Hey all,
This is just a quick shot quesiton because I don't want to ressurect an old thread. I was looking at the MyRef approach and was wondering if the configuration can be adjusted to use any configuration of tubes as a gain stage instead of an OPAMP. Admitedly I don't know enough about the LM3875 to know if it can be used as a buffer.
Thanks,
Tim
This is just a quick shot quesiton because I don't want to ressurect an old thread. I was looking at the MyRef approach and was wondering if the configuration can be adjusted to use any configuration of tubes as a gain stage instead of an OPAMP. Admitedly I don't know enough about the LM3875 to know if it can be used as a buffer.
Thanks,
Tim
...if I had wings...
I'm afraid you're wishing too much....
My_Ref relies on a close integration of the devices used, and the techniques employed to achieve its permformance are closely realted to the internal structure of the components, so closely that Mauro even specifies the brand (National) of the LM318 (because of different compensation masks used by other manufacturers).
Moreover it is NOT a buffer but gives voltage gain and drives (quite hardly) the LM3886....
If you want to use tubes with this amp I've heard that someone used Joe Rasmussen's basic buffer in front of a my-ref with good results.
Cheers
Andrea
I'm afraid you're wishing too much....
My_Ref relies on a close integration of the devices used, and the techniques employed to achieve its permformance are closely realted to the internal structure of the components, so closely that Mauro even specifies the brand (National) of the LM318 (because of different compensation masks used by other manufacturers).
Moreover it is NOT a buffer but gives voltage gain and drives (quite hardly) the LM3886....
If you want to use tubes with this amp I've heard that someone used Joe Rasmussen's basic buffer in front of a my-ref with good results.
Cheers
Andrea
Ah, thanks for setting me straight
I'm just curious though. I know the LM318 is just there for voltage gain. However I rwad in the MyRef documents that he chose the LM318 mostly because it's a nice compramise between price and sound out of many choices for OPAMP. Also it looked like the LM3886 is acting more like a current source that has its currents varied by the voltage, ie it "senses" the dynamic load of the speakers and attempts to compensate. Is this the case? If so that was why I thought a tube would work in the first place. What I don't know is if the LM3886 is implemented in a way that HAS to take advantage of an OPAMP's characteristics.
I'm just curious though. I know the LM318 is just there for voltage gain. However I rwad in the MyRef documents that he chose the LM318 mostly because it's a nice compramise between price and sound out of many choices for OPAMP. Also it looked like the LM3886 is acting more like a current source that has its currents varied by the voltage, ie it "senses" the dynamic load of the speakers and attempts to compensate. Is this the case? If so that was why I thought a tube would work in the first place. What I don't know is if the LM3886 is implemented in a way that HAS to take advantage of an OPAMP's characteristics.
Hi Tim
I think your idea is viable but difficult. Assuming the tube works as common cathode and NFB is applied to the cathode you will obviously need some readjustment of dc levels in order to directly couple the anode to the 3886 current pump. More importantly, the dc stability of the output will now be dependent upon the tube dc stabilty which doesn't seem like a good idea.
I think your idea is viable but difficult. Assuming the tube works as common cathode and NFB is applied to the cathode you will obviously need some readjustment of dc levels in order to directly couple the anode to the 3886 current pump. More importantly, the dc stability of the output will now be dependent upon the tube dc stabilty which doesn't seem like a good idea.
indeed this is an idea I was wondering about since quite some time now. Sadly, I never had the time to develop it.
BUT, It's not goin' to be a trivial thing! This is not like sticking a tube stage in front of a gainclone, here you have to really carefully redesign the whole thing knowing what you're doing.
BTW: to refer to such a thing as to a "MyRef + Tube" would be rather improper... one should rather say "Hybrid MyRef -like amplifier" or something like that.
I guess the easy way would be to AC couple the tube "VAS" to the current pump and care about 3886 offset locally (either statically and/or with a DC servo).
Otherwise (as you said) one may take care of DC levels (likely using a bipolar supply) to keep the DC coupling and use a servo (as in the Mauro's my_Ref "Evolution") to keep DC out of the way. Of course care must be taken to keep the tubes operating within proper range...
At least for a first try, I would likely go for the AC coupling.
I would also lower the overall NFB to get less stability problems, of course at the expense of a reduced nominal DF. But IMHO that may only improve rather than worsen the results soundwise, at least as far as DF is kept no less than a few units.
Mmmhh... this last sentence may sound heretical, isn't it?
Well, let' see. Perhaps it deserves some explanation.
IMHO, a good audio amplifier should be able to put out so much current as to be able to "drive it's own output impedance". That is, the overall (closed loop) output impedance of the amplifier should be such that it can keep working "normally" (producing a linear output current, without clipping or going wild) even with a 0 ohm (short circuit) load.
(this is to make sure it will never misbehave under any condition with a real load, leading to possible overload/clipping, slew rate limiting, etc).
For a single 3886, we have ~ max 10A typical. Assuming supply rails at about +-35V, that would imply that we would have a Zmin = Vsupply/Imax = 35/10 = 3.5 ohm (!).
Given that of course we don't care to drive an actual short circuit at the output and assuming perhaps ~ 1 ohm of "stray" DC resistance (voice coil, xover, cables, etc.) to be always present in the output circuit, we would have to have a minimum Zout of about 2.5 ohm, for a "nominal" DF of ~ 3 on 8 ohm and ~ 1.5 on 4 ohm (I said "nominal" 'cause effective DF must take into account the "stray" resistance too, thus would be ~ 2 and ~ 1, respectively).
Worried about this low DF figure?
Well, that's about the maximum "real" DF you can get out of a single LM3886 anyway!
Well, at least @ +-35V; from the above, higher supply rails would make things even worse, while reducing them would allow for a smaller Zout -> higher DF... e.g. at a mere +-12V supply you could go as low as 0.2 ohm for a nominal DF of 40 on 8 ohm.
(BTW: maybe that's why some Class D with "virtually unlimited" output current capability have an higher "perceived DF" than just about any other amp out there?!
)Notice that no matter how small is the amplifier Zout, the actual DF would be limited by the "stray" resistance in the output circuit and thus would be no more than a few units anyway (unless you go to negative Zout, you can't get a DF higher than that).
This is to say that it make no sense to hunt for ridiculously low Zout... while a not-so-small, "properly chosen" Zout will provide just about the same "real" DF while assuring that the amplifier will always behave properly and correctly under any (dynamic) load condition.
You have allready been advised that the myref is not a simple 2 stage amp, but a current pump...
I was the person to use the rasmusen buffer, it is very warm sounding and too mellow to say its a good allround solution...certainly nice for some jazz, and olde worlde stuff...
I eventually matched the myref to a pedja rogic buffer... which I think is the best thing after flavoured milk.
You can search the solidstate forum, for my stripboard layouts etc... Come to think of it , I should upload the new PCB files for it...
I was the person to use the rasmusen buffer, it is very warm sounding and too mellow to say its a good allround solution...certainly nice for some jazz, and olde worlde stuff...
I eventually matched the myref to a pedja rogic buffer... which I think is the best thing after flavoured milk.
You can search the solidstate forum, for my stripboard layouts etc... Come to think of it , I should upload the new PCB files for it...
Nordic said:
Ahem... first of all, it's NOT a current pump in itself as a whole, it does USE a current pump. That's not the same...
Though the "my_Ref" proper integrates a number of different ideas and techniques to achieve certain design goals, the very basic idea behind it is that to wrap a transconductance amplifier (a voltage controlled current source, the LM3886 op-amp based "current pump" in this case) within a voltage NFB loop.
(this offers a number of advantages with respect to a more "classic" design which typically use a voltage amplifier within the loop and opens up a wide range of opportunities to control the intimate operation of the whole. But this is a long story, which have been already explained many time elsewhere and is not the case to repeat here).
Back to the "my_Ref" design, it's perfectly correct to see it as two stages, a voltage amplifier (let' call it "VAS" for short), which is based on an LM318 followed by a transconductance amplifier which is based on an LM3886 configured as "current pump". Those two stages are wrapped within an overall voltage NFB loop. That's it (what's tricky about the my_ref proper is how many details have been carefully thought out, but that's another story).
Once you have understood how it works and you know what to do, nothing prevents you to design an amplifier based on (some of) the same concepts and ideas using different technologies.
You may replace the LM3886 current pump with a different transconductance amplifier, such as e.g. one using discrete MOSFETs configured in common source... (that way basically you'd be cloning the Tim De Paravicini's Musical Fidelity A370 design).
Or you may keep the 3886 current pump and replace the LM318 "VAS" with a different gain stage using discrete components or even -why not?- tubes, as we (at least analog_sa and myself) where talking about here.
Of course you could even change both stages to end up with an "all discrete" design. The SKA GB150 may be an example of something like this (or is it using ICs somewhere? I heard a lot 'bout it, but unfortunately didn't had the chance to see its schematic diagram).
Eventually, you may even make up an all tube design... and likely you will end up with nothing but a typical OTL tube amplifier!
Yes, if you think a bit about it, the De Paravicini / Penasa idea may be seen as nothing else than a typical tube OTL amplifier "transposed" to sand state!
(N.B.: this is NOT intended to take any merit off Mauro or Tim, but just to note how breaking new ideas are often old concepts and ideas put together to new life in a somewhat different context).
UnixMan said:
Ahem... first of all, it's NOT a current pump in itself as a whole, it does USE a current pump. That's not the same...
(N.B.: this is NOT intended to take any merit off Mauro or Tim, but just to note how breaking new ideas are often old concepts and ideas put together to new life in a somewhat different context).
This is very true, you covered some things I wanted to point out.
Since Mauro showed this design to us I have tried various permutations of the Transconductance Amp + Voltage Amp design, including discrete fully symmetrical versions (with super symmetrical voltage stage, balanced input, bridged output). This has proved very successful, but the amp much more complex then the two op amp single ended version Mauro presented. Definitely not a beginners amp. A beginner can easily build a MyRef.
The wonderful thing about what Mauro did, is that he took a proven concept, and made it very simple and elegant. This is one reason why it is so special.
It is relatively simple (if you understand the amp) to duplicate (or even improve) the results we get from Mauro's great design using other parts. But you will have to "design" it. Mauro already did that work using easy to get parts. Mauro put a lot of research into what he showed us, and got a lot of us thinking. For that I at least owe him a great debt of gratitude because it really started me down the path of understanding many electronics concepts I had not really understood before. Like "what is transconductance anyway?"
So lets just understand the terminology:
A simple transconductance amp in the feedback loop of a voltage amp is not necessarily a MyRef... though it may be "MyRef like." Only if it follows Mauro's compensation schemes (any of the three major versions, even including feed forward) and uses the LM3886 and the LM318 is it a MyRef.
Also I would go so far as to say that the power supply (shunt reg for V stage) is also a part of the MyRef signature, some would say for the better, some would say for worse, but all would agree its a fine amp altogether, and very cheap to build.Anyway, I hope, if Mauro sees this, he is proud of all the good he did. Thanks Mauro!
Cheers!
Russ
Russ White said:
that's quite interesting...
have you presented these nice works here or elsewhere?
of course, only a "my_Ref" is a "my_Ref"!
If you change anything it's no longer gonna be Mauro's design anymore, but something different. It may be better, worse, just plain different or even apparently undistinguishable from the original, yet it will NOT be it and may not be called "my_Ref". For quite obvious reasons only Mauro has the right to use that name for his own designs and developments.
That's why I previously cared to hint about proper naming...
And of course it's not just a matter of name/"trademark": the name here represents the designer and the many concepts, ideas and goals (and taste and skill ...) that he have put into his own design(s). Different designers will always have different tastes, ideas and point of view, and in the end different goals too (or at least different priorities).
Anyway, I hope, if Mauro sees this, he is proud of all the good he did. Thanks Mauro!
100% agreed...
UnixMan said:
that's quite interesting...
have you presented these nice works here or elsewhere?
I have posted bits and pieces here and there. But I am holding off on the details until I have really proven the concept. I also need time to finalize the project (many other things competing for time).
Here is the general idea.
The "master" voltage amp has been tested as both a discrete super symmetric amp (ala XBOSOZ), and a THS4131(which is also super symmetric).
One note, my best results have been with no R/C filter at the inputs of the Gm stages.
Attachments
Floric said:
Ahem... sorry, I would not call that a "half Rev C". It is a power op-amp based Stasis implementation. It's a completely different beast!
indeed, and it would be interesting (and even relatively easy) to try that out.
But again, it's nothing like a "my_ref-like" design, it's a Stasis. That's a completely different thing!
It's rather kinda like a "solid state OPT" (that is, an impedance matching device) capable of power gain on its own.
Coupled with a good tube amplifier - with such design you won't need just a voltage gain, you also need some current output capabilities too! - it may indeed give some VERY interesting results.
BTW, personally I would reduce the huge current gain of the Stasis "follower" (or "impedance multiplier", or whatever you'd like to call it) as presented by Mauro and use a stronger tube amplifier in front of it, perhaps to the point that the tube directly provides a somewhat significant fraction of the output current to the load.
if you have a well sounding small SET or tube headphone amplifier at hand, try using that one as the front-end!
In the case of the SET amplifier (or if your tube headphone amp uses an OPT), you can either use it "as is" taking the output after the OPT or modify the circuit to get the output before the OPT (maybe still using it just as an inductive load for minimal circuit modifications).
In any case, the current gain of the Stasis circuit must be properly adapted in order to properly match impedances. That is, Stasis circuit values must be adjusted to have a proper impedance presented to (seen by) the front-end.
(notice that in the case of a small SET with the output taken directly off the OPT, the gain must be reduced by a couple orders of magnitude and stability problems may occur with the 3886).
I would also make sure that ANY possible overload condition (clipping, overcurrent, etc) is always handled by the tube front end. The circuit should be designed so that the Stasis must NEVER ever "run out of steam" whatever the input signal and the output impedance are! That is, the tube front end must clip (and/or limit current) long before the Stasis does.
That way you can keep enjoying the great overload handling (and instant-recovery) capabilities of a good SET but with much more power out and no need for an expensive OPT.
Mmmh, the more I think of it, the more I like the idea... it MUST be tried out!
(but yet once again, that would NOT be anything like a "my_ref" design, this is a whole different thing).
Thanks, Unixman for your answer,
you are right, I have been a little lax with my formulation.
But the idea of fireing the proposed Stasis buffer with a tube has been in my since I read the thread about it.
In addition I am not shure if the approach of modifying the Rev C with a tube stage could be successful. Mauros design is really good, but it is a very special one that is customized for the parts he used.
Best regards
you are right, I have been a little lax with my formulation.
But the idea of fireing the proposed Stasis buffer with a tube has been in my since I read the thread about it.
In addition I am not shure if the approach of modifying the Rev C with a tube stage could be successful. Mauros design is really good, but it is a very special one that is customized for the parts he used.
Best regards
Mauro's design is actually very simple to modify(even using most of the same parts) its just that it is no longer a "My-Ref". But hey, individuality is a good thing. And I also think it is intriguing the many ways you can use the LM3886 and similar as "stasis" or "current pump" components.
Mauro's "Stasis" schematic is actually very similar to one in Nationals own application notes. And Analog Devices has a similar application note.
Cheers!
Russ
But hey, individuality is a good thing. And I also think it is intriguing the many ways you can use the LM3886 and similar as "stasis" or "current pump" components.Mauro's "Stasis" schematic is actually very similar to one in Nationals own application notes. And Analog Devices has a similar application note.
Cheers!
Russ
oh my...
oh my...
OK, I was in a hurry during the lunch break in the middle of a long meeting, but was I drunk
(without drinking...) or what?!? 
maybe I have found the "OPT with gain" analogy so sexy I ended up fooling myself?! ...perhaps my brain was just off duty. 
Anyway, I have written TWO big stupid things up there... sorry. An errata corrige is definitely required.
1) While a tube headphone amp (as long as it can provide some 30-35 Vp before clipping) can indeed be used as a front-end for the Stasis (and may be a good choice too), for quite obvious reasons an unmodified SET would NOT be such a good choice...
first of all 'cause it could not provide enough voltage swing to produce much more power than what it would produce without the Stasis! The effects of the added Stasis stage would be a reduced output impedance and an increased output current capability. If the output stage uses triode(s) or "pseudotriodes" it would also enjoy a much better linearity, at least to the extent the OPT is able to reflect to the primary the much higher impedance seen by its secondary (but if it use Pentode or Ultralinear mode, distorsion would rather get worse instead!). As said, the power output would be limited by the max voltage swing and will remain just about the same at best. Not very appealing...
2) The Stasis circuit does indeed increase the impedance seen by the front end (by means of "bootstrapping" the input current sensing resistor), but obviously it does NOT change the input voltage the same way an OPT would do. Thus, the whole idea of connecting to the primary side of the OPT of a tube amplifier was just a complete nonsense!
My apologize.
(BTW: I wonder how nobody noticed and pointed that out before... and fired a flamethrower on me thereafter!
)
oh my...
OK, I was in a hurry during the lunch break in the middle of a long meeting, but was I drunk
(without drinking...) or what?!? maybe I have found the "OPT with gain" analogy so sexy I ended up fooling myself?! ...perhaps my brain was just off duty. Anyway, I have written TWO big stupid things up there... sorry. An errata corrige is definitely required.
1) While a tube headphone amp (as long as it can provide some 30-35 Vp before clipping) can indeed be used as a front-end for the Stasis (and may be a good choice too), for quite obvious reasons an unmodified SET would NOT be such a good choice...
first of all 'cause it could not provide enough voltage swing to produce much more power than what it would produce without the Stasis! The effects of the added Stasis stage would be a reduced output impedance and an increased output current capability. If the output stage uses triode(s) or "pseudotriodes" it would also enjoy a much better linearity, at least to the extent the OPT is able to reflect to the primary the much higher impedance seen by its secondary (but if it use Pentode or Ultralinear mode, distorsion would rather get worse instead!). As said, the power output would be limited by the max voltage swing and will remain just about the same at best. Not very appealing...
2) The Stasis circuit does indeed increase the impedance seen by the front end (by means of "bootstrapping" the input current sensing resistor), but obviously it does NOT change the input voltage the same way an OPT would do. Thus, the whole idea of connecting to the primary side of the OPT of a tube amplifier was just a complete nonsense!
My apologize.
(BTW: I wonder how nobody noticed and pointed that out before... and fired a flamethrower on me thereafter!
)
Floric said:
indeed, it's appealing. Try it out!
As said, the easier way would be to use a tube headphone amplifier capable of delivering some 30-35V peak (even better if it delivers just that much at clipping...) and the highest current output capability. You have a HUGE range of DIY designs and kits (as well as commercial products) to choose from. And should you be not satisfied with the results of the Stasis, you could always keep using it by itself for what it is...
(moreover, having the chance to listen to it's "sound signature" with the headphones will also give you the opportunity to verify whether the Stasis does really preserve it or not).
Once you have chosen one such amplifier, note its max output current capability. Divide 10A (3886 max out current) by that much. Take just a bit less than that and you've got the (IMO) optimal current gain required for the Stasis.
As for the the current pump output "sense" resistor, I would use something more than 0.1 ohm (let say 0.22 - 0.27 ohm), and smaller valued resistors for the bridge too (also Mauro have done that on his later my_ref revisions, and there are some good reasons which are valid also here). Of course, as for the my_ref (and any bridge) those must be closely matched and stable.
Now multiply the chosen sense resistor value times the required gain and you have got value for the input sense resistor. Choose the closer smaller value from the standard range. Note its max power dissipation == R*(Imax)^2 and chose an adequately sized one...
eventually, add some suitable PS for the 3886 (again you have countless proven designs to choose from here around) and you're done.
Have fun!
Floric said:
it all depends on what you mean (and what you want). Building an hybrid amplifier with a my_ref like structure is not that hard at all. Making it behave "exactly" like a Rev.C in the bench tests (like the reverse driven test, etc) is trickier and require quite more work and skill.
Call me biased, but I bet that from the "sonic" point of view any decently done hybrid would win hands down over any all-chips based one...
After thinking a bit about it, I would say that I would probably begin trying with a classic Williamnson-like topology where the LTP driver is AC coupled to both inputs of the current pump (in place of the PP output tubes grids...) so that the current pump is driven differentially. This way both the AC and DC impedances seen by the 3886 inputs are perfectly balanced; the pump works better and you should get little offset too.
Use something like an ECC82 or a 6N1P for input & phase splitter and a somewhat more "rugged" tube (like e.g a 5687, 6N6P or 6N30) for the LTP (you won't need much voltage swing, but quite a bit of current from the LTP). If you want to go fancy, add another double triode and make it SRPP.
That's it. Simple, effective & proven... ad nauseam.
If Mauro is right when he says that the sound "character" of the my_ref is dictated essentially by the voltage gain stage, then this one may be the next Killer Amplifier...
2nd thought...
Thinking again about it, the Williamnson-like structure I proposed in the previous post is definitely overkill here. Too much gain and unneeded complication. We're not gonna drive a PP output stage directly so a "perfectly" balanced drive is not required (balanced impedances is all we really need) and -given that the current pump offer some gain on its own- we don't need so much gain either.
if I remember correctly, the original my_ref requires only about 8 or 9 Vp at the current pump input for full out swing. That's not much. And driving the c.p. differentially we gain another factor of two with respect to that.
Thus a simple LTP with any mid-mu triode should be more than enough for the task. And with such a small voltage swing it's gonna be practically distortionless even using simple resistive loading with relatively low supply voltages.
...that's gonna be really easy, and cheap too!
I would probably use something like the ubiquitous 6922 or the heftier 6N1P, but even an ECC82 or similar should do. I guess just about anyone here have all the required parts laying around in some junkbox... c'mon, let take 'em out of that box and try it out!
Thinking again about it, the Williamnson-like structure I proposed in the previous post is definitely overkill here. Too much gain and unneeded complication. We're not gonna drive a PP output stage directly so a "perfectly" balanced drive is not required (balanced impedances is all we really need) and -given that the current pump offer some gain on its own- we don't need so much gain either.
if I remember correctly, the original my_ref requires only about 8 or 9 Vp at the current pump input for full out swing. That's not much. And driving the c.p. differentially we gain another factor of two with respect to that.
Thus a simple LTP with any mid-mu triode should be more than enough for the task. And with such a small voltage swing it's gonna be practically distortionless even using simple resistive loading with relatively low supply voltages.
...that's gonna be really easy, and cheap too!
I would probably use something like the ubiquitous 6922 or the heftier 6N1P, but even an ECC82 or similar should do. I guess just about anyone here have all the required parts laying around in some junkbox... c'mon, let take 'em out of that box and try it out!
Unity gain output
The "current pump" has unity voltage gain, and 1000:1 current gain as implimented in My_Ref. With the 34 -35 volt rails, it will take more than 9 volts to get full power output.
Overall, the tube input stage looks nice. But I am too happy with the design as is to add a tube stage to warm up the sonics.
That is one area that surprises me to no end. Used to reject IC's and opamps in general. But when properly used they can sound very good. And no heater hum, micrphonics, and heat generation.
If someone cooks a good one up, I would love to see the details of the tube input/current pump. I am sure it will be very nice sounding.
George
The "current pump" has unity voltage gain, and 1000:1 current gain as implimented in My_Ref. With the 34 -35 volt rails, it will take more than 9 volts to get full power output.
Overall, the tube input stage looks nice. But I am too happy with the design as is to add a tube stage to warm up the sonics.
That is one area that surprises me to no end. Used to reject IC's and opamps in general. But when properly used they can sound very good. And no heater hum, micrphonics, and heat generation.
If someone cooks a good one up, I would love to see the details of the tube input/current pump. I am sure it will be very nice sounding.
George
I am very interested in this project. But I have the straight plan to finish some of the other things around here first. So my time is very limited at the moment.
In addition I have to say that my knowledge concerning tubes and designs with tubes is very (VERY) limited. But as I sayd before, the idea is very appealing, it could be the best of two worlds.
Unixman, I think the 6922 you proposed is a good idea. As far as I know it is a very good tube. As a second advantage it opens the path for a cheapo-solution with the ecc88 so the project would not be limited by financial skills too much.
Best regards
In addition I have to say that my knowledge concerning tubes and designs with tubes is very (VERY) limited. But as I sayd before, the idea is very appealing, it could be the best of two worlds.
Unixman, I think the 6922 you proposed is a good idea. As far as I know it is a very good tube. As a second advantage it opens the path for a cheapo-solution with the ecc88 so the project would not be limited by financial skills too much.
Best regards
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.