DAC Filtering - the "Rasmussen Effect"

Status
Not open for further replies.
IOW, it's reducing distortion, in the D/A process. I've been "hearing" this behaviour for 30 years now, achieved using different techniques - for the first 25 years or so of that period I found the quality of normal digital replay, on other systems, pretty dreadful - only recently does the overall standard seem to have been raised significantly ...
 
I read through Joe but couldn't for the life of me pick out what exactly is the 'Rasmussen hypothesis' as to the mechanism of the effect (which incidentally I don't dispute). I searched but could not find a testable hypothesis. Do you have one? 'The cap has an effect on the DAC' does not count (as I've pointed out previously) because its not falsifiable.

If you consider my OOB noise filtering hypothesis has been falsified by your experiment with the tube amp then do please explain why. I'm not attached to my hypothesis, happy to give it up when nature requires that of me. But just not when Joe requires it 😀
 
IOW, it's reducing distortion, in the D/A process. I've been "hearing" this behaviour for 30 years now...

I think you are right and that is the point I believe I have been making as well. I have known for a long time that a filter like this has an audible effect - and one day I followed my instinct and kept going without actually measuring how far down it was at 20KHz. And it sounded so much better and then I measured the FR.

I shall continue to repeat it: It was down 1.3dB @ 20KHz relative to 1KHz.

for the first 25 years or so of that period I found the quality of normal digital replay, on other systems, pretty dreadful - only recently does the overall standard seem to have been raised significantly ...

I think a lot of us have had that experience.

Mind you, as you would know, many of the original Ladder DACs were not that great (some had so much post-DAC ringing due to brickwall type filters, if I remember correctly - and then some of these multi-bit DACs came good to a certain extent.

I know people, some for 15 years plus, and they won't touch d-s DACs and stick only with Ladder DACs and/or NOS-DACs.

Here is an hypothesis - what they are hearing (hating?) in d-s DACs is what we are dealing with here... I believe you are right, what we are perceiving is some form of distortion in the D/A process that can be improved by a more aggressive non-ringing 1st order filter.

An hypothesis is just that - an explanation, or a possible explanation, for a phenomenon that has been observed. It may be no more that a tentative explanation.

Cheers, Joe
 
I like the fact you call it PDM, which I kinda prefer over PWM. If this is a demodulator/integrator issue, then D for density is better for me pictorily, the way I view it in my head. Nut maybe that is just me?

Technically PDM is the correct general term, PWM is a special case of PDM.

I am Northern European as well. It was an evening we first heard that Kennedy had been taken to hospital and about an hour and a half later they announced he was dead. I was in my bedroom shared by my brother and the radio was on in the adjoining lounge room. We lived in Mørkhøj, a suburb of Copenhagen.

Ah, the days - I grew up in Helsinki, but spent my childhood summers up in countryside in Finland, stringing a long wire between two trees to be able to receive Radio Luxembourg... 🙂
 
Technically PDM is the correct general term, PWM is a special case of PDM.

Good to know as I like that and will use correct term from now on.

Ah, the days - I grew up in Helsinki, but spent my childhood summers up in countryside in Finland, stringing a long wire between two trees to be able to receive Radio Luxembourg... 🙂

OK, I get from that you are no spring chicken either. I believe about half of the population is Swedish speaking/background, right?

Julf sounds kind of Swedish?

Sibelius' native language was Swedish, or I have read, I have a photo of him in Copenhagen with Carl Nielsen - so Danes understand Swedish and vice versa. I don't know, maybe trivial but it's the kind of info I get into. 😀

Cheers, Joe
 
OK, I get from that you are no spring chicken either.

Just a man in his prime 🙂

I believe about half of the population is Swedish speaking/background, right?
Only about 5% is officially Swedish-speaking (remnants of first the Vikings - some of the dialects of the smaller villages on the west coast still resemble Old Norse - and then landowners from the time of the Great Swedish Empire in the 17th and 18th century) but Swedish is still the second official language of Finland.

Julf sounds kind of Swedish?
Well, originally it started from a pun involving both Finnish and Swedish - my first name on my passport is still "Johan" - how more Swedish can you get? 🙂

Sibelius' native language was Swedish, or I have read, I have a photo of him in Copenhagen with Carl Nielsen - so Danes understand Swedish and vice versa.
Turns out Finnish Swedish is the one language all the Scandinavians understand - as they have adopted the total lack of intonation from the Finns. I have no problem with Stockholm Swedish, but some of the Southern Swedish dialects are a different story - and while I have no problem with written Danish, spoken Danish is a different story.

Cheers,

Julf
 
I think we are dealing with OOB effects. And I think what we may be hearing is an audio signal that has not been totally extracted as an audio signal yet and the horrible and almost inconceivable thought that we need a filter that is an incursion on the sacred 'flat to 20KHz' rule - and that is why it has been overlooked?
Joe, on the "flat to..." part.... I was thinking in a different way. I don't know much about the mechanism by which the human ear "hears" depth in sound. But, could it be possible that the - 1.3 db @ 20 Khz has some relation with that mechanism?

As mentioned before, there are signals that higher frequencies than 20Khz also influence our perception of sound and direction.
 
Joe Rasmussen;3803750I think a lot of us have had that experience. Mind you said:
My experiences with the BB PCM63: The default DAC & I/V itself will ring or overshoot. A filter R/C was used over the feedback resistor. I set the values of the R/C filter as such, that the bit change ringing where completely removed using a scope and a 997Hz signal where more or less all bits of the DAC where altered/applied. Then I connected the same network to the + of the Op Amp to drive the I/V Op Amp symmetric.


After this I applied a Bessel alike analog filter to have a very low 20Khz roll-off... this was my old DAC DIY...

I mean, the old and new DAC's in terms of allowed output load and bit I/V circuit & overshoots are may different, but should be checked how, how they deal with the bit change ringing...

Just my 2 cents

Hp
 
Not really. These days cheap USB-based analog front-end devices combined with fast PC processing gives us very affordable and capable measurement systems, doing everything that we used to need very expensive scopes and spectrum analyzers for.
Okay for behavior of dampning I guess we need some kind of step-response and different difficult wave structures in audio band frequencies to be visual compared between signal input and output for device under test, and under this test the -1,3dB at 20KHz is expected to show very good, relative to other values.
I am hobbyist diy and don't at present have knowledge of such setup, of course I have the hardware in form of desktops and different soundcards, Jult can you recommend a setup, with reasonable costs or freeware.

Joe Rasmussen and Jult, expect you understand below after reading your smalltalk.
Rødgrød med fløde.....😀
God jul og godt nytår.....😀
 
Last edited:
Okay for behavior of dampning I guess we need some kind of step-response and different difficult wave structures in audio band frequencies to be visual compared between signal input and output for device under test, and under this test the -1,3dB at 20KHz is expected to show very good, relative to other values.
I am hobbyist diy and don't at present have knowledge of such setup, of course I have the hardware in form of desktops and different soundcards, Jult can you recommend a setup, with reasonable costs or freeware.

You do want to use a measuring sample rate well over 40 kHz. Even a 192 kHz sample rate sound card isn't fast enough, so you would need one of the USB oscilloscope front ends (such as one of the Picoscopes). Once you have captured the data, any math/signals package (such as gnu octave) can be used for the actual data analysis.

Joe Rasmussen and Jult, expect you understand below.
Rødgrød med fløde.....😀
God jul og godt nytår.....😀
Thanks! Been quite a while since last time I had rødgrød. No way I will try to pronounce it - I understand the phrase was used just like the Dutch used "Scheveningen" - as a shibboleth to detect non-native spies... 🙂
 
You do want to use a measuring sample rate well over 40 kHz. Even a 192 kHz sample rate sound card isn't fast enough, so you would need one of the USB oscilloscope front ends (such as one of the Picoscopes). Once you have captured the data, any math/signals package (such as gnu octave) can be used for the actual data analysis.

Thanks would be future then, will remember tip 🙂.

Thanks! Been quite a while since last time I had rødgrød. No way I will try to pronounce it - I understand the phrase was used just like the Dutch used "Scheveningen" - as a shibboleth to detect non-native spies... 🙂

Yes pronounce it is kind evil for non danes, therefor good anti spy tool 😀.
 
Joe, on the "flat to..." part.... I was thinking in a different way. I don't know much about the mechanism by which the human ear "hears" depth in sound. But, could it be possible that the - 1.3 db @ 20 Khz has some relation with that mechanism?

As mentioned before, there are signals that higher frequencies than 20Khz also influence our perception of sound and direction.

Please don't misunderstand me - I never said that being down -1.3dB was a good thing, I even called it a 'penalty' and a 'marker' - perhaps some are not familiar the last term, but 'penalty' should be obvious.

The question I am raising is, at least in part, that d-s DACs with PDM output have a peculiar limitation that intrudes in a negative way on an accepted rule, that we expect flat to 20KHz to be the rule or convention. Then, as they say "Houston, we have a problem".

I want a large bandwidth, but if a larger bandwith opens the door to other gremlins, then we have to take action. It is not a good compromise, but the only one that may work. Then what so you do?

But we need people to actually try this, to overcome skepticism, and then it will become real for them as well, and that will drive the conversation forward and lead to things I may not even yet thought about. But I do know that the effect is very real and to many instantly obvious. It will take it beyond the pure hypothetical.

Again, as Niels Bohr once said:

"How wonderful that we have met with a paradox. Now we have some hope of making progress."

We need to get a few more soldering irons heating up and that might cool the conversation.

Cheers, Joe

PS: By 'marker' I mean something along the lines of an artifact - not one that is necessarily acceptable in a conventional sense, but cannot be ignored.
 
Last edited:
Thanks! Been quite a while since last time I had rødgrød. No way I will try to pronounce it...

The full expression is "rødgrød med fløde" and you have to learn to say it as a toddler or you will never get it right. All four "d"s are soft and that is as difficult as the "ø".

Yes, I am able to say it - not exactly proud of it as it sounds so stupid to non-Danes. 😀

Cheers, Joe R.

PS: Another thing that is incredibly hard is to say Bjørn - and that is my misfortune. No prices to guess why.
 
Can somebody write down a detailed measurement procedure? How to be done in fact? What we will looking for? What signal in, what to be seen (looking for) on outputs?

There is a possibility to be used for measurements a sound card so as it is, and we will have some results, and is possible to be used an Sabre DAC, in a player some of us already have, and then we will have some other results... We may have so some facts to discuss on.

A sound card it may be limited in speed. This is true. A Sabre DAC system is quite fast. I have run through it 80Khz without any problems.

Only speculate about some or another in theoretical spheres, or simulate it on computer will not help to much to go further in all this. We need some facts, because the positive impressions about this effect it were already expressed.

From a practical point of view, I can see two actions to be taken so far: they who own a DAC system, no matter what kind (but a S-D one) it may solder the caps in place, and register its own impressions. The ones who have both the device (with the caps on), and some measurements tools, can proceed also to some measurements. In such case one may know precisely what to be done (measurement procedure).
And then one may come here to show its results, snap shots, etc.

BTW, I just wonder if Joe, who have done already some analyses and own some measurement tools, is willing to share some of his scope results, pictures, or so...
 
BTW, I just wonder if Joe, who have done already some analyses and own some measurement tools, is willing to share some of his scope results, pictures, or so...

Hi Coris. My main (and expensive) analyser is not really intended for this kind of work. I use it for speaker measurements and preparing data for modeling Crossovers in SoundEasy, and amplifier measurements, RIAA measurements and so on in that vein. I don't had a lot of experience measuring HF and VHF stuff - simply lack of opportunity and experience, but not lack of desire. But it comes down to my nature of work - but as I am increasingly working with digital playback systems, I do see a need to get into that - and I am not above saying that in this particular area I am a newbie, whereas in loudspeaker design I am actually regarded by a number of people as an authority.

My Cliofw analyser can be set to 24/192 and I have files and disks with sine waves at 0dBFS and -20dBFS (1KHz and 20KHz - but can get others) and I can certainly connect up the output via the output of my Oppo 105 here, with my own zero feedback post-DAC circuit and add and take away the filtering and see what it does up to 80KHz. The Clio has 90KHz bandwidth when set to max resolution 24/192.

I don't have any sound cards and not well versed in them - but I do see increasingly less expensive equipment available, can anybody make some comments about what is available - and cost, pros and cons. I think that could be shared and would be appreciated by others.

BTW, the Clio have heaps of stimuli it can generate - but not that easy to insert into a digital stream as its output is analog - unless burnt to disk I suppose from files - but I should be able to generate many different frequencies and even combination of frequencies, but burning to disk I would have to rely on a soundcard and ADC - better if somebody had files that they could make available here or post them on a server - or I could even store it on my website and supply URL to download them.

Can anybody make such files available - or is there a better way?

Also, my work on digital equipment and making them all sound better, has all been about finding analog weaknesses and analog solutions. Clocking is analog, power supplies are linear/analog (yes, even SMPS are analog), slew rate problems are analog and so on. And... post-DAC circuits and post-DAC filtering is... analog.

So if anybody has suggestions how I should use/set up my Clio, then I am very interested. I can set up a scale up to 90KHz, but what stimuli do I feed it?

If there is indeed something significant (despite the doubters - which is to be expected), then this is a side-issue to me. I will simply continue to use the filter no matter what the outcome is here, because I have many clients and everybody is going Wow! when they are hearing it. No problem. Right now, I am working on something much bigger than this and the next few months am employing mathematicians that can figure out how we can drive a 2nd Order Butterworth box alignment from any source impedance, drive any order filter, whether high or low pass, from any source Z, whether mechanical or electrical filters or any resonant filter, and that is a far bigger thing than anything mentioned here. If you think this d-s/d-d filtering is far fetched, then that is nothing compared to what a small band of us are working on here, as jitter is a huge problem in dynamic speaker systems and there is a key here to suppress jitter in loudspeakers. This is a much bigger deal to me than this is - I will just continue to use the filter - that has been done, so I can go past that and just keep going with it. I just threw it out there for others to try. And Coris, Ken are at least enjoying it too.

Cheers, Joe
 
Last edited:
Status
Not open for further replies.