Practical Implementations of Alternative Post-DAC Filtering

[Mark Whitney]

I am not trying to be mean.
I start from a position that all the claims are true and that the circuits do something. Then I look for the connections: If it works for current and voltage DACs, what can be concluded from that? Is a voltage DAC not a current DAC with an op-amp build in?
Somebody claims the filter also works as post op-amp, another clue?

 

[Joe Rasmussen]

I am not trying to be mean.
I start from a position that all the claims are true and that the circuits do something.

Indeed, can't fault that. Same here.

If it works for current and voltage DACs, what can be concluded from that? Is a voltage DAC not a current DAC with an op-amp build in?

This is where I am more comfortable with current DACs. This is much trickier with voltage DACs, but when you get it right, the effect is just as strong. We are talking about switched capacitor output stages - actually, the best results here does not come from a filter, but a Zobel, one that has little affect above 100KHz. At this stage we will try to solve this with current DACs first and then see what that reveals and take another look at voltage outs. Somehow, there has to be a common denominator. The effect is not mild, you have to hear it to appreciate this. I really mean it, the effect is not subtle. BTW, I do have an R&D background going back to the 70's and have some experience - looking for commonality in behaviour is important.

Somebody claims the filter also works as post op-amp, another clue?

No, I think that was some kind of 'noise' - staying with current, we know it must be applied to the pins of the DAC, something that is not buffered. Once it is buffered, as after the I/V, then it cannot affect the DAC, stating the bleating obvious I know. But at that point and after, it can be used to correct FR, if so wished. It just becomes the same a tone control after that, as you would expect.

There is something else, that I was expecting somebody to ask the question and have not. Note the large input capacitance on the power 0.33 Farad. I believe this is important as a clue, that they don't affect FR and yet the improvement is actually similar to that of the post-DAC filter. Like, more of the same. One affects FR and the other does not. Following my instincts, this tells me that the improvement we hear is in the time domain - it is a jitter reduction mechanism, both of them.

That latter point, I am not the only one that has made that postulation. It also is encouraging then, that a physical measurement can be made.

Now I am not sure if you are an objectivist or subjectivist? I have clashed with both camps - but this is one that I dearly want the measurement. Some here don't seem to hear me, I have stated enough times.

Cheers, Joe

 

[SY]

...but this is one that I dearly want the measurement. Some here don't seem to hear me, I have stated enough times.

Then get off the Internet and do it. This is perhaps an hour or two of work, exclusive of soldering caps time.

 

[madds1]

For what it is worth - I did these changes about a year ago... ES9018 differential into a 1:4 Lundahl BUT with a silver mica cap across the pins on the primary resulting in a 3db at 23khz

I have had several of my friends listen to this combination and it has a very "noticeable" effect.

Yes.. the effect is similar to a better clock (low jitter) but that is the subtle part of it. The primary effect is better layering and drastic improvement in instrument placement and listener involvement.

I tried multiple caps (Polyester, Polypropylene, Teflon etc) but Silver Mica (Russian) seemed to be the best.

I can affirm that this effect is most pronounced for sigma delta differential current dacs. Doing the same for single ended (say ES9023) or after the conversion to single needed produces no effect. AK4399/AK4396/CS4398/WM8741 (voltage output dacs but differential) do benefit but not to the extent observed with ES9018

I have made many changes to my system over the last 10 years and this change is the one after which I stopped (for now until I get bitten again)

As an aside - I first noticed this mod here http://www.raylectronics.nl/pdfs/Mauros_CD6367_tweak.pdf

Hope this helps others looking to solve this mystery

Peace!

 

[Greg Stewart]

Joe,

How does this affect DSD played back through a multiple-mode DAC. I have a Sony HAP-Z1ES I've been molesting with signal path and power supply mods. It uses a pair of PCM1795 DACs into an analog output setup very close to your first example. It can of course playback both PCM and DSD, but also has an option to convert all to DSD. How is your experience with this setup and DSD playback?

Also, it already has significant capacitance at the 3.3v & 5v inputs to the DAC, 3-sizes of capacitance at each right at the DAC with the largest between 50uf and 100uf. Would the 22uf SMD caps still apply?

TIA!

Greg in Mississippi

 

[plasnu]

I am not trying to be mean.
I start from a position that all the claims are true and that the circuits do something. Then I look for the connections: If it works for current and voltage DACs, what can be concluded from that? Is a voltage DAC not a current DAC with an op-amp build in?
Somebody claims the filter also works as post op-amp, another clue?

This is exactly what I have been thinking...

 

[plasnu]

Doing the same for single ended (say ES9023) or after the conversion to single needed produces no effect.

Can anyone explain why no effect on SE configuration?

 

[madds1]

I don't know why it doesn't work for a SE output (even after the conversion from differential via opamps or transformer although with transformers some of the effect is still there)

I suspect that is what Barbara E. Gerhold (Author of DAB), Mauro Penasa and Joe discovered independently is by experimenting and listening and each have their theories.

I think Joe is trying to find an explanation via measurements.

I can tell you this much - I fed PCM 20Khz sine through these DAC and checked the diff output with and without the cap. Couldn't see any difference visually on the scope. I don't have a spectrum analyzer to check with

In the same vein - I have found that choke loaded supplies with dacs and preamps shine, ES9018 likes a 1+1+1+1:4 transformer after, Silver mica sounds good for filtering, Serial crossovers in some speakers sound more powerful (Magnepans especially), Single ended JFET stages (or tubes) as preamps which introduce 2nd and 3rd harmonics (like -75db and -90db) bring life to music

I have no explanation for any of the above although my ears (which I understand are subjective) can vouch for. And none of the above are my claims - I simply tried out whatever anyone suggested and verified for myself based solely on listening (not "double" blind but I close my eyes while listening)

Not to digress from the topic but I think it behooves any audiophile who is in pursuit of perfection to try this irresistibly simple tweak

Cheers

 

[Joe Rasmussen]

.

Salute the silent majority out there, we have bit 500+ views per day.

.

 

[Coris]

I can confirm too, that a simple measurement with/without cap it shows no any differences. Nor on a spectrum analyser... This is intriguing.
We may find a method to measure what is going on here, but first we should find out where to look for changes/differences with/without cap, or what is to be measured. A frequency response modification by introducing this cap is normal and is measurable, but in my opinion this is an adjacent consequence of this cap in place, but not the main cause of the improvements it introduce in the system, and do not define the exactly functional mechanism. Please keep in mind that here is about a drastic improvement in the precision of the sound elements in the space of the sound scene. How is to measure this???
I think there is not quite appropriate to call it the result of this cap approach as an effect. In my opinion what happen here is not introducing of an effect by using this cap, but an improvement in the signal processing, which we for moment do not know exactly where it happen and what actually is.

 

[Zoran]

I think that the filtering should be start earlyer in F domain. But with softer slope to keep the phase in minimum drift at the top...
here is one simple circuit. can be made from ferrite beads 22.2mH=2 x 10mH + 2.2mH
10mH usually have DC res of 50ohm cca so total DC res have to be substracted from R followed also for other R in the chain.

this filter have -1.25db @ 20KHz -14deg Phase.

 

[Zoran]

this is example with simple JFET driver-buffers but it could be done with OPs too...

 

[Zoran]

I try to see what is happening beyond 20K
And That is what should look like with applied filter
Sin(x)/x with 44.1K Fs based

 

[Zoran]

It is Very hard to acheive that kind of propper filtering WITHOUT disturbing phase in audio
domain. And Should be 2 filters one for 44.1KHz base other for 48KHz base...
Some passive examples You can find from Zanden papers. And that could be acheived with FDNR filters with L emulated from OP amps...
All that sections should be inside of -6db or so (take a look at the graph)

 

[SY]

I can confirm too, that a simple measurement with/without cap it shows no any differences. Nor on a spectrum analyser...

Can you give me a link to these measurements? Joe says they haven't been done yet and that measurements will show significant differences between filtered and unfiltered as well as filtered at one point versus filtered at another. That's in lieu of any valid (ears-only, no peeking) listening tests, which he's been clear that he wants to avoid.

 

[Joe Rasmussen]

...in lieu of any valid (ears-only, no peeking) listening tests, which he's been clear that he wants to avoid.

Just not a 100% believer in them - call me agnostic. I have my reasons and I believe they are every bit as valid as yours - that's all.

But yes, let's find a measurement - quite willing to let somebody better than me figure it out.

 

[Max Headroom]

Joe, you state 1uF cap....what kind ?....and data sheet specs please ?.
Ditto the supply bypass caps please.

Dan.

 

[SY]

Just not a 100% believer in them - call me agnostic. I have my reasons and I believe they are every bit as valid as yours - that's all.

Yes, you've been clear that you need to peek. That's understandable- neither you nor I are youngsters. Perhaps you can find someone else who actually can hear these claimed differences without peeking to run an ears-only listening test?

 

[Mark Whitney]

I can confirm too, that a simple measurement with/without cap it shows no any differences. Nor on a spectrum analyser... This is intriguing.
We may find a method to measure what is going on here, but first we should find out where to look for changes/differences with/without cap, or what is to be measured. A frequency response modification by introducing this cap is normal and is measurable, but in my opinion this is an adjacent consequence of this cap in place, but not the main cause of the improvements it introduce in the system, and do not define the exactly functional mechanism. Please keep in mind that here is about a drastic improvement in the precision of the sound elements in the space of the sound scene. How is to measure this???
I think there is not quite appropriate to call it the result of this cap approach as an effect. In my opinion what happen here is not introducing of an effect by using this cap, but an improvement in the signal processing, which we for moment do not know exactly where it happen and what actually is.

Simulation of the schematic posted at the beginning of this thread for the PCM1794 shows an attenuation of the output by 20% (500mV->400mV) and a phase shift of -55 degrees at 20 KHz.

 

[Joe Rasmussen]

Perhaps you can find someone else who actually can hear these claimed differences without peeking to run an ears-only listening test?

I honestly don't understand? You are basically saying we are all deficient and can't trust our ears? I know there are pitfalls, but there always is. Even in double-blind tests.

Perhaps the biggest objection I have, and this is an opinion, and in my mind as valid as yours, is the concept of differences or listening for differences.

Let me explain respectfully, when I sit down I never listen for differences, it is not natural for me, I don't ever do it. Is this just me? I don't think so - but a double-blind test is is all about listening for... differences.

I think we have a much better handle now on how we listen. One thing that has emerged (and double-blind enthusiasts are yet to catch up) is that when a person sits down to listen to a piece of music, he decides, without even thinking about it, very subconsciously, what to listen for at that particular moment. This will depend on his mood, perhaps even the time of day or what kind of day it has been. Will he be drawn to rhythmic pattern on this occasion, will it be that lovely tone of the trumpet on the left channel? Could it be some aspect of the imaging that captures his attention this time? It could be any element of the music performed and have so many variables.

Yet this is not even at the heart of the problem. The real problem comes next time he listens to the same music. This time, will he be listening to the same element as before? If this a listening test he will be forced to, so now he no longer makes the decision subconsciously. The truth is that naturally, we are inclined when listening again, to explore something different in the same piece of music. This time we may be drawn by a much different element, the phrasing of the piano on the right channel? Indeed, if we did not explore something else in that same piece of music, we would become bored - so we don't. I know that when I get bored I am disinterested, no longer focused. Concentration lapses.

I really could go on, there is so much more to this topic - indeed I should write an article about this - there are layers I have not even touched on.

Does this mean I am a total disbeliever in blind tests - not at all - I only think they should be reconsidered as the current orthodoxy is problematic in light of better understanding. There has to be a better way that it could be done - one that gets away from merely trying to detect differences, because we end up asking what differences?

For example, do double-blind tests that only feature one musical instrument. It would not be perfect, but it would be an improvement. I suspect that music for double-blind testing have to be specifically recorded and the music should not be chosen ad hoc. Much more considered thought about the source material - it is crucial. A solo drum kit recording, stuff like that. Then gradually expose the listener to recordings with only two elements. Reduce the need for the listener to make subconscious decisions, even though he does not even realise himself that he is asking.

I do think you have prejudged me somewhat - just a bit too quick to draw conclusions about me. Have I once said I was against double-blind testing? Not once!

I can see both sides of the fence, I can see the pros and cons, and that means I can see cons in both, but that you can only see cons in one.

Bottom line? Double-blind tests can be made better.

But it isn't going to be easy. You see, even in controlled double-blind listening tests, our minds can still play tricks on us.

Respectfully, Joe

.