EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: Yansi on January 24, 2016, 07:50:45 pm
-
Hello,
I am currently working on some simple design consisting of multichannel audio DAC. The DAC provides differential outputs at +10dBu level. I'd like to interface it to a symmetrical audio cable, with the same nominal level of (up to) +10dBu.
The question is, how to do it, efficiently?
The DAC we are talking about is CS4365. Here is the datasheet https://www.cirrus.com/en/pubs/proDatasheet/CS4365_F2.pdf (https://www.cirrus.com/en/pubs/proDatasheet/CS4365_F2.pdf)
and here is an evaluation board schematic for a similar DAC, CS4362: http://www.cirrus.com/cn/pubs/rdDatasheet/cs4362eb-1.pdf (http://www.cirrus.com/cn/pubs/rdDatasheet/cs4362eb-1.pdf)
Most probably, two things are required to do: Lowpass filtering and impedance matching.
The filter I'm considering should be a bessel type (due to impulse response characteristics) and two pole. But how to choose the cutoff frequency? There is no guide from the manufacturer, the datasheet is pretty basic, not providing any additional info about the required analog design. There is only statement that a 50kHz on-chip filter is present (including all the characteristics). It seems the filter is not an analog one, it is a digital one on the DSD direct inputs. (no idea about DSD interfaces).
So do I need to build another external filter? (Probably yes?) What cutoff? (bessel, 2 pole) Is there a rule of thumb for the cutoff related to a sampling rate?
Impedance matching. I need to drive a length of symmetrical audio line. The signal level required is up to 10dBu (2.3V~). The DAC already provides such output level. So it seems only a unity gain differential buffer will be needed. (plus the filter).
Something like OPA1632 would be nice, but due to the number of channels and overall cost, OPA1632 is currently not the option. Can someone please suggest suitable cheaper alternative of fully differential OPamp for this audio application?
The design already uses a buttload of NE5534/5532 opams. Is there a way those could be used instead?
Thank you for suggestions.
Y.
-
A 5532 is THE standard jellybean opamp in pro audio, and yea, depending on exactly what performance you need they will do just fine for this.
I recommend a trip to your local library and specifically commend a copy of 'small signal audio design' by Douglas Self to your attention.
Also look at AES-48 for some recommendations as to what to do with the screen on shielded cables.
Regards, Dan.
-
The NE5532 should be fine.
If you really want better, without too much expense, there are plenty of alternatives, such as the LM4562.
http://www.ti.com/lit/ds/symlink/lm4562.pdf (http://www.ti.com/lit/ds/symlink/lm4562.pdf)
Unless you're using it to amplify a very small signal, such as from a microphone, then I doubt you'll notice any difference.
-
I have nothing against using 5532 opamps, there is currently 16 of them in the design. The only question is how to use them correctly for the DAC buffer.
The best would be some FD opamp, as both the input and ouput required is differential. Or possibly to "partialy replace" the FD opamp by two standard ones (single 5532 package).
Does anyone have a tip for a FD opamp for audio, that would not cost a fortune like the OPA1632? I have googled a bit, but haven't discovered any useful or cheaper to use. (Going through the TID's website reveals no other than the OPA1632.) Seems like nothing else exists.
'small signal audio design' by Douglas Self - I know that book (already have a copy!) but never got time to actualy read it :-(
I've just opened it and listed line output driver section. Will go through and then I'll add a report what I've found there. :-+
//EDIT: Wow! Heavy stuff there... I definitely must take some time to read all of that...
-
Surprise, surprise!
Product evaluation board from cirrus... :-DD
-
http://www.analog.com/library/analogDialogue/archives/39-05/op_amp_applications_handbook.html (http://www.analog.com/library/analogDialogue/archives/39-05/op_amp_applications_handbook.html)
sections 3 on ADC/DAC interface, 6 has subsection on audio line drivers
you do need to spec cable length, load, decide bal/unbalanced - and cable impedance matching really isn't a audio frequency thing - although balanced driver/receivers have controlled impedances to maintain balance, but source/load can differ by orders of magnitude
you do usually see some defined series output Z to decouple the op amp/buffer from the cable Cload
most audio DAC today have internal upsampling to >64x and implement a steep digital filter internally for the the 1st Nyquist zone of the incoming sample rate, you still need to attenuate clock feedthru and images at the upsampled rate Nyquist and above
but the digital filter lets you make the analog filter be comfortably above the 1st Nyqust to reduce its impact on audio frequency accuracy
Bessel sucks for audio accuracy, low order Bessel is rolling off over much of the passband
Butterworth maximizes frequency flatness - which is clearly audible, so Butterworth is the most commonly used alignment in audio
there is much unjustified hyperventilating around phase accuracy in audio
phase shift that matches in all channels is poorly audible, above ~ 4 kHz direct phase discrimination of human hearing is pretty much gone