Wow! So much to respond to, and I appreciate the spirit that you are seeking and the civility of tone.
That configuration you describe sound very similar to the one I suggested on the Transformers/Buffalo thread, summing the two phase(s) of the Sabre DACs and enabling a subsequent single-ended tube circuit.
Re snubber, I think you are referring basically to a Zobel? I have done this to a number of voltage DACs, Cirrus Logic, Wolfson, AKM etc, and it works very well - then add a Zobel on the secondary, typically overcooking the value of the cap (say 50% higher) and then addjust the resistance to give the required slope. The output impedance of the voltage DAC and the parameters of the 1:1 transformer, like the DCR on the Primaries and the Secondaries are ALL effectively in series with the output Z of the voltage DAC, typically 20 Ohm per phase. With the transformer I use I end up typically with 15n and 1K2 to 1K5 and I aim to be lower that -1.3dB down. This was with straight voltage DACs and the ES9018's high output Z straight into 1:1 txs is not ideal, unless you parallel a LOT of phases together, like all 8 - some have done that, but you will need to use 2 Sabre DACs in mono L and R.
Re whether we can hear the roll-off by applying (or maybe "correcting" for the Rasmussen Effect), this is becoming a rather mute point now. Since the world is now transferring to listening to files and feeding their music into USB DACs where the bit-rate and frequency is set by the Music Server like JRiver, then it is a simple matter to correct for the non-flat 20KHz response and maybe even aim at -1.9dB to -2.0dB and correct for it.
In JRiver, you have a 64bit Parametric Equalizer, open it. Enter the following:
Select: High-shelf
Frequency: 40000
Bandwidth (Q): 0.7
Gain: 4dB
In the above, if you have set it to -1.3dB @ 20KHz, then enter:
Gain: 2.6dB
ALL ABOVE = FLAT RESPONSE
This way you can set the @ 20KHz response anywhere you want it.
What if we are not using a Music Server and spinning disks and not using USB DAC? We may still have a solution.
Let's face it, the most common way is that most market units use opamps and Virtual Earth or voltage DACs into oppamps as well.
Well, here too we can deliver solutions to give us back out flat @ 20KHz response again, that can restore it.
At this point I want to use the description that the post-DAC filter is also a De-Emphasis Filter, and the following will explain why. In the above USB DAC, we are really talking about a De-Emphasis Filter and then subsequent De-Emphasis EQ, dialed into the Parametric Equalizer of JRiver.
We can do the same with opamps (and also ways found for discrete non-opamp circuits too), and below I shall furnish such an example.
Now with opamp Virtual Earth, the output of the I/V is buffered and low Z, the second stage usually employs subsequent HF filtering and the final response is almost flat @ 20KHz and typically -3dB @ 70-80KHz.
Let's look at the recommended post-DAC ciruit of PCM1794A:
I can tell you that two 3R resistors and 1uF is inserted as the De-Emphasis Filter and that the De-Emphasis EQ is achived by adding two small 12nF caps.
Now take a look at the below responses generated from the PCM1794A sample above:
I think the above is self-explanatory.
BTW, it can also be achieved with voltage DACs, but it is only slightly more tricky to do, but I had no problem coming up with a circuit that did exactly the same.
The point is, with these techniques we can set the 20KHz response where we want it and if we want to correct for it.
BTW, a website is being construction re the Rasmussen Effect and the above is but an example. It will be singular focused on just the one topic and especially a bent towards promoting manufacturers to take it up.
Also, one manufacturer is gearing up a production run of a USB DAC and he is but the first one.
Also, a technician I now work with has been on advisory boards to Texas Instruments (who make among other things Burr-Brown DACs and semiconductors) has commented that the Rasmussen Effect is in his mind about "damping" something related to delta-sigma modulation and subsequent architecture that is common in all delta-sigma modulated DACs, as the filter has the same effect no matter who the manufacturer is. Note some way back when I also, on this thread, suggested that some kind of 'damping' was going on, as making the filter more agressive have similar qualities in other audio where playing around with damping (Moving Coil cartridges comes to mind, lowering or changing the load resistance can improve almost any moving/magnetic coil product).
May I recommend that you go a bit more aggressive than just -1.3dB with your 'snubber' and even as high as -2dB and then report back? You are playing around with damping at the expense of HF loss, but one you may have the option of adding De-Emphasis EQ further down the chain - and may possibly prefer not to? Whatever you like.
Finally, there is something else happening, related to the Rasmussen Effect, that will be later covered on the website, that there are two techniques that comes together. More about that, as it too will be covered on the new website. Delta-sigma modulation is really a 'clipping' engine that introduces another problem directly converts into jitter at the point the digital content becomes analog, that on/off state where the 'on' state is somethings akin to a clipped rail. Something needs to be done about this but rarely is because it is not understood or not aware.
Cheers, Joe
Last edited:
over a decade old news - actual delta sigma ADC/DAC designers, sophisticated engineering users knew even earlier than the audio community that single bit DSD had a fundamental problem
http://sjeng.org/ftp/SACD.pdf
most flagship audio DACs have been multi-bit delta sigma for a while now - hugely improving stability and increasing the applicability of linearized modeling in modulator design
http://sjeng.org/ftp/SACD.pdf
most flagship audio DACs have been multi-bit delta sigma for a while now - hugely improving stability and increasing the applicability of linearized modeling in modulator design
May I ask, have you read earlier parts of this thread?
The Lipshitz/Vanderkooy paper is something we already know about here. Like you say 'old news'.
But that isn't quite the topic here.
Pure DSD is used by PS Audio in their DirectStream DAC designed by Ted Smith and Ted did not use the added low-multibit processing that modern chip DACs do. John Atkinson @ Stereophile measured the pure DSD DirectStream and based the resolution measurement on THD+N (according to standard thinking), then the DirectStream DAC only had 17 bit resolution. This was predictable. But what does it mean? The DirectStream still sounds very good and does not sound low bit. So go figure.
The issue here is quite different, it's about the sensitivity of delta-sigma DACs to the load imposed on it externally via a passive filter, has a definite influence on the characteristic nature of the delta-sigma DAC's sound (consistently across different manufacturers) and is something that can be applied and then corrected in terms of amplitude response, as an option. You are quite free to enter that discussion, because that is definitely not 'old news.'
Cheers, Joe
Interesting...is it a coincidence that the filter response you show is about dead-on to the original "Rasmussen Effect" recommendation (1.3dB down at 20kHz)?
Yes Joe, I meant a Zobel, commonly called a snubber when used on the output of a transformer.
What was interesting was that I could apply a single zobel, 3nF and 1k to the secondary, and dial in whatever rolloff I wanted by placing a 5k pot on the primary. 3k3 turned out to be the spot where I reached -1.3dB at 20kHz. Since I'm using a Lundahl xformer, I can conveniently put the pot between the series primary windings. The LL1676, with its cobalt core, has a large amount of primary inductance (>200H) and the bass response is very punchy indeed. I think it sounds excellent on the output of the Buffalo IIIse wired 2:1 with primaries and secondaries each in series. The stepdown is used because the gain of the 26 tube is 8X, and I like to keep the volume control at least halfway up.
Interesting that you seem to be commercializing the "effect". I look forward to following the evolution of this.
What was interesting was that I could apply a single zobel, 3nF and 1k to the secondary, and dial in whatever rolloff I wanted by placing a 5k pot on the primary. 3k3 turned out to be the spot where I reached -1.3dB at 20kHz. Since I'm using a Lundahl xformer, I can conveniently put the pot between the series primary windings. The LL1676, with its cobalt core, has a large amount of primary inductance (>200H) and the bass response is very punchy indeed. I think it sounds excellent on the output of the Buffalo IIIse wired 2:1 with primaries and secondaries each in series. The stepdown is used because the gain of the 26 tube is 8X, and I like to keep the volume control at least halfway up.
Interesting that you seem to be commercializing the "effect". I look forward to following the evolution of this.
PS. Here are a couple pics of the DAC/Preamp. They are, in order, the completed unit, the DAC/preamp chassis open for testing, and the power supply chassis.
Output tubes are late '20s vintage balloon 26 DHTs. The tubes in front are gas voltage regulator tubes (0D3 and 0A3, one pair per channel to supply 225V to the 26 plate). All PS are choke input with a 5Y3G tube rectifier for the 26 and Schottky diodes for the others. Tube mounting uses vibration-absorbing bushings, mass and dissimilar materials such as wood, aluminum and sorbothane to tame the 26's noted microphonics.
Kind of neat to combine a very modern DAC (with the hot news "Rasmussen Effect") with 1920's-era tube technology. They go very well together!
Output tubes are late '20s vintage balloon 26 DHTs. The tubes in front are gas voltage regulator tubes (0D3 and 0A3, one pair per channel to supply 225V to the 26 plate). All PS are choke input with a 5Y3G tube rectifier for the 26 and Schottky diodes for the others. Tube mounting uses vibration-absorbing bushings, mass and dissimilar materials such as wood, aluminum and sorbothane to tame the 26's noted microphonics.
Kind of neat to combine a very modern DAC (with the hot news "Rasmussen Effect") with 1920's-era tube technology. They go very well together!
Attachments
Last edited:
PLEASE: Attention to the Moderator (perhaps Dave?). Now that the earlier heat has been drawn out of the earlier discussion and indeed the "Rasmussen Effect" as named by Ken Newton and seconded by Coris (the starter of the thread) is now being taken increasingly serious by some serious people (who they are will come to light in time), can I appeal to the Moderator for the original title of this thread to be restored.
It was called DAC Filtering - the "Rasmussen Effect".
There are now several manufactures who will be bringing out USB DACs with the "Rasmussen Effect" incorporated.
Different ways of implementations has also been conceived, where the possible option of creating a "Pre-Emphasis" in the digital domain which has an exciting possibility: The response of the output of a Delta-Sigma DAC can be set at any point and the digital "Pre-Emphasis" simply mirrors the response and hence still kept flat to the degree desired (one commercial USB DAC has the facility already, just writing the code that can be downloaded). That way the optimum response of the Delta-Sigma DAC can be found, whatever and wherever it might be. This is very exciting. We may not have found it because we have wanted to limit dulling the top end, but this need not be the case. Wherever it may be, it can be found without any complaint leveled against the frequency response.
Also, on this weekend an experiment were made with a high-end USB DAC that omits the multi-element matching after the Delta-Sigma Modulator, and implementing the "Rasmussen Effect" was instantly obvious in the very way it was expected to be. A near two thousand kilometer round trip was made to do this trial. Some may realise why that is so important (Ken?). The proof is the omission and elimination. Think about it.
The number of people are growing who are not only hearing this, but also enthusiastically supporting, it is not going away. Pandora's Box has been opened wide enough and it will not be shut.
Best Regards to ALL, and many heartfelt thanks to ALL of you who have supported me, here on this threads they are of course Coris and Ken, but the others know who they are.
Cheers, Joe
.
It was called DAC Filtering - the "Rasmussen Effect".
There are now several manufactures who will be bringing out USB DACs with the "Rasmussen Effect" incorporated.
Different ways of implementations has also been conceived, where the possible option of creating a "Pre-Emphasis" in the digital domain which has an exciting possibility: The response of the output of a Delta-Sigma DAC can be set at any point and the digital "Pre-Emphasis" simply mirrors the response and hence still kept flat to the degree desired (one commercial USB DAC has the facility already, just writing the code that can be downloaded). That way the optimum response of the Delta-Sigma DAC can be found, whatever and wherever it might be. This is very exciting. We may not have found it because we have wanted to limit dulling the top end, but this need not be the case. Wherever it may be, it can be found without any complaint leveled against the frequency response.
Also, on this weekend an experiment were made with a high-end USB DAC that omits the multi-element matching after the Delta-Sigma Modulator, and implementing the "Rasmussen Effect" was instantly obvious in the very way it was expected to be. A near two thousand kilometer round trip was made to do this trial. Some may realise why that is so important (Ken?). The proof is the omission and elimination. Think about it.
The number of people are growing who are not only hearing this, but also enthusiastically supporting, it is not going away. Pandora's Box has been opened wide enough and it will not be shut.
Best Regards to ALL, and many heartfelt thanks to ALL of you who have supported me, here on this threads they are of course Coris and Ken, but the others know who they are.
Cheers, Joe
.
Last edited:
I think the usual criteria for naming an effect after the discoverer requires at least a rigorous description of the effect (along with empirical evidence) in an academic, peer-reviewed journal.
Can we not revisit this, please? I happen to know a bit about peer-reviewing and the assertion is not correct. I happen to have close friends for many decades who are scientists.
.
To me, it seems you are the one keen to revisit this.
That is nice for you. I just work with them.
For humorous contemplation:
The Streisand Effect is the phenomenon whereby an attempt to hide, remove, or censor a piece of information has the unintended consequence of publicizing the information more widely, usually facilitated by the Internet.
It is named after American entertainer Barbra Streisand, whose 2003 attempt to suppress photographs of her residence in Malibu, California inadvertently drew further public attention to it. Similar attempts have been made, for example, in cease-and-desist letters to suppress numbers, files, and websites. Instead of being suppressed, the information receives extensive publicity and media extensions such as videos and spoof songs, often being widely mirrored across the Internet or distributed on file-sharing networks.
Smile, OK?
.
The Streisand Effect is the phenomenon whereby an attempt to hide, remove, or censor a piece of information has the unintended consequence of publicizing the information more widely, usually facilitated by the Internet.
It is named after American entertainer Barbra Streisand, whose 2003 attempt to suppress photographs of her residence in Malibu, California inadvertently drew further public attention to it. Similar attempts have been made, for example, in cease-and-desist letters to suppress numbers, files, and websites. Instead of being suppressed, the information receives extensive publicity and media extensions such as videos and spoof songs, often being widely mirrored across the Internet or distributed on file-sharing networks.
Smile, OK?
.
More details on this omitted multi-element matching please Joe as this sounds to my ears rather confusing. In general multi-bit S-D modulators rely on not-so-well-matched elements (at least in comparison with prior art in terms of multibit DACs) and attempt to overcome the mis-match by turning errors into noise. So are you saying the circuits which implement the randomization of the elements were turned off in this particular high-end DAC?
Only by name and reputation, but then he lives in Hichtum, almost 75 miles away, on the other side of the Great Closure Dike, in Friesland, where they even speak their own language (West Frisian, a close relative to Old English). Amsterdammers go there very rarely 🙂
Here in Australia 75 miles is spitting distance. I forget how small some European countries are and how compact, you don't have to travel very far and people speak differently. Earlier today, I drove from Coffs Harbour to Sydney South-West, nearly 600 miles and still the same state/province. But I do remember watching a series of programs about the English language, and Frisian was indeed mentioned. Language has always intrigued me.
Cheers, Joe
.
Yes, Australians and even Americans have a different perspective on distance. The Netherlands is small, but densely populated, so I think of distance not in terms of miles, but in hours.
Then I am sure you will appreciate the fact that Frisian (not to be confused with Saterland Frisian or North Frisian) has eight officially recognized dialects. And just to confuse the outside world, Frisian is only referred to as "West Frisian" outside the Netherlands (to distinguish it from Saterland Frisian or North Frisian), but within the Netherlands "West Frisian" actually refers to a dialect of the Dutch language (not Frisian at all).
Let's not go into the Netherlands vs. North and South Holland issue - or that the highest mountain in The Netherlands is in the Caribbean.... 🙂
Oh, I forgot to mention, Menno's AES 2007 paper on "Low Level Audio Signal Transfer Through Transformers Conflicts With Permeability Behavior Inside Cores" makes use of experimental data I supplied on using high frequency bias in output transformers to boost Permeability that increases dramatically the low level resolution, especially of silicon steel cores (much cheaper than amorphous cores). Alas, no commercial manufacturer makes use of high frequency bias, even though there is no real technical difficulty. It worked with tape heads, why not output transformers? Several other scientists with whom I have had this discussion, totally agrees. But the 'commercial world' is not always sane. Missed opportunities abound.
Cheers, Joe
.
Cheers, Joe
.
Ah yes, the Danes (in Denmark 🙂) have a similar joke, I can't exactly remember it. Flat as a pan cake.
A bit beat and late here, done a lot of driving in the last 4-5 days, so let us resume hostilities or friendlinesses later.
Don't sell yourself short, I may yet be confident that the end you will come aboard. There are interesting things happening behind the scenes here, the US and Europe. One day you will be listening to music and you will have a smile on your face. 🙂
In fact, it is a fascinating process, in some cases it's become a bit like a chess match, in a psychoanalytical sort of way. For example, using the phrase "post-DAC Filter" does not go down well with a particular crowd or persons, but explaining that it is a "De-Emphasis Filter" can really change the attitude, because (especially in an analog sense) the idea of pre-emphasis and de-emphasis used in many form of 'transmission' through different formats, becomes to them a more acceptable discussion. We are all human and even the way we take on knowledge is as much to do as to how that knowledge comes to us as well as exactly what the knowledge is - and knowledge is what science is, it really is at its heart the basic idea, and of course the process to acquire it, which can mean a bumpy road.
Cheers, Joe
PS: It just struck me; definition of science? A bumpy road? 😀
Or as Niels Bohr said "An expert is a person who has made all the mistakes there is to be made in a narrow field" and "don't take anything I say as an assertion, but as a question." Worth reflecting upon.
Off to zzzz land.
.
This word playing game makes things a lot worse in my opinion. A de-emphasis filter is something clearly defined from the start, the corollary of a defined pre-emphasis filter. If we're using these words, we need to know where is this pre-emphasis filter and how is it specified.
I still haven't seen anything in this thread which can't be described as a poorly defined low pass filter.
- Status
- Not open for further replies.