GoodSoundClub - Romy the Cat's Audio Site

Melquiades Amplifier
Topic: AC line stabilization

Page 1 of 1 (20 items)

Posted by N-set on 09-22-2012

This is something I've been chewing on for quite some time and finally decided
to make it happen: a serious Stax headphone amplifier.
I hope I can use this space as a sort of a blog-report on the progress...if not feel free to remove.

I've been "forced" to listen on headphones due to real estate constraints and it
seems those constraints will stay for few more years. At the moment
I'm using Stax Omega 2 MkI powered by a solid state amp Stax 717. I've always felt
it's barely up to the task of driving those little monsters and, apart from many other factors,
it sounds anemic. Adding Romy's EAR834 down the chain openned the whole new universe
(I could not listen to EMT155st for more than 10min...even after heroically changing Elko's to Elna's)
and 717 is now the weakest link.
I'm also mounting a second combo on my EMT930--SME3012R+whatever interesting cart I find,
which, if successfull, would too demand
more advanced amplification up the chain. Hence the need for a new amp. The most praised (at least by some groups
on head-fi.org and head-case.com) amp comes from a certain Stax mafia and is called Blue Hawaii.
I've never heard it, but for $5000 and 2yrs waiting time I can cook something myself...apart from that
looking at BH's schematics makes my eyes bleed...so I tried to design myself.

I must admit I have no idea how to properly power electrostatic headphones. I'm improvising from a few obvious facts:
*el-stats are capacitive loads (O2 is 110pF), very current hungry
*this current is better delivered fast: high slew rate
The very nature of el-stats forces the design to be Pi-Pi,avoiding the transformers. Another
axiom was to make it DC coupled so that there is nothing but tubes in the chain.
The inspiration comes from Milq, Axiom and Lynn Olsons Karna (reading Olson on PiPi amps has been very educative).
I want it to be flexible to some extent to try
different output tubes: EL34 (std. "Stax tube"), 2A3, 300B, 842, 801A
After endless hours of simulations here is the v1 design:

Staxquiades.v1.bmp

Milq's driver with it's bias but PiPi'ed, DC coupled with stacked PS (inspired by Chimeralabs' Axiom), the simplicity of the
signal chain...but a complicated PS. The most cruicial is the biasing of the driver as it sets the operaiting point of the output.
Hence I want to be able to change this bias and for each tube individually.

I ended up with a design which is a monster, a mastodont. Might be the most ridiculous amp to
power...mere headphones...well, maybe...one never knows until one
tries. A good point is that changing the output chokes to PP transformers would make it a medium-power speaker amp,
which may be usefull...somehere.

I'm slowly starting to get the parts: Lundahl chokes, PSU transormers, big caps, etc.
Cheers,
N-set

Posted by Paul S on 09-22-2012

No doubt you have reverse-engineered the amp's output plots from the Stax' input requirements/limitations; right?

The Chimera Axiom is a tank circuit, with a zillion taps off a giant, complex tranny. Looks like your sections are ?stepped in terms of values, but physically discrete before ground?

Is that bypassed trannie amorphous? If yes, why?

Why the AO2, etc. instead of SS?

Best regards,
Paul S

Posted by N-set on 09-22-2012

Paul,

I know the capacitance -- this is an analog of impedance of ordinary speakers.
I also compare to known serious designs, like mentioned Blue Hawaii.

The multiple taps on PS iron is to accomodate different output tubes.

The "bypassed transformers" are in fact PP chokes. At this stage, the one loading the driver
is a off-shelf steel Lundahl. The output one is more complicated: I have huge amorphous cores
whic I plan to use. Why amorphous? And why people use amorphous? "Speed, clarity, blahblahblah"
Here I need a fast, high freq. core because I need as tight coupling between both halves
of the choke as possible.

Why not HV Zener? It's said to be noiser and the noise is nasty (Olson writes about that on his pages) and this diode
sits in the driver grid. VR in turn has a few mV of a broadband noise.

Cheers,
Nset

Posted by N-set on 09-23-2012

N-set wrote:
Another
axiom was to make it DC coupled so that there is nothing but tubes in the chain.

Of course I was not very precise here: not pictured but a 1:1+1 phase splitting input transformer
is inevitable...still I believe it is easier to wind a line level transformer than an output one
(here the voltage swing at the output, ~1.4kVpp, rather than a current would be the limitting factor)

Posted by N-set on 09-23-2012
fiogf49gjkf0d ...and the heaters: CCS chain on 6E5P + AC on the output. The Stax pro bias (580V) is coppied form the Auridux amp http://www.dddac.de/tp08.htm

Posted by Romy the Cat on 09-24-2012

It is unquestionably an ambiguous project. The topmost question that I would ask myself after this project would be the following: can Milq-biased 6E5P to be as good (sound-wise) DC driver as some vintage direct heated tubes. I did ask thin question myself after I implement my MF DSET but I do not have the answer.

In regard the project as a whole. I wonder if it worth to built up something as elaborate as it proposed just to drive headphones? I do not want to belittle headphones. I like headphones myself but I can’t not lose the feeling that headphones do not need elaborate or sophisticated amplification. I had 3-4 dedicated headphone amps myself in past and have built one on my own. None of them sound good and I very mach appreciate to drive headphones from some kind of chip-based headphone jacks. Perhaps it is me but I am very forgiving to the quality of headphone drive and I would question if it worth to create a federal case about headphones.

The Cat

Posted by N-set on 09-24-2012

Romy the Cat wrote:
It is unquestionably an ambiguous project. The topmost question that I would ask myself after this project would be the following: can Milq-biased 6E5P to be as good (sound-wise) DC driver as some vintage direct heated tubes. I did ask thin question myself after I implement my MF DSET but I do not have the answer.

I do not know and I don't know any-trusthworthy-body who may know. Having very very few (if any) my own reference points I proceed
by the method of KWS -- Known Working Solutions. The design is elastic, I've spent great time simulating and then specifying PSU's,
so putting a DHT driver should not be a problem.

Romy the Cat wrote:
In regard the project as a whole. I wonder if it worth to built up something as elaborate as it proposed just to drive headphones? I do not want to belittle headphones. I like headphones myself but I can’t not lose the feeling that headphones do not need elaborate or sophisticated amplification. I had 3-4 dedicated headphone amps myself in past and have built one on my own. None of them sound good and I very mach appreciate to drive headphones from some kind of chip-based headphone jacks. Perhaps it is me but I am very forgiving to the quality of headphone drive and I would question if it worth to create a federal case about headphones.

I've been wrestling with the same question for 2-3 years myself without any better solution than to try it myself.
I have now a SS Stax amp, supposedly a very good one, but with the increasing source sophistication I feel it stands in the way.
Few more points

* I do not like listening on headphones but due to real estate constraints I have no other choice in the forseenable future, so the investment
comes out of the real need
* Staxes, esp. TOTL such as Omegas are quite outstanding and special headphones,
supposedly very amp sensitive and demanding; they may, or may not, be worth the effort--the only solution I've reached is to try;
people do use quite sophisticated amplification (mostly EL34 surrounded by tons of SS currnt sources, sinks, servos and other electronic engneering bullshit) and claim it does open new doors (Sennheiser's flagship Orpheus,the mentioned Blue Hawaii amp, Stax T2, SET's with 1:1+1 output trans,etc)...but I've never seen a no-compromise design, without all that EE bullshit and all my inquiries about the subject never got any reasonable answer apart from some sort of a fear and a sheep crowd instinct of following the old Stax designers and their modern analogs like Kevin Gilmore (the EE behind blue hawaii and Stax T2 revision...with 250 SS pieces in the design...). I think I'm on a +/- terra incognita here...
* As I mentioned, changing the output choke to an output transformer will easily transform the Staxquiades into a speaker amp with quite a few Watts
of output power, so the project can be recycled and incorporated in e.g. a dedicated multi channel amplification
* I do have some serious concerns with respect to my design, the LF stability beeing the most important one (before we even approach talking about the Sound...if any); you may rememebr I was asking you about the stability and reliabilty of 6E5, as the driver is the most cruicial element here
(for example it provides the bias to the output via the voltage drop on the RDC of the interstage choke). Again, the sims show it can work but I think I've reached the point the sims will give me no further insight

Posted by Romy the Cat on 09-25-2012

N-set wrote:
I do have some serious concerns with respect to my design, the LF stability beeing the most important one (before we even approach talking about the Sound...if any); you may rememebr I was asking you about the stability and reliabilty of 6E5, as the driver is the most cruicial element here
(for example it provides the bias to the output via the voltage drop on the RDC of the interstage choke). Again, the sims show it can work but I think I've reached the point the sims will give me no further insight

Well, in those DC-couple applications the stability of the driver is very important and bias of output stage sits on a head of the plate current of the driver. In your case you a choke between but it is essentially the same thing. I do not have problem with 6E5P and I do run my MF DSET in exactly the same application, driving way more expensive and hard to get tube then 300B. The 6E5P might have low plate current from start (as any other Russian tube) and I did come across 2-3 tube that did not give me enough current and I trashed them but if it give the right current at nominal bias then it will hold the current very stable. At least I did not have any problems with 6E5P/6E6P that I accepted. If you want to be even more confident then get 6E6P-DR. They are even more reliable then 6E5P, particularly at low voltages you use them at.

Posted by N-set on 09-25-2012

Thank you Romy! Actually I've realized that when I wrote it I rather ment dynamical stability then
frying the outputs. But in proper PP the even order harmonics are strongly surpressed, so
there should be no or very little change of the output OP due to the amplified signal.

Romy the Cat wrote:
driving way more expensive and hard to get tube then 300B

I'd actually never consider 300B--I don't like it's "smokey" character (also I've heard only few specimens
with it and long time ago). But I was looking for a triode able to give substancial voltage swing,
comparable with that of El34 triode connected and surprisingly 300B gives it (on paper).
No idea of the quality of this quantity, but just for a comaprison I've included 300B in my list.
I'd actually be much more interested in hearing 842. Although it gives only a bit more than a half of
300B/EL34 swing (800Vpp vs. 1400Vpp), the simulated FFT spectrum (if it has any connection to the Sound...)
looks very clean.

Posted by scooter on 09-26-2012

N-set wrote:

I've been "forced" to listen on headphones due to real estate constraints and it
seems those constraints will stay for few more years.
At the moment I'm using Stax Omega 2 MkI powered by a solid state amp Stax 717. I've always felt
it's barely up to the task of driving those little monsters and, apart from many other factors,
it sounds anemic. Adding Romy's EAR834 down the chain openned the whole new universe
(I could not listen to EMT155st for more than 10min...even after heroically changing Elko's to Elna's)
and 717 is now the weakest link.
I'm also mounting a second combo on my EMT930--SME3012R+whatever interesting cart I find,
which, if successfull, would too demandmore advanced amplification up the chain. Hence the need for a new amp. The most praised (at least by some groups on head-fi.org and head-case.com) amp comes from a certain Stax mafia and is called Blue Hawaii.
I've never heard it, but for $5000 and 2yrs waiting time I can cook something myself...apart from that
looking at BH's schematics makes my eyes bleed...so I tried to design myself.

N-Set,

in Tokyo there are a virtually unlimited number of headphone combinations to try out here as you might know. I have read all the hype about those Headphones and expected something special. I have listened to them in at least a dozen places with that 717 amp. A few times with different amps. Incredibly the variation in sound quality went from quite terrible to quite good. In all cases I felt fatigued / bored (whatever the word is) of listening after a few minutes.

I don't know what to attribute the variance to; key suspect is differing quality of electricity going into the 717. Could have been the source CD players I suppose. Like you, I also thought the headphones could do a lot more, although that may be biased from reading to much stuff. Fact is, after many listening sessions I never was satisfied with those headphones.

Based on this experience, I would recommend you go to some headphone "meet" or "store" and listen to these things through a few sources to get a better view on if the project is worth pursuing.

Posted by N-set on 09-26-2012

Scooter, thank you for the warnings. I already own O2 MkI for +/-2 yrs and got to know them a bit. As I said I do not like headphones in general
but I have no other choice. I own 2 more staxes: Sigma and Lambda Pro and Omega can do things others cannot. But so far in my limited exerience this depends a lot, a BIG lot on the source. With some sources O2's are "trying" to ameliorate it's shortcommings and make it listenable (my old Sony 707 CD) with others are very brutal for some reason (EMT155).

scooter wrote:
In all cases I felt fatigued / bored (whatever the word is) of listening after a few minutes.

I had exactly the same feelings when I tried to listen to my EMT930 with it's stock 155s amplifier. After 10mins I was switching the system off.
Change to DIY Pussyfied 834 did change the situation completely (although my 834 is more time on the workbench...). Quite a few times I got already completely lost and submerged in music.

scooter wrote:
I don't know what to attribute the variance to; key suspect is differing quality of electricity going into the 717. Could have been the source CD players I suppose. Like you, I also thought the headphones could do a lot more, although that may be biased from reading to much stuff.

Firs of all I've never thought that headhones can do a lot more. I'm not a headphone type of guy. But the change in the phono pre did rise my apetite.
I'd put forward the thesis that this is the source which matters in the leading order. I've never been on CanJam or any other meeting, but if you look at the pics what "level" of sources are used there, no wonder the sound is crap. Same with all that head-fi and head-case crowd. I've never seen anybody with a sophisticated source there. Not even the Stax mafia like Spritzer or KG.
So, having a +/- sophisticated source and a mediocre amp I'm wondering what is behind the curtain of poor amp. Maybe nothing....
I may try to look around and borrow a BH or Woo but I doubt that there are many of them in Spain and the owners would be willing to borrow.
But frankly I just don't get Kevin Gilmor's EE style. Maybe it all sounds good, but in the speakers amp what he uses is a well known no-go crap.
It seems that almost nobody (apart from one guy at head-case building SET's) have tried to execute a Stax descent amp....maybe not worth it?

Posted by N-set on 09-26-2012

Actually I do not have the instability issue sorted out completely...
I'm talking about circuit instability now, not the tube instability. Let us imagine all tubes are rock stable.
As I see it there are at least two major sources of instability, in the sense of changing the OP of the output tubes,
in the DC scheme I try to implement:

A) Signal-related instability.
A1) Even harmonics generated by the driver will momentarily (at the timescale of the period of the fundamental)
rebias the output stage. This effect I belive is easy to minimize. First of all by the use of a linear driver (Hagtech 6E5 curves published here look
as straight as legs of a supermodel). Linearity of the output will also help here as it will be less sesnitive to small OP variations.
For PiPi amp this effect is moreover strongly supressed by even-harmonics cancellation
A2) Cathode blocking. If the driver is cathode biased (a very natural choice here, see below) and the cathode R decoupled, a strong signal will
momentarily charge the capacitor, it will hold the voltage for the time ~1/RC, moving the driver towards cutoff and by that rising the current of the output. So I'd say cathode decoupling is a no-no.

B) Supply line instability. This I belive is much more serious.
Change of the supply line will change all the voltages in the system but with different effects as driver and output will react differently.
The standard thing is to use cathode biased driver as it will be self-balancing. This is what's done e.g in the Axiom

http://www.chimeralabs.com/images/axiom.sch.jpg (with diodes instead of R)

but is really an old approach: http://dev.emcelettronica.com/amplifiers-builder-guide-10-watts-direct-coupled

I'm perhaps shooting my own foot badly trying to put the driver on the fixed bias as then the supply line variation is greatly
increased (!) and I go very much agoinst the current. Few ad hoc rescue scenarios:

* That the driver should be in a high-enough-current point to accomodate 10% line variation is clear (hence my 6E5 current is highish).
But one can hope to minimize the OP change effect in the output too, by using i) linear tube and ii) in the most linear region; the last is of course high current, so if the output is close to its max it can be fried. With headphones one does not need in theory much more current than to drive the C at a given f and V. My crazy high output satge current comes in the 1st order from trying to get a big A1 voltage swing, but the secondary effect is that the tube is far away from the cutoff but yet safe enough to accomodate the 10% line variation, so at lest in terms of survival and not cutting off it should be OK (the sims say so). If this will work sonically...no idea.

* Driver power supply passive. Huge cap and some resistance will create a mechanism to smoothout the line variations. And this is perhaps my main hope. Huge, HV caps were not available to the fathers of DC coupling, so the only way for them to stabilize the circuit was cathode bias.
I've put on paper 10.000uF cap and some effective 200R from the line side, giving the line time constant of 0.5s in the driver PS. Will this be enough...no idea

* Driver power supply active. Active regulation is another option. There are voices that particularly shunt regulators are sonically least intrusive or even beneficial. Adding a 6E5 as a shunt shoud not be a big problem.

* Line stabilization...this is a very dangerous territory as the "mad electricity" thread shows. It's very difficult to imagine what makes all those line regenerators audible even given the Romy-style PS with a lot of decoupling. THe one who resolves it will get an audio nobel!
One ad hoc idea is that as those regenerators seem to carry some sort of switching power supplies (loooking at PP2000 size it's way too small to have a linear supply of thet power), perhaps line dirt uses the switching freq. as the carier and gets spread all around and then is picked by some high imp. (esp. grid) circuits inside the amp....ony a naive theory, but checking the line before, after and the air with a spectrum analyzer can shed some light. I'm wondering what happens if one forgts slim, compact, attractive shape and tries to design a regenerator that would not use switching...

Summarizing, there is a serious point regarding the stability (esp. short-time scale as it's most audible) of my circuit. I keep thinking....

Posted by Paul S on 09-26-2012
fiogf49gjkf0d N-set, are you pretty sure that the "direct drive" tube amp need not be PP? Does the Stax amp have +/- typical RC output? Would it perhaps make more sense to "update" the Stax amp? ATH-50 cans would sure make your life easier... Best regards, Paul S

Posted by N-set on 09-26-2012

Paul S wrote:
N-set, are you pretty sure that the "direct drive" tube amp need not be PP? Does the Stax amp have +/- typical RC output? Would it perhaps make more sense to "update" the Stax amp?

ATH-50 cans would sure make your life easier...

Best regards,
Paul S

"Need to be PP" you ment? Well, yes, by the very nature of elctrostatic transducer.

Stax amp is some SS crap. ATH-50...Paul, please...table radio would make our lives easier too

Posted by N-set on 09-27-2012

N-set wrote:

* Line stabilization...

I took a quick look at the industrial stabilizers (Purepower and alike look quite suspicious to me given that they carry AC/DC and DC/AC converters
and claim 2kW or whatever with a very slim and compact case...). They all look like a small fridge, OK, I'd expect that if it is to carry a linear supply inside.
To my surprise stabilization principle is very simple and funny and probably dates back to the early 20th century: a servo motor
controlling an autoformwer and a buck-boost transformer. Very simple!

http://www.ashleyedison.com/voltage-stabilizer-design.htm

Of course this is only a stabilizer: even with fast and precise modern motors,
it'll only eliminate the low freq. variations, and will do nothing
to the shape of the AC. But still it can be a basis of a more sophisticated
conditioner (as I understand one of the the problems is to make the RF filtering sound-invisible)

Posted by Romy the Cat on 09-27-2012

You need to be very carefully with them as they might be anything. I very much doubt that we need any AC line dedicated stabilizers; even though I do feel that it is good to have more or less stable line. The mechanical line stabilizers might be a bit too sensitive. If you have a voltage spikes for a few does of person you do not want the unit to react to it immodestly. The AC regulation has to be very gentile and you do not want to yank voltage too much by switching the transformer turns. Also, the transformers that those units have are usually very bad for sound. I did use a number of manual stabilizers and they all were sub-acceptable. There is another twist. Most of the mechanical line stabilizers has common mode and differential filters that made not for the “injected” voltage but for full load. Most of those filters not good, I mean were not good for sound. I agree that the mechanical solution is very simple and elegant, this is basically taping the transformer but the valuable transformers that I used did not sound right. I still have a few of them in basement and I do not see any use of them outside of my testing lab.

Posted by N-set on 09-27-2012

Romy the Cat wrote:
I very much doubt that we need any AC line dedicated stabilizers; even though I do feel that it is good to have more or less stable line.

Well, this is a very general statement, perhaps too general? I'd say that *purely theoretically* DC coupled designs cry for a line stabilization, but what's on paper and in the real life can be two very different things...

Romy the Cat wrote:
The mechanical line stabilizers might be a bit too sensitive. If you have a voltage spikes for a few does of person you do not want the unit to react to it immodestly. The AC regulation has to be very gentile and you do not want to yank voltage too much by switching the transformer turns. Also, the transformers that those units have are usually very bad for sound. I did use a number of manual stabilizers and they all were sub-acceptable. There is another twist. Most of the mechanical line stabilizers has common mode and differential filters that made not for the “injected” voltage but for full load. Most of those filters not good, I mean were not good for sound. I agree that the mechanical solution is very simple and elegant, this is basically taping the transformer but the valuable transformers that I used did not sound right. I still have a few of them in basement and I do not see any use of them outside of my testing lab.

I think those points can be addressed...with a DIY design of course as I do not believe the industry is able to produce anything reasonable these days.
The speed of reaction can be easily controlled in the servo by some sort of integrating circuit with variable int. time.
Common mode and diff mode can be left aside. The autoformer and the buck-boost transf. can be made custom to good specs.
Then putting it all together nicely would be a challenge.

Posted by Romy the Cat on 09-28-2012

Yes, in case of tube DC coupling line stabilization is desirable but it would serve only practical aspect of locking the operation point of output tube. The way HOW the line was stabilized very much might affect negatively the way how the whole amplifier sounds. So, here we have a dilemma. From one side the output tube might run out of the optimum operation point and it might impact sound negatively. From another side putting stabilization also might impact sound negatively. So how to find a good balance of negatives and positive and how to evaluate what negative impact is uglier?

I do not have the answer. I do not look for one either as I am sitting behind PP that has the line stabilized. In my house the line might run from 167V to 127V and for all intended purposes it is not too horribly impact the operation of the amp. If I was in the world where my voltage swing too widely over the course of a day and if I do not want to put AC stabilization, then I would probably implement some kind of threshold guard with one single step. Let say if the voltage go over 125V then I would hard re-tap the transformer with a powerful relay. I would certainly avoid a taper that all time runs across the wires of the transformer and constancy adjusts voltage. I do not insist that it is the best way to do the things but my experience with variac was not so encouraging.

Posted by N-set on 09-28-2012

Romy the Cat wrote:
I would certainly avoid a taper that all time runs across the wires of the transformer and constancy adjusts voltage. I do not insist that it is the best way to do the things but my experience with variac was not so encouraging.

And you are probably right--maintaining the high quality of the wiper-winding contact under 1kW+ load is difficult.
But there seems to be a way around it--I've found a somewhat cooler idea-- the magnetic induction AVR:

http://www.ustpower.com/Support/Voltage_Regulator_Comparison/Mechanical_Voltage_Regulator/Magnetic_Induction_Voltage_Regulator_Operation.aspx

Quote from the same site: "...the servo induction voltage regulator also provides a very smooth output voltage compared to the discrete switching that occurs in tap changers".

Attractive! Smooth operation, no mechanical contacts to the windings, by construction there will be a high leakage between the primary and the secondary--good, less to no coupling at HF. I'm trying to see if there is any reasonable commercial unit available--DIYing sth. like that would be more difficult.

One must keep in mind that all those stabilization divagations come from my stupid desire to combine seemingly uncombinable: DC coupling + fixed bias of the driver....With cathode biasing the circuit would be self-balancing and would survive an atomic blast, but ...

Posted by N-set on 09-29-2012

Romy the Cat wrote:
It is unquestionably an ambiguous project. The topmost question that I would ask myself after this project would be the following: can Milq-biased 6E5P to be as good (sound-wise) DC driver as some vintage direct heated tubes. I did ask thin question myself after I implement my MF DSET but I do not have the answer.

N-set wrote:
The design is elastic, I've spent great time simulating and then specifying PSU's,
so putting a DHT driver should not be a problem.

After a second thought on the subject, I realized that if the Stax people are right and Omega's need around
800Vpp-1kVpp swing to sing, I need a driver with mu of at least 25 or so, esp. when power are 2A3.
Difficult to get witha DHT keeping a simple 2 stage topology.

Page 1 of 1 (20 items)