deHavilland Electric Amplifier Company
GM70 SET Monoblock Amplifiers
Supreme smoothness, or in corresponding musical term,
Review By Dick Olsher
here to e-mail reviewer.
Over the past 15 years
single-ended triode amplifiers (SET) have gained a measure of
respectability in high-end circles, which is rather ironic considering
that about 60 years ago such an output stage topology was only considered
suitable for inexpensive designs. Times have changed. It is now recognized
that directly-heated SET designs are capable of superbly fleshing out that
first watt of power, which is responsible for getting a couple of things
right in reproduced music: a sense of 3-D spatiality and a closer approach
to the vitality of the real thing. Flea-powered SET designs based on the
2A3 or 45 triodes have demanded ultra high-sensitivity matching speakers.
However, even a typical 300B-based SET sporting 8 wpc will thank you for
mating it to at least a 95dB/W/m sensitive load.
A few designers, not content with the power and headroom
limitations imposed by traditional SET designs, have taken to explore the
possibilities offered by large power triodes such as the 211 and 845.
Apparently, deHavillandís designer, Kara Chaffee has taken a liking to
the Russian Gm-70, in reality a beefier version of the RCA 845. This is a
massive power triode whose graphite anode is capable of dissipating a
maximum of 125 watts. Throw the power switch and the thoriated tungsten
filament lights up like a light bulb sucking 3 amps out of the wall. For
the record, Kara is very proud of the GM70 amplifier and believes that it
maintains all the virtues of the Aries 845G with nearly double the power
output. Well, I do have to quibble with the 50 wpc spec, because at
maximum rated power output the total harmonic distortion (THD) is already
well above 10%. Indexing the power output to a more reasonable 5% THD,
gives a power rating of about 30 wpc.
input stage is DC coupled and is based around the 12SN7GTA/B dual triode,
which is a 12-volt version of the 6SN7. The 6AU5GT pentode, a TV tube
originally designed as horizontal deflection amplifiers, is used for the
driver stage. There is only a single cap in the signal path, which
fulfills the mantra that Ďless is more,í at least when it comes to the
signal path. A soft start circuit allows the filaments to warm up before
applying plate voltage. The nominal plate voltage is 1,000 V, so please do
me a favor and do not operate this amplifier with the bottom plate
removed. Plate current is adjustable and should be set to 80 mA, which
corresponds to the red mark on the bias meters. I can report that bias is
quite stable and once tweaked after 45 minutes or so of play time stays
stable for many weeks.
No global negative feedback is used. As with other
design choices, there is no free lunch, and as a consequence the output
source impedance (not to be confused with transformer impedance taps) is
about two ohms. This limits the amplifierís damping factor and can
impact the speakerís frequency response with loads whose impedance curve
is non-uniform. Typically, a speakerís impedance curve can vary
significantly, by a factor of a few or more between the minimum and
maximum values. Some ESLís are far worse in this regard, varying by a
factor of 30 or more, and dropping to 1 ohm or so in the extreme treble.
The amplifierís output impedance acts as a voltage divider in concert
with the speaker impedance, resulting in reduced output as the load
impedance approaches the amplifierís output impedance. My recommendation
is that the GM70 should not be mated with speakers whose impedance dips
significantly below 4 ohms.
A single set of output taps is provided, which are
optimized in terms of power transfer for a 4 to 8 ohm load. Note that a
16-ohm configured output is available on request. The output signalís
polarity is inverted, so be sure to compensate for it. I simply reversed
the speaker cable leads at the output binding posts. And by the way,
Iíve always been a fan of Cardas binding posts, so be sure to opt for
this option at an added cost of only $100.
During the design phase, deHavilland discovered that the
Gm-70 Russian tube socket was poorly made and caused plate current instability
due to poor socket connections. The decision was made to custom engineer a
new matching tube socket to get the best sound possible out of this tube.
This is no ordinary socket: the base is machined out of a bar of 7/8"
thick Teflon, while the terminals are machined out of 3/8"x 3/8"
OFC type copper billet. The socket terminals float slightly in the Teflon
block so that the tube pins fix their location. As the tube and socket
heat up and expand, the copper terminals maintain an intimate
spring-loaded contact with the tube. Kara says that the new socket is
responsible for better bass, enhanced dynamics, and tighter midrange
One of the advantages of hanging on to a review sample
for several months is the opportunity to audition it in multiple contexts,
and in the case of the GM70, that amounted to several loudspeaker loads.
In the end, it became a complex tale of amplifier-speaker interactions,
making it difficult to offer a simple sonic assessment. In fact,
auditioning any high source impedance amplifier is a tricky business as
its tonal balance and bass performance are very much load dependent. But I
see that I am getting ahead of myself...
In the beginning, life was good. Hooked up to the DIY
OB2 open baffle loudspeaker, the GM70 literally astonished me in several
performances aspects. In an audio review world overrun by hyperbole, it is
difficult to communicate its soundstaging and imaging prowess. Let me make
this perfectly clear: its spatiality was not on par with that of other SET
designs. Au contraire, it raised the bar higher than Iíve previously
experienced. Yes, image outlines were palpably fleshed out, but they were
so solidly etched within the confines of the soundstage, that when I
closed my eyes accepting the spatial illusion became effortless.
Ordinarily, when I listen with eyes shut and attempt to localize
instrumental outlines, there is a residual imprecision, a spatial flicker
if you will, that distracts from the illusion. Driving the OB2, image
outlines felt like they were built out of brick. Soundstage width and
depth perspectives were well extended and layered, and on natural
recordings, hall size was clearly communicated.
Midrange harmonic textures were lusciously sweet
sounding. But make no mistake about it: this is not your grandfatherís
SET amplifier. Many SET amplifiers constrict the power bandwidth,
struggling to get past 20 kHz, and as a result make textures sound overly
liquid and soft and turn transients mushy. Since the DC bias current flows
through the transformer primary of an SET design, special air gap
techniques are required to prevent core saturation, and in the process,
bandwidth ends up well below that of push-pull output stages. In contrast,
the GM70 exhibited quickness of attack and transient definition that in
fact rivaled that of push-pull designs.
The tonal balance was full bodied, offering a slice of
vintage tube sound. No doubt the presence of octals contributes to the
perceived full tone character. But I would also have to point a finger at
the harmonic signature of the Russian Gm-70 as being mostly responsible
for the ampís tonal authority. It seems to me that the 300B has been
overly exploited in SET designs. Its sound tends toward the civilized and
refined, which might lead someone solely exposed to 300B sound to presume
that the entire SET universe sounds more or less that way. Leave it to
Kara Chaffee to yet again follow the road less traveled and explore new
sonic possibilities. The sound of the GM70 was not only refined but also
dynamically bold. Microdynamics bubbled to the surface with almost no
effort so that musical lines ebbed and flowed with dynamic conviction and
plenty of drama. And despite its modest power rating, it shifted gears
from loud to very loud much like a muscle car going through its paces. I
was simply having a blast!
only proverbial fly in the ointment was bass performance. In general, bass
lines sounded under damped, lacking adequate tightness and definition.
There was plenty of magic in the range covered by the D5nf 'Naturflux'
driver. Being a 16-ohm nominal impedance driver, the D5nf appeared to be
ideally suited to the needs of the GM70. On the other hand, the GM70
failed to adequately damp the 4-ohm Augie 15-inch woofer. Even given its
imperfections, I was so blown away by its sound that I actually considered
spending a chunk of our life savings towards purchasing this amplifier.
The second chapter unfolded with the Micropure
ĎKotaroí minimonitor loudspeaker (review in progress). This Japanese
phenom features a 4-inch full range augmented by a Murata ceramic
supertweeter. It unveiled its absolute best when driven by the GM70
monoblocks. The purity of the midrange was almost electrostatic in
character with squeaky clean textures and exceptional retrieval of
low-level detail. Harmonic colors were imbued with satisfying warmth and
natural vividness. And above all else I experienced this feeling of
supreme smoothness; the corresponding musical term is legato. It felt as
though the great Mozart himself had descended from the spirit world to
take control of this session. Unfortunately, there is nothing to report on
with respect to deep bass reproduction, as the Kotaro has no deep bass
response to speak of.
Having recently acquired a pair of refurbished QUAD
ESL-57s via Electrostatic Solutions and QUADS Unlimited, I decided to try
my luck with the GM70. The QUADs, represent a difficult load with a
drooping impedance magnitude in the treble that approaches short circuit
conditions, well almost, but around 1.5 ohm is low enough. Presumably
Peter Walker was already aware of all the problems posed by electrostatic
speakers when he designed the QUAD II amplifier, which works
well enough with the ESL-57. But, in general, it is safe to say that the
Quad ESL is a bad choice for SET amplifiers. There was still magic in the
midrange, but it was clear that the GM70 was having a tough time at the
frequency extremes. The speakerís measured response shelved 2dB in the
extreme treble and bass lines sounded distinctly flabby. And the last
thing I was looking forward to was less treble output.
It was a similar story with the Swiss Piega TC-10X.
Relative to the response obtained with the E.A.R 890 tube amplifier, the
GM70 caused a 2dB drop in treble output and a similar decrease in upper
bass output. These tonal balance deviations were noticeable. Since the
Piega is a 4-ohm load in the midbass and only of moderate efficiency, I
was able to drive the GM70 out of its comfort zone with orchestral music
and did not like what I heard. Distortion levels rose rapidly, well above
acceptable levels of tube euphonics, even tapping into odd-order harmonic
harshness. This was a low point for the GM70 and highlights the need to
carefully match it with a compatible load.
operated within its comfort zone, this amplifier is adept of conveying a
sense of suavity, legato style, and imaging prowess that is nothing short
of stunning. Bandwidth is reasonably extended to maintain transient
realism and it retrieves harmonic colors with vibrancy and dynamic nuance.
As with other SET amplifiers it does impose its personality over the
music, but what a personality!
By virtue of its low damping factor, that is to say
highish source impedance, the GM70 is far from the ideal of a universal
amplifier. To be fair, this is a common problem with most SET amplifiers.
The GM70 appears to work its most compelling magic when partnered by
highish impedance loads and moderate to high efficiency speaker design.
Therefore, to be on the safe side, it would be imperative to audition this
amplifier with a prospective load to ensure compatibility. Its fortunes
can swing considerably from a state of sonic bliss with a synergistic load
to frustration with an incompatible load.
Type: Single-ended triode monoblock
Frequency Response: 20Hz to 35kHz
Power Output: 50 watts rms into 8 ohms, Class A
Tube Compliment: Power triode Ulyanov Gm-70
Driver tube Type 6AU5
Input tube Type 12SN7 GTA/GTB
Input Impedance: 50k ohms. (Values from 10k to 500k available on request).
Residual Hum: <2mv.
Dimensions 12 x 18 x 10 (WxDxH in inches)
Weight: 60 lbs.
Price: $9,995/pr (optional Cardas Gold/Rhodium binding posts - $100)
AndeHavilland Electric Amplifier Company
2401 NE 148th Court
Vancouver, WA 98684