[PlanetCCRMA] Re: HDSP9652 muted on system start

Michele Spinolo michele.spinolo@tin.it
Sun Aug 28 10:12:01 2005

Hi Mark,

you pointed out some nice points, which I will go through.

First my setup is: RME HDSP9652 connected to ADI-8 Pro via ADAT, ADI-8 Pro
analog out to ADI-8 Pro analog input, ADI-8 Pro ADAT output to HDSP9652 ADAT
All stuff is running at 44.1Khz: RME HDSP9652 is set as master, while ADI-8
Pro as slave, getting clock signal from ADAT input.

Now lets go through clock settings: I noticed that setting a different clock
source for RME ADI-8 Pro inputs and outputs (i.e. ADAT input for analog
outputs, while internal clock for analog inputs) gives really really bad
I also tried to set ADI-8 Pro as master clock, connecting HDSP9652 by
wordclock (BNC 75Ohm cable) and results where really bad (not so bad as in
first set up described).
The best setup I can get is using HDSP9652 as master clock and slaving ADI-8
Pro reading clock signal from ADAT input (both for ADC and DAC).

I went through the problem posted in this thread this afternoon, and made a
new measurement: looped output 1 with input 1 and output 2 with input 2:
Results are relly bad in latter case, while in the former are comparable
with RME declared ADI-8 Pro performances.
So the problem is releated to a different behaviour of DAC and ADC on
channel 1 with respect to DAC and ADC on channel 2, NOT to hdspmixer
I find this really annoying.
Next week I will go through the problem, measuring all loops and also
crossing them, to understand if it is an single problem of a DAC or ADC
section or not.

Concerning Edward's program it a simple tool which performs these
-playing a logaritmic sine sweep signal while recording it (you can assing
sample frequency, starting and ending frequency)
-convolving the recorded logaritmic sine sweep with a inverse logaritmic
sine sweep extracting an impulse response: this is the impulse response
which represent your chain (in my case AD and DA conversions performed by
ADI-8 Pro)
That's all.


P.S. Z-sys unit is actually not used: it has some design problem which I am
trying to solve with Z-sys

P.P.S. PDF files measurements report is generated by an automatic script
which I use to analyse speaker performances: than only fig. 1 to 13 and fig.
33 to 40 are of some interest in this case. Also do not read captions which
do not describe those figures

Hi Michele,
   It may be because reading technical documents and looking at graphs
is always a bit difficult at 7AM on a Sunday morning, or simply that
I'm too dense to understand what the data is telling me, but so far
I'm not understanding the exact problem. Maybe you can clarify that a
bit more for me?

   I am interested as I have an HDSP 9652 also and I've had a problem
with it for over 3 years that the developers have never solved. I'm
wondering if my problem and what you are seeing have anything to do
with each other.

   If I understand what you are doing then you have Edward's program
generating a sine sweep, going to the HDSP 9652 which is attached to
the ADI-8 Pro and is converted to analog. That analog input is looped
back into the ADI-8 Pro, sent back to the HDSP 9652 and then to
Edward's program again. Is this the setup?

   Where is the clock master in this setup? Is it the HDSP 9652? Is it
the ADI-8? Is it something else? (You had a Zsys, correct?)

   The problem is that with hdspmixer closed you get some clipping,
but the clipping only happens when you start the program and then goes
away, while with hdspmixer open you do not see the clipping. Is this
the case?

   If I'm understanding both the setup and the problem correctly, then
please do some simple experiments with hdspconf. You may actually be
seeing a version of the problem I've had for years. With hdspmixer not
running open hdspconf and cjeck the frequency that the HDSP 9652 is
running at. Now run Edward's program. (You may need to run it a few
times.) Do you see the frequency of the HDSP 9652 changing? Do you see
the clock source of the HDSP 9652 changing? Note that you may need to
run the test a few times to see the frequency change in hdspconf. It
happens quickly. The indicator in hdspconf will move and then move
back and it's not always easy to see.

   I do see these change, and I See them change at strange times. I've
found that things are much better when running Jack since Jack locks
the device down to a specific frequency and Jack apps cannot change
it. However, with OSS and Alsa apps I see the freqency change at odd
times. It is worse with OSS emulation. These frequency changes are not
well managed, either in Alsa or OSS, and the driver does not ensure
that ''zero' volume data is going out the port when it changes
frequency, so when it happens I get horrible popping/scratching sounds
that will sometimes make me think my speakers have blown up.

   My setup here is similar to yours, assuming that I've described it
correctly above. If you see the same frequency changes then maybe we
can get some attention from the developers on this issue. I've found
it's worse using external clocks where the HDSP is being asked to sync
a better clock.