TPC DAC-1...TNES
Pete Goudreau
TNES...The Never Ending Saga...:-)
Finished another modification and test sequence over the weekend that
proved to get more noise out of the unit since the last test. It
still isn't there but the last mod (which involved bypassing the
output RCAs, rewiring the output with GS-4, and adding a clamp-on
ferrite bead to the GS-4) and the correction of ground plane to
chassis bypassing went a long way towards reducing the noise in the
output. This time around, it was virtually impossible to tell the
difference between the modified DAC-1 and the No.36 on first listen
unlike before where the noise floor obscured low level detail enough
to be obvious on first listen.
This modification involved removing the existing chassis bypass caps
at the DC return of the power supply output and at the two ground
planes at the analog outputs. Copper tape was used to add a chassis
ground copper fill at both locations with two being added at the
former in order to link the common of the main filter caps to the new
bypass caps. Additionally, the previously added copper fill near the
digital input plane (for connection of the digital input pulse xfmr's
shield pin) provided a location for local bypass there and another new
copper fill had to be added in the remaining corner of the board so
that the +/-5V regulator common for the DAC ICs could be bypassed as
well.
Using 22nF/100V/X7R/1206 chip caps, one each was installed at each of
the five bypass locations. The sound of the unit was improved only
very slightly over that of the standard bypass arrangement but it was
noticable. Went back in and added another cap at each of the
locations bringing the bypass total to 44nF at each location; this is
close to the 47nF value of the existing bypass caps in place in two of
the locations in the stock unit.
The improvement with the addition of the extra caps was really
amazing, definitely better than it had ever been. For the first time,
imaging was clearly enough defined that one didn't have to mentally
strain to visually it, although the strain had been growing less and
less with each modification, it was still there; with this level of
modification, it was no longer necessary to do so. My immediate
perception of the improvement was that it had edged into the realm of
true high-end sound for the first time, albeit just barely.
Apparently enough noise reduction had been accomplished with this mod
that the low level detail available in the signal could be more
readily discerned indicating that it was possible to reduce noise
further without resorting to jitter reduction techniques yet although
that is coming as there is little else possible in the DAC-1 at this
point that isn't a function of the board layout, chassis design, and
power supply topology. These issues aren't really addressable in this
unit and will have to wait until I can get a correct board layout done
and fabricated, etc...yeah, right...anybody know where a spare pile of
money's laying about...;-)
In order to get a more accurate take on the modified unit's sound, I
enlisted the aid of a friend who works part time at the hifihut where
the testing was done as his ears are definitely solid gold. He was
able to get in a short critical listen to the unit after it had been
warmed up for a few hours. He was able to detect a noise floor in the
DAC-1 that is still higher than the No.36 which obscures some low
level detail leading to a greyness of background and a slight
obscuration of the space between instruments. He also felt that the
unit sounded a bit dark and when combined with the good dynamics of
the unit, sounded a bit forward and somewhat "hi-fi-ish". I'm sure he
will jump in and correct or add to any of my comments here but that is
his decision. This result wasn't a real big surprise as there is
still no reclocker in the unit. BTW, I didn't listen to the unit this
time around at the store relying instead on the objective reports of
others.
One of the rahe denizens here (he'll weigh in if he wishes, I'm sure)
also took time to drop by and take a listen after the unit had had a
day to warm up and reported that there was still a bit of shrillness
to the response and a bit of excessive resonance in the bass which
distinguished the two units in much the same way as the visual
difference between good laser printer output and a photosetter output
in that if you look closely you can see the jaggies and once noticed
it is impossible not to see them anymore. An excellent analogy, wish
Id thought of it...:-)
Anyway, it was an interesting set of tests and indicated that there
are some subtlties in the current state of the unit that may or may
not be related to jitter in the bit clock. The next step will then
have to be adding a reclocker to minimize this as the unit is
currently using the master clock output of the 8412 rcvr IC which is
about 200ps which, I think, corresponds to a noise floor at the 14 bit
level...someone please correct me if I'm wrong on this. With any
luck, this should be complete in a few weeks as all of the parts are
finally here to make a stab at it. It should be interesting to hear
what differences this will make and whether there are any subtle noise
and distortion issues hiding underneath.
As a side note, I went back into the unit, after the testing was
complete, to add more of the 22nF caps to bring the total local bypass
value up to near 100nF where I thought it ought to be. That had only
a slight effect but nothing dramatic which indicates that a) 40-50nF
is probably plenty of chassis bypass and b) since this is about what
the value was before, the inductance of the chassis bypass is more
critical than the value...hardly a surprise there. It's clear at this
point that leaded components simply will not work in this environment
but that shouldn't come as a surprise either as high speed digital
circuitry in the presence of highly sensitive analog circuitry
virtually demands SMT and multilayer design to minimize noise
generation and coupling, shame it is so expensive to do for small
companies...:-(
As far as I can tell, there is little else to do to the unit to
further improve performance other than making an attempt to reduce
jitter and am still amazed that it could be improved this much just by
applying industry standard engineering practices aimed at reducing
noise and EMI. A better board layout with optimum separation of left
and right channels' DC/low freq power and ground impedance as well as
improved shielding of the digital section and reduction of RF ground
loops would likely make a significant improvement and along with
jitter reduction might push the performance very close to that of the
big buck stuff but the reconstruction filter and analog output stage
might remain limiting factors.
A 7th or 9th order passive RLC quasi-bessel filter in a sectionalized
shielded can along with a '744/BUF-04 or '744/'811 composite buffer
might also go a long way towards finishing off the unit but the cost
of both a new layout/chassis and a custom built filter are quite high
and that defeats the purpose of low cost modifications. The cost of
even having these fabricated is sufficiently high so as to make it
uneconomical as well so they will likely remain on paper as if the
money was available to do this (>$5K) I would just buy a No.36 or a
comparably priced Wadia and be done with it. As it is, I suspect the
addition of a reclocker will be more than sufficient to make it
perfectly acceptable for my use for the forseeable future given the
limitations of my current system. It does sound pretty acceptable now
even though some of the shrillness and forwardness are occasionally
audible but it is, by far, much, much better than it was in stock
condition.
BTW, attempts to modify the xport and UltraJitterbug to reduce EMI
were unsuccessful other than removing the jumpers on the AES/EBU
output of the UJB to disable it and reduce EMI somewhat. These
modifications to either unit were clearly audible and were for the
worse and were then removed from the units. Also, removal of the UJB
from the system was clearly a bad thing and although subtle, was quite
audible as a reduction of timbral resolution and a flattening of the
soundstage...sounds kinds like jitter problems to me...:-)
Just for reference, my system is as follows:
EAD T-1000 (modified) on a CD Sink
SF UJB (modified)
TPC DAC-1 (modified)
Tensolite/Amphenol BNC digital interconnects with ferrite beads
Stock power cords with ferrite beads
Reference Line Audio Preeminence One, Series Two passive controller
Krell KSA-150
JMlabs 715 Oriane K2 speakers (modified with Tc120Tdx tweeter)
Monster 8-cut turbine RCAs/Canare GS-6 interconnects
Homemade low inductance/high capacitance spkr cable
Magnum Dynalabs FT-11 tuner
Hope this was of some interest,
Pete
brad sanders
Pete Goudreau <pgou...@lsi.dsccc.com> sez:
>A 7th or 9th order passive RLC quasi-bessel filter in a sectionalized
>shielded can along with a '744/BUF-04 or '744/'811 composite buffer
>might also go a long way towards finishing off the unit but the cost
>of both a new layout/chassis and a custom built filter are quite
high...
Jeez, Pete, I'm a little disappointed. Get out the copper clad, a 50W
Ungar and some 2% silver, and start making boxes! If it's passive,
does it *really* have to be inside the "main" box? Give it a 75Ohm
input and output, and...
---------------------------------- Inspired by actual events ----
> http://radioactive.home.ml.org <-------------------------------
Pete Goudreau
> On Tue, Dec. 17, 1996 radio...@geocities.com (brad sanders) wrote,
>
> Pete Goudreau <pgou...@lsi.dsccc.com> sez:
>
> >A 7th or 9th order passive RLC quasi-bessel filter in a sectionalized
> >shielded can along with a '744/BUF-04 or '744/'811 composite buffer
> >might also go a long way towards finishing off the unit but the cost
> >of both a new layout/chassis and a custom built filter are quite
> high...
>
> Jeez, Pete, I'm a little disappointed. Get out the copper clad, a 50W
> Ungar and some 2% silver, and start making boxes! If it's passive,
> does it *really* have to be inside the "main" box? Give it a 75Ohm
> input and output, and...
Me too, actually. I've built sectionalized shielded boxes before for
RF applications but it is the design of the filter that I don't have
the tools for. Saber on our workstations here doesn't have a filter
synthesis package and I guess I'm just too lazy to go to the library
and find the books most likely to have this filter in canonical form.
Besides, the cruel part is figuring out the appropriate input
impedance so that the I-V stage op-amp can drive it easily without
being pushed too hard and the actual inductor design so as to minimize
distortion components arising from the nonlinearity with branch
voltage (not the easiest thing in the world to do).
I'd want it directly on the ground plane between the I-V stage and the
output buffer for no other reason than to avoid the possibility of
radiative coupling around the filter and to simplify current flow in
the regulator/bypass system on board. Just can't see it doing much
good hanging out the unit where it can only affect conducted noise...
OTOH, it *could* be fun...:-)
Cheers,
Pete