ein wenig knif-philosophie:
© Jonte Knif 2011
Quality Factors in Knif Pro Audio Equipment
I'm lucky to be in mastering equipment business, one of the last resorts of real
understanding in quality. How ever, I do feel that there is never too much info
about those things that affect longevity and sound quality among other factors.
With this essay I wish to shed bright light on my own thinking and what you
should be aware of when comparing gear and trying to understand pricing of (my)
products. I will cover some technologies or building practices that surpass mine in
some ways to give an idea what kind of compromises even expensive gear
sometimes includes. It is also useful to think what factors and techniques do not
justify their price.
I like to keep things as objective as possible, but heck, when writing about my gear
that is kind of difficult, so excuse me. Also excuse my English, I don't care to give
this essay to someone to correct as long as it is readable.
I will cover mostly those things that have some relevance to Knif Soma, since it is
my most recent and perhaps most complex design.
Case
Lets face it, it is not common to find equipment that fails in the rack because of
badly built case. But there are some factors worth to study a bit. Gear will
eventually be shipped from place to place and connectors will be pulled and
pushed etc.
Case can flex. If there are more than two or three components sharing same the
substructure in the equipment any flex will try to move them in respect to each
other. If they share the same circuit board and are attached to the case the situation
is bad. Any strain will be transferred to the most vulnerable part, namely solder
points. My policy is simple: I do not do it. The maximum amount of components
attached to the same sub-PCB and case is two. Second problem with big
substructures is the difficulty in repairs and modifications. There is no limit to my
irritation when someone brings me brand X device with one broken switch that
requires the dis-assembly of the whole front panel in order to be replaced.
All switches and connector must be screwed to case. A connector or switch relying
on soldering to PCB is a weak spot. One kick and you possibly ruin a whole circuit
board. It is cheap though.
What about case material? I use aluminum cases with milled parts from thick
plates because of many things.
One is heat conduction. I can safely use any part of the case as heat sink for
regulators etc, since the heat will be effectively dissipated to the whole case.
The other is electrical conductivity. Thick aluminum is pretty perfect for a back
plate. There is no need to hard wire every earth potential to one spot only. I can
assume that substantial area of the back plate is at the same potential and attach
different grounds to it locally and have effectively perfect star grounding. For
example safety ground is firmly attached to the proximity of power inlet and signal
input and output grounds close to their connectors.
Third is the fact that aluminum is not ferromagnetic. One could err to think that it
is nice to have steel case in order to shield components from external magnetic
fields, but thin steel sheet is not very effective as a shield but still effective in
transferring fields inside the equipment. Therefore I think that steel is a definitive
no-no in quality equipment.
Soma has sides and front panel “frame” milled from 10mm thick aluminum and
threaded for all necessary screws. With 4mm thick bottom and back plate the
structure is extremely stiff, even without lid on.
Magnetic shielding
I often get requests for separate power transformer. It certainly isn't a bad idea, but
so far I have decided to solve possible problems with magnetic interference with
quality components and proper shielding. Why? Of course to cut cost, yes, and
because it is safer, but also because it is possible and most importantly because it
makes my gear less prone to pick up stray fields from other gear. It is surprising
that most mastering quality EQs which use coils do not have proper shielding for
them. Therefore they buzz. Soma has proper mu metal shielding for the coils.
Let me tell an example of bad designs which illustrates the two sides of magnetic
interference. I would very much like to mention the very famous brands involved,
but must contain myself. In a local studio here in Helsinki there was a problem
with an EQ buzzing a lot. Too much for even pop music. We started to isolate the
problem. The source for the magnetic field was from ca 5 rack units above, a
certain tube compressor with inferior mains transformer. And that was picked up
by EQ coils, which did have shielding, but unfortunately not from proper mu
metal. Since the source was so bad and receiving end so sensitive there was no
way of using the gear in the same rack. Not very professional.
Power transformer must be of toroidal type and have magnetic shielding. In Soma
it has two layers, one composed of silicon iron and another from mu metal.
Signal switching
The best way to switch signals is with relays. There are again three reasons. Relays
can be placed anywhere and thus signal path is foreshortened. Good relays are
hermetically shielded, normal switches are not. Relays last very long. Typical life
for a front panel switch is 50 000 operations and those sub-miniature relays I use
in Soma last 100 000 000 operations according to the manufacturer. That is right, a
hundred million operations. On first reading this doesn't make any sense, but
remember that these are used in telecommunication equipment that switch
constantly and one million operations is easily reached in a year. Also, 100 million
operations estimated life means that they very probably don't break before one
million, and we can forget the risk that even one of the ca 140 relays in Soma's EQ
circuits breaks. That is a relief since those relays are not available everywhere and
desoldering them is very difficult and it would be difficult for an average tech to
find the broken one.
For many adjustments like gain trims we have to use either potentiometers or
rotary switches. Since no respected mastering device uses potentiometers we are
left with the other option. 23-position high quality rotary switches are expensive.
There is no way around it. The equation gets still more expensive since every
rotary switch must then be equipped with resistor ladders, either directly to solder
lugs (a royal pain) or small, single, local PCBs (as in Soma) which is still a lot of
work. All respected mastering equalizers therefore have eight, ten or even twelve
23-position, expensive gain switches and (sometimes) a bit less expensive
frequency switches.
If you are reading this you probably know the difference between stepped
potentiometer and rotary switch, but if you don't do find out and never mix them
because there are some equipment which use stepped potentiometers and pretend
that it is almost the same thing. Of course it isn't.
Signal connection and connectors
It is very simple. Never use connectors in the audio path if you don't have to.
Obviously you sometimes have to use them, but inside equipment their amount
should be minimized since they can fail. There are some good gold plated
connectors, yes, but you can never beat a solder joint. In Soma and in Knif gear in
general the only connectors inside are tube sockets. They are pretty impossible to
avoid of course. Luckily (and perhaps this is the only time one can say so) tubes
have limited life span, and therefore the connections will be “rubbed” every now
and then.
Modern high end manufacturers have not been able to resist the temptation to use
a lot of connectors. Almost equally disturbing is the use of million pieces of cheap
ribbon cables. You can think what ever you want about sound quality differences
between short pieces of cables or wires, but there is no way around the fact that
ribbon cables either require a connector (bad) or a solder joint. When the day
comes that you should desolder a ribbon cable you will regret that they exist. It
probably doesn't make any sense to be too delicate then, just clip them into pieces
and desolder one thread at a time. OK then. I use them too. Shame. But not for
audio and never for big part sets. In Soma there are straight, short, high quality
ribbon cables for every bands frequency selector switch, all of them without
connectors. Furthermore it is highly unlikely that they ever have to be desoldered,
since the rotary switch is easy to desolder from the small sub-PCB.
For all power and signal wiring I use only teflon insulated silver plated copper
wire. Sounds very exotic and waste of money, but actually it is not very expensive
and has substantial benefits. Teflon is the best “sounding” dielectric and
mechanically strong. The best part is that it does not melt or deform when
soldering and thus any mods or repairs can be done multiple times and everything
still remains neat and safe. I never use wires directly soldered from both ends to
circuit boards, and quite rarely even from one end. Instead I use turrets soldered to
PCBs, which serve as soldering points for wires. In this way when soldering or
desoldering there is not risk of damage to PCB or wire and everything remains
easily reassembled.
When it comes to XLR connectors there is only one right brand and it is Neutrik.
When it comes to precision and quality, certain countries seem to excel and one
has for centuries been above other, Switzerland. I use their gold plated connectors
but the regular sort is fine too if you ask me. The difference is mostly cosmetic but
if you remember to use gold plated plugs too you are free from corrosion for ever.
If only on side is gold plated it has no use.
Audio transformers
Tube equipment usually requires some audio transformers, in pro audio mostly
both input and output and sometimes, like in Knif Vari Mu II also interstage. Since
I design gear to be sonically neutral and care for good specs too one of the most
important factors is transformers. Good transformers are made by a handful of
companies only, and transformers suitable for Soma by Lundahl only. Making
small signal, low impedance transformers is easy. Making high impedance
transformers suitable for MS coding is hard, and making high impedance single
ended tube output transformers suitable for MS decoding is very hard. Choosing
suitable transformers is hard and finding the right connections and impedances and
terminations to make them work both in non-coding and coding mode is very hard
even when the starting point is Lundahl quality. Furthermore, to keep the
component costs of Soma increasing it became evident after scientifically valid
AB-evaluation that only amorphous core material is suitable for output
transformers when the ultimate low level information recovery is needed. No need
to hate normal iron though, it sounds fine, but because of hysteresis it imposes
kind of gray veil over delicate information. I have also been able to repeatedly
measure less memory effects in amorphous cores, effects that lack nice and solid
explanation at the moment but are in agreement with AB tests and null tests. Once
again you get what you pay for. Amorphous output transformers from Lundahl
cost a lot because it is not a mass product and the cores are made to order in the
USA.
Circuit boards
Thicker is better, especially in tube equipment which needs to be sturdy. There is a
limit often determined by those modern components with the shortest legs, but so
far there has been no problem in Knif equipment to use 2.4mm laminate. I find this
to be the minimum for high end equipment. It is very scary to put a tube in socket
on a regular 1,5mm thick board flexing like crazy. PCB failure in any equipment is
a major pain and should never ever happen. Thick laminate helps to keep
capacitance between top and bottom layer low. This is important in high
impedance tube circuits.
Copper on PCB in Knif products is 70u thick which is twice the normal. This is
good for heat transfer from hot components like power resistors and tubes. It also
helps to wire tube heaters in restricted space and keeps ground potentials low.
But then, the best PCB is no PCB. Point to point wiring and turret boards are
superior in most respects. They are easy to modify and repair and have superior
electrical performance in many ways, but unfortunately have to be reserved for
fairly simple gear that does not use modern components. Building Soma without
circuit boards would be next to impossible. How ever there are many ways to
utilize PCBs. The worst and very often seen is the way to fill everything on one
big thin PCB and hope that nothing ever falls apart or no one ever wants
modifications to be done.
Passive components
I don't think the sound quality between polyester and polypropylene capacitors is
worlds apart, but why go cheap here? I'm totally amazed that most high end
equalizers use polyester filter caps. They might cover some hard sounding
artifacts, but they do sound veiled and boring too. Polypropylene caps are not the
biggest investment at this level anyway, so why save those 20 dollars?
Also power supply filter caps, specifically the last ones in the filtering and
regulating “chain” must be polypropylene if possible. They are in the signal path
as much as any other capacitor at least in any single ended amplifier. Do not get
fooled to think that power supply is somehow separated from the signal, because it
isn't – current draw from power supply caps varies according to the signal and if
the caps are not ideal, their behavior will affect the signal. From all passive
components electrolytic capacitor is worst in objective terms and also the easiest
to “hear”. There are audio grade types around, but they are still miles behind film
caps. Fortunately tube circuits can use polypropylene capacitors in power supplies
quite easily. Knif Soma has dual power supply for the make up gain amplifiers, so
both channels have their own regulators and last filter caps, which are from
Mundorf, Germany.
Integrated circuits
To continue the casual style let me tell you about one comment I read today on a
certain forum. A person tried to explain others that paying big money for premium
quality gear makes sense “because it will last least 10 years” Now, please, even the
crappiest stuff can last 10 years, we are not talking about televisions or computers
here. High end audio gear is designed to last 20 years minimum, and I like to think
much further. That's why you have read my comments about possible repairs and
modifications so many times. Components do age, we replace those aged or
broken components, if (and this is the big if) they are available. I'm not claiming
that I could somehow know some part's availability after 30 years, no, but I can
say that if you have gear with programmable IC or Eprom, you will never find the
same components again and these components do break. If the manufacturer still
exists you can perhaps have them program a new chip for you, but if the original
code is lost, it is lost and somewhat tedious to find out. Possible, yes, cheap, not.
In Soma I have to use digital circuits to make the real Q -agenda possible. It was a
hard choice whether or not to use something programmable. It would have allowed
some more tricks, but then I decided to find out what can be done with decent
amount of the most basic CMOS logic. If any IC is going to be available in some
form in distant future then I think it will be these logic ICs, if not new, then at least
as surplus components. No need to worry. I have used IC sockets to make eventual
replacement possible without soldering, and in any case it is likely that CMOS
logic is more robust than programmable stuff anyway.
Thanks for reading. In this partially flaming “essay” I've covered many of the
things I find important in quality design at the moment, namely in
May 2011
Jonte Knif
