Teardown of DSD Recorder

Head over to Soomal for a teardown of Korg’s DSD recorder [link]

 

Musiland’s Upcoming Audio Processor Chip

Here is an English version of Musiland’s recent press regarding the development of a new processor for Audio:

A famous philosopher once said: “Today’s stillness is for tomorrow’s outburst”.

5 years ago, the release of Musiland’s first FPGA-based product, the LILO V ENJOY USB sound card, marked the beginning of Musiland Audio Labs involvement and mastering of chip-level programming technology.

For the last 5 years, Musiland Audio Labs has continue to invest in chip-level programmable solutions by gradually introducing FPGA-based solutions to its entire product line and by the using larger size FPGAs to implement increasingly more complex capabilities.

Four years ago, Musiland Audio Labs recognized that the advances in embedded 32 bit microprocessors, led by companies such as ARM and MIPs, far surpassed the advancements in desktop processors. At that time, Musiland made the decision to use FPGAs (as a bridge) to develop its own general purpose 32-bit processor.

Three years ago, the MD11 was born, followed by the HP11 and MD30. Musiland Audio Labs started to use early versions of its general purpose processor (implemented in FPGA) to handle other functions such as user interface, LCD control and storage. Adding a good graphical user interface to HIFI equipment was just a small step but demonstrated the continued investment towards a general purpose processor technology

Of course the goal to develop a 32-bit general purpose processor is not to just deal with the user interface and other minor tasks. Our vision is to develop a complex, large-scale processor to do more complex audio processing functions such as decoding FLAC. This requires a plan with a long time horizon to develop a processor that not only meets current audio processing needs, but also keeps up with the advancements of computer technology and meets future needs.

On deciding on the architecture for the processor, Musiland engineers had different views on what to do. The technology camp felt that they could develop the processor from scratch (call it “M-CPU”) and thus not be subject to any external constrains or patent protection; but more seasoned engineers knew that a development from scratch was a long and sky-high expensive proposition. After exploring different proposals and after much discussion among the different teams, a rational, consistent and scientific decision was reached: Use ARM+DSP dual-core architecture.

Industry leaders had same strategy. At the same time Musiland Audio Labs decided to create a 32-bit processor based on ARM+DSP dual core, we learned that Creative Technology set up Zii Labs for the research and development of multimedia processors. This was a great encouragement for our Musiland team and allowed us to understand the competitive environment with more clarity and also to strategize on how and where to differentiate our products. We decided to abandon complex video functions and to focus instead on the field of audio.

Later on we lamented the divesture and selling of Zii Labs but realized that our decision to focus on audio was the right decision. We at Musiland Audio Labs will always respect the people at Creative, and would like to take this opportunity to appreciate Creative’s contribution to the industry and in making multimedia and audio applications and products so pervasive. We must always remember the name Zii Labs for their truly innovative processors.

For the past two years, Musiland Audio Labs engineers worked night and day to integrate the ARM core and audio DSP functions and developed an ARM-based 32-bit processor with 32-bit floating point DSP with an internal  unique data bus, the “MP-Bus”. We will call this new processor “SuperDSP” and will produce a family of products: SuperDSP100, SuperDSP200, etc.

A family of devices will be produced to meet different applications. The different versions will differ on the size of on-board and external memory, the type of packaging (QFP and QFN). The chips will support USB 2.0 high-speed and optionally USB 3.0, external mass storage support (SD, MMC), external bus support (PCI) and other communication options such as SPI, I2C UART, etc.

It is worth mentioning that the 32Bit floating point audio DSP unit can handle up to 64Bit/768kHz sampling rate of the audio data, support multichannel or DSD decoding.

According to our product roadmap, the first SuperDSP processors will be implemented for the personal multimedia sound card/integrated player market and will be released towards the middle of this year and will provide unprecedented audio capabilities and sound experience. Soon after, we will focus on the HIFI and audiophile products. So stay tuned!

If you have any questions, requirements or expectations about this new product, please use our forum (bbs.musiland.cn- look for SuperDSP) to post your queries. Meeting your requirements is our responsibility; be sure to let us know!

The SuperDSP interface, decoding algorithms and Library API will be open to third party developers. We welcome industry colleagues to discuss OEM/ODM arrangements with us. Musiland Audio Labs looks forward to this cooperation and is prepared to help you innovate your products

TPA3122D2 Kit

TPA3122D2 AMP Kit on sale: [link]. I got two for balanced-in, dual mono operation (BTL configuration) for $25…

MarkAudio Alpair Speaker Drivers

There is an inherent attraction to single driver, full-range loudspeakers in that it is fits with the “straight-wire” camp of thought where less is better. The amplifier output is directly connected to the windings in the voice coil, and the power-train responds gracefully to all the frequencies in the audio signal.

In the design of the full-range driver, there is also a “purist” approach of using a single surface to cover the entire frequency range. (Many full-range drivers utilize a “whizzer” or a plug – which constitutes a second surface, to handle or enhance the higher frequencies).

The challenges in this approach become increasingly difficult as the designer tries to extend the low frequency response by increasing the size of the emitting surface and at the same time improve the high frequency response by reducing its mass. One can readily appreciate that size and mass not only go hand-in-hand (bigger size, bigger mass) but the structure must also have enough strength and rigidity to  keep its form through repeated motion for endless hours.

MarkAudio, a small company established in Hong Kong, has mastered and conquered these challenges and has delivered its most up-to-date technology in the Alpair 12p driver.  With constant dialog with end users gathering requirements/wishes from the audio community and by utilizing the latest materials, manufacturing technologies and unconventional design ideas, MarkAudio has produced a stellar product.

(BTW, sounds like a marketing, but I am not associated with MarkAudio except as a customer and I am indeed impressed at the technology and capabilities of this driver)

True to the theme of this blog, the Alpair driver is another “best-bang-for-buck” product. Aimed at the diyaudio crowd, one needs to build a cabinet. For those of us with only rudimentary woodworking skills, there is a new thread in the forums documenting such a project [link]

THE ALPAIR 12P

As it apparent, I have become the proud owner of a pair of MarkAudio’s Alpair 12P full range drivers.

The Alpair 12P is the latest generation of the Alpair 12 series. It is designed base on customer feedback and requirements and incorporates the most advanced ideas from MarkAudio to date. The driver was released in mid-June of 2012 and although it is classified as “Generation 2″, time-wise it coincides with “Generation 3″ technology.

I enjoy learning and appreciate the technology invested in a product and about the people behind their designs. Luckily, Mark Fenlon, the founder of MarkAudio shares a lot of information in his diyaudio forum. The information presented here is gathered from the forum, but the nicer photos are mine

MARKAUDIO

First, some history on MarkAudio (from Mark Fenlon himself):

(Mark Fenlon started with E. J. Jordan Loudspeakers)…

I much enjoyed the time I spent with Ted Jordan. I have much admiration for his stamina and spirit. Ted is of the traditional school of British design, something sadly that is fast disappearing. Ted and me had our good times and our not so good times, sadly our association was not destined to last. We had markedly differing ideas on cone and other key component design. To be fair to Ted, I was the one who eventually pushed for radical development of lower mass power-trains using new materials and shallow profile cones, with all the inherent production, operational and financial risks. It became inevitable that we part our ways and so Markaudio was born in its own right.

For the last 2+ years, Matsubara san, the farther designer of the Fe series has been my great supporter. I well remember our first meeting where I showed him the Alpair 12 prototype, with its ultra thin front suspension, incapable of supporting the cone, he told me I was absolutely nuts to attempt a single mechanical suspension driver of this size! But he listened to my plan, poured over my drawings and we came up with the mother of all rear suspensions. A few weeks later, he came to a Markaudio listening event in Hong Kong and was bowled over by what he heard. The rest is history as the saying goes.

As important is Evan Yu. This guy has been my right-hand man for the last 5 years. He’s one of these guys you need in any factory to get things done. He also a descent conventional driver designer in his own right. He also thinks I’m nuts at times, but he knows me well enough to give pretty much all my ideas a fair go. More recently, I’ve had the pleasure to work with Jeff Tanigichi san and Kitagawa san, 2 first rate audio product engineers. Many an hour is spent debating over all sort of developments, from custom connectors to changing over to fused spider mounting systems, you may imagine some of the daily conversation, but its worth it in the end.

I hope our pioneering spirit continues and home custom speaker builders continue to invest their confidence in our work. My hearty thanks to all the guys that have bought and used Markaudio drivers, as without this support, Markaudio couldn’t exist. My hope is that we gradually spread more confidence in Full-Range drivers. I believe there is a big future for single point source, single cone full-range emitters.

I’m glad of all the help, support and encouragement as we’re still a young company with hopes to make into main-stream audio some day in the future.

Thanks
Mark.

[link]

FULL RANGE DESIGN CONSIDERATIONS

In order to appreciate the advances embodied in the Alpair 12p, it is important to understand what are the things that would make a good full ranger.

These are the main design elements that are in play for a single Full-Range cone design:

1. Rigidity relative to flex (of the power-train: suspension, cone, coil, etc)
2. Mass relative to dimension (of the cone and related)
3. Profile of the cone element
4. Oscillatory properties
5. Resonance properties
6. Micro-resonance properties

Number 1: A relatively rigid cone form is needed to maintain the LF (low range) oscillatory function (stability) of the power-train (front suspension, cone, cap, coil and rear suspension). Some flex in a low-mass cone is inevitable and needed for the resonant functions. Control and balance between these 2 functions (rigidity and flexibility) is critical, material selection plays an important role in this design element.

The second and third design elements are particularly critical. The larger the cone gets, the greater its mass, the harder it gets to make it resonate. Mass is the limiting factor at play for primary mid-high emittance. Hence why it takes me longer to modify and improve the larger single cone drivers, and why most designers give up and deploy whizzers and phase plugs.

The third element is where I’ve spent allot of design time these last 3 years. The cone profile is not only critical for oscillation, but more so for resonance. The ability to allow the passage of a wave signal is fundamental. The profile (shape) is also critical to the design of the dispersion characteristic. Some of you may have noticed Markaudio driver cones are becoming more shallow (lower profile). This design factor improves mid-high primary emittance and increases dispersion. However, there are limits as the stability of the power-train remains important.

The 4th, 5th, and 6th design elements are sub factors specific to the design processes relating tooling and press production.

Summarizing:

Its takes allot of time and effort to get all the elements working together to make a single cone emit full-range (to 20-kHz+). Mass is the enemy of making a driver go “full-range” yet it is needed in order to maintain mechanical/operational stability.

Also

Effectively the cone cannot act as a rigid piston at all frequencies, so “breaks up” and effectively becomes smaller at higher frequencies. This breakup will show as resonances which will result in peaks and troughs in the frequency response, as can be seen in any driver, not just wide range units. This is a very simplistic statement, and the challenge to the designer is to minimise the unwanted effects that result. The normal approach is to avoid using the driver in the range where this happens.

To get bass, a large cone must be used, so for these another way is to create a discontinuity in the cone so the cone becomes “smaller” at a frequency chosen by the designer. Hence the use of whizzers, but these have their own unwanted resonances. An extreme case was used by Hartley, who placed a flexible coupling between the inner and outer sections of the cone. This prevented the whizzer type resonances, as the edge of the treble section was now supported, and damped, but introduced a deep narrow trough in the frequency response.

For a single cone wide-range driver, the designer does not avoid this breakup but controls it and makes use of it to extend the frequency response. As Mark indicates, the main ways of doing this are through cone material, cone thickness and cone profile. It becomes easier if the requirement for bass is given away, and the cone limited in size. But that is only the start and Mark indicates some of the other problems that are created as a result of optimizing these three factors. [link]

And further

…Designing an electro-acoustic driver power-train is very complex, much more so that simply thinking in terms of excursion V the mass issue.We have to remind ourselves that the input to the driver’s power-train is non-linear, varying in frequency and amplitude. Traditionally, high SPL drivers with short coils have a short usable X, relying on the design of the box to extend LF emissions by wave lengthening. For LF emittance, this driver type is mostly limited to use in LF extended gain box designs (horns etc.), most quite physically large, not particularly practical for many hobbyists. Such drivers also are vulnerable in terms of their linear excursive capacity; Said capacity easily being exceeded even by a moderate increase in input signal strength, the result is increased distortion at close to limit of the driver’s in-gap coil stroke.

I realized along while back that there was potential for creating a power-train with a longer throw (X) that could mitigate the limitations of short X drivers provided the relative low mass of the power-train components could be retained.

Having worked on this concept for several years, we are now at the point where relatively long stroke power-trains with low mass wide long-wound coils give the driver greater linearity when driven by LF inputs. The concept is more flexible, allowing for less critical box designs; And the greater use of a wide variety of box types. This concept is still low-power. Its not designed to be a replacement for a woofer, but it is designed to more accurately emit non-linear LF loads that drivers with short coils cannot.

Historically, this was a major part of my efforts to extend the bandwidth of single point source Full-Range drivers.

Thanks
Mark.

[link]

THE FIRST GENERATION ALPAIR 12 (METAL CONE)

Developed towards the end of the “generation 1″ cycle, he following features were notable:

(I am summarizing here)

1. New standard hole M4 metric hole size for easy installation
2. Custom anti-resonant frame to minimize resonance that can be transmitted to the cabinet
3. The front suspension is a new advanced molded design, made only for this driver. The mass of the piston section has been reduced. It’s ultra-thin compared to other driver
4. All new Multiform cone. The low mass, mixed alloy material is very thin, less than 0.11-mm on the critical sections along its profile. To our knowledge, the Alpair 12 is the only current production driver of single cone, single cap 8″ design, capable of reaching 20-kHz at 87dB.
5. The cap is now directly bonded to the coil former using a A/B bonding method. Strong and efficient method for the transmission of middle and high frequency responses.
6. The spider is a custom design. Its profile and the suspension shapes are specific to this driver. This spider makes a significant contribution to the fast response of the Alpair 12.
7. New connector system has been design for easy installation and optimized to provide the damping of the leads that run to the coil.

THE NEXT GENERATION: GENERATION 2

Right after the release of the metal cone 12p, the company started R&D on the next generation. These were the design goals (and customer ask) for the next generation Alpair12

1. Raise SPL above 90dB
2. Increase frequency range (where possible)
3. Increase Vas
4. Reduce Mms
5. Improve the power-train suspensions
6. Produce a paper cone alternative to metal

Results

First, this is a radical departure from the Gen. 1 driver:

• The 12P has an Mms of 10g, down from 16g on the old driver, a radical reduction is mass of the Power-Train
• The New paper cone is emitting 22-kHz @ 90dB before falling away, some 4-kHz greater range than the Gen. 1 driver
• SPL is up at 92dB, an increase of 3bD from the old driver

For me personally, its difficult to know where to begin as I personally took some radical steps on the design of the cone profile and the coil, in order the achieve the desires given in past feedback. This driver goes allot further than anything I’ve made to date, that includes the current Alpair 7. [link]

PAPER CONE

Type: the paper cone is the same as the one developed for the Alpair 6.

The paper is of Japanese origin with Taiwanese production (KK and BKH mixed fibre papers). The Multiform process has been taken a stage further so much of the forming is now double pressing. Note the color. The dyeing process is critical to rigidity of the cone, along with reduced mass.

Back side of the cone: the front and back sides of the cone have different texture. It is a “press point” pattern on the underside of the cone, again to aid stability [link]

Profile: Shallower

The change to a custom paper cone with a more shallow profile reduces mass and increases dispersion performance. The new cone is only 21-mm deep compared to the metal unit at 26-mm.

Response: more mellow

The emittance properties in the upper-mid to high range on larger paper cones drops off more than metal. The bigger the paper cone, the more challenging getting it to emit high range. On the plus side, paper is often considered to be more “mellow” in the low-mids while retaining detail if the designer sticks to a low-mass design.

I’ve spent years “tweaking” metal to sound more warm and paper to offer more bandwidth. Much a subjective situation, beauty being in the ears of each listener. [link]

VOICE COIL

The 12P employs a new custom CCAW coil with an aluminum/paper wrap, vented, low mass thin-walled body. Notice the aluminum coating on the inside of the paper body. The holes reduce the mass of the voice coil

Here it is compared to the voice coils for generation 2 Alpair 12s and generation 2 Alpair 10s. Notice the new coils are longer and have more holes for lower mass.

Here is a photo of the voice coil in the Alpair 10 (1st generation) [link]

Two types of voice coil were tested before selecting the aluminum body type:

TIL (a type of fiber/resis) coil prototype: The driver’s T/S data is looking quite good. SPL is +92dB while Qts is 0.3 which hopefully will make it easier for a variety of box designs. However, typical of TIL bodied coils, the transfer properties to the cone neck are more variable compared to Alu and paper coil bodies. The result is a frequency spectrum is more variable with some larger peaks coming in the upper mid-band. This coil type can exert increased micro-reasonant extension and compression. I haven’t had time to run my tests, but looking at the frequency, I suspect this driver could sound a tad shrill. Can’t really tell at this stage as I wasn’t at the factory today. At least we are at the stage where this prototype was gone almost full-range (-7dB from mean @ 20-kHz), pretty good for a single paper cone of this size with no whiz or phase plug assist. I’ve not had time to generate the Impedance graphs but observing the peak at resonance, it reaches 145 Ohm, high and typical for this coil type, that might not suit some tube amps.

Alu coil prototype: This driver gives me more excitement. Normally, I wouldn’t expect an Alu coil body to generate enough transferable resonant energy to get a single paper cone design of this size over 12 to 14-kHz. Effectively, we might say this driver’s going full range as it gets to 17-kHz before falling off. Alu coil bodies have a higher material damping factor, so we’ve the better looking frequency response of these 2 prototypes. Overall, its frequency spectrum is the most controlled with SPL peak-dip +/- variations of around 4dB, good for drivers of this size and type. Also of interest is the mms, at 9.775 grams its close the much lighter TIL proto than I anticipated. As a result, the Alu prototype’s SPL gets above 92dB. Also, its Impedance peak at resonance is 60 Ohm and relatively flat, should suite allot of lower power tube amps. However, its power-train does have some increased damping so QTS is lower than I hoped @ 0.277. Depending on which box calculation software used, something around a 15 litre BR looks feasible. At this early stage, this prototype looks the most promising.

[link]

FRONT SUSPENSION

The front suspension is made of a rubber material and it feels very thin to the touch. It is bonded/glued to the paper cone and the frame. The small mass of the suspension is in line with the goal of reducing the overall mass of the power-train. I think the front suspension is similar to the one used in the 12 (metal cone).

…These new suspensions are so low mass, they can’t support the cone’s alignment in the power-train during the assembly process. I’ve developed a series of tools to do this work, this pic is my latest design…

Typically, front suspensions have enough rigidity to level and support the cone when being installed into the power-train. But there’s a catch. Thicker front suspensions add mass and often introduce increased non-linear motion. This is due to process and heating differential limitations when making suspensions that require larger volumes of material. Often, the outcome will be variations in the flex-stiffness ratio in various locations within the piston wall of the suspension (rounded section of the component). Such variation aren’t normally an issue for commercial grade drivers, or Full Rangers with limited Xmax. However, for drivers with low mass power-trains, wide frequency generation and larger excursions, the tolerance and mass of this component becomes service critical.

More significantly, thicker front suspensions limit a can a cone’s ability to resonance at higher frequencies. For those of you who own Alpair 12′s, you can observe just how thin the front suspension is by very gently touching it from the back of the outer frame. Please Note the word “gentle”! The thin suspension forms part of the Alp12′s ability to go +20-kHz on a single cone design, without the use of whizzers, plugs or co-axial assistance. [link]

BACK SUSPENSION: SPIDER

The spider of the 12p is an improvement over the spider design in the original Alpair 12 which itself was a unique design for the driver.

The mechanical properties of the convention spider delivers a particular resistance proportional to a given X load point. This has some drawbacks. The first is a non-linear damping effort while overcoming the inertia of the power-train, particularly at the initiation of a LF oscillation. Makers jargon sometimes names this effect “rolling bounce”. Second and more important is the design may offer little control towards the limit of excursion. Many of us have driven a woofer to a point where the coil has hit the back the the yolk plate. Most makers try to overcome these challenges by simply making a stiffer spider but this approach while partially solving the resistance challenge, creates others (sorry no time to go into this now).

Here is a photo of a “conventional spider [link]. Notice the uniform rib profile and separation.

The Alpair 12’s spider is different. It is designed with a specific resistance profile. There is moderate resistance on short X-LF loads. Going through the normal operational range, the resistance profile continues to be proportional to the loads, then stiffens near the limits of the excursive load. Since the front suspension is extremely soft compliance, the spider operates the significant part of the load damping and maintains the oscillational stability of the power-train.

Here is a photo of the spider design in the 12 [taken from here]. Notice the non-uniform rib profile, with the ribs becoming “larger” away from the center

The Alpair 12p spider adds radial “bridges” to the concentric rib profile of the Alpair 12, in order to  increase its rigidity. The 12p weaving seems finer as compared with the 12.

FRAME

Made of “anti-resonant” material to minimize transferring vibrations to the cabinet. An incorporated rubber gasket further helps minimize vibrations.

…The use of polymers in frame design/making is strictly a design issue based on engineering research. Low mass drivers present more design challenges, one being isolation between the power-train and its mounting frame. The last thing a low mass driver needs is a resonating frame (ringing) transmitting specific frequencies back to its power train. Of the 32 steel and cast frames I’ve measured to date, all emit a prime resonance between 800Hz to 3-kHz depending of size and design. 22 of them emitted secondary resonances. By contrast, the resonant characteristics of polymer frames are much better damped. [link]

Reclocking with 100 MHz Clock

It has been discussed that one can use “asynchronous” reclocking with a high frequency clock, that is a clock frequency not a multiple of the source sample rate frequency. This method has been used in the past and good results have been reported. See for example reclocking the TDA1545 (Monica DAC) with an 80 MHz clock [link]

Many have planned (including myself) to use the on-board clock of the Buffalo DAC (and other Sabre32-based DACs) in order to achieve “ultra low jitter” results.

But, there are some problems with the ES9018 on-board sample rate converter when dealing with this kind of clocking.

Russ just reported the following:

A quick update – I will state up front – I was completely wrong.

I found the re-clocking via flip flop to the 100Mhz clock was actually destructive to a jittery signal. Not in the way you might think. It actually can cause a very slight error in the data. It definitely will not help. Someone called this a decimation error, but I am not sure that is the correct term.

After a while I noticed that the sound was actually worse… less precise, kinda murky. It sounded like jitter… ouch.

This is because the original bit clock reference which the ES9018 uses to tune the DPLL is completely lost. It becomes “aliased” or masked if you will with the master clock.. In other words It sees the damaged data as perfectly good defeating the super cool algorithm ESS designed to fix jitter… yikes not what we wanted at all now is it.

This mod makes it super easy for the DAC to lock, but now every 20 or so samples (I am not sure exactly how many – it depends on the rate) the data contains errors.. bummer. We shifted clock errors into the data domain… Only now – to our dismay – the DPLL is missing the source clock context it needs to *fix* the data! Bummer. We effectively defeat the DAC at one of the things it does best, Rejecting jitter.

If you refer to the ESS DPLL/ASRC patents it starts to make perfect sense.

Further – It was explained to me that the idea itself completely redundant. The reason is that the DAC *already* time aligns all inputs with the master clock during the process of re-sampling. and even better when it re-samples it does so in a loss-less way. You cannot make the flip flip circuit loss-less – thus the data error.

So chalk it it up to a fun, but tragically failed experiment. A valuable lesson learned.

It was an interesting idea, but alas, not a good one. My advice is to feed the signal to the DAC – and let it do what it does best. Don’t try to fix what ain’t broke.

This video gives a good glimpse into why such a mod is redundant:

It also very interesting in regard to digital volume control.

Yes! Analog is technically better , but only if your implementation can manage a -133db noise floor. Good luck with that. It highlights just how good the volume control (and analog noise floor) of the ES9018 actually is.

Other DAC chips would be far better candidates for various re-clocking schemes.

Cheers!
Russ

The mental picture for this kind of reclocking was that the clock would be modulating the I2S signals with the 100 MHz frequency. The new signals would be like “pulse width modulated”. The question was how would the chip respond to that.

Therefore, the only viable solution at the moment is to use the source clock to synchronous reclock the I2S signals and to use the Sabre32 DAC in asynchronous or synchronous operation (which for a “low speed clock” results in the maximum sample rate that can be reclocked and the drawbacks of feeding the SabreDAC chip a “low frequency” clock)

ASL Wave-8: More Mods

I hope there are tube fans reading the blog. If not, do not worry. This is the only tube device I have . You can read part 1 here: [link]

In reality, I am not really that much of a fan of tubes but this amp has served me well in my secondary system and consistent with the theme here (“lot of value, little money”) the ASL Wave-8 is great value in audio (unfortunately no longer manufactured). But the mods are a lot of fun to do and improves this already good amp up “a notch or two”.

Even compared with the current, higher power version of this amp (the Wave-25), the Wave-8 still has “sweeter mids” according to the designer. The bigger brother of course has more power and extends further in its frequency response, but with mildly efficient speakers (the KEF Q15.2 I use are 91 dB), this is the amp to have.

Two primary mods for the ASL Wave-8 were described in the previous post, along with rearranging the ground connections:

• Better regulation of the power supply
• Better coupling capacitors

The PS mod is the most elaborate. Based on simulation, the ripple is reduced from 2V to apporx 20 mV, an improvement of 100X. A we know,  the quality of the power supply is perhaps the most important factor for good sound.

Based on simulations (Duncan’s PS Designer II -Scott sent me the simulation files) we can compare the output of the two power supplies:

Original PS: ripple ~ 2.4 volts

Modded PS: ripple ~ 27 millivolts

HUMMMMM

Hum has been a problem with this Amp since day one.  Many people have reported hearing hum and mine are not an exception. However it is low level and only audible if you put your ear close to the speaker. After the mods, which included rearranging the ground connection, I can report that the hum is much, much reduced. With a RS sound meter and only after putting the meter inside the port, I can measure 57 dB on the modded amp and 66 dB on the original, un-modded amp. If I put the meter right next to speaker surface (or anywhere near), the meter (which has a sensitivity of  50 dB) does not measure anything on the modded amp.

This is a reduction of 9 dB which translate to about 80% reduction in sound power. Very nice. According to some, it may not be possible to totally remove the hum in this design:

My experience has been that getting all the hum out of a push-pull amp can be difficult. When one tube of the push-pull pair is drawing more current than the other tube there will be a hum. I can hear this hum from about 1 foot from the speaker. I usually set the DC balance to try to minimize it. The Wave8 does not have a DC balance, so all you can do is to swap output tubes around such that each tube of the push-pull pair is producing equal amounts power. [link]

But more can be done…

There is an excellent tutorial on heater wiring. Essentially: heater wiring – the Good the Bad and the Ugly

• Tight twisting of the wires -to cancel the magnetic field (shielding reduces electric field but no magnetic field)
• Move wires away from circuitry

I redid both the AC wiring and the heater wiring. The AC wire goes from the back of the amp to the front where the switch is located. Originally I used stranded zip-wire with only 6 turns along the length of the wire. This time I utilized18 gauge solid core and twisted it very tight (at least 6 turns per inch). Solid core is better because it keeps its shape. [Tip: you can buy thermostat wire which has two 18- gauge wires for 24 cents/foot at the Home Depot]. Same for the heater wire. The first photo is the mains AC switch, the second photo is the filament/heater wire (which is also AC)

I measure the sound level as before and now I read about 56 dB. We have an improvement of about 1 dB. The original wiring was pretty good. In any case, a reduction of 90 % of hum (10 db) is already pretty good.

HEATER WIRING: we can do more

Here is the original heater wiring:

Here is a diagram and instructions from http://www.el34world.com/charts/commonhookups.htm:

If you do not have a heater center tap on your power transformer, you must run two 100 ohm 1/2 watt resistors to ground to create an artificial center tap. If you do not have a center tap, you will get 120 cycle hum. Each 100 ohm resistor is soldered to one of the heater wires. The other ends of the 100 ohm resistors are twisted together and then soldered to ground.

The heater wires are usually run up in the air, above the tube sockets in a twisted pair. Twisting the heater wires cancels hum. This is why phone line wires are run in twisted pairs. The twisted pair wires drop down and get soldered to the tube socket pins. The twisted pair continues down the line to every tube in the chain.

Keeping the wires in phase helps with hum sometimes. In other words, pin 7 on one power tube goes to pin 7 on the next power tube. Pin 9 on a pre amp tube goes to pin 9 on the next pre amp tube. EL84 power tubes heater connections are pins 4 and 5. Most other 8 pin power tubes use pins 2 and 7.

Also, there are some discussions on heater wiring with PCB (as opposed to point to point)

There is NO good way to get low hum with PCB heater wiring.

Live with it, or fix it. First verify that 6VAC flows in PCB “wires”, simple circuit tracing. If so, cut the heater lines AT the socket pins, leaving only the pin solder-blob. Get some hookup wire and run heater power OFF the PCB, in twisted pair cable AWAY from all audio points (basically everything except the heaters). Study some of the excellently-wired amps posted here. [link]

The original heater wires are connected “out of phase” (don’t know why. Perhaps to keep the wires as short as possible). Also, the artificial center tap is done at the end of the wire run with the ground connection at the end of the ground line. This means the current travels the whole length of the ground trace to the star ground. Will try implementing the artificial center tap near the power ground and also connect the wires in-phase.

Notice that the artificial center tap in the PC board has been removed (R18  and R19) and the heater wire  has been shortened, more twists and connected “in-phase”:

The artificial center tap is now implemented towards the back of the amp near the transformer, with direct ground wire to chassis:

Results: I measure 55 dB, an improvement of 1 dB.

ELEVEN dB REDUCTION IN HUM

All the AC and heater wiring modification resulted in about 2 dB reduction of hum. Not too bad. The total reduction combined with the previous mods is in the order of 11 dB. Remember that this measurement is sticking the meter into the port of the speaker. If we measure the sound pressure at the speaker, we measure 57 dB with the un-modded amp and unmeasurable for the modded amp (which should be 46 dB, 11 db lower).

How loud is that? I can start hearing the hum at about arm length (~32 inches) from the speaker with the unmodded amp and at about 7 inches with the speaker with the modded amp. This is with good ears and in the quietness of midnight… In the morning hours, the distances are about 24 inches and 3 inches respectively.

Happy New Year!

On OSF Bypass, DSD, 352.8K and 384K

I’ve settled (for now) on the Amanero interface with the reclocking mod and can manually change from the following options:

• Normal bitclock straight out from the Amanero device (supports every sample rate including 384KHz)
• Re-clocked non-inverted bitclock (supports sample rates up to 192KHz)
• Re-clocked inverted bitclock (supports sample rates up to 96 KHz)
• DSD 2.8 and 5.6 are supported by all configurations

With this board I decided to test oversampling bypass as indicated in the diyaudio thread [post 1061]

Previously using a Musiland Interface and with a Buffalo II DAC with the 80 MHz clock, bypassing the oversampling filter did not work for 352K material.

OSF BYPASS (OSF OFF) WORKS!

This time in this configuration, OSF bypass did work. The DAC played both 352.8 and 384K material perfectly. It was theorized that OSF bypass would work only with synchronous operation. Not so, it works even in “normal” asynchronous operation such as this case even with an 80 MHz clock

HIRES OVERSAMPLING (OSF ON)

The oddest thing is that now, both 352.8 and 384K material works with the oversampling filter ON! Although occasionally it would cease to work and exhibit the same “stuttering” behavior I experienced before. Here (with OSF on), switching in and out of 352.8 and 384K material and between these sample rates would result in a loud noise and glitches.

Thus in general OSF does not work with 80Mhz clocks and 352/384K material. In such situations you can bypass the OSF.

DSD and OSF BYPASS?

If the source material is DSD, bypassing the OSF would result in no sound.

NOS?

Not only does it work with the high frequency material, but the oversampling filter bypass also works with everything else. I tried the OFS bypass setting on 44.1K and it works. How can I tell?  It sounds different. To my ears, bypassing the oversampling filter removes the “sparkle” from the DAC. Is this the “NOS” sound some people prefer?

USING AMANERO I2C TOOL TO INSTRUCT THE DAC TO BYPASS OSF?

The Amanero board can issue I2C instructions to any slave device. It can for example (and as proposed in the diyaudio discussion thread) that when the sample rate is 352 or 385K, the Amanero board can send the instructions to the Sabre32 DAC to bypass the oversampling filter. Unfortunately, it is not so simple as the register for the OSF also takes care of other parameters as shown below:

 |1| | | | | | | | Mono Right (if set for MONO)
 |0| | | | | | | | Mono Left (if set for MONO) (D)
 | |1| | | | | | | OSF (Oversample filter) Bypass
 | |0| | | | | | | Use OSF (D)
 | | |1| | | | | | Relock Jitter Reduction
 | | |0| | | | | | Normal Operation Jitter Reduction (D)
 | | | |1| | | | | SPDIF: Auto deemph ON (D)
 | | | |0| | | | | SPDIF: Auto deemph OFF
 | | | | |1| | | | SPDIF Auto (Only if no I2S on pins) (D)
 | | | | |0| | | | SPDIF Manual (Manually select SPDIF input format)
 | | | | | |1| | | FIR: 28 coefficients (D)
 | | | | | |0| | | FIR: 27 coefficients
 | | | | | | |1| | FIR: Phase invert
 | | | | | | |0| | FIR: Phase NO invert (D)
 | | | | | | | |1| All MONO (Then select Mono L or R)
 | | | | | | | |0| Eight channel (D)

In order to send the correct register value, you need to know the current setting of the register (in the simplest configuration, you can assume that the values are fixed, but in a general configuration, the values would not be fixed).

One way to do this is for the Amanero board to send a value to another I2C device. For example a port expander, turning high or low certain pin. Then the Arduino can read these pins and turn the OSF on/off.

SOFTWARE SUPPORTS OSF BYPASS

The ^ mark indicates that the oversampling filter is on. In select mode, you can turn-off the OSF and you will see a “.” in that position

Note: A reader reported a bug on the OSF bypass code. If you have a dual mono configuration, take a note at the following section of the code:

else {
  bypassOSF=true;
  bitSet(reg17L,6);
  bitSet (reg17L,5);
  writeSabreLeftReg(0×11,reg17L);
  delay(50);
  bitClear(reg17L,5);
  writeSabreLeftReg(0×11,reg17L);
  #ifdef DUALMONO
  bitSet(reg17R,6);
  bitSet (reg17R,5);
  writeSabreRightReg(0×11,reg17R);
  delay(50);
  bitClear(reg17L,5); //“reg17L” should be “reg17R” here.
  writeSabreLeftReg(0×11,reg17L); //“reg17L” should be “reg17R”
  #endif DUALMONO

How Good is ES9018 in Voltage Mode?

I have been using my BII DAC in voltage mode since day one (that is no output stage, the output of the DAC is connected straight to UCD-180HG amplifier modules).

JACQUES HIFI PAGES has a a comparison of the Sabre32-based Buffalo II DAC used in voltage mode compared with AKM AK4397-based Xindak DAC. The reading is at least entertaining…(note: I have edited some of the content I pulled from the website for clarity):

I eventually could buy two 32bit dacs which are scarce , expensive and hard to locate…

I don’t think you need presentation for the 2nd (DAC) one which is the famous Twisted Pair Audio Buffalo II featuring the not-less-famous ESS Sabre 9018 “where-are-you” dac.

This chip seems to be the hardest to locate on the planet and I really can’t understand why. Usually , once a chip is invented, it costs nothing to make a million of them, as it is just a photocopy on carbon wafers. But in this case, it is just like somebody makes them one by one in a small workshop. For example, the Oppo DVD players which supposedly use this chip are always discontinued or not yet available (out of stock?). And Twisted Pair Audio is constantly putting people on waiting lists, while other DACs are available by kilos. Great mystery.

The other one is a masterpiece crafted by small chinese hifi factory Xindak for their 20th anniversary. Quite expensive but made on very high standard as you will see. Let’s start the fight !

The contenders:

THE XINDAK 20TH ANNIVERSARY DAC

This monster cannot sound bad and I really enjoyed it from the 1st minute, with its always musical silky precision.

THE BUFFALO II DAC

In comparison the Buffalo really looks like nothing: a small circuit just like the 25$ generic dac kit from China. It seems well implemented and it has no output section so  you hear what comes out from the DAC—(the DAC is) not visible on the picture; it is under the improved analog supply (the AVCC board) according to Twisted pair.

They say the DAC is I (current) out but works great like this….nothing like the classic I-out DACs like the TDA1541 where there’s not sound at all with direct output.

Here is a better photo of the PS used for the Buffalo II setup:

Note the following:

• SPDIF input
• Sabre32 is operating in voltage-out mode
• Only half of the output is used (“single ended” output connection: + and GND): internally, the Sabre32 has 8 pairs balanced dacs (16 dacs in total), each pair provides the in-phase and the anti-phase output. Only taking the in-phase output means that only half of the DACs are being used.
• Based on the photos, the PS is very likely based on run-of-the-mill 78XX linear regulators.

RESULTS

Well I’m sorry to say the Buffalo II DAC beats the Xindak DAC. Sorry because it cost me more money. The Sabre32 DAC circuit does not feature up-sampling but straight out of an old Philips player (with CDM 4/19 of course, you should know my drive taste by now) it really brings CD sound to something I have never heard before.

Of course the Xindak sounds great, silky, analytic but the Sabre, plugged crudely from its output to my amp, is 10 times more analytic… and it was really what I looked for.

2 years in searching a super-precise source for my super-precise Elipson 4140 speakers and it’s done at last. Funny how 40 years separate those speakers with their electronic companion.

Sorry again because to improve my system I should buy another pair of big 70′s Elipson monsters. And again because those buffaloes are as hard to buy as a real buffalo from USA….

The fact that the Buffalo II DAC is not only used in voltage-out mode but it is “crudely” connected to the amp as the reviewer said, says a lot about the capability of the Sabre32 DAC even in voltage mode

According to the Sabre white paper:

The highest performance in terms of THD is via the current mode, but both voltage and current mode provide about the same DNR.

• DNR better than -132 in both voltage and current output modes
• THD: -108db in voltage output mode, -120db in current output mode

THD in the Stereo current mode is limited by the external components and measurement equipment. We recommend using an extremely good op-amp for the highest performance but even an excellent op-amp is the limiting factor in the THD.

MY SETUP

Like I mentioned, I’ve been using  the DAC in voltage mode since day 1 basically for several reasons:

• Don’t feel I am missing anything. Have been impressed at the clarity of presentation that this DAC provides. (I guess you don’t miss what you don’t know?)
• Lack of motivation (lazyness? ) to build-up my Legato 3.1 kit which is still in the box.
• The fact that an output stage has the additional requirements of a bulky bipolar PS and transformer also adds to that “lack of motivation”
• Simplicity of a direct connection from DAC to amp. The “nothing in between” concept sounds really appealing
• And now, in an A/B comparison the Sabre32 DAC in voltage mode beats a very capable DAC

Further, I am still not convinced if an output stage is a “step up” or just a “side step”.

First, the two output stages offered by TPA have their own pluses and minuses. Between the IVY and the Legato,  the Legato output stage seems the favorite with respect to sonics. However, the Legato differential input is unlike that of a differential opamp (of the IVY) where there is rejection of common mode noise. According to Russ [post 602]

Legato is simply an I/V stage (actually 2 pair of independent non-inverting I/V stages) output is essentially the same as the input just converted from current to voltage. Pretty much exactly what you would get with a plain resistor at the output pins to GND – only the output of the DAC does not modulate to the same degree because of the low impedance. As with the simple resistor it will come down to the output levels/value matching. If there is a slight mismatch there will be some differential error given a common mode signal. So essentially any common mode signal is still there at the balanced outputs. But that’s what’s wonderful about balanced signals, it does not matter in the least. On IVY-III the output common mode is dictated in large part by the device. That is why it is not a straight up comparison…

One nice thing is because most of the common mode noise on the ES9018 is extremely high frequency (far above audible) the passive filters actually get rid of almost all of it before it can really come into play. So practically the “CMRR” (if you want to call it that) is superb. That’s why its important to have those low pass filters. Even if there is a small error between the balanced halves those very high frequencies in play would have been filtered out.

Second, the THD of the entire audio chain is determined by the THD of the amplifier. According to the specifications of the UCD180HG the best number for THD is 0.008% which translates to -82dB. Better THD measurements at the DAC stage may or may not translate into any perceivable difference after the amp. In addition, the UCD amp is very easy to drive due to its high input impedance.

Third, the very capable Wolfson WM8741 DAC and the PCM1792A, both top of the line from their respective companies, have  THD specifications of -100 db. The ESS DAC already beats them by -8 db in voltage mode.

And fourth, I can expect even better performance that the DAC in the review because of the configuration I am using:

• I2S input with lower DPLL bandwidth setting
• Full balanced connection to the amp
• Digital volume control
• Placid shunt regulator powering the DAC

So, as a whole, is there any more real improvement to be had by running the DAC in current mode?

Daily Free HiRes Until Christmas

Here: http://christmas.linn.co.uk/

Join the discussion here: http://forums.linn.co.uk/bb/showthread.php?tid=19936

Free hires tracks are in different sample rates, from 48 KHz to 192KHz…

 

Thank You, Linn ! ! !