaviour of the mosfets. It is hard to get, an still, when I am tierd in my head I don´t quite get it.
It is like walking bacwards. |
|
|
| stefanobilliani |
Thank You Henrik
:) |
|
|
| Henrik |
Stefano
You are more than welcome, You were actually the one to start all this, GREAT!
:yummy: |
|
|
| Nelson Pass |
If all you have is a diff pair biased by a current source,
and there is no other source of noise and distortion (including
the load), then the feedback does not lower the distortion;
it only controls the gain and output impedance.
This is because any AC current flowing through the diff pair
is already identical.
:bigeyes: |
|
|
| eLarson |
Right... which is why I was wondering about the need for the feedback path.
I thought I had gotten it. But I apparently didn't quite "get it" all.
Erik |
|
|
| Nelson Pass |
But do you get it now?
:confused: |
|
|
| Henrik |
I for shure need to study this once more!
Nelson, what You are saying is, that it is only the distored and noised part of an amplified signal that will be cancled through the X-feedback? |
|
|
| eLarson |
I believe so. If not slap me upside the head. (But gently... I have a cold. ::sniff::)
How's this?
The drain-to-gate feedback path as shown in Henrik's circuit above provides the normal benefits of feedback (lower output impedence, for instance) even if you don't have a subsequent gain stage from which you are trying to collect distortion products to feed to the mirror image.
E. |
|
|
| Nelson Pass |
Yes. Anything but the diff pair itself. There, we only
get simple cancellation from matching. |
|
|
| eLarson |
Thanks for your kind patience, Nelson. :)
Erik |
|
|
| Selfmade |
Hello Henrik,
it is really a very interesting thread.
Now a maybe silly question.
Why not using an active current source like the Aleph CS instead of the resitors? A ZEN V4/XSOZ without input puffer?
regards
Günter |
|
|
| Henrik |
Günter
Not at all a silly question!
I already have builded the SOZ so it is relativley easy to change it into a X-SOZ. But I also like the passive currentsources, although I haven´t heard the SOZ with an active currentsource. For me this simplicity is the SOZ, and I am quite happy with the efficiency since I only needs the 8 watt output. Also Gray Rollins "has ben there" in a thread about this issue, and his conclusion was in favor of the resistors by sonically reasons.
If You need the higer effiency the active currensources provides, they are easy to add, and I am shure they will sound nice.
The thread by Gray Rollins was named something like "Passive versus active currentsources for the SOZ".
By the way, I woun´t be able to answer any questions before wensday next week, until then I am travelling in Sweeden. |
|
|
| Ian Macmillan |
I too have wondered about this, though I suspect that Henrik was following a slightly different path - one of true minimalism. I note that the distortion of Henrik's latest incarnation of the X-SOZ exhibits fairly high distortion at the 10v pk to pk output shown (about 2.8%) thought of course this may not matter in practice. Personally I am more concerned at the very low efficiency of this circuit, particularly if scaled up for increased power and for this reason would consider active current sources for both loads and diff pair tail.
Henrik, I too would like to add my thanks for taking this thread in a new and interesting direction, namely one of optimising the pre-power amp interface. This is something that is much easier for us DIYers than it is for commercial products that must cater for a variety of equipment combinations. Even so, I haven't seem much in this vein on the forum.
Ian. |
|
|
| Variac |
Yep, it is kind of a philosophical issue.If you add an actice current source to an X-soz in the forest, is it still an Xsoz? (referring to the koan "if a tree falls in the forest and no one hears it etc.....)
The SOZ has traditioally been as simple as possible, always using resistors instead of active stuff. The lack of efficiency is actually another historical characteristic of SOZ.Otherwise, why not just do an XAleph?
I do think the pre/power interface discussion is really exciting!!
Thats where the X-soz is going to be made to shine. |
|
|
| Nelson Pass |
Of course the Zen and SOZ were made as simple as
possible partly because that creates an attractive
starting point for DIYers. After we beat that design
up enough, we start adding stuff...
The XSOZ can be made pretty efficient with a single
polarity supply, a constant current source biasing the
diff pair, and Aleph current sources. In my test amps,
I get about 30%, meaning that a 30 watt output will
idle at 100 watts. A lot better than the 6% or so of
the original.
Still, resistive loading has it's charms, which is why I still
enjoy the lightbulb zen. And it has a nice warm glow.
:idea: |
|
|
| Henrik |
I am not quite gone yet, so:
Ian
If I needed more power/effiency even I would add some active currensources, so on that one i totally agre with You, but scouts honner, I am playing very loud music with this amp, and never at full volume, pleanty of headroom, so I don´t need more, but I am also shure that others wants more power.
I also think, that it is very easy to add those active currensource, and as I said before, I do think they also will sound good, I am not that religious in those matters.
I am glad that You like the pre- and power-amp "interface", it was nesesary to do something like this to avoide buffers, though it was very educational for me to work with them and You.
Impedances like those is what DIYers can do, but it is only the the X-SOZ that will suffer when connected to an other sources, the X-BSOZ will fit anywhere, even to headphones.
Thanks for Your support along. |
|
|
| Henrik |
Almost gone, but:
I really have enjoyed making this as simple as possible, and still I havn´t got that feeling, that I have to "start adding stuff" as Nelson does.
So I will stick to the true spirit of ZEN. :angel: |
|
|
| jh6you |
| quote: | Still, resistive loading has it's charms, which is why I still
enjoy the lightbulb zen. And it has a nice warm glow. |
Is your lightbulb zen unbalanced or balanced? |
|
|
| jh6you |
| There are three. One unbalanced. Two balanced. |
|
|
| Nelson Pass |
I run them as balanced monos, with 6 300 watt bulbs
per channel. This is bigger than the published project,
although I did mention it, and runs at the edge of hotness.
Another possibility that bubbles to the surface is to combine
the active current source and resistive load in parallel. This
way we could look at some of the charm of the resistive
bias, say at 16+ ohms, and still use an Aleph source to
raise the efficiency.
I'm sure that one of you bright bulbs could cobble this
together for the thread.
:idea: |
|
|
| jh6you |
Thanks, Nelsen Pass.
I will try to be a bright bulb. But, my head is still in bed (7:30AM). |
|
|
| jh6you |
| quote: | | I will try to be a bright bulb. |
Idiot.
| quote: | The XSOZ can be made pretty efficient with a single
polarity supply, a constant current source biasing the
diff pair, and Aleph current sources. In my test amps,
I get about 30%, meaning that a 30 watt output will
idle at 100 watts. |
What about this...? |
|
|
| mrothacher |
It looks nice, but I think this would be more akin to a balanced ZV2. If I'm not mistaken (and that's a certain possibility), in order to X-ify it you'll need to pin a tail on the long-tailed pair in the form of a normal CCS.
Mike |
|
|
| lieven |
henrik, ever thought of making an additional tail current source (connecting r16 and r17 to negative supply).
So q4 and q8 floating between 2 active current sources and perhaps double feedback loops (kind of r36 and r38 on negative side).
Curious wath the effect could be.
In that way we could play with different values for feedback from pos and neg side (if there's a difference, it must be caused by q4 and q8) to add to the diff. pair.
just an idea, perhaps not a good one |
|
|
| till |
| am i stupid - i can´t see the resisistive load, the bright bulb? |
|
|
| jh6you |
Would this work?
"The XSOZ can be made pretty efficient with a single
polarity supply, a constant current source biasing the
diff pair, and Aleph current sources. In my test amps,
I get about 30%, meaning that a 30 watt output will
idle at 100 watts." |
|
|
| gnomus |
Henrik,
Geat work by Henrik and others in the thread. Mr. Pass has a nack for giving the right information at critical points.
Are you able to measure how much feedback is being used in your X-BOZ and X-SOZ?
I have seen feedback commonly given in db.
How can it be calculated?
Thanks |
|
|
| lieven |
general idea behind previously mentioned post: 'balanced' feedback-loop on an diff. pair.
Could be also a way for better tracking current draw of load.
Remember, just an idea, perhaps not even worth thinking about. |
|
|
| Nelson Pass |
You need some input resistance for this to work the
way you want. This could be provided by the output
impedance of whatever will be driving it.
A pair of .5 resistors at ground is a little low for the X
effect as it is as low as the transconductance of the
gain devices, and only a percentage of the distortion
will be nulled. A high impedance current source, by
contrast, will give the whole effect.
Nevertheless, it is now technically an X amp.
To measure the amount of feedback being applied to
this circuit, drive it with a load and a source having very
low impedance and measure the gain. Then drive it with
the finite impedance you will be using in real operation
and compare the two figures.
Divide the lower figure into the greater, find the log10
of this and multiply by 20 and you have the decibels.
:wiz: |
|
|
| leiade |
Hi,
Once in a mag called TAA they often say, make it like the the big boys do it.
So... I make a pair of full range quasi ribbon loudspeakers I don't know that the bass panel was ribbon so I did it my way and... it works even better than the original just my own design.
I bought a MLSSA and som other measurement equpiment and started to measure, one thing is obvious you can more than you happens to know, keep o trackin' |
|
|
| jh6you |
Thanks, Nelson Pass.
Every word seems like a thousand to me.
Being really encouraged, I will take this X as my new project.
Before I light its fire, however I¡¯d like to come up with a further
developed circuit, with a brief description.
I'm with you, leiade, about the philosophy.
By the way, I wish to see someone tackling this:
"Another possibility that bubbles to the surface is to combine
the active current source and resistive load in parallel. This
way we could look at some of the charm of the resistive
bias, say at 16+ ohms, and still use an Aleph source to
raise the efficiency."
Best Regards |
|
|
| jh6you |
Updated Revision 1
An identification name, XSOZV2, is given to the circuit.
If this ID name gives you a confusion, I will change it at your request.
A brief description about the circuit is as follows:
- MOSFETs are to be properly matched.
- Bias current is about 1.32A.
- R20 is set to 3 Ohms, which could provide the tail of current source reasonably stiff.
- The input resistor R2 is set to 1k Ohms. This could be adjusted later, if required by the source output.
- The drain voltage of Q1 is set to +18V for an AC voltage swing of about +/- 14V. These closely depend on the supply DC voltage and the size of the tail resister.
- R16 is set to 1.5K for a 30% AC fraction. This could be adjusted later for a proper speaker damping.
I am thinking about a system set unbalanced up to the preamp input, and balanced behind it.
JH |
|
|
| stefanobilliani |
| jh6you ... looks really nice ! :cheerful: |
|
|
| Selfmade |
Is C2 (10µF) really necessary?
Why not using the input-circuit of Henrik? It looks more SOZ like.
How much output power do you expect from the schematic above?
regards
Günter |
|
|
| Variac |
Nelson wrote:
-Of course the Zen and SOZ were made as simple as possible -partly because that creates an attractive starting point for DIYers.
Well.... you are RIGHT! that's what got me into this crazy world
-In my test amps, I get about 30%, meaning that a 30 watt
-output will idle at 100 watts.
This really does appeal-house wiring becomes an issue if higher power is needed using the resistive loading!
I've studyed my motives and realized that I am prejudiced in that what I have accomplished to this point is mounting 32, 50 watt power resistors to heatsinks! Thanks to Henric's wonderful test diagram I can see how to make the various versions. He MUST make a diagram if any other scheme becomes preferred!
I have now decided that I can do my SOZ and later add the constant current sources with a big switch to change from one to the other depending on my mood.
Go for it!!, but don't forget the preamp issue! |
|
|
| lieven |
| sustained bias plateau on xsoz? |
|
|
| jh6you |
XSOZV2 R1- Stefanobilliani, I am happy with your praise.
- Selfmade, C: Yes or no. It seems up to the personal preference.
- Selfmade, Power: I hope 25W/ch with 8-ohm load.
- Selfmade, Input circuit: I am notoriously ignorant about the input buffer. Inned more study.
- Lieven, Sustained bias plateau of xsoz?: Probably.
Sorry for a short writing. Whew…busy day with my profession and
occasionally busy night on duty for wife’s ecstasy… :bigeyes: |
|
|
| Nelson Pass |
C2 is essential as a practical matter.
Driving this circuit single-ended is not recommended, but
it will work, sort of. Optimally the diff pair is biased by a
current source. You can do this with just a couple of volts
loss, so it's practical.
:cool: |
|
|
| jh6you |
XSOZV2 R1
Yes... Nelson Pass, C2 is essential for the practical feedback.
By the way, Nelson, XSOZV2 is one of your properties.
It looks like a simple circuit, but the following three are there all in one.- Zen V2
- Pass Patent #
- Pass Patent #
Meaning I’m afraid of the lawyers (three for each) supporting you, I presume| quote: | | as a matter of policy we do not concern ourselves with DIY efforts |
is still valid.
Otherwise, it is too heavy to me because I am already wanted, $$$, by 6 girls.
:cool: |
|
|
| Nelson Pass |
I'm sure it will amuse you to know that I have 5
attorneys. Whenever somebody wants me to do
some work, I charge the same rate as the most
expensive of them.
;) |
|
|
| jh6you |
Yeah, it is amusing enough. :usd:| quote: | Driving this circuit single-ended is not recommended, but
it will work, sort of. |
I meant converting unbalanced signal to balance thru (X)SOBOZ
and feeding the balanced signal to XSOZV2. Now quite acceptable...? Please...don't say no. |
|
|
| stefanobilliani |
Five mosfets par channel :lickface: if we bias the diff pair with a current source.
Heat sink(S) !
:hot: |
|
|
| jh6you |
BOSOZ + XSOZV2
The following BOSOZ is to convert the unbalanced signal to
balance. I need this because of all my low-end source gears
having only unbalanced output.
I expect this BOSOZ would work properly.
:) |
|
|
| grataku |
jh6you
This topic has been popping up about once every four months for the past 2 years so you maybe aware of it. You shouldn't expect miracles from this circuit. The BOSOZ is not a great unbal-to-bal converter, actually a user has already posted simulations of the circuit that quite clearly show different voltage levels from the hot and the cold outputs when the cold input is unused, or grounded, in SE operation. The same guy has already posted a version with current source which performs better "instrumentally". If you want a Pass original circuit you could go for the AlephP that has a full complement of current sources. |
|
|
| jh6you |
Grataku
Well noted your comment. If my post escalated a monotony and boredom, I am very sorry for it. All right, I will trace back the old issues carefully for a study.
Please note that with the post I am not arguing BOSOZ is a miracle or not. I have no ability or know-how to do such big thing. My interest is in a small thing. I would like to know which preamp is having acceptable function as a converter for unbalanced input and balanced output. I just wanted to know whether BOSOZ is acceptable or not. The main reason why I attached the circuit was not to be proud of it, but for an easier communication.
I do not want Aleph P. I want a simpler one. What do you recommend for my purpose?
Please come back. |
|
|
| Nelson Pass |
As you shrink the value of R15 on your circuit, it
will become a better unbalanced to balanced converter,
and you only need deal with the additional gain
produced.
Or you can use a current source. Or you can increase
the negative supply voltage and use greater resistor
values biasing the diff pair. |
|
|
| jh6you |
Nelson, please let me walk out of the issue about BOSOZ after this.
Your three alternatives explain me that the current source is the key.
I would try all three someday. Today, however, I am engaged to
the 1st alt only. Here is my new plan.- DC supply kept of +/-30.
- Bias current reduced down to ~20mA.
- The values of tail resistors increased.
- The value of R15 decreased for a convincingly stiff current source.
- The high gain attenuated at the input with a voltage divider.
- The volume planned at the output.
Question: If I shrink the value of R15 further down to 47 ohms,
would it be even better? If yes, I should also change the input voltage
divider from 10k/4.75k to 12k/2k. Is the 2k, if connected to the ground,
too small? Thanks.
JH |
|
|
| Nelson Pass |
This will make it better in terms of the output
balancing, but I think at some point you have to
look at why you are trying to achieve a perfectly
equal output amplitude on both halves. If your
amplifier has an ordinary balanced input, it is not
of any particular value in terms of improving the
noise rejection, since the noise will be identical
on both sides regardless.
Unless you go to a constant current source for this,
you will continually be trading off some other
performance parameter such as noise and distortion
against the degree of output level matching.
If you want to continue to use resistors for biasing, you
probably have made as much improvement as you
need or want. |
|
|
| jh6you |
When I evaluate deflections and stresses of ships for my
profession, the first thing I do is checking balanced force at the
one reaction point whether the reaction force is zero. It is to be
zero. If it is nonzero, means unbalanced and there are always
certain degrees of error in the results.
All right, I will go for the constant current source tails,
based on your advice and to have the strong foundation.
I will come back for your positive feedback. Takk! |
|
|
| Henrik |
You shoud check the relation between outputimpedance of your BSOZ and the inputimpedance of your X-SOZ. I havent simulated or calculated your suggestions, but i beleve that you will get a highend roll off at -1db at around 20kHz, wich is too low in my oppinion.
I coud be wrong, but check it! |
|
|
|
