Sorry we\'re later than usual, but what a ride the last months have
been!
Our official website was hacked by an anonymous party, hiding behind
various IP addresses. No personally identifiable information was
disclosed, so the privacy of our customers is guaranteed. A writeup of
what happened can be found under the Blog button on our new website:
[Metrum Acoustics Security Incident Report -- February 2026](https://metrumacoustics.com/blog/20260218-incident?utm_source=Email&utm_campaign=news10).
The rebuild took a while, but now we\'re very sure that the same entry
door can\'t be used again. After \'failure\' we rebuild and make it
better. The result is our new website that is safer and loads a LOT faster.
Online ordering and paying is now possible as well!
Our previous newsletters and other information will be added shortly.
Will keep you posted\...
## A change of pace
Large parts of the Netherlands, where Metrum is built, will celebrate
Kingsday. Most cities will be full of colorfully dressed people (mostly
orange, commemorating a part of our heritage, the house of Orange). Some
of them lavish themselves with spirits and beer, effectively turning
them into [Zombies](https://www.youtube.com/watch?v=JftjI_mu4Sw)\... We
guess "[It\'s the time of the season](https://www.facebook.com/reel/1549198382771602)". At Metrum
headquarters, we\'re only \'high\' on the love of Music, and working on
this newsletter, among other interesting developments\...
As announced in [our previous newsletter](https://metrumacoustics.com/blog/newsletter-april-2025?utm_source=Email&utm_campaign=news10),
the new I2S-LVDS modules are out! The first module was ordered just 2
minutes after we went live, and orders kept coming in after that. The
new modules are received well by our customers:
"At 24 hours burn in, symphonic and operatic chorals are gorgeous, vs,
the former chain is not very listenable actually. I\'m basing this on my
memory of live orchestral performances, which has not only the huge
dynamic range, but also the gorgeous tones at any volume, never piercing
or fatiguing. So with that as the golden rule, the LVDS chain is the
closest BY FAR of any system I\'ve owned, and it\'s dramatically
improvement over the USB chain, with no other changes of equipment."
:img[Metrum & Sonnet LVDS I2S modules on grass, with logos superimposed]{src="/images/I2S-HDMI_grass_cropped_logo.avif" width="12.76cm" height="5.747cm"}
:img[Metrum I2S LVDS connected to power inside a Metrum Jade]{src="/images/MetrumLVDS-pwr-diagram.avif" height="11.555cm" float="right"}
A final \'touch-up\' we added is for the I2S-LVDS
module version that fits the Metrum Acoustics Jade, Onyx and Amethyst
DACs that don\'t have the 4-pin power connector. We made a small plug
available to plug into the DAC\'s power without the need for a
modification at our factory. We already made mention of this on our
[Facebook page](https://www.facebook.com/MetrumAcoustics/posts/pfbid0iZWZDbVXDCARhFdd2pxFKsSLUnSGUaZpw11vyqQEGP9hDiQyKeEgGiETYgPPPZPVl),
but we want all our subscribers to know we now have an elegant solution
for our customers. Before you order, please check which version of
motherboard you have! You can order the I2S-LVDS module in our web shop
and add a remark to your order, or send an email to add the power wire.
We will then send you a module with the power wire attached,
plug-and-play!
The new I2S-LVDS modules are available in our webshop:
[Sonnet (all models)](https://metrumacoustics.com/product/sonnet_i2s_lvds_module?utm_source=Email&utm_campaign=news10)
[Metrum Adagio / Pavane](https://metrumacoustics.com/product/metrum_i2s_lvds_AdagioPavane_module?utm_source=Email&utm_campaign=news10)
[Metrum Jade / Onyx / Amethyst](https://metrumacoustics.com/product/metrum_i2s_lvds_JadeOnyx_module?utm_source=Email&utm_campaign=news10)
## Why is it called LVDS and what does it do?
LVDS is the abbreviation of Low Voltage Differential Signalling. This
tells us that it uses a low voltage, which prevents radiated EMI
(interference to other parts of the circuit board). Differential
signaling means that the connection is balanced instead of single-ended.
Balanced connections have several advantages, the largest advantage
being reduced common-mode noise. Without going too technical (read up at
the Techie Section below), a signal uses 3 conductors instead of 2. This
clever trick makes sure that you have more signal and are less bothered
by interference from the outside world. There are several ways to reduce
interference, of which screening can also be very effective.
Furthermore, galvanic isolation can help to remove ground loops and make
a device pretty much impervious to external interference.
The new modules function up to 384kHz sampling rates, just like our
DACs. Of course, PCM is our game, so DSD (Direct Stream Digital) and DOP
(DSD over PCM) will not function. Please note that the module uses the
HDMI plug, but is not compatible with your favorite console, receiver or
TV. *It works for digital-audio-specific I2S transmission only!*
## Tech Section for LVDS
In a single-ended connection, one conductor is for ground reference
while the other is the signal. In an ideal world, the current needed for
the transmission is a round loop through the 2 conductors, making the
current through both conductors equally large. However, a disturbance
from the outside (real) world can induce a voltage spike on the
conductor(s), which is called interference.
In a balanced connection, one conductor is again the ground reference.
The other two are carrying the same signal, but opposite in phase. Now,
if a disturbance from the outside world creates a voltage spike on both
cables, that spike has the same direction for both conductors. When
going from balanced to single-ended, one of the signals is inverted
(also inverting the voltage spike), and added to the other
(non-inverted) signal. The end result is twice the signal (because we
added two now in-phase signals), and a cancellation of the voltage spike
(because we inverted the phase of signal + spike). Clever, isn\'t it?
## PCM and DSD coding
PCM is an abbreviation of Pulse Code Modulation, coding a signal using
\'pulses\' with fixed time intervals between the pulses (or samples). An
analog value can be translated to a binary code in an Analog to Digital
Conversion process. On the other end of the recording chain, this code
is translated to an analog voltage by our Digital to Analog Converters.
So, the code stored in a sample is a representation of an analog value
at a specific time. A signal can be positive or negative, depending on
the pressure on the membrane of a microphone, or the position of a
string with regard to a pickup. The designation for positive or negative
is stored in the sign bit. A 1 is positive, a 0 is negative. PCM is
"sign and magnitude", so one bit tells us the polarity, where the rest
of the bits tell us how large the signal is. DSD is a totally different
animal, consisting of a 1-bit signal that can be \'high\' or \'low\',
coded as a pulse density. More \'high-time\' gives a higher output
voltage. A negative voltage is coded as more \'low-time\'. Having an
alternating pattern of 0, 1, 0, 1 with exactly 50% duty cycle gives a
zero-volt output. DSD is another way of representing an analog signal,
not inherently better or worse than PCM in capturing and storing the
data. It is just different.
PCM is a multibit \'thing\' (24 bits in the Metrum and Sonnet case),
where DSD consists of one data line per channel that can only be \'ON\'
or \'OFF\'. Feeding DSD signals into a PCM DAC would make no sense from
an electrical point of view because the DAC element doesn\'t
\'understand\' what it\'s supposed to do with the 1-bit information.
Luckily, most digital audio streamers are created with a decimation
filter inside, which can convert DSD to PCM \'on the fly\'. Yes, that is
a conversion, but this process, called decimation, is rather benign. Our
[Metrum Ambre](https://metrumacoustics.com/product/ambre?utm_source=Email&utm_campaign=news10)
and [Sonnet Hermes](https://metrumacoustics.com/product/hermes?utm_source=Email&utm_campaign=news10)
streamers can also do this with the proper settings in the Roon player
software. This means that the DAC can have a PCM architecture, but you
can effortlessly play DSD files because of smart and flexible software.
An advantage of doing the conversion in the streamer is that more
processing power is available, and it is upgradeable in the future. With
a hardware-based approach, this flexibility is not possible. Moreover,
adding DSD architecture to the hardware design of our DACs would imply
building two physical DAC devices into one chassis. This adds
significantly to the cost. Another inherent \'issue\' with DSD is that
you can\'t digitally equalize it or use plugins like room correction.
DSD should be taken "as is", although some manufacturers convert DSD to
PCM, do processing, then convert back to DSD, and then convert to
analog. If you use a PCM DAC like ours, and this path is chosen, you
eliminate conversion steps by staying with PCM after processing. So,
inherently, the signal chain is simpler when PCM is used (in multitrack
recordings).
Some manufacturers market DSD as pure, untouched and better. We\'ve been
doing the same with NOS PCM for years. Every processing step introduces
quantization (rounding) errors that lead to a loss of resolution, but
can also introduce pre-ringing or post-ringing. Jitter can be cleaned
up, but it is hard to eliminate the low-frequency fluctuations. Metrum
Acoustics designs products with a good clock as a base, then the signal
that is offered to the DAC doesn\'t need cleanup.
## Tech section: PCM R2R NOS DACs
The simplicity of R2R NOS DACs helps to avoid phase distortion and
jitter interactions introduced by complex digital processing. Because of
this, the time-domain behavior of Metrum Acoustics and Sonnet Digital
Audio NOS DACs is very clean, which is very audible! The single-pole
analog filter implemented in our DACs is deliberately chosen at a high
frequency, in the sweet spot that prevents high-frequency roll-off (in
our DACs the -3dB point is at around 80kHz!) but also doesn\'t introduce
too much high frequency aliases. Aliases are mirrors of the original
signal, so they are harmonic in nature. Harmonic 3 is at -80dB, and
harmonic 2 is at -100dB. In practice, this is so low that these
harmonics are not coloring the analog signal. Moreover, amplifiers
usually also filter their incoming signal, leading to even fewer issues
with high-frequency content. Given the noise floor values for a lot of
amplifiers (larger than -100dB), we can safely argue that high frequency
content is not an issue.
Audible time-domain problems like **pre- or post-ringing** that are
introduced by (overdriven) oversampling/digital filters are simply not
introduced in NOS DACs. Latency is low because the digital signal is not
filtered (or processed). Asteeper filter, or heavier (more complex)
processing, introduces more latency because a longer filter (with a lot
of filter coefficients) is needed. Also, digital filtering is usually
implemented for only one sampling rate, because the filter coefficients
need to be adapted to the incoming sample rate. The easiest way to do
this is by sampling all incoming data to one sample rate; the one the
digital filter operates at. This adds yet another step of processing,
and for us this is another reason to not implement sample rate
conversion.
**Sample rate converters** that are sometimes implemented in
Oversampling DACs are either re-clocking or re-synthesizing new sample
points. Jitter on the input clock can modulate the interpolation
process, resulting in timing errors spreading across the signal. NOS
DACs do without these steps, so jitter can\'t be recirculated,
re-interpreted or multiplied by recursive filters. No oversampling
filter = no jitter multiplication. If the input clock is jittery and
oversampling is used, all the \'extra samples\' can inherit the
jitter-related timing errors.
\
A **PLL** can help to clean up jitter, but its output always deviates
from the frequency it tries to lock on to. It is like the music is
playing faster and slower, which can be very tiring to listen to. You
could even perceive so-called \'beat-frequencies\' if the difference
between the PLL frequency and signal frequency is within a certain
bandwidth (usually not in high-end equipment).
:img[Picture by 愚木混株 Yumu]{src="/images/愚木混株Yumu_Notes.avif" width="9.394cm"
height="9.394cm" center}
> Picture by [愚木混株 Yumu](https://www.instagram.com/cdd20)
Timing errors, ringing and interference on the line are "no bueno"
No high-frequency processing implies less **RF interference** to account
for. Like in older oversampling designs, separating the digital filter
from the DAC element was recently reintroduced in newer oversampling DAC
designs to prevent interference. If oversampling is not introduced, the
design can be without shielding for extremely high frequencies in mind.
The design can be with timing in mind. This implies that jitter-induced
phase noise is not even introduced.
NOS DACs are not bothered by **inter-sample overs**, a digital form of
clipping, because they natively convert the digital signal at their
input. Inter-sample overs are prone to emerge when digital interpolation
filters are fed with very loud and dynamic recordings; some recordings
are so \'hot\' and \'compressed to death\' that the meter reads 0dBFS
during most of the track.
In a NOS DAC, the bits that go in are the bits that reach the DAC
element. If an analog transient would have led to clipping, our DACs
have more than enough analog headroom to accommodate that musical peak
(the supply voltage of the IV-stage is much higher than the analog
output voltage). So, loudness war, bring it!
Inter-sample overs can also be problematic in DACs with digital
reconstruction/interpolation filters that can generate peaks higher than
the original samples (overshoot). Digital filters multiply and add
digital values (sample x filter coefficient) to create their filtered
output. If the end result of that calculation is bigger than the size of
the register that holds the calculated value, a digital form of clipping
is introduced, leading to distortion. The digital filter should
therefore be capable of storing more bits than the original audio
signal. In the data sheet of digital filter chip SM5842 (introduced in
the early 2000s), we find the text \"Overflow limiter built-in\", and in
the datasheet of the SAA7220, we find \"Overflow protection\",
indicating that this was a known caveat. Please remember that the
loudness war started earlier, with the well-known Oasis album referenced
by many as the start.
Some oversampling DACs are known to implement a digital -3dB trick to
create headroom for the prevention of Inter-Sample-Overs, but in the
process create quantization errors because of the volume change in the
digital signal. A -6.02 dB trick could also be done, but would lead to a
lower resolution. In technical terms, a 6.02 dB change of volume is done
by shifting the signal 1 bit. Does that solve the problem, or does it
decrease resolution (and in effect add noise) to circumvent the original
problem?
**R2R NOS DACs are direct and simple:** with each sample, the whole R2R
ladder is set to the correct configuration, then with the switch of the
sample clock, the digital signal is converted. Our NOS DACs avoid the
problems described above by simply not participating in them.
Moreover, you can always choose to add processing at the source and
decide what you like better. In the case of Low-frequency room
correction, this can be very beneficial. In an inherently oversampling
audio DAC chip, there is no possibility to turn the processing off
because the filters are embedded in the silicon of the chip. You\'re
always listening through the same filter (giving a coloration), there\'s
no way of circumventing it and you would never hear the difference i
makes. In some off-the-shelf audio chips, the user can select which type
of filter to use (fast, slow, NOS). This is a step forward, but after
the digital filter there is always a Sigma-Delta modulator. This
modulator can\'t be turned off because it is inherent to the way that
DAC chip works. It makes sure that the incoming bits are made usable for
the DAC element inside the DAC chip.
Very technical and long story short: we\'re happy to have a very True to
Nature \'analog\' sound coming from our \'simple\' R2R DACs. Our DACs
honor the NOS principle and are truly bit-perfect, where oversampling
DACs -- de facto -- are not. This is the main reason why NOS DACs sound
more natural when compared to their oversampling counterparts.
## Where DACs meet speakers
:img[Acelec Model One with Stereophile 2023/2024/2025/2026 Recommended Component Awards superimposed]{src="/images/Acelec_stereophile_award.avif" width="7.53cm" height="9.123cm" float="right"}
Our Acelec Model One Speakers were crowned with
[Stereophile\'s Recommended Component award for the fourth year in a row](https://www.facebook.com/MetrumAcoustics/posts/pfbid02HNPPmeMXHW7E61Ya6x8q4DjBdtpjMhWdhiRTVgL6tSxTiMN4b5tpdrrrqWFxSS9yl)
and deserve more attention. At 17 kgs (that\'s 37.5 pounds) per speaker,
this \'little heavyweight\' measures, sounds and
[reviews](https://www.stereophile.com/content/acelec-model-one-loudspeaker)
exceptionally well. Placement is not very critical, and it doesn\'t need
a power amplifier that can deliver prodigious amounts of power. Since
the panels are glued together rather than screwed, cabinet resonances
are the lowest encountered in 30+ years of speaker measurement,
preventing the cabinet from sounding like a church bell when an impulse
hits it. Moreover, the drivers are well-damped, resulting in an
exquisite cumulative spectral-decay plot. The waterfall plot is short,
[no unnatural ringing, just like our DACs](https://metrumacoustics.com/blog/newsletter-november-2024?utm_source=Email&utm_campaign=news10)\...
Does the Model One speaker swear to speak the
[truth](https://www.stereophile.com/content/recommended-components-fall-2023-edition-loudspeakers),
the whole
[truth](https://www.stereophile.com/content/recommended-components-fall-2024-edition-loudspeakers)
and nothing but the
[truth](https://www.stereophile.com/content/recommended-components-fall-2025-edition-loudspeakers)?
[I rest my case](https://www.stereophile.com/content/recommended-components-2026-edition-loudspeakers)!
[John Atkinson\'s measurements](https://www.stereophile.com/content/acelec-model-one-loudspeaker-measurements)
on our Model One speakers speak volumes!
Fun facts:
- The first I2S-LVDS module was ordered 2 minutes after it was
available in our webshop.
- [PCM coding](https://en.wikipedia.org/wiki/Pulse-code_modulation)
was invented long before the first R2R DACs emerged. PCM was
patented in the year 1952. Thank you, Claude Shannon, Bernard Oliver
and John Pierce!
- Yes, this is the same Shannon as the one mentioned in the commonly
known [Nyquist-Shannon sampling theorem](https://en.wikipedia.org/wiki/Nyquist–Shannon_sampling_theorem).
Mr. Whittaker beat them to it in 1915, but is much less known.
- PCM is an uncompressed (thus lossless) way of storing an analog
signal.
- FLAC is a form of lossless audio data compression, making an audio
data file smaller while still being able to revert to the original
data stream.
- SATA, HDMI, PCI Express and Gigabit Ethernet are all examples that
make use of differential signalling. Computers only do "Beep Boop"
internally, you still need one of our DACs to let it play Music ;)
- [LVDS](https://en.wikipedia.org/wiki/Low-voltage_differential_signaling)
emerged in 1994, although the idea of a [balanced connection](https://en.wikipedia.org/wiki/Balanced_line) is much
older and dates back to good old phone lines in the early 1900s.
- For a line (other word for cable in this context) to be balanced, it
needs to transmit the signal differentially (2 times the same
signal, in opposite phase), and the impedances of the wires and
receiving end must be equal for a pair. Do we hear someone say
twisted pair?
- In the days when electronics were not widespread, a transformer was
used to go from single-ended to balanced and vice versa. Modern
electronics mostly use operational amplifiers for this.
- The XLR connector \'for audio\' was created by Cannon in the 1950s.
L is for Locking, R is for Resilient, X implies that Elon\'s X-type
is not very unique.
- Our newsletters are received and enjoyed by a lot of Audiophiles.
The counters for our [online newsletter archive](https://metrumacoustics.com/News.html?utm_source=Email&utm_campaign=news10)
reach more than 18000 for some publications, and some were even
republished by our [Brothers in Arms](https://metrumacoustics.com/News/metrum-newsletter-january-2024-b50.html?utm_source=Email&utm_campaign=news10).
Hope you enjoyed this edition of the Metrum newsletter.
Keep listening to beautiful Music, we'll talk soon...
**- Team Metrum Acoustics**