Beyond the Ariel

I find it unremarkable that amps and speakers you have designed yourself sound better than everything else.

Really? Have I claimed that my amps and speakers "sound better than anything else"? That's a clever rhetorical tactic, but it's not what I said, in these or any other postings.

What I actually said is that I design my equipment for my own tastes. Not the tastes of an employer, my neighbor, a forum, or a magazine reviewer. If it sounds good to me, I consider it good. I have no way of knowing what other people hear; for that matter, I've only met a few audiophiles whose sonic and musical tastes are similar to my own.

I would imagine nearly everyone here on DIYaudio does pretty much the same thing, since we don't have to answer to a manager or a marketing department. That is one of the great luxuries of DIY; not to save money, but build what is not on the market. What's the point of copying commercial designs, which are almost always compromised to reach a price point, or aimed at a certain narrow demographic?

The majority of high-end products, in the USA at least, are designed for the tastes of a small circle of reviewers who write for the Big Two magazines. These magazines, even today, have the power to make or break a manufacturer. This is not a fairy tale; I've met the people whose businesses were destroyed by a bored reviewer, or an editorial rewrite. (The tell-tale signs of a clumsy editorial rewrite is when the review suddenly changes tone in the last three paragraphs, or when the final conclusion in the review contradicts the middle part of the review. That's a quick-n-dirty rewrite by an editor in a hurry.)

The DIY community is free of these commercial pressures, or the need to satisfy a reviewer. We can design products as weird as we like. They don't need to be manufacturable, or aimed at the tastes of the Big Two magazines.

It is only ethical to warn people who are curious about building the Ariel, Amity, or Karna that they do not sound like the majority of equipment at a high-end show; if they expect to replicate what they hear at hifi shows, they will be disappointed.
 
Last edited:
Riippppppp

A more exotic version would use a 1:2 Input Transformer -> PP 45, 2A3 -> 1:2 IT Transformer -> PP MOSFET followers -> Output Transformer. This driver would have (much) lower distortion and more dynamic drive characteristics. The B+ would be raised to the 200~250V range for the best possible dynamic range and slew rate from the balanced pair of direct-heated triodes.
<snip>
That's a complete hybrid amplifier; anyone interested?

You here that noise? It's me tearing up the current set of plans and starting all over again. Moving the 2A3s back to the driver stage....
 
Modern single-plate 2A3's have substantially lower upper-harmonic distortion than the historic bi-plate 2A3's, and really behave more like a souped-up 45.

When you ignore the 2nd-harmonic distortion (which is nearly inaudible), there are quite substantial differences between tubes. Gary Pimm and I look at the 5th-and-higher harmonics; you'll see differences of 10~20 dB in these high-order harmonics, depending on type and vendor.

Changing the quiescent current by 50% will only shift the upper harmonics by 3 dB or so, but swapping tubes can result in a 20 dB difference. Within vendors, distortion is tightly grouped, so the differences are clearly the result of proprietary manufacturing processes.

This is the area where the best DHT's pull ahead of the indirectly-heated driver tubes, and not by a small margin. For example, as good as the 5687/7119/ECC99 family is, the 45 DHT has 5th-and-higher-order harmonics that are 20 dB lower. Gary and I looked at the -130 dB noise floor of the instrumentation, and we could not see any evidence of 5th or higher at all. These things are clean - and to me at least, they sound the way they measure. The modern Chinese meshplate 2A3's are very good as well, and will cheerfully operate at substantially higher currents than a classical 45.

Be careful about microphonics. DHTs are pretty sensitive to vibration thanks to very large grid structures, which are not as rigid as conventional pentodes or indirect-heated triodes. PP operation does not cancel vibration all that well, thanks to the large spacing required by the physically large envelopes (and an additional inch or so for adequate ventilation). In more technical terms, there's a phase difference between the vibration experienced by each grid, and that results in loss of PP cancellation.

I am not familiar with the spectral signature of MOSFET followers, particularly what happens with 5th-and-higher order harmonics. If the upper-harmonic levels are moderate to high, there's no point in chasing after DHT's and their microphonic requirements, as well as hassles with good-sounding DC heating. The whole point of DHTs is their very low upper-harmonic content; if that is masked by the upper-harmonic distortion of the MOSFET followers, a triode-connected EL84, 6L6, or EL34 might be a better choice as a driver. Build it and see for yourself.
 
Last edited:
With the output transformer wired in 4:1 mode, the output power into a 4 ohm load is only 27 fast Watts. Damping factor is 29. That's the mode I was thinking mated well with this new speaker.

Oops. Above I specified a damping factor of 29 for the OPT wired 4:1 stepdown. Actually, the damping factor in this mode is 16 for the load of the Beyond the Ariel speaker. (Output Z is 0.25 ohms.)

In case it matters....
 
I am not familiar with the spectral signature of MOSFET followers, particularly what happens with 5th-and-higher order harmonics.

If you poke around Susan's www.audiophonics.com website, you will find actual fft measurements. Look at the 4:1 modes, which we'd be using for this speaker.

I also look at harmonics 5 and higher. I would not have built the amp if the spectral signature were not favorable. As I recall, distortion was low and concentrated at 3rd harmonic (transformer-dominated distortion).

Noise is in the range of -150db. Yeah.
 
I would imagine nearly everyone here on DIYaudio does pretty much the same thing, since we don't have to answer to a manager or a marketing department. That is one of the great luxuries of DIY; not to save money, but build what is not on the market. What's the point of copying commercial designs, which are almost always compromised to reach a price point, or aimed at a certain narrow demographic?

True of the boutique transformer designers you can find on diy Audio too. Tribute, Intact Audio, O-Netics, Lundahl and now Susan have all moved beyond the typical simple model hobbled recipe. Each of these designers provide extremely good sounding transformers. Each of them emphasize quite different characteristics, in how much of what, they allow the transformer to actually transform.

The range available is quite wide and deals with internal gradient structures in tone and transients and overall timing issues more than it does with low distortion or tonal purity, those being a given. I am sure there are plenty of other manufacturers that denizens here will be eager to list, I will also point out that I know these few to be at the top of that heap. How high you rate one of them over another will be determined by those characteristics Lynn points to.

I would suggest that the output transformer will have an even greater impact upon the amplifiers eventual sonic character and how you respond to it than any other component. A world renowned designer of tube amplifiers I know says that the output transformer is 90% of what is wrong with a given amp and 70% of what is right. Of course, he designs guitar amps, but nearly all of the professional musicians we love to listen to own his amps, along with a huge number of just normal folks who haven't got the god light shinning on them.

Bud
 
If you poke around Susan's www.audiophonics.com website, you will find actual fft measurements. Look at the 4:1 modes, which we'd be using for this speaker.

I also look at harmonics 5 and higher. I would not have built the amp if the spectral signature were not favorable. As I recall, distortion was low and concentrated at 3rd harmonic (transformer-dominated distortion).

Noise is in the range of -150db. Yeah.

Good website. Interesting that Susan was working on the Zeus at the same time I was working on the Amity amplifier, and had access to the same test gear (my Tektronix pals have superb HP and Tek instrumentation). Back when I was using the Amity amplifier, the transformer-coupled solid-state Rowland linestage was one of the very few preamps that could pass a direct-bypass test; most of the solid-state preamps added a layer of grit-n-grain and a metallic-spunding brightness, and the many (but not all) of the tube preamps were mushy and dull-sounding.
 
Complementary AB Mosfet common source

Thanks Lynn. Well, Nelson's Class A MOSFET amps take the output from the drains. Per Susan, drains are a bad place from which to take output because 1) gain is involved and transistor gain colors sound, and 2) the drain capacitance is large (amplified by Miller) and non-linear if the voltage swings are large (which adds a kind of "jitter" so to speak).

As part of my studies in the eighties I was looking at MOSFET complementary common source amplifiers - in class AB - to improve the efficiency of the amplifiers by dropping the rail voltage by the 12V of gate voltage.

Suffice to say that it was a very pointed lesson in the limitations of feedback, small signal linear models and even the Spice models we had then (30 years ago). The non-linear capacitances around turn off made a mess of any attempt at stability.
 
I'm very interested in your notion that the Ariel is, er, rubbish with a transistor amp, Lynn. Especially as I've decided that the Vifa P13WH is about the best polycone in town and want to buy some! I'm also planning on adapting your crossover for a small "BBC LS3/5A" type 2-way. :D

Surely that is just tonal differences with a valve (and significantly high output impedance) optimised speaker? But we are currently interested in the very symmetrical Susan Zeus power Mosfet amplifier, so I'll go with that for now...:)

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


Now never mind discussion of harmonic distortion artefacts and Class A or A/B, what leaps out at me is the evidently superb common-mode rejection from those expensive centre-tapped transformers. That along with whatever output impedance it has must define it. For those who are unfamiliar, Common-mode is noise picked up from the earth rail. Earth is in fact a vague concept, it is better to think of it as a reference voltage. You must then ask yourself what reference the output signal is relative to. Sometimes it is the power rail, sometimes it is the conventional earth.

Common-mode rejection is a huge topic, made worse by the need for volume controls and circuit input/output protection. For interest I include a Balanced Differential amplifier circuit which is happy as long as both arms are symmetrical reactively and driven by zero source impedance.

You also ask yourself whether an op-amp circuit driven by a long tailed pair is going to be better or worse than one of John Linsley-Hood's simple little one transistor Class A line amps. I know which I'd soonest have. JLH. :D
 

Attachments

  • Susan_Zeus_Power_Amplifier_Schematic.JPG
    Susan_Zeus_Power_Amplifier_Schematic.JPG
    32.8 KB · Views: 959
  • balancedopamp.JPG
    balancedopamp.JPG
    29.3 KB · Views: 1,146
  • Long_Tailed_Pair_Amplifier.JPG
    Long_Tailed_Pair_Amplifier.JPG
    21.9 KB · Views: 1,141
  • J. L. Linsley Hood Amplifier.JPG
    J. L. Linsley Hood Amplifier.JPG
    24.3 KB · Views: 939
I'm very interested in your notion that the Ariel is, er, rubbish with a transistor amp, Lynn. Especially as I've decided that the Vifa P13WH is about the best polycone in town and want to buy some! I'm also planning on adapting your crossover for a small "BBC LS3/5A" type 2-way. :D

Surely that is just tonal differences with a valve (and significantly high output impedance) optimised speaker? But we are currently interested in the very symmetrical Susan Zeus power Mosfet amplifier, so I'll go with that for now...:)

Not really a matter of output Z; it's easy enough to add a 1~2 ohm series resistor to a conventional transistor amp - and all it sounds like is an even worse transistor amp, with degraded dynamics and more edge and grit than usual.

Change to nearly any valve amplifier, and the feeling is immediate relief, along with far more vivid tone colors and more interesting dynamics. Not a small difference at all.

The Ariels, for better or worse, reveal substantial differences - sometimes amazingly large - between amplifiers. I certainly didn't plan on this, but that's how it turned out. I've heard transistor amplifiers sound everything from murky and dull, with unintelligible vocals (Krell), to unbearably bright and zippy (some prosound amp with internal fans). Most just sound somewhat bright, with noticeably flattened tone colors, like MP3 at 128kbps.

Valve amplifiers restore musical tone colors to real-life levels, but many have lots of obvious added-on colorations of their own. SET amplifiers can sound tubby, lush, or edgy and hyper-detailed, depending on how the driver stage and power supply were done. PP amplifiers are typically more neutral, although higher-watt (6550, KT88) PP pentode amps can have a fatiguing, grainy quality, particularly if 6DJ8's are used as drivers.

Some of the SET amplifiers combine neutrality with vivid tone and dynamics, but they tend towards the exotics - 845's and larger tubes, which require heroic drivers and power supplies. It's easier to get there with moderate-power pentodes and direct-heated triodes. Small details of phase-splitters have a large effect on PP sonics. Not a fan of the "concertina" or split-load phase splitter at all.

One thing I can tell you; despite moderate efficiency, if you use the Vifa 5.5" driver in MTM pairs with a good crossover, you will hear very subtle differences between amplifiers that are simply not audible with most commercial speakers. What would be a slightly noticeable difference on say, a B&W high-end speaker, will be glaringly evident on the paired Vifa's. I surmise the very good objective performance of the Vifa - which has superb, near-electrostat time response - allows the listener to hear subtle differences more easily, while the time-decay crud (resulting from out-of-band breakup) of most midrange drivers masks amplifier differences.

Strictly a guess, but drivers with very good time-decay performance let your hear "deeper" into the amplifier, revealing subtle problems with balance under dynamic conditions, or nonlinear capacitances that affect forward-path time delays (in other words, stability margins that are program-modulated).

Although it's kind of a mixed blessing, the Ariels do make good analytic tools to evaluate small changes to amplifier topology, or audition different flavors of DACs. You will hear the difference between delta-sigma and ladder DACs, for example.

As for the Susan amplifier, I have no idea at all what it would sound like. I've never heard one. It is radically different than a complementary-symmetry Class AB feedback amplifier, so it's a safe guess it sounds different as well. In what way, though, I have no idea. When I first auditioned the Amity amplifier, I had no idea what it would sound like - I just knew it would sound different than a Williamson, Dynaco, Fisher, or Marantz, and it did.
 
Last edited:
I'm very interested in your notion that the Ariel is, er, rubbish with a transistor amp, Lynn. Especially as I've decided that the Vifa P13WH is about the best polycone in town and want to buy some! I'm also planning on adapting your crossover for a small "BBC LS3/5A" type 2-way. :D
The "susan" is going to be brutally honest about the qualities of the output devices.

As with anything driving a 90db plus efficient speaker the first watt is going to be the most important - where any heavily "A" oriented amplifier is going to "win". Beyond that, with out feedback you're in a trade off between microdynamics (my words for Lynn;s "can you hear the room") vs macro dynamic - where the chorus and the orchestra and the organ get serious. And I/M and its friends enter the room. Think about 6 minute into the adagio for spartacus or any massed chorus (eg Carl Orff's Carmina Burana )
 
Thanks for the clarification, Lynn. I think you're just saying the Ariel WORKS BETTER with valve amplification. :cool:

On a rare foray into the Solid State section here, we had a look at Mark Houston's DIY Class-A 2SK1058 MOSFET Amplifier.

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


Quite a valve-like 8 ohm output impedance and Class A current drive via Mosfet with a mere 5W available. Nelson Pass would look for something more special in output devices, but it's the efficient loudspeaker and tonal response that needs some thought with this sort of amplifier. As thoglette remarks, some speakers like the low inductance Fostex full range can lose it a bit with complex music.

Enter the WLM La Scala again. I find this speaker fascinating. 2.2mH inductance Visaton W200S 8" bass and of all things, a Visaton TW-70 CONE tweeter. Almost a relic of the valve driven sixties.

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


The only filter I could get working properly was a simple one, quite Robin Marshall in fact. I used the old trick of adding a 6 ohm resistor to estimate the difference when driven by a valve amp. The bass loading that was best and least affected with either type of amp (Solid State or Valve respectively in the diagrams) was stuffed port aka aperiodic, here with 30 litres box. I suppose it is possible to use Zobels and small bass coils to get the impedance flatter so it is tonally similar with all amps, but I hope you can see how it works anyway. :)
 

Attachments

  • WLM_LaScala_Clone_Circuit.JPG
    WLM_LaScala_Clone_Circuit.JPG
    22.1 KB · Views: 207
  • WLM_LaScala_Clone_Transistor_Drive.JPG
    WLM_LaScala_Clone_Transistor_Drive.JPG
    60 KB · Views: 161
  • WLM_LaScala_Clone_Valve_Drive.JPG
    WLM_LaScala_Clone_Valve_Drive.JPG
    61 KB · Views: 139
For anyone interested in a PP source follower power stage, including Lynn's suggested hybrid, here are some interesting graphs.

First the Current through the Mosfets in my hot-rodded version of Susan's amp. Shown are the push and pull Mosfet currents with the output transformer configured 4:1 delivering 5.5v peak into 4 ohms. This is close to 3.8 Watts which is much louder than needed for the Beyond Ariel speaker, I think.
The currents swing from around 1.1 to 1.75 amps.

Next is the capacitance curve for the Fairchild FQA34N20 Mosfet. To address the non-linearity of MOSFET capacitance, it is desireable to pick a part that has nearly non-changing capacitance in the operating region. Note the bottom right of the graph where I have annotated the 40v bias point for the amplifier. The controlling capacitance for this follower topology is CRSS (reverse transfer capacitance). CRSS is found in the lowest line in the graph and is specified as 55pf typical, which is where I operate it. Notice how this capacitance is effectively flat over the entire range, which range is +/- 11volts about the 40v operating point. The +/- 11 volts are need to provide the 5.5 volts at the output side of the OPT in this example, the condition for the current waveforms of the first graph.

This Fairchild part if effectively flat for a range that is double that, +/- 22v (11v peaks at output). Even +/- 30volts isn't bad (15v peaks output).

How loud are a pair of Beyond Ariel speakers at two meters with +/- 15volt peaks at the terminals?

The area to avoid on the second graph is below 10v. Note the Y-axis is linear.
 

Attachments

  • PP Mosfet currents OPT4to1 delivering 5.5v peak into 4R.jpg
    PP Mosfet currents OPT4to1 delivering 5.5v peak into 4R.jpg
    63.9 KB · Views: 141
  • FQA34N20 capacitance.jpg
    FQA34N20 capacitance.jpg
    146.6 KB · Views: 116
Last edited:
Interesting how this thread is turning in an amp thread.:D
Lot's of problems with amplifiers is that the amplifier and power supply are treated as separate components. Secondly, whenever we talk about the source end of either amp output or power supply output, we ignore the return aspects of the current flow. These ignorance lead to a trend where devices that have more separation between the high current loops and low current loops tend to sound smoother. Simple device have simple current loops, which make optimization much simpler.
 
The new loudspeaker is around 99~100 dB/metre/2.83Vrms, with a load between 12~20 ohms over the working range.

The bass driver is an Altec/GPA 416B (Alnico magnet, 16 ohm version), the HF horn is an Azurahorn AH425, the large-format compression driver is either a Radian 745Neo (Neodymium magnet, 16 ohms) or Altec/GPA 288 (Alnico magnet, 16 ohms), and the passive crossovers are 3rd-order lowpass at 700 Hz (transitional between Butterworth and Bessel) for the 15-bass driver, and a 4th-order highpass at 700 Hz (transitional between Bessel and LR4) for the HF driver.

If the Radian 745Neo is used, it needs to be attenuated by 14 dB, which is accomplished with a custom autoformer with a 16-ohm shunt resistor between the autoformer and the lowpass filter. The highpass filter is designed for a 16-ohm resistive load.

The woofer uses a standard Zobel inductance compensator between the lowpass filter and the woofer, and the lowpass network is designed for 12-ohm resistive load.

No inband equalization is required for either the woofer or HF horn system. The system is time-aligned with all drivers in-phase; the lip (outer edge) of the AH425, by lucky coincidence, is about 1/2" forward of the front face of the bass cabinet, with about 2" of vertical spacing between the top of the cabinet and lower edge of the AH425. The bass cabinet is about 4.5 cubic feet and uses a resistive vent.
 
Last edited:
Sounds like a sexy design and perhaps a worthy alternative for those who have aspirations to collect the bits and pieces to build a M19.. GPA has now just released their own crossover for the M19, making it a one stop shop for all the bits except the horn etc..

I read somewhere that Bill at GPA says there is really no audible difference between the Alnico 416 and the Ferrite version, so maybe that could help shave the cost down a bit for some?

:cool:
 
The new loudspeaker is around 99~100 dB/metre/2.83Vrms, with a load between 12~20 ohms over the working range.

Regarding the new speaker being nominal 16 ohms... If that's the case then 4:1 output mode on any Susan Zeus-type power output stage may not be optimal. Instead, 2:1 might be optimal.

Either way, 16 ohms reflected back on the primary side of the OPT will look like either 256 ohms or 64 ohms. Either are a piece of cake to drive and will be very low in distortion.