Lyra Connoisseur 4-2L SE: What a masterpiece!

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hello Everyone,


there is a review of the 4-2LSE in the German magazine Stereoplay (August 2005) that has some superb photos of the inside of the line amp, and also has plenty of informatio about the circuit topology, and also about its components.

The article can be downloaded here: http://www.stereoplay.de/sixcms/media.php/188/stp0805Lyra.pdf

MRupp wrote "I could translate the relevant passages in a few days but maybe somebody else is quicker? In the meantime you may try to get a reasonable translation from babel fish."

I will do a translation/resumé of the important stuff from that article:

* The power supply PCBs are right on top of the audio signal PCBs makeing the PS wires connect directly to the signal amplifiers

* A differential cascode on the input

* output stage with linear bipolars, powerful MOSFETs (IRF 510) and super fast JFETs are used to have a high bandwidth and a current potent amplifier

* dual mono power supply all through the amp.

*special iron plate circular transformers

* huge 20 000 uF caps in the PS

* Carr uses no ICs for the power supplies. Instead he uses 24 discrete regulators

* important signal wiring is done with wires and not on the PCBs

* Schottky diodes in the PS (20A diodes)

* the 20 000 uF caps are mounted in a way to minimize microfony effects

* input rotary switch and stepped attenuator with specially made Shallcos - Carr do not find potentiometers sonically good enough

* it is not balanced inside the line amps, but it can take balanced connectors

* star ground


That was it. If I have missed something, I apologise.


Photos in the next post.


Regards,
Sigurd
 
This is taken from another thread (I hope you do not mind Mr Carr):

"hesener:

If you want to use a similar topology to the 4-2SE line amp, for the input I'd recommend a dual-JFET with a high Vds (40~50 volts), in addition to good noise and linearity. The 4-2SE can produce about 11V output, and the input JFETs should preferably work well with a healthy amount of Vds across it. The input JFETs' load is the low impedance of a common-base stage, so small capacitance is perhaps of less importance.

The input JFETs should also have low gate leakage currents with the operating Vds applied. Regarding gate leakage currents of JFETs in general, P-channel JFETs tend to behave better than N-channels, but FWIW, I've chosen N-channels for the inputs of the 4-2SE.

>The topology I am thinking of is a input differential pair, followed by source followers.

Hmmm. Although the 4-2SE does operate differentially, it is essentially a much-augmented folded-cascode topology, loosely similar to what the AD829 would be if it were made from discrete semiconductors and had a JFET input. And a bit more complex than what you have in mind.

How informative was the article about the circuit topology, componentry or physical structure of the 4-2SE?

regards, jonathan carr"



I found the article in Stereoplay very informative! :)


Here is the AD829 schematic:



An externally hosted image should be here but it was not working when we last tested it.



Regards,
Sigurd
 
Sigurd, thank you for the translation. I was wondering if the only thing that I could do with the article was look at the pretty pictures, like a comic book. :D

>The power supply PCBs are right on top of the audio signal PCBs makeing the PS wires connect directly to the signal amplifiers<

Actually the regulator boards sit right under the audio signal boards, and the two were designed as pairs so that the audio signal boards "plug into" the regulator boards. The mental image that I had for this structure was how a CPU plugs into a computer motherboard, but with proximity regulation, i.e., each individual regulator sits directly underneath the circuit section that it needs to supply power to. And to tie the two boards together, rather than using additional wires, I have employed the components of the circuit and their leads. In other words, the component leads and components on one board have matching pads on the other, all of which need to be physically aligned when mating the two boards together, and then these multiple connections between the boards are soldered.

>important signal wiring is done with wires and not on the PCBs<

The signal wiring isn't made using wires, nor circuit traces, but rather with component leads which are kept separate from the board (air dielectric). And in order to increase the density of the air dielectric parts of the circuitry, the air dielectric exists in nominally 5 layers which are axially symmetrical to the board, with the center (3rd) layer running through the board (via oversized holes so that the component leads have good clearance to the board), the second and fourth layer spaced away from the top and bottom (respectively) of the board, and the first and fifth layers spaced somewhat farther away from the top and bottom of the board. Conceptually imagine a structure which looks similar to how a sewing machine stitches cloth. I can run signal leads parallel to each other or I can cross them without fear of short-circuits by choosing from any of the 5 air-dielectric layers. I call this type of structure "boundary-layer air-dielectric".

Regarding the AD829, yes the overall topology is similar, but some sections like the output stage are totally different. Rather than the AD829, the output stage of the 4-2SE shares conceptual similarities to a White follower, except that it is a fully active, complementary P-P design. However, it is a hybrid design, so FETs are not necessarily matched to FETs, nor bipolars to bipolars. You could just as easily call it a mongrel as a hybrid. :D

There's more but that should be enough for now.

kind regards and many thanks for the translation, jonathan carr
 
No problem, JJonathan. My pleasure. I am a German with a Polish name living in Sweden, and I am fortunate to master many several languages :)

I just translated the technical stuff as I figured that this was what people wanted to read. There is also a short hiistory of Lyra, and a short review of the phono amp 4-2 P SE (which Stereoplay sounded the best that they have ever heard (together with the Lyra Titan) but it will not be their reference as they want more impedance choices for that). I did not translate what they thought about the sound of the Lyra. It goes something like this:

"...that the Lyra did not have a remote control for volume attenuation was quickly forgotten once we started listening to the Lyra. The sound from the Lyra was breathtaking. Such precision had the reviewers never heard before. Voices were extremely natural sounding. Also, the treble was produced with exactness and it even cared for the smallest details, and it was easy to follow how cymbals died away.The huge soundstgae that it produced was also unique. Even our long time reference Naim Audio NAC 552 could not match the Lyra. Stereoplay have a new reference, and this proves that enthusiasm in designers pays off."


Thanks for sharing some design details with us. You are really teasing us, and challenging us :D
I like that!

<but rather with component leads

I do the same just because it is the simplest way, but have recently been thinking about what are the component leads made out of? Some are pure copper with gold plating, but most are made out of lesser highend audio quality metalls.
Have you thought about using super short component leads and use for ex solid silver wire instead?
Or will the extra soldering joint destroy any poaitive gain one gets?

I am somewhat puzzled that bipolar transistors can give such good treble. I though that JFETs was the way to go a la Borbely's newer designs.

A question: the power transformers you use. What are they? and in what way are they better than toroids?




Sigurd

Sigurd
 
Sigurd Ruschkow said:
I am somewhat puzzled that bipolar transistors can give such good treble. I though that JFETs was the way to go a la Borbely's newer designs.
If I understood Jonathan correctly, J-FET's are used at input and BJT's for the folded cascode. For a cascode, BJT's seem to do pretty well....but I never tought about J-FET's as cascode devices in a folded cascode. Hmm....

Sigurd Ruschkow said:
A question: the power transformers you use. What are they? and in what way are they better than toroids?
I believe these are R-Core transformers.

For John Curl's explanation not to use toroids for line level amps, read page 8 + 9 of this document:
http://www.parasound.com/pdfs/JCinterview.pdf

Jonathan Carr suggested R-Core transformers here:
http://www.diyhifi.org/forums/viewtopic.php?p=2785#2785

Maybe they have some R-Cores left:
http://www.hifituning.de/html/r-core_trafos.html

Ciao, Tino
 
In the latest issue of Hi-Fi + (issue 43) they have an 8 page review of the lyra 4.2 se with a LOT of content about the design and construction side. They claim it is the best componant that has ever come along.

Magazines always give lavish praise in reviews so we tend to take it with a grain of salt, but trust me, you have never seen anything like this Hifi Plus review.
They talk of the Lyra like the second coming of god, with the reviewer mortaging his life to get one and the publisher about to as well. They talk of it being a pre amp that is on a level so far above all other products seen yet. And considering the exotic/expensive gear reviewed in Hi-Fi + this is saying something.

So well done Jonathon Carr!!

How is the more reasonable 6.0 coming along? (for those not in the country club)
 
In the next latest issue of the British magazine HIFI+, there is a review of the Lyra 4-2L SE. The comments in the review were very, very positive.

You can find the review at http://www.hifiplus.com/issue43.html

I thought that SE stood for Single Ended, but it stands for Special Edition, and the 4-2L SE is an upgrade to the 4-2.

Jonathan, it would be interesting to know what you changed between the older and the new SE versions?



Best
Sigurd
 
Agisthos said:
In the latest issue of Hi-Fi + (issue 43) they have an 8 page review of the lyra 4.2 se with a LOT of content about the design and construction side. They claim it is the best componant that has ever come along.

In emails, Chris (the reviewer) and I were talking about being able to guess the exact make and model of the instruments being played on a recording, and also of recording engineers, who, after listening to the 4-2SE, spoke of wanting to revisit some of their mixing and production decisions. It would have been an even nicer read if some of these things had made it into the published article, but it was a very kind, generous piece of writing nonetheless, and I am indebted to Chris Thomas and Roy Gregory for making the article happen.

Agisthos said:
How is the more reasonable 6.0 coming along? (for those not in the country club).

The 6 phono stage has been through a couple of name changes (most likely it will be called the 206-P or 306-P, depending on how much I ultimately decide to put into it and what that means for production costs), and it's been played in public a few times (at audio shows in Japan and the US). It's also been delayed somewhat because of other priorities - launching new cartridge designs, and the latest being a revision of the 4 to fit ROHS mandates. But I have been using these delays to steadily update the design concepts of the 206-P, and in fact introduce some more adventurous ideas that haven't been possible with the 4 series. Although the voltage regulation will be discrete (and I'll probably use 6-layer circuit boards), the amplification circuitry will almost certainly be opamp-based, and it will be a fun challenge to see how much performance I can extract from monolithic circuitry. There will also be a companion line stage, and this is intended to address many of the attenuator contact issues that I mentioned in the "Blowtorch" thread.

regards, jonathan carr
 
Sigurd Ruschkow said:
Jonathan, it would be interesting to know what you changed between the older and the new SE versions?

Sigurd: The 4-2SE has an uprated power supply with lower noise and less crosstalk. Inside the main unit, some of the semiconductor and passive component choices are new and so are some of the key operating points. The intrinsic DC offset stability of the amplification circuits has been improved and part of the increased margin used to extend the low-frequency response. The impedance of the grounding system has been reduced, the chassis mechanicals have been revised so that there is less resonance and less vibration transmitted to the circuit structures, the attenuator switches are disassembled, blueprinted and rebuilt, and so on.

The unit reviewed in Germany contained some of the 4-2SE features, but judging from the photos that you posted, it had a more standard power supply, and AFAIR the circuitry that enabled the improved offset and increased low-frequency extension wasn't included, nor was the lower-impedance grounding network.

regards, jonathan carr
 
Jonathan,

the changes between the two versions of the 4-2L are some major changes! If I had an older unit, I would definitively ask Lyra to make the upgrade - if possible.

I am not sure how you deal with DC offset and its temp. drift, but it sounds like you use a DC servo, and with the newer version you have adjusted the freq where the servo starts to work, and thus improving the bass. Or am I am totally of track?

Some say that the operating point of any transistor has a sweet spot where the semiconductor sounds the best. I have not done any experimenting with operating points, but I look at the datasheet and try to figure out where the semiconductor is the most linear, and then I set the operating point accordingly. Useing cascodes to lock the operating voltage for the semiconductor and resistors to set the operating current. Some times a CC is used.

Would you say that it is possible to generalise and say that each semiconductor has a sweet operating point (soundwise)?

When it comes to vibration dampening I would say I am a bit uncertain whereto stop dampening the cabinett and its contents. With my headphone amp, I dampened top and bottom with Antiphon ("lead" mat) and I use rubber feet for all the PCBs. SOmetimes I even use clay to dampen parts. It doesn't look good but sounds good.

The 4-2L SE looks very good inside! and I do not see any ugly Antiphon or clay. What I see is thick Al plates on the sides, thick hard wood sides and front, and I think you have used Teflon plates under the big caps in the PS box.

Would you say that this is how you have dampened the preamp and PS?


Best regards,
Sigurd
 
Sigurd Ruschkow said:
the changes between the two versions of the 4-2L are some major changes! If I had an older unit, I would definitively ask Lyra to make the upgrade - if possible.
The subjective sonic differences from the 4-2 to the 4-2SE were at least as large as the differences to the 4-2, so yes I would agree.

I am not sure how you deal with DC offset and its temp. drift, but it sounds like you use a DC servo, and with the newer version you have adjusted the freq where the servo starts to work, and thus improving the bass.
All of the 4's have used offset servos, but with any offset servo, the topology and settings are influenced by the behavior of the main amp circuit. If the offset behaviour of the main amplifier is not so good, the servo needs to have a higher corner frequency and be set for greater injection. With the 4-2SE the offset behaviour of the main amp circuit was improved, and this allowed the servo to be redesigned / readjusted to a lower corner frequency and less injection.

Some say that the operating point of any transistor has a sweet spot where the semiconductor sounds the best.
This depends on the circuit and designer as well as the semiconductor. The various curves (noise, ft, Cob, current linearity etc) of a given semiconductor usually peak out at different points. If you think that the circuit would work best with optimized noise, you might run the semiconductor differently than if you thought the circuit preferred maximal ft. Pperhaps the gestalt is that there is a certain operating point that subjectively sounds the best (and maybe measures well, too), but the reality is that you need to line up your priorities for the circuit and choose operating points that fit.

Using cascodes to lock the operating voltage for the semiconductor and resistors to set the operating current. Some times a CC is used.
Yes, I also prefer to have good control over the semiconductor's operating voltage, OTOH, I nearly also use a CC.

Would you say that it is possible to generalise and say that each semiconductor has a sweet operating point (soundwise)?
Please see two paragraphs above. And when we say "soundwise" we are now talking about something subjective, and therefore affected by personal differences among individuals. I am sure that you have heard amplifiers and audio systems that the designer/owner was proud of but you really didn't like, and you can bet that said designer/owner would feel the same about the kind of sound that you might produce :). Ditto for the choice of transistor operating points. There are many different operating environments and many, many people in the world, and given their varying likes and and dislikes it isn't possible to satisfy all of them.

More later.

regards, jonathan carr
 
How important is the air dialetic construction method to the performance of this product Jonathan? It seems to be something which would be very labour intensive to produce.

If you had to just use a normal pcb type design would you be at 70-80% perfomance or would it just not be possible to do at all?

I think I read somewhere you used to work at Spectral? Or someone at Lyra did anyway.
 
Hi again Sigurd:

In general, for resonance control, I find that there is an optimum degree of coupling assemblies and components together - neither too tight nor too loose. Too loose and the individual assemblies can resonate, too tight and the entire stucture will resonate as one.

The chassis of the main unit is made from machined and shot-peened aluminum plates, while the wood cabinet is made and mounted so that it constrains and partly damps the remaining resonances of the aluminum (what you think are aluminum side plates are in fact wood).

Rather than trying to damp the electronic assemblies themselves, or rather than mounting the assemblies directly to the chassis, I now use various types and forms of plastic to provide greater isolation between assemblies as well as to the chassis (similar in concept to the plates under the caps in the PS).

I have experimented with rubber and lead damping, but don't use them because I don't necessarily like all of what these do to the sound. Also, lead in electronic components runs afoul of ROHS regulations, so for a product in thsi day and age it is a non-starter.

best regards, jonathan carr
 
Hi Agisthos:

Air dielectric has various performance advantages (such as keeping high-impedance nodes at high impedances across a wide bandwidth), and some of this is, I feel, essential to how the 4-2SE performs and sounds.

But at least according to how the air dielectric is done in the 4-2SE, it also gives major advantages in packaging density. There are two stacked boards per amplification module, one primarily for the regulator, and the other primarily for the amplifier. Each circuit board consists of 4 layers, and there are components on both sides of each board, so right there I have 8 layers for traces, and 4 layers for components. But on the upper (amplifier) board, this is augmented by 5 more layers of air-dielectric construction, and with air dielectric, I can stack components as well as traces.

If I were to transfer the 4-2SE circuit as is to a conventional board layout, the physical size of the circuit would balloon, enough so that I would need additional phase compensation and/or bandwidth limiting to keep everything stable. The larger circuit would also represent a better receiving antenna for RFI garbage, and I'd have to go to a bigger chassis, too. None of this do I find appetizing.

In general, I am a firm believer in designing the circuit and the physical construction as two inseparable sides of the same coin. When I design I hop between board layout and schematic editor and continuously alter both the schematic and layout to better suit each other. In effect the physical structure dictates the schematic constraints, and the schematic in turn dictates the requirements for the board structure (and most of the components on it, including things like resistors and capacitors).

The 4-2SE circuit works well in air-dielectric, but if I had to do it on a conventional board layout - I'd end up with a different circuit. Actually, I'd prefer to start over with a clean-sheet design, something that was conceived of from day one to be built on a conventional board structure. But one gains fresh insights and learns new things all the time; I hope that some day I'll be able to do a conventional board-type product that exceeds the performance of the 4-2SE, air-dielectric notwithstanding.

regards, jonathan carr
 
Those are terrific posts, thanks! If I understand correctly, the air dielectric construction (sounds like stacked super point-to-point) simplified managing high frequency circuit stability and appears to have enough intrinsic RF rejection to permit a great deal of latitude in chassis design. Am I on the right track?
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.