3
\$\begingroup\$

I'm trying the ADAU1701 (from Analog Devices) with an external SPDIF receiver (TI's DIX9211). I've noticed that if I change the sampling rate, the amount of bass at my device's output changes. The best results I get with 32KHz/16bit. I get good, full bass with this. But as I go up in sample rates, I lose more and more. At 48KHz I notice it's almost as I'm playing music with a portable stereo (this is a 2.1 system with 10" subwoofer, enough to shake things around).

I tried changing the sampling rate in SigmaStudio but it has no effect.

Measuring the master clock, with a 48Ksamples/sec input signal gives the correct 12.288MHz, LRCK is 48KHz and bitclock is 3.072MHz. This, of course, changes with the configured sample rate.

\$\endgroup\$
  • \$\begingroup\$ I know nothing about either of those parts, but I'll bet anything the clocks are not synced somewhere. \$\endgroup\$ – Matt Young Aug 28 '14 at 18:47
  • \$\begingroup\$ @MattYoung: the DIX9211 is the clock master, because the ADAU MCLK needs to be in sync with LRCK and bit clock. So the DIX generates all clocks. ADAU1701 doesn't have its own crystal in my setup. \$\endgroup\$ – hjf Aug 28 '14 at 19:04
  • \$\begingroup\$ I see you're all about that bass. \$\endgroup\$ – ACD Aug 28 '14 at 20:15
3
\$\begingroup\$

I'm not familiar with the parts, but some devices include built-in high-pass and low-pass filtering with characteristics that are all affected by master clock rate. Doubling the clock rate will double the cutoff frequency of all the high-pass and low-pass filters. In some cases, there may be some options to e.g. control whether cutoff frequency for the high-pass filter should be at clock/256, clock/512, or clock/1024, or clock/2048. In other cases, the high-pass and low-pass filters may have a fixed frequency relationship.

I wouldn't expect a 1.5x change in the high-pass filter cutoff to turn a signal all the way from "big and boomy" to "wimpy", but without knowing the spectral content of the sound you're listening to and the behavior of the parts, it's hard to say for sure.

\$\endgroup\$
  • \$\begingroup\$ heh well i suppose i should make more accurate measurements with fixed frequencies. I've checked the datasheet for filters but I don't see any mention of that. There is a PLL mode setting, which is set to 256*fs. That's correct. \$\endgroup\$ – hjf Aug 28 '14 at 23:19
0
\$\begingroup\$

I finally found the problem. It seems if you make any changes to the sampling rate options in Analog's SigmaStudio, you need to go into Actions -> Propagate sampling rate, so it will recalculate filter coefficients for all modules in your schematic.

After doing this, I got my bass back. (And if I go below 48KHz, I get too much of it).

\$\endgroup\$
  • \$\begingroup\$ Switched capacitor filters will do that. Unlike a resistor-capacitor filter in which the fraction of a capacitor's charge that gets transferred in a given time will be proportional to the product of that capacitor and a resistance, in a switched-capacitor filter the fraction of a capacitor that gets transferred on each cycle of a clock will be proportional to the ratio of two capacitances, and the ratio for a given amount of time will be the per-cycle ratio raised to the number of cycles in that amount of time. \$\endgroup\$ – supercat Aug 29 '14 at 2:09
  • \$\begingroup\$ you mean as an output filter? The DAC output for this circuit is a series 47uF capacitor (not filter but DC blocking capacitor) with an RC filter (560 ohms/5.6nF) \$\endgroup\$ – hjf Aug 29 '14 at 13:50
  • \$\begingroup\$ Many clocked chips that include filters internally implement them using a mixture of switched and unswitched capacitors, rather than resistors and capacitors. I would expect that the chip you're using is one such. The filter coefficients you specify effectively control the values of switched and/or unswitched capacitors [part of the device might e.g. have 0.1pF, 0.2pF, 0.4pF, and 0.8pF caps which may be individually configured as switched or unswitched, connected to a 100pF unswitched cap. The fraction of charge transferred each cycle would thus be adjustable from 1/1014 to 15/1000. \$\endgroup\$ – supercat Aug 29 '14 at 16:55
  • \$\begingroup\$ It's much easier for semiconductor manufacturers to include on-chip capacitors and control the ratios among them (without having to worry about the proportionality constant) than to include various sizes of resistors and have to achieve a particular RC time constant. Switched-capacitor filters work very well, and yield very nice consistent behavior, but a major difference compared with filters that use resistors is that--for better or for worse--cutoff frequencies scale with the supplied clock rate. \$\endgroup\$ – supercat Aug 29 '14 at 16:58

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.