You are not logged in. Please login or register.
RME User Forum → FireWire & USB series → Babyface sampling frequency problem
Hello,
I use the Babyface at 192 Khz / 24bits.
I measured the sampling frequency recording a PPS (pulse per second) signal delivered by GPS card, and the measure gives 192001.73 Hz !!
For me it's a problem of calibration of the internal clock.
Is this normal ?
Is there a way to recalibrate the clock ?
Thanks in advance.
It is normal. Please try to calculate how many ppm deviation your value equals. It has zero meaning in real world. You can also play around with the pitch slider if you want to change the frequency (and to find out what meaning it has to change by 1 Hz @192 kHz).
It is an amazing feat that today you can even confidantly measure to that accuracy, because GPS gives you access to flying caesium clocks. At my entry to electronics the most accessable time and frequency standard was WWVH!!
But to put the accuracy in context. Let's say you are using your BabyFace to record a sound file in double system to be synchronised to a video running exactly to speed and expecting the file to be exactly 192000Hz. (I've done this recently but not at 192K). SMPTE standards say that sound and picture should be synchronised to within 50ms (slightly more than one frame). So this would mean, that if picture and sound were synchronised at the start of the file, it would be over two hours before there was a danger of noting that picture and sound were not in synch. So doing a 'wild' recording of a concert (two halves, each of one hour) and synchronising picture and sound at the head of each half, no other synchronisation action should be required (video cameras have very accurate crystal control these days).
Commercial time code accuracy nowadays is expected to hold to one frame within 24 hours, but they use special crystal oscillators that can cost as much as a BabyFace alone.
And even if you have absolute pitch, a one cent shift in pitch (one hundredth of a semi-tone) would require a deviation of 100Hz in the sampling frequency. And only machines can detect a one cent shift ....
OK, but I absolutly need a good sampling frequency precision because I use the number of sample to timestamp pulses in the signal.
In comparison, I did the same test with a SoundDevice 744T and the measured sampling frequency is 191999.97 Hz !
I chose the Babyface because it's easier to use and I can, in real time, measure sound level and process the input signal.
So if I want to solve this problem, can I use an optical external clock ?
If yes, do you know product that can do this ?
Thanks.
If you run multiple tests, is the outcome the same?
You could maybe use the pitch slider to tune it exactly right.
I did the same test several times and I have the same results (in the same conditions). But the clock drift with temperature.
So I will do a test with the Babyface after a trip in a refrigerator, and I will check if the frequency shifts.
After 4h in a refrigerator at 5°C (roughly), I measured 192002.19 Hz.
I did the same test with a SoundDevice 744T and the measured sampling frequency is 191999.97 Hz !
The 'T' in 744T means 'timecode' so see my remarks re this in previous post - and it is remarkably accurate - and 5 times the price of a BabyFace.
Perhaps you could use the digital out from a 744T as a sync input to the BabyFace :-)
Are you saying that in your application you are trying to accurately time (recorded) pulses by counting the number of samples between them? To what accuracy do you need to make the measurement?
Of course the 744T is more expensive and no so convenient, because you have to record and after that you can analyze the data.
With a USB card I can do the same thing in real time.
I need an accuracy of 10-7 s/s for the clock.
After 4h in a refrigerator at 5°C (roughly), I measured 192002.19 Hz.
That´s not helping, but did you try the pitch slider and set it to 191999 or 191998 and see which one is closest?
Coarse modification in steps of 50 Hz
is done by clicking with the mouse to
the left and right of the fader knob.
Fine
Fine modification in steps of 1 Hz is
done by using the left/right cursor
keys.
Reset
Ctrl key plus left mouse click.
I cannot control the ambiant temperature so I will always have different frequency. So calibrate with the pitch is indeed not helping. 
and your 744T is always the same at all temperatures? Are you not taking this a little to far for consumer equipment? Then you will need a wordclock with oven and optical spdif out to use your babyface. Or get an electrical to optical spdif converter and clock to the 744T.
Still I would try at -2hz and see what it does after being on for an hour (will smooth out temperature fluctuations, because of internal temperature).
For the 744T, yes the clock is stable because it is compensated in temperature (we have ultra-stable clock doing the same).
In fact I'm looking for a way to use clock with the optical input using an external ultra-stable clock.
I'm no expert when it comes to clocks, but my thoughs:
Put the Babyface on something like this to have better control over its temperatures (random pick, not a specific preference).
You should at least try if using the Pitch shift makes the clock run closer around your needed frequency. 192.000 +-1 is better than 192.002 +-1.
Forgive me but I´m getting really curious what you are doing that needs this.
I need an accuracy of 10-7 s/s for the clock.
So you need a 10MHz reference accurate to 1Hz. Could you not record the 2-second pulse from GPS as a reference and calibrate to this? Even if you used the 2-sec reference, there is no guarantee about the consistency of the rise-time of the 2 second pulse to this accuracy.
Of course there is no indication whether this is a short term (say 1ms) or long term (minutes) requirement. If you want an accuracy of 0.1 ppm in your measurement, then you have to have at least 10 million samples to be accurate to one sample. Therefore your recording needs to be at least 53 seconds long to resolve to this accuracy, ie., plus or minus one sample.
The BabyFace is a capable and flexible tool, but I think it might be stretching things a bit far to put it into this role. Likewise I am curious as to what you are trying to measure ....
O)n the other hand, maybe (some parts of) the audio world needs accurate sampling rate. I see a product opportunity for a GPS based LTC/AES generator that is highly accurate. This is what your application seems to require - synchronise the BF using this into the SPDIF input. Any
So you need a 10MHz reference accurate to 1Hz. Could you not record the 2-second pulse from GPS as a reference and calibrate to this? Even if you used the 2-sec reference, there is no guarantee about the consistency of the rise-time of the 2 second pulse to this accuracy.
I will probably do this.
About the SPDIF input, I can use ultra-stable clock to generate sync signal but I ignore what is the type of signal needed for this mode.
Can you help me on this point : web site, literature.... ?
Thanks.
Ideally it would need an SPDIF (or AES-3) compliant waveform for the BabyFace to sync to.
However your requirement is merely for the BabyFace to lock to the bitrate of the input. Perhaps RME may be able to advise what is the minimum feature requirement for the input waveform to lock. I think it would need to be something more than just a square wave at the appropriate multiple of the clock rate (no framing structure or the like). At best you would have to design an SPDIF generator (there are chips out there to make it easier) whose clock is derived from the stable reference frequency.
Pretty esoteric, but I expect there are some gear-nuts out there who could advise you better than I.
RME User Forum → FireWire & USB series → Babyface sampling frequency problem
Powered by PunBB, supported by Informer Technologies, Inc.