The Quad ESL-63 and Gradient SW-63 Combination
Reviewed by Mats Törnqvist, November 1999.

Now the time has come for me to review my speaker combination. I think it could be interesting to read my opinion about the tweaked ESL-63/SW-63 combination since there is a significant difference in performance when the tweaks are implemented. A difference from regular reviews is that I have not done A/B testing with other speakers, I have lived with the loudspeaker for many years playing all sorts of music. This is a test of time that reviewers are not capable of and it takes some time, often months, to spot the failures and added sound that HiFi equipment are responsible for. The first reactions are sometimes wrong, and A/B testing certainly favours equipment that add the little extra that makes it stand out in front of the others. My ESL-63s have serial number 16185 and 16186.

Technical Description Quad ESL-63

The Quad ESL-63 is an electrostatic loudspeaker with a unique feature in the concentric rings that are delayed in a manner that makes it work like a point source positioned 30 cm behind the membrane. The membrane in electrostatic loudspeakers are usually considered to be lighter than air, meaning that all the resistance made to the output signal when it moves the membrane comes from the air itself. This is what makes the electrostatic speaker response so exceptionally fast and controlled. The light membrane is also the reason to why a single element can be used to reproduce the whole frequency range, and this is where the concentric rings comes in and makes the response free from beaming found in all other large panel speakers, and the off-axis response is unusually even for a dipole speaker. No filters are needed and this is a major advantage since such are difficult to make perfect and there is no interaction between filter components and the amplifiers output impedance (includes the wires). The other underlying physical element that makes electrostatic loudspeakers interesting is that the moving force is distributed throughout the entire membrane, making it controlled and free from break-up found in all cone speakers.

Delay principle

The delay line is made up of six concentric segments, each delayed with 24us which is the same as if the segment was positioned 8mm further away than the previous. The speaker is equipped with four elements, the outer two elements are not segmented and are used as (delayed) bass reinforcement panels.

Panel layout

The delay principle really works which is obvious if you look at the polar plot.

ESL-63 polar plot

At 8 khz the response is within -3dB at angles ± 25 degrees and this is a lot better than the Quad ESL-57's 7 degrees. The directivity is only 5dB up at 8 kHz which is remarkable for such a large membrane. The -6dB points are at 35 Hz and above 20 kHz. Sensitivity is 86dB at 2.83V input, maximum undistorted input is 40V, maximum output pressure is 2N/m² at 2m distance.

Estimated HF-behaviour of Quad ESL-57

The polar plot above is an example which probably is similar to the Quad ESL-57 bahaviour at 8 kHz. The old ESL has a very narrow sweet spot. At the end of this page there are some dispersion examples and a directivity curve for a 12" speaker.

Technical Description Gradient SW-63

The Gradient SW-63 subwoofer is designed as an add-on unit for Quad ESL-63 and the combination is actually better looking than the Quad ESL-63 on its own. The SW-63 is a dipole subwoofer with two 12" long-throw (± 12mm) cone speakers pointing in opposite direction to reduce distortion levels. The ESL-63 is placed under a lid on top of the subwoofer with an opening for connections. I have placed my ESL-63's on sorbothane feet and clamped them with the lid, making the Quads very firmly held in place. The ESL-63 sounds quite a lot better on its own when mounted in the Gradients because of the higher position and better stability. The overall stability of the Gradient subwoofer is pleasingly good, the subwoofers are not cheap but very well made. The dipole action gives a 6dB/octave reduction of level and this is corrected by the Gradient filter unit giving a flat response with -3dB at 28Hz (I have changed this to -3dB at 22Hz). The filter unit has two outputs, SW with a 3rd order filter at 115Hz and ESL with a 2nd order filter at 115Hz. There is also a mono switch and two attenuation settings giving a slight reduction in the low midrange. The set-up requires a separate amplifier for the subwoofers and the two 8 Ohm units in each SW-63 can be connected either in parallel (recommended) giving a 4 Ohm load or in series giving a 16 Ohm load. The combination is able to deliver an SPL around 110dB with good bass extension.


The dipole action (sound is emitted in anti-phase from the behind of the speaker) actually makes the set-up quite easy, the only thing to remember is not to place the speakers too close to the wall behind (at least 1 m distance) because if the reflection from the back wall is not delayed enough it will alter the frequency response. If the distance is large enough the brain can sort things out and practically ignores the reflected sound (nice brain). It is the early reflections that make up the alive feeling in good recordings and the more delayed sounds are heard as echoes rather than ambience. In my living room the distance to the rear wall is more than 3 meters. The reflections from the side walls is of no major importance with dipole speakers and neither is the ceiling reflex, the floor reflex is damped well enough with a thick carpet. The standing waves are only excited in the speaker direction dimension. Even if the reflexes and standing waves are damped the positioning and angling of the speakers is important, I have spent much time trying to find the optimum position within the limits of the rest of the furniture and the overall looks of the living room. Even if the positioning is important the ESL-63/SW-63 combination will never sound bad, you can put them in the corners if you have to but it is not optimal. I use a tube amplifier (Jadis DA-30) for the ESL's and a sand amp for the subs. I think that the ESL-63 combined with a tube amplifier is one of those happy marriages and it seems as if the new Quad II-forty combined with ESL-989 raised the Quad performance quite a bit at the London HiFi show, possibly the valve amplifier is a major part of that.

Sound Quality

I am very sensitive to equipment with a sound of their own, at first it may sound nice giving a more exciting sound than the sound-free equipment but in the long run it is simply boring when all records sound similar and the sound also masks the important details that makes music listening such a great occupation. This is why I like the ESL-63/SW-63 combination, the ESL-63 has absolutely NO sound of its own and the SW-63 is the most sound-free bass units I have heard.

The Quad ESL-63 sounds very good without any tweaking, but with the tweaks the resolution and clarity of sound is improved to an astonishing level. Take a look at the 'Quad Tweaks' page. The tweaks I have performed are:

  • Replaced the input electrolytic capacitor with two Solen polypropylene 100uF/250V capacitors in parallel and replaced the 1R5 resistor with a higher rating (50W) resistor of better quality.
  • Removed the metal grids.

  • Removed the dust covers.

  • Power line filters are applied.

The SW-63 is also an exceptional add-on without tweaking but a lot is gained (more than the ESL-63 tweaks) by the tweaks I have performed:

  • By-passed the high-pass section and use a polypropylene capacitor on the input of the main amplifier (DA-30) for a first order hi-pass.
  • Altered the frequency characteristics of the low-pass section giving a flat response almost down to 20Hz.
  • Increased the input impedance to 500 kOhm, making it a suitable load for my tube pre-amplifier.

If the Quad ESL-63 in original shape can be described as a window to reality with some limits in the sight and slightly grey (not coloured) and slightly distorted glass, the glass is changed into an almost clear glass with no distortion (the electrolytic capacitor distorts) when the tweaks are performed. The limits of the sight is also reduced a little.

Now we add the original Gradient set-up and the window is increased in size, the clarity of the window itself is also better and more details are recognizable in the soundscape, but there is also some filth appearing on the window making it obvious that it is a window not reality itself. Responsible for the filth is the Gradient high-pass filter which is not quite good enough to be invisible in my set-up. The bass performance is exceptionally good but there seem to be some difficulties around the cut-off frequency and this problem is addressed in the same way as the high-pass filter problem. By-pass of the high-pass filter is the solution (read more on the 'Quad ESL' and 'Quad Tweaks' page) and the problems reported above are no longer. The window size is now even larger and the filth is gone, reality is well within reach and the soundscape appears very detailed and in proportions rarely found. The bass is much more even and controlled, tuneful and heavy when it shall be, almost unrecognizable when it shall. This is landmark bass performance, but with the alteration of the low-cut frequency to -3dB at 22Hz instead of 28Hz the bass reaches an even higher performance level, the window size is by now very large almost surrounding you and the window is nearly invisible, rewarding you with a palpable soundscape as realistic as it can be, you can almost feel the presence. The last tweak is the increase of the input impedance which left me gasping in surprise of the increase in performance level, the reason is that the output impedance of my pre-amplifier turned out to be several kOhms and loaded with only 22 kOhm (original value) the loadline of the pre-amplifier circuit is altered considerably, reducing the level of performance. The window is still present, but you will only notice it when you tap with your finger upon it or try to walk through it (a slight exaggeration of course).

Given the performance raise of the described tweaks, what does it sound like then?

It does not sound at all, it reproduces the program material. The character of the reproduction is similar to that of the original Quad ESL-63 on their own, but with more detail and improved clarity. The dynamic capabilities are very much better as is the bass extension which also makes up for a much larger and more lifelike soundstage where you can feel the air moving, combined with the ease and freedom of coloration often found in speakers with less bass (large boxes add coloration). The sound is perhaps not as exciting at first as some of the popular high-end speakers of today, but it is much more musically rewarding. Most modern high-end equipment is peaked in the treble giving some more detail and initial immediacy, but no other speaker I have heard can reproduce the sound of real instruments as convincingly as the tweaked ESL-63/SW-63 combination.

At demonstrations snap and slam is over-emphasized and the program material used are often studio recordings with transient-clipping (almost always present in modern close-mike recordings) and with fast speakers such as ESL's the clipped transients are clipped, but with ringing dome-tweeters some energy is added and it can actually sound more lifelike at such a demonstration. I can not accept speakers that sound better on clipped program material than on good recordings (they will ring then), let us instead hope that the recording quality is increased. With real stereo (two-mike recordings) the Quad/Gradient combination sounds marvellous with all the treble extension and energy you could wish for and when the microphones are used at some distance the early reflexes which make up the 'alive' sound are not damped, giving you a soundscape as real as reality.

The most impressing kind of recording on ESL-63/SW-63 is a LARGE choir recorded in real stereo, you can actually spot the individuals and the voice reproduction is exceptional as is the sense of the room where it was recorded. Reproduction of voices is the most challenging task for loudspeaker constructors because the brain is so good at recognizing voices, you can actually recognize your friends voice in an environment full of other voices and this is by no means an easy task, you can also easily hear if this person holds the hands before the mouth, try to make a measuring unit that can do that! This means of course that the speakers that can handle the human voice best are the ones with the lowest coloration. Bass and treble is actually much easier to reproduce in a way that fools the brain that it is reality, still this is the areas of most interest in many reviewing articles. These areas are of course important, but it shall never be on the expense of the voice reproduction capabilities.

Other areas in which these loudspeakers excel is the even response with no emphasis on specific elements or frequency ranges, the treble that does not make any
sound of its own and is just right and incredibly 'short', the tuneful bass with its depth and low coloration and of course the exceptional imaging qualities.

Of course the sound is not perfect with all recordings, but that is due to the recordings not to the speakers. Studio recordings with no room information are reproduced in the same way, resulting in a lifeless flat sound in between the loudspeakers. Recordings using equalisation and compression will sound compressed and with a non-flat frequency characteristic. There is not much to do about the fact that most recordings are far from perfect, but one thing is certain, the ESL-63/SW-63 combination never sounds bad. An interesting aspect of high resolution, low coloration reproduction is that added sounds like metallic noise from microphones is separated from the voice, making it much more easy to listen to bad sounding recordings. The loudspeakers shall be used for music listening and I listen to the records with music I like rather than the ones that sound good.

This does not mean that I would prefer other loudspeakers for the non-perfect recordings. I think that the Quad/Gradient combination is equally well fitted to play Led Zeppelin or Metallica as grand scale orchestral pieces or small scale acoustical music. The music always comes through in great detail without any fuzz and listener fatigue is an unknown phenomena. The only limitation is the maximum SPL which with my tweaks probably is around 107dB, but seriously how often do you play that loud except when you are demonstrating the system to your friends.


There are some measurements on the 'Quad ESL' page that I have made on my loudspeakers, these are amateur measurements but they show some of the qualities of the reviewed loudspeaker combination, for instance the excellent impulse response, the even waterfall plot or the indecently low distortion.


The Quad ESL-63 is a remarkable design that will hold the opposition away for many years to come. The design is so brilliant that any major improvements are unlikely. At the moment I am anxiously awaiting the new Quad ESL-988 and ESL-989 and I seriously hope it will be a major upgrade, but I doubt it. I am sure that they will blow away an un-tweaked ESL-63, but a tweaked ESL-63/SW-63 combination is not easy to better. When the new ESL series arrives I will come back to this.

It is no secret that I like the ESL-63/SW-63 combination, but there is no such thing as a worlds best loudspeaker. Other individuals may not agree with me and they are not wrong in doing this, music reproduction is to a large degree down to personal taste and the Quad/Gradient combination seems to attract the more mature music lovers with a great interest in acoustical music. Young hip-hop lovers or head-banging heavy-metal fans would probably not make it their first choice, so it is really up to the reader to make up his or her own mind what kind of sound is the most preferable. The only thing for me to say is that I am satisfied and that I can recommend the loudspeakers to those interested in low coloration, high resolution equipment and that the performance in the end is the result of pure physics and ingenious design. You will probably never reach this level of performance with ordinary methods such as dynamic elements and boxes.

More About Directivity

Below are some examples showing the directivity behaviour of loudspeakers, a not well known phenomena.  





Polar plots for a 12" speaker






 Directivity curve for a 12" speaker






 Polar plot for a curved multicell speaker




The behaviour of the multi-cell horn speaker above is quite similar to that of the Quad ESL-63. In the horizontal plane it is also similar to the ESL speakers that use a curved membrane (for instance Martin Logan), but in the vertical plane these speakers will suffer from beaming. If you have a loudspeaker of this kind, don't sit to close to them and at a height where your ears are in the vertical centre of the membrane.





Directivity index













This site uses frames, but if you have found this page via a search engine and do not see any frames, you can use the button below to get to the main frame of the MT Audio Design site.