Practical Implementations of Alternative Post-DAC Filtering

Status
Not open for further replies.
This is exactly what cables like MIT do. They have in addition to complex parallel filters, simple series filters which roll off slightly around 20kHz. They do the same for interconnects, speaker cables, power cables, etc.

I can also hear this difference and I think it sounds great! My guess is that it has more audibly to do with time/phase distortion than just a bit of roll-off.

I've been using MIT for ages and don't plan on ever stopping. DACs are not the only components which benefit 🙂

They are not just cables they are cables with extra filters added to make them different, more audiophile foolery...
 
Hi Joe, thanks for answering.
Ok, my take on what is possibly going on.
The 1uF is indeed causing smoothing/slew reduction of the DAC outputs, but at the expense of increased power supply transient currents if indeed the DAC output is Pulse Density Modulated or Pulse Width Modulated.
You state that the supercap has internal resistance of around 75R, so it could be argued that the supercap is a zobel damping network covering down to very low frequencies, in addition to it's high energy storage capacity.
What happens if you connect only the 20uF ceramic cap...does this actually damage the sound due to power supply ringing perhaps ?.

Dan.

I may suggest you try by yourself this about large decoupling capacities, in any configuration you may consider. Then you can have your own conclusions about the benefit (or not) of such approach. And maybe share it here....
If you/one may want to try caps over 0,5F, please note that their discharging time is quite long, after power off. The rush in current at power on is fixed by the protection mechanism of the used regulators (must have such).
 
Hi Coris.
In other audio applications I have experimented with adding power supply caps of varying values.
In so doing I have effected improvements, but not always.
The cause of non beneficial change is supply resonances due to parasitic inductances.

I do not have a suitable DAC chip, so I am unable to try Joe's experiment for myself.
Are you able to provide waveforms of the output pins to establish the modulation type ?.

Dan.
 
Still haven't listened to the modded player (more mundane things are getting in the way) but I have been able to confirm that the latest component values I've used for the mod (post #743) give a truly flat response as measured at the line output sockets. The deviation amounts to no more than 0.05db at 19kHz (referenced to 1 kHz 0db and checked at 2, 4, 6.4, 10, 16, 18.1 and 19 kHz). The de-emphasis is also correct (that is done in the digital domain but its good to confirm it) when tested at 1kHz, 4kHz and 16kHz and at levels on the disc of -9.63, -6.46 and -0.963 db respectively. All output correctly as a -10 db playback level.

So its looking good to go when I get a chance.

Did anyone listen to the first set of files in post #711 ? Did any audible differences show ?

http://www.diyaudio.com/forums/digi...ternative-post-dac-filtering.html#post4473822
 
Hi Coris.
In other audio applications I have experimented with adding power supply caps of varying values.
In so doing I have effected improvements, but not always.
The cause of non beneficial change is supply resonances due to parasitic inductances.

I do not have a suitable DAC chip, so I am unable to try Joe's experiment for myself.
Are you able to provide waveforms of the output pins to establish the modulation type ?.

Dan.

I think it may be a misunderstanding here about these large capacities caps. There is not (as principle) about large filtering caps used in the power supplies, but large decoupling capacities for the targeted circuits. In this case, caps placed right or very near, on the power pins of the DAC chip or on the post DAC processing chips/circuits. Well, in some cases these decoupling caps it can be in the same time filtering caps, when the PSU or regulators are very near the targeted chip.

Do you think at 1khz or something on inputs and then getting the waveforms of that on outputs of a circuit, when decoupled with such capacities?
Please precise exactly the setup you are thinking about, and I may try to scope it. Right now I do not have too an accessible working circuit, but I may fix that in very near future.
As i could see in my previous measurements, there is not noticeable differences (with/without) to be seen on outputs, but the audible differences are quite obvious. Especially the low end domain is (perceived) much improved. In the case of this filtering cap effect, large decoupling capacities it accentuate even more the perceived improvements.
Quite difficult to see such things in measurements...
 
Last edited:
The use of a 'supercap' for decoupling or 'reservoir' duty is something that doesn't sit easily with me. I was in on the ground floor when these devices first became available and they have (or at least had) a poor track record for reliability and longevity. You have to understand that their primary function is to maintain a supply voltage while the cap itself experiences essentially no demands on current outflow, typically just maintaining CMOS volatile memory. The E.S.R. of these caps is high even when new (and we are talking many 10's of ohms) and only deteriorates from then on. Even when it deteriorates to many k ohms, it doesn't affect their intended use.

So the cap aspect of all this leaves me unconvinced.
 
Measurements and proof are always hard, aren't they?

The first three are design patents. The next describes a cap spanning the distance from amp to speaker with its leads. Parasitic L? The technical content here is mostly gibberish and nonsense you can patent just about anything no proof necessary. In either case the stated goal is flattening frequency response not exactly -1.5db at 20k.
 
Last edited:
Using such big cap not affect if the high perfomance of a shunt regulator like Salas Reflektor-D is used?

Agreed, can be a problem with shunt regulators since it can create a time constant that is too long and hence upset the fact that ancilliary circuitry may not see the DAC yet functioning on turn-on quickly. Using voltage regulation has been proved to work reliably. The ESR of say 75R means that peak current is 44mA and 67mA @ 3.3V and 5V, worst case.


 
So the cap aspect of all this leaves me unconvinced.

Until you try it - and listen. I do think this is definitely about reducing noise at a critical point, where that noise become jitter a la Ted Smith's explanation (on YouTube) - and that as noise descends and jitter it induces, becomes worse. Low frequency noise/jitter is worse than the side-bands we normally see - they are less objectionable sonically.

This alsmost invisible jitter was revealed by Paul Miller of Hi-Fi World some years ago using a very expensive piece of testing gear (I think he went to a university contact) and it shows via an overlay that it can be seen as +40dB or more relative to the noise floor:

Standard.gif


Blow up the Blue area x100, then this shows up - see Red:

100xStandard.gif


See how far the Red clings up realative to noise floor? Easily +40dB.

Paul Miller called it "Low Rate Uncorrelated or Noise-Like Jitter".

Could be called a number of things - but I believe that is the kind of jitter that makes things sound 'digital' and the others don't, they are less obnoxious to the ear.

Now that is simply expressing an opnion based on what I see.


 
Last edited:
Using such big cap not affect if the high perfomance of a shunt regulator like Salas Reflektor-D is used?

These caps have to be used with voltage regulators. One can try such approach with shunt regulators, but a current limit protection have to be present somehow into the regulator`s circuit. However, shunt regulator and very large decoupling capacities I see it as redundant approaches.
One should appreciate the whole system before deciding using of large cap for decoupling, when the PSU is a shunt regulator. If the shunt regulator is quite far from the chips to be decoupled, then one should appreciate the appropriate value for a decoupling cap. If the shunt regulator it may be very near the targeted chip, then very usual decoupling caps are enough. In my opinion, a shunt regulator or a voltage one + decoupling cap should be placed right on the power pins of the target circuit.
A large decoupling capacity should be always paralleled by good quality (lowest ESR/ESI) small value caps. Also a large decoupling capacity cap should be chosen for lowest ESI.
 
The use of a 'supercap' for decoupling or 'reservoir' duty is something that doesn't sit easily with me. I was in on the ground floor when these devices first became available and they have (or at least had) a poor track record for reliability and longevity. You have to understand that their primary function is to maintain a supply voltage while the cap itself experiences essentially no demands on current outflow, typically just maintaining CMOS volatile memory. The E.S.R. of these caps is high even when new (and we are talking many 10's of ohms) and only deteriorates from then on. Even when it deteriorates to many k ohms, it doesn't affect their intended use.

So the cap aspect of all this leaves me unconvinced.

Well, since these components appeared on the marked, it evolute it quite much. Their quality and stability increased significantly. I would want to use however good quality ones, and even lowest possible ESR/ESI ones for such purposes. There are today Farad range caps with ESR of few hundreds mOhm, for not very high prices.
In my opinion, even though obvious improvements these large decoupling caps it bring to the involved circuits, to work out on this filtering cap, and realize the improvement effect, there may not be absolutely necessary.
My suggestion to those interested to experiment with this filtering method, is to not care so much firstly about large decoupling capacities. This effect it is obvious without that large capacities as decoupling, but it accentuate/improve it even more if used. After a measurement approach for this effect it may be founded, it will be very interesting to see how such large decoupling caps in the circuit it may influence that measurements.
 
Last edited:
Did anyone listen to the first set of files in post #711 ? Did any audible differences show ?


Well, I did. There are (as I can perceive) audible differences in between the files in both first and second set.
I listened the files in Foobar, digitally processed as network input and optical outputted. Then optical input, DAC processed (with cap filter in place), class A analogue output.
A1 I hear it with a quite poor sounscene, where the elements are concentrated in a small central area in between the speakers, not very well definition/resolution of the sound elements (voice in centre).
A2 it have a larger soundscene, with more space, and much better definition/precision of the sound elements in the created audio space. The voice is much better reproduced with particular presence (still be in centre).
The B set I perceived as better reproduced quality than A, where B2 is better than B1. B2 it have a larger and more pleasant audio scene, the voice it have more presence, the sound elements definition into the audio space is better in B2.

I just forgot what was the meaning you`ve sent it out these files. I was not going back to read your that post before writing this actual mine. Just played back the previously downloaded files, and wrote these observations. So, I think I`m quite objective in my subjective appreciations...😉
 
Last edited:
Super Capacitor can not be used for applications in AC circuit such as ripple absorption because it has high internal resistance
(several hundred mΩ to a hundred Ω) compared to aluminum electrolysis capacitor. Thus its main use would be similar to that
of secondary battery such as power back-up in DC circuit. The following list shows the characteristics of Super Capacitors as
compared to aluminum electrolyses capacitors for power back-up and secondary batteries.

http://www.nec-tokin.com/english/product/pdf_dl/supercapacitors.pdf

1.3 Circuits though which ripple currents pass
When using a capacitor in a circuit through which ripple currents pass, please note following matters.
(1) The internal resistance of electric double-layer capacitors is higher than that of electrolytic capacitors. Electric doublelayer
capacitors may generate heat due to ripple currents.
(2) Please do not exceed the maximum operating voltage when the voltage changes from ripple.
(3) Because internal resistance is high, the gold capacitor is not basically suitable for the absorption of ripple current.

http://industrial.panasonic.com/lecs/www-data/pdf/ABC0000/ABC0000PE41.pdf
 
Agreed, can be a problem with shunt regulators since it can create a time constant that is too long and hence upset the fact that ancilliary circuitry may not see the DAC yet functioning on turn-on quickly. Using voltage regulation has been proved to work reliably. The ESR of say 75R means that peak current is 44mA and 67mA @ 3.3V and 5V, worst case.

Super Caps don't all have high ESR and many have very low internal ESR in the
order of tens of milliohms.

Check a few data sheets out.
 
A conceptual basis for how these supercaps might produce some positive result in an audio DAC does seem elusive.

Does anyone know the source impedance of the self-noise generated by a typical IC regulator, such as the LM317? Since self-noise is effectively outside the feedback loop, the impedance of that noise might be significantly large. Should that be so, then, these Farad-scale supercaps may be usefully voltage dividing the regulator's noise, especially the problematic 1/F noise located at deep infra-sonic frequencies (0.1Hz to 10Hz). Which then could be reducing the close to the carrier phase-noise produced by the oscillator circuit, resulting in reduced close-in clock jitter.
 
Last edited:
Well, I did. There are (as I can perceive) audible differences in between the files in both first and second set.
I listened the files in Foobar, digitally processed as network input and optical outputted. Then optical input, DAC processed (with cap filter in place), class A analogue output.
A1 I hear it with a quite poor sounscene, where the elements are concentrated in a small central area in between the speakers, not very well definition/resolution of the sound elements (voice in centre).
A2 it have a larger soundscene, with more space, and much better definition/precision of the sound elements in the created audio space. The voice is much better reproduced with particular presence (still be in centre).
The B set I perceived as better reproduced quality than A, where B2 is better than B1. B2 it have a larger and more pleasant audio scene, the voice it have more presence, the sound elements definition into the audio space is better in B2.

I just forgot what was the meaning you`ve sent it out these files. I was not going back to read your that post before writing this actual mine. Just played back the previously downloaded files, and wrote these observations. So, I think I'm quite objective in my subjective appreciations...😉

:up: Thank you.

Is anyone else going to give these a try ? I won't reveal anymore just yet because we might need the original files to use as a comparison for the final modified version... all of which is going to roll over until at least tomorrow before I can give it a serious listen.

There are today Farad range caps with ESR of few hundreds mOhm, for not very high prices.

Super Caps don't all have high ESR and many have very low internal ESR in the
order of tens of milliohms.

Check a few data sheets out.

I would really start to worry about using large value low E.S.R. caps as a general panacea. The two big big worries would be,

1/ It would (could) prevent the initialisation of the logic. You could have the uprocessor or syscon waiting for a response from a chip that hasn't even got power yet. Also, any reset pulse could be over with long before the chip voltage had reached a workable level. That's a very real possibility.

2/ A player like the Sony I'm using uses a small discrete series pass regulator for the DAC supply. A genuine low impedance cap of the size we are discussing would appear as a short to the reg for a significant time.

Don't let any of that detract from listening to the results. Trust your ears.
 
Given the low power demands in a DAC and that your usual DAC PSU will inherently be overbuilt for DC needs, isn't high frequency, extremely local, decoupling going to be the most important place to put our energy (I pun, I pun)?

Supercaps are generally used for local power redundancy in electronics to give the component enough time to finish its present tasks and shut down with minimal data loss. Given our application is non-mission critical, a simple 330 uF polymer-aluminum cap is far lower ESL/ESR and pretty small. That with a ~100nF 1206/805 X7R MLCC sitting right on the pins power will do far, far more good than a supercap.

I looked at Murata's datasheets for this--they make very good, no nonsense stuff. Their Supercap is around 50-60 mOhm fresh (and degrades from there) and doesn't even specify an ESL. A 330 uF 10V electrolytic is ~10 mOhm with low impedance out to 1 MHz+.

A 100 nF cap like this will finish things off well:
Capacitors | details for LLL215R71A224MA11# | Murata
 
Last edited:
Status
Not open for further replies.