This is the 1kHz square wave from a Lundahl LL1690, primaries connected to a ES9008 DAC with 195Ω source Z, secondaries into 8.9kΩ (pot before buffer).
Connection schema:
Is that slight overshoot (or is it ringing?) something to care about?
An externally hosted image should be here but it was not working when we last tested it.
Connection schema:
An externally hosted image should be here but it was not working when we last tested it.
Is that slight overshoot (or is it ringing?) something to care about?
Based on purely unscientific information, when i tested a friends buffalo DAC, iirc, the square wave showed some of the same ringing... personally I would do something to damp it myself...
Check the xfmr with an analog square wave - maybe the calibrator signal from the scope?
Also check ur scope + probe using the calibrator and look to see if it has overshoot or not, then adjust the comp trimmer in the probe or scope for best squarewave.
_-_-bear
Check the xfmr with an analog square wave - maybe the calibrator signal from the scope?
Also check ur scope + probe using the calibrator and look to see if it has overshoot or not, then adjust the comp trimmer in the probe or scope for best squarewave.
_-_-bear
Lesson learnt!
Google is my friend ...
http://www.robots.ox.ac.uk/~sjrob/Teaching/SP/l6.pdf
Thank you very much, guys!
Google is my friend ...
http://www.robots.ox.ac.uk/~sjrob/Teaching/SP/l6.pdf
Thank you very much, guys!
Last edited:
SY and RDF, et al, quite so, the chipzet is doing the ringing, but if the probe is off, the ringing may seem greater... as I said at the top, I have seen it before with this chipset.
Groundloops, you need the right rather small value of cap... use ur scope's split timebase function to measure the period of the ringing wave, invert to get freq, and then pick a cap that is suitable given the impedances... I might put it on the primary side and see if that works better or not.
Alternately a small series inductance might also prove useful, perhaps with a resistor in parallel...
And, actually, please do report back what freq you measure the ringing to be, as the xfmr seems to be doing an excellent job!
_-_-bear
PS. if you have the circuit, maybe a better place to address the ringing can be discovered. Post if possible.
Groundloops, you need the right rather small value of cap... use ur scope's split timebase function to measure the period of the ringing wave, invert to get freq, and then pick a cap that is suitable given the impedances... I might put it on the primary side and see if that works better or not.
Alternately a small series inductance might also prove useful, perhaps with a resistor in parallel...
And, actually, please do report back what freq you measure the ringing to be, as the xfmr seems to be doing an excellent job!
_-_-bear
PS. if you have the circuit, maybe a better place to address the ringing can be discovered. Post if possible.
There is no ringing to address. This is Gibb's phenomenon, caused by a brickwall filter. The 'ringing' is the inverse of the sum of all the square wave Fourier components which have been removed by the filter. Of course, you can add some extra filtering to smooth out the edges but then you are modifying the sound too so you no longer hear what came out of the anti-aliasing filter before the ADC.
You can disconnect the transformer altogether, and hook a 1 Mohm 10:1 or active probe directly to the DAC output. This will show you what the digital filter effect looks like. It will likely look the same as you see now.
I have a bunch of scope photos here showing the same thing, but they all look a little different - it depends on the filter and DAC used. That one is not too bad, I've much worse.
My reason for asking is to raise the issue of anti-alias filtering in the generator. Although a pure square wave can be generated in a computer, it would contain illegal values that would cause aliases, so filtering occurs there too.
All good fortune,
Chris
What is "flawless"? A perfect square wave will require anti-alias filtering before sampling in any digital storage system and more before any sampling rate reduction, as is done for the common (currently almost universal) oversampled recording systems.
Square waves are an abstraction from the analog world that may or may not tell us anything useful in the digital world; correlation should be proven first.
Thanks,
Chris
Square waves are an abstraction from the analog world that may or may not tell us anything useful in the digital world; correlation should be proven first.
Thanks,
Chris
Well there is a fairly clear correlation between leading edge overshoot vs. damped leading edge in terms of how it sounds to the ear.
Some people believe that this is one indicator as to why most digital reproduction sounds different than optimized analog reproduction.
If you work with power amps for any length of time, I think this relationship becomes quite obvious and self-evident. If there are papers or the like that test or "prove" this idea, I do not know.
_-_-bear
Some people believe that this is one indicator as to why most digital reproduction sounds different than optimized analog reproduction.
If you work with power amps for any length of time, I think this relationship becomes quite obvious and self-evident. If there are papers or the like that test or "prove" this idea, I do not know.
_-_-bear
A digital generated square wave can have a step response, right? You can step DC amplitude between 2 samples. This is what most test CD's with square waves do. The problem is in the reproduction side, at the DAC, digital filter, analog filter. And the DAC can have DC response.
But you can use a non-oversampling DAC, (no digital / oversampling filter) and no analog filter, and my bet is you'll get a near perfect square wave. This setup may have other problems, but there are many fans of NOS, and this is part of their argument.
Yes, this is all correct. The issues lie in anti-aliasing and in adherence to Red Book. Reconstruction filters are a fundamental part of the D/A process and can't really be wished away, just as anti-aliasing filters are fundamental to A/D conversion and to sample rate (down)conversion.
The "NOS" argument is generally misguided, but that's pretty far off topic.
Thanks,
Chris
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.