The First Watt Model F1 Current Drive Power Amplifier
by Dick Olsher
And now for something totally different. This report is unusual in that it combines a product review with a Do-It-Yourself (DIY) project. There is a very good reason for doing just that , as you will discover shortly. Also unusual is the amplifier itself, the brainchild of Nelson Pass, and as far as I know, the only current-drive design currently available to the consumer. Before you get the idea the F1 is in fact an ordinary consumer product, take heed of the following excerpts from the F1 owner’s manual, penned by none other than Nelson himself.
“Who is First Watt? That would be me. I’ve been building audio for about 37 years because that’s what I do. While I have a nice business (Pass Labs) to tend in the daytime, sometimes I come home and can’t resist making something. First Watt is a kitchen table operation – the projects that are not appropriate for my day job. I like power amplifiers best, and I have the manufacturer's mentality - never build one when you could make a hundred. The inspiration for the company name and type of product comes from Dick Olsher. Dick is a long time hard-core audiophile reviewer whom I respect greatly, and he once said, ‘The first watt is the most important watt.’ I have more recently heard it expressed as, ‘Who cares what an amplifier sounds like at 300 watts when it sounds like shit at 1 watt?’ (At my kitchen table I can say any word I like as long as it’s OK with my wife). This philosophy dominates First Watt, as does the belief that simple elegant designs are the most worthy of my time and your money. There is a vast range of possibilities that fit that description - I am ferreting out some of the most interesting.” Note that there will only be 100 F1’s offered for sale to the public, and then Nelson intends to make a different design. He says that he mostly built this amplifier himself, and that if you find any fingerprints in it, they’re most likely his.
Thanks, Nelson, for the kind words. First Watt is indeed a cool name, and the F1 is Nelson’s first “kitchen table” design in the First Watt series – a 10 wpc, stereo transconductance power amplifier; a fancy name for an active current source. The F1, unlike a voltage-controlled amplifier, attempts to sink a constant current into the loudspeaker - irrespective of variations in the load impedance. There is only one gain stage, operating in pure class A mode: 100 watts per channel are dissipated to deliver a precious 10 watts to the speaker. The amplifier runs hot to the touch and requires plenty of ventilation. And, of course, no feedback is used to reduce distortion. Despite the lack of feedback, this is probably the quietest amplifier money can buy, with a typical figure of about 100 pico-watts noise (pico-watt is one trillionth of a watt). Low-power and low noise, would appear to mark the F1 as an ideal candidate for driving a high-efficiency loudspeaker. Unfortunately, as Nelson himself readily admits, that it will “not sound good with about 98% of the loudspeakers on the market.” The problem is that loudspeakers are typically designed to work with ordinary, voltage controlled, amplifiers, and will most likely exhibit significant bass and treble response boosts when driven by the F1. Nelson wants you to know that “This is a tinkerer’s amplifier, and experimental tool. If you just want to be a consumer, then buy something else, otherwise be prepared to patiently experiment with your system to get the best performance.”
This brings me back to the Leopard and Panther variations on the BassZilla theme. I was so fascinated by the sonic potential of the F1, as well as being painfully aware of the practical problems faced by intrepid experimenters, that I plunged head first into developing two new speaker designs – specifically optimized for the F1 and other (future) current-drive amplifiers. A few months later, it is now possible to say that there are at least two loudspeaker that you can build, that work splendidly with the F1.
Current Drive Demystified
Mills and Hawksford coined the term current drive in 1989, in two essential papers published in the Journal of the Audio Engineering Society, dealing with Current Drive Technology. To understand the beauty of current drive, one needs to first understand how a conventional voltage-driven, moving-coil, loudspeaker operates. The current into the speaker is limited primarily by the voice coil resistance and inductance. The voice coil current interacts with the magnetic field generating a motor driving force. As soon as cone motion occurs, an opposing motional impedance is generated, which further limits current flow. From the amplifier’s perspective, it is doing its job, since it strives to maintains a constant voltage at the loudspeaker terminals for a given volume level. It should be obvious that linearity between the driver’s cone velocity and the applied signal is dependent on the impedance of the various electrical elements in series with the voice coil. And that distortion would therefore result if the signal itself happened to induce an impedance change. A major culprit turns out to be the voice coil resistance, which is not a constant, but actually varies with the audio signal.
How is that possible you ask? It is well known that most of the power delivered to a speaker is dissipated as heat in the voice coil. Speakers are poor transducers of electrical energy into acoustical energy. A typical conversion efficiency for a domestic speaker is about 1% to 2%. That means that up to 99% of the signal is spent heating up the voice coil. Voice coil hearing has been estimated as being on the order of 0.4% per degree centigrade for copper wire. It is not uncommon for voice coil temperatures to spike up to 200 degrees centigrade, at which point a 6-ohm nominal voice coil resistance increases to 10.3 ohms. As the voice coil heats up, the power delivered by a voltage-controlled amplifier decreases. This leads to loss of sensitivity with increasing drive signal, also known as power compression. Since the loss of sensitivity would be most prominent in regions of minimum impedance, shifts in frequency response would also occur with drive signal level. For a woofer, the impedance minimum corresponds to the lower midrange, which implies tonal balance shifts as the vice coil heats up and cools down. In addition, the increased resistance reduces driver damping and leads to misalignment of crossover networks.
Current drive offers a major alternative to the speaker-amplifier interface. A transconductance amplifier is operated as a current source with a high source impedance relative to that of the speaker. The output impedance must also be linear and frequency independent. An ideal current source would possess an output impedance so large as to be considered infinite relative to that of the speaker. The F1’s 80-ohm impedance is certainly adequate in this regard. As a result, as Nelson puts it: “the amplifier delivers a precise current to the voice coil of the loudspeaker driver, ignoring the series impedance elements in the circuit, including the wire, the inductance of the voice coil, the resistance of the voice coil versus temperature – all that stuff.”
Current drive offers the potential for dramatic reduction in moving-coil speaker distortion due to power compression and other voice-coil non-linearities. However, there are two basic practical problems to overcome. On the amplifier design side, it turns out that it is difficult to design a high-power current source. The high-power designs investigated by Mills and Hawksford are rather complex. The simplest approach lends itself to low-power designs, which fortunately fits nicely into the high-efficiency loudspeaker scene. On the speaker design side, things are also different. For example, the ploy of reducing a driver’s rated impedance from 8 ohms to 4 ohms in order to gain 3 dB in sensitivity does not work with a current source amplifier. Instead of doubling the current draw, as would be the case with a voltage-controlled amplifier, the current draw remains the same, which means that sensitivity is actually lost. On the other hand, increasing impedance leads to increased sensitivity with current drive. Consider that a woofer’s resonance region is no longer current limited and sinks the same current as the frequency band above resonance. That makes for a huge boost in bass response around the resonance region. The bottom line is that using drivers of differing nominal impedance but identical sensitivity ratings will result in output mismatches. Finally, conventional parallel type crossover networks do not work well with current sources. I spent a weekend convincing myself of that. Instead, series type networks are necessary for optimal performance. No wonder then that current source amplifiers require matching loudspeakers, designed for a current drive interface, and capable of taking full advantage of their distortion reducing potential.
The BassZilla Felines: The Leopard and Panther Loudspeakers
The BassZilla loudspeaker family embodies a strongly held design idea that a midrange should operate as a dipole radiator. For the past several years, I have sought to place the midrange in an open baffle. Maybe it is a case of cross pollination, the result of a life-long love affair with planar speakers. The old QUAD ESL transformed many of my perceptions about reproduced sound, and many years of exposure to the Sound Lab A-1 have convinced me of the superiority of a dipole midrange. In addition to the avoidance of boxes, I also object to placing a paper-cone driver in a horn’s throat/compression chamber, where it is subjected to severe buckling pressure.
The BassZilla definitely thinks outside the box! Good things happen when the midrange is freed from its enclosure (prison if you will). First, the power response into the room is more uniform through the upper midrange. It's a law of physics that an 8-inch driver beam severely around 3 kHz, and radiates much less energy off axis. Dipole operation more uniformly disperses midrange energy within a room. Why should you care? Well, at least half of the acoustic energy at a typical domestic listening seat is made up of reflected energy. Thus, if you color the off axis sound, you impact the total sound.
Second, harmonic distortion is reduced by eliminating standing waves and cavity resonances, which enhances clarity Finally, soundstaging becomes much more spacious and dimensional. All of these benefits are directly attributable to the manner in which the midrange is configured in a BassZilla design, and this is a common thread throughout the series.
The BassZilla loudspeakers are high-efficiency designs intended for use with low-power amplification. Center stage is given to high-efficiency full-range drivers such as Lowther and Fostex. Whenever deemed necessary, a super tweeter is used to fill out the treble range. The open baffle is positioned on top of a matching direct-radiator woofer enclosure. The woofer chosen for this project is the Eminence Magnum 12HO, a made-in-the-USA, and superbly built 12-inch woofer. The 12HO rated nominal impedance is 8 ohms and its sensitivity specification of 95 dB is obtained the old fashioned way – by using a 109 ounce magnet. It is not surprising then that it is overdamped with a total Q of 0.29. The 12HO’s Efficiency Bandwidth Product (EBP =Fs/Qes) is 138.8, suggesting that a vented bass box would be optimal.
The Pass Bass Box
Current drive makes possible bass alignments not otherwise available. A case in point is the possibility of using the 12HO in a sealed enclosure. Why a sealed box? There are two very good reasons for preferring such an alignment. First, the quality of the bass is much improved relative to a vented box, being more phase coherent and more tuneful. Second, the woofer is loaded by the box’s air spring even below the system’s resonance frequency. Vented designs are notoriously susceptible to overload in the deep bass, as the cone decouples from the interior air spring. Cone excursions become extreme below system resonance. This is a basic reason for avoiding vented designs with short-voice coil, limited excursion drivers, such as Lowthers. In contrast, cone excursions are nicely controlled in a sealed box design - down into the deep bass range.
The Pass Bass Box denotes a new class of sealed box alignments made possible by current drive amplifiers such as the F1. The name was chosen to honor the man who has re-awakened current drive technology from its long slumber and has followed a road less traveled. On the First Watt website, Nelson Pass offers an overview of current drive technology and offers guidance on using a variety of full-range drivers in small sealed boxes, including the necessary EQ networks (http://www.firstwatt.com/current_source_amps_3.htm).
The Pass Box design for this BassZilla project is shown in Figure 4. The physical volume of 3.72 cubic feet was chosen so as to keep the system resonance within 20 percent of the woofer’s free-air resonance frequency of 43 Hz. The bass boost derived from current drive operation was used to extend the bass response. The overall Q of the alignment is controlled by the use of a shunt resistor across the woofer terminals.
The near-field response of the Magnum 12HO in the Pass Box is shown in Figure 5, using a 25-ohm shunt resistor.
The green curve gives the system response when driven by the F1. The purple curve illustrates the frequency response when driven by the Pass Aleph 30 – a conventional voltage controlled amplifier. The bass extension under current drive conditions is obvious, resulting in an F3 of 45 Hz. The fine tuning of the overall Q was done by ear to obtain the most satisfying in-room bass quality.
The BassZilla Leopard: Open Baffle
The Leopard’s “heart” consists of the affordable Fostex FE206E , an 8-inch twin-cone full-range driver, with a reputation for being a sonic giant killer. The paper cone is said to be made of banana plant fiber, which strikes me as a good thing for the environment. The free air resonance is 39 Hz, making it possible to cross over the FE206E at around 150 Hz.
I found the treble range to be coarse and unsatisfying. A command decision was made to investigate a matching super tweeter. Fostex’s own FT17H horn tweeter left me cold; it measures well, but lacks speed and resolution. I eventually selected the Fountek JP3.0-R tweeter, available from Madisound. This is a true ribbon tweeter with excellent speed and resolution. Its limited vertical dispersion makes it more difficult to integrate into a system relative to a dome, but ultimately its rewards are worth the effort. I would suggest a listening axis slightly below the ribbon to obtain the best overall balance.
A schematic drawing of the open baffle is given in Figure 6.
The BassZilla Panther: Open Baffle
It was Lowther-America’s Jon Ver Halen and Nelson Pass who introduced me to the Lowther DX55. In fact, Nelson was kind enough to send me a pair of DX55’s mounted in 0.5 cubic foot sealed boxes and EQ’d with his favorite network (see his article on the First Watt website). This proved to be a wonderful starting point, as it showcased the DX55’s strengths in the midrange. “As good as it gets on female vocals,” according to Nelson. But there was no bass extension and I felt that the driver struggled as a full-range driver. I hope that no one is seriously suggesting that the DX55 be operated full range. While the treble is adequate – at no time did I fee the urge to introduce a super tweeter into the mix – with a nominal free air resonance of 80 Hz, it definitely needs help in the bass. An ideal crossover to the Pass Bass Box would be at three time resonance or at a frequency of about 250 Hz.
A schematic drawing of the open baffle is given in Figure 7.
The Crossover Networks
Richard Small, of T-S parameter fame, gave his seal of approval in 1971 to first order crossover networks. To quote from his paper in the January 1971 issue of the Journal of the Audio Engineering Society: “The most desirable crossover network for general use would seem to be the simple first-order network. This network provides both constant voltage transfer and constant power transfer, the least phase difference of any network design, as well as economical simplicity of construction.” Interestingly enough, Neville Thiele, the T in T-S parameters, disagreed with Small, arguing that a third order Butterworth type was ideal in that it was far more selective than a first or second order network, and thus offered better protection for the drivers. I think that both T and S may be right, providing that the underlying assumptions are clearly understood. Small had based his analysis on the major assumption that the drivers were closely mounted. By that he meant that the spacing was a small fraction of the wavelength at the crossover frequency. He also assumed that the drivers were of similar type and resistance load. And if you throw in the additional caveat that the drivers are of sufficiently full range that they can tolerate gentle roll-in and roll-off slopes, then I would tip my hat to Richard small in total agreement. I attempted, to the extent possible, to keep things simple, and the drivers in general cooperated. There are occasions, however, when one has to take into account the acoustic slopes of the drivers and the need to reduce overlap in the midrange. This translated in the case of the 12HO woofer to a second-order low-pass network.
The crossover schematics for the Leopard and Panther BassZillas are given in Figures 8 and 9, respectively.
During the development phase of the Panther, I discovered that the treble response of my DX55 samples varied by as much as 4 dB above a frequency of 10 kHz. I obtained another pair of well-broken in samples from Lowther America for further measurements. And Nelson Pass also contributed measurements. A summary of my measurements in a Panther system is given in Figure 10.
The mic was on-axis relative to the DX-55, at about 5-feet in-room, which explains the shelving in the bass region. The driving amplifier in all cases was the First Watt F1. It's clear that no two DX'55 measured exactly alike in the treble. The treble efficiency is a variable, as is treble extension (Green curve). Still, it is fair to say that in the range from about 8 kHz to 16 kHz, the treble response falls in within a specification of about +/- 3dB. Taken together with Nelson's measurements, I'm inclined to think that this probably represents state-of-the-art performance for a twin-cone driver. Since the treble output is driven by complex cone breakup behavior, slight variations in paper thickness and stiffness could translate into differences in treble output. To put this into perspective: common dome tweeters are rarely matched to better than 2 db. And I also agree with Nelson, that these variations are not a big deal. I would be much more concerned with 3 dB variations in the midrange or presence region. In practical terms, it means that that there may be an occasion when you feel that the treble balance is a bit too hot for you taste. Be sure to break-in the drivers for at least 100 hours before making a definitive judgment. In that case, feel free to experiment with resistor R1 in the Panther network, by reducing its value to as low as 6 ohms. Another possible solution would be to slightly change the listening axis.
Putting it all together
The Impedance magnitude of both the Panther and Leopard is shown in Figure 11 (the Panther is in green). Similarly, the frequency response is shown in Figure 12 (again, the Panther is in green). Keep in mind that these are in-room measurements, at a distance of about 5 feet from the front baffle. The mic position in the case of the Panther was on-axis with respect to the Lowther DX55, and slightly below the ribbon tweeter for the Leopard. The differences between the two curves below 100 Hz are due to the height of the mic above the ground plane. The SPL axis was compressed to allow the eye to easily assess the overall tonal balance. Thus, voicing differences become apparent. It is clear that relative to the Leopard, the Panther exhibits a bit more personality in the upper midrange and presence region. In contrast, the Leopard sounds more neutral in the presence region, and slightly laid back through the midrange, and although it is not obvious from the curves, its treble range is more refined.
Figure 11 - Impedance Magnitude
Sharing a common bass platform, both of these big cats pack a strong bass punch. In particular, the upper bass range sounds full-bodied and authoritative. The full magic of a cello projects with power and control. Audiophiles often complain about lack of bass. The typical scenario involves a two-way, stand-mounted speaker, with quick bass lines, but without bass authority and punch. The common misconception is that the addition of a sub 100-Hz subwoofer will cure a case of lean bass. This has cost more dollars and probably led to more frustration than any other purchase. The problem is that the addition of a subwoofer does nothing to address deficiencies in the upper bass, the octave from about 120 Hz to 240 Hz where help is really needed. Well, with the Leopard or Panther driven by the F1, help is on the way!
In keeping with the big cat analogy, these speakers are adept at transient attack. Speed complements power. The ability to go from zero to several G’s of acceleration in the blink of an eye is a hallmark of high-efficiency full-range drivers. To be honest, however, the Panther, powered by the Lowther DX55, is quicker out of the starting blocks. It pounces on transients with primal energy.
First and foremost, what the F1 brings to the table is control over the voice coil. Sonically, this translates into extreme clarity, transparency, and a dramatic reduction of inter-modulation distortion. Imagine cleaning a dirty window so that you can see through to the inner recesses of the soundstage and effortlessly focus on instrumental outlines. The F1 can do all that. In addition, resolution of complex passages is no problem. The ability to resolve ambient clues and follow transients into the noise floor of a recording can be absolutely breathtaking. There is lots of low-level detail to focus attention on. Should you decide to follow a particular thread in the music’s harmonic tapestry, the F1 facilitates that without artificially brightening or etching textures. Its presentation is consistently organic, natural, and good for you.
Having said all that, it should be recognized that the F1 is still a solid-state amplifier. When it is pushed hard, which should not be very often in the context of a high-efficiency speaker, its distortion spectrum sounds third-order (dissonant) rather than second-order (euphonic) in character. This means that it redlines less gracefully than does a single-ended triode amplifier (SET). And if you’re hopelessly hooked on lush and romantic SET sound, you should know that the F1 is quite neutral in its midrange perspective. The way around this, for an incurable romantic such as yours truly, is to install a tube-based line stage or preamplifier ahead of the F1 to sufficiently fatten out harmonic textures. Of course, my preferred medicine is to add vintage 6SN7 sound to the mixture. Keep in mind that the F1 sports a rather low input sensitivity and needs to be driven by a preamplifier with at least 10 dB of gain. It is revealing of front end shortcomings and demands a high-quality front end to reach its full sonic potential.
In the context of the Leopard and panther loudspeakers, the F1 does not sound like a low-power amplifier. It exerts powerful control over bass lines, and most remarkably, it unleashes an amazing dynamic range. It kicks into second, and then third gear, and all of a sudden it engages a fourth gear that you did not know was there. This ability to expand from soft to loud is as convincing as I have ever heard from any amplifier, irrespective of type or power rating.
In summary, I have enjoyed the F1 to no end over the past several months and have no plans to liberate it for the foreseeable future. It redefines what the loudspeaker-amplifier interface is all about and offers dramatic proof of the superiority of current drive over conventional voltage-controlled amplification. Kudos to Nelson Pass for turning at least a few of us onto the joys of current drive.
The First Watt F1: Nominal Specifications
Price: $2,500 suggested retail [But as I understand it, Mark at Reno Hi-Fi (firstname.lastname@example.org) is giving a 10% discount to anyone with a pulse.]
Input Impedance: balanced 100 Kohms, 8 ohm load, unbalanced 50 Kohms, 8 ohm load
Output impedance: 80 ohms
Output power 8 ohms: 10 watts @ 5% THD 1KHz
Output power 4 ohms: 6 watts @ 5% THD 1KHz
Output power 16 ohms: 9 watts @ 5% THD 1KHz
Gain: 14 dB @ 8 ohms
Maximum output voltage: +/- 20 volts
Maximum output current: +/- 1.75 amps
Frequency response: -3 dB @ 3 Hz, 100 KHz
Noise: 30 µV unweighted, 20-20 KHz, 8 ohms
Power consumption: 200 watts (100 watts / channel)
Fuse: 3AG slow blow type, 4 Amp for 120VAC (2 Amp for 240 VAC)
Warranty: Parts and labor for 3 years, not covering shipping or consequential damages.
Madisound Speaker Components: http://www.madisound.com/