No response. I am done with responding to this kind of thing.
But you just confirmed exactly what I meant, can't you see that?
.
For all others coming at me with this kind of stuff - sorry, but from now on I will only respond to positive posts. I am quite open to any discussions, but not in a bad spirit. Fair enough? We can ALL do better than that. We have YouTube for the other stuff.
.
.
Indeed, we just established that together. You meant "let's use words out of context to make things unclear". And indeed, when asked to clarify your terms, you refused.
The first step towards any discussion is to establish a common understanding of the words used.
Well, it means the same it meant earlier, back to posts #2 and #10 of this thread.
You say that having a first order LPF creating a 1.3dB drop at 20KHz is audible and subjectively better after a DAC. No need for creative terminology here.
You say that having a first order LPF creating a 1.3dB drop at 20KHz is audible and subjectively better after a DAC. No need for creative terminology here.
Then let me try. Really trying not to be negative in any way.
We established early on in this thread that the "Rasmussen Effect" is basically a first order LPF with a 1.3 dB drop at 20 kHz. It appears that some people find that that drop-off creates a pleasing effect.
Would it perhaps be more helpful for us to talk about it in those terms, instead of terms like pre- and post-emphasis that some people find possibly misleading?
Let's me phrase it in another way. This LPF at -1.3db @20KHz is the main thing of this thread.
On top of this issue, there are (apparently successful) attempts to correct that 1.3db drop by using digital EQ, as described in post #461. It's a bit secondary but seems to point out that the "effect" isn't related simply to frequency response (which would be surprising at 20Khz anyway).
edit: crossposted with Julf, agreed.
Good. But the answer is a bit yes and a bit no. Things have progressed somewhat beyond that. If I may explain, there are several issues and they need a fair airing as there is no exact answer, or at least no consensus (and we are talking about among those who are taking the topic seriously), whether the correction should be applied or not.
This is not easy to put concisely, but the question is, what frequency slope (or how far down @ 20KHz) do we extract the best sound out of a delta-sigma modulator and if that is found to be severely rounded, then most certainly it would be advantageous to put in a correction. But that correction may or may not be flat again @ 20KHz. How the correction is to be applied is a logistics matter, it can be done via digital pre-emphasis before the d-s stage, or it can be done in the post d-s stage, at which point we are talking about an analog correction on the analog side. I gave one such simple example in post #461.
If no correction is made in a USB DAC, then Media Centers like JRiver can use its 64 bit parametric equalizer to correct it. Turn off 'clipping' - put in 40KHz as frequency, then the Gain in dB +4dB will equal actual gain of +2dB (if that was the choice taken) @ 20KHz.
What is clear is this, we can see that we have a powerful tool to set the response to where we believe, for whatever reason, where the response of the d-s ought to be.
So... Nobody is demanding it to be down by any amount dictatorially or arbitrarily.
Right now I favour -2dB @ 20KHz uncorrected, as it barely sounds rolled off at all and using listeners who were unknowingly used (as blind as it gets) have provided reasonable confirmation of this.
Please, for those who read the above, please try not to be picky about any single phrase or sentence, it is hard enough to be concise and clear enough to be understood.
Of course, the above means nothing unless one takes the attitude that there may be something to all this, to take it seriously, or as very skeptical Ken Newton said earlier, in the very end he just got curious and had to find out for himself. I give him credit for that, and I feel I have a rightful expectation that his testimony and others, should lead others to do the same. And some have, some who are not visible on this thread.
More are coming along with that view.
I will make an admission here: What is the cause by which this "effect" takes place, is not entirely understood. But I spoke to a guy who has worked on advisory boards to Texas Instruments (makers of Burr-Brown) and he said he suspected that some kind of damping was taking place, which he said without prompting from me and in an earlier post I expressed a similar thought, it has that tune-able character.
There is also something else, and this is truly curious and seems to support a 'damping' theory: Take a "voltage" DAC and fit a 1:1 transformer to sum the two phases. This works very well and I have done this many times, and 2V RMS can be had a lot of the time, so full output. But transformers need a Zobel network in most cases, to even to make the flat to 20KHz and beyond, in fact they will have a pronounced rise.
So we design a proper Zobel and make it flat - and it sounds OK. Then we can 'overcook' the Zobel, generally by increasing the capacitance by 50-100% and now trim the R value and give us -2dB @ 20KHz. And we hear the "effect" very clearly. I have done this in blind tests and the listeners heard it very clearly and very positively. Now why is this important? Because the Zobel will have virtually no effect at extreme HF - the "effect" is related to frequencies much lower than that. The Zobel is actually not a low pass filter - think about it. It seems to exclude EHF as the source.
But there is unfinished business here - and I have only given you some of the highlights of what has been going on lately.
Keep any comments positive, please.
Cheers, Joe
.
Please read the data sheet carefully. An op-amp can not behave as an integrator or first order low pass with only a single feedback capacitor due to finite gain bandwidth. Eventually open-loop gain becomes 0db and you get a zero in the response which not longer filters high frequency noise. That circuit forms a critically damped two pole response and has a large C across output to ground even when the op-amp runs out of gas. This circuit is not designed as a simple lowpass of <20kHz BW.
It worked very well and I proved it with spectrum analyzer plots but received brick bats from the "not invented here" crowd.
EDIT - When this was done there were no DS DAC's 🙂
Last edited:
I appreciate the time you took to answer and restate your position in clear terms. My current unease with this whole thread is how we have a single "effect" in extremely different setups. I'm no engineer but as part of my studies in both social sciences and philosophy, I've had my fair share of methodology classes. When it comes to research, a good principle is to always try to isolate the problem, to reduce as much as possible the possible causes at work and their interactions. Anything looking like a blanket solution makes me skeptical.
I'm no fan of the damping explanation because how can it apply to both current out DACs and voltage out DACs ? Their analog output stages are very different.
It would be also nice to clarify where does this effect happen or not:
- does it happen if we use digital EQ to emulate it ?
- does it happen if we add a further stage with a simple LPF opamp filter?
- or does it happen only if we use analog (R)C filtering right at the output pins of the DAC?
As it is, I'm under the impression that you're saying that there is an optimal analog filter for DACs, a filter which should be more aggressive than usually advised. If that's so, it would be better investigated by using a voltage out DAC with single ended outputs, followed by a as simple as possible parametric LPF made of opamps in cascade.
I'm no fan of the damping explanation because how can it apply to both current out DACs and voltage out DACs ? Their analog output stages are very different.
It would be also nice to clarify where does this effect happen or not:
- does it happen if we use digital EQ to emulate it ?
- does it happen if we add a further stage with a simple LPF opamp filter?
- or does it happen only if we use analog (R)C filtering right at the output pins of the DAC?
As it is, I'm under the impression that you're saying that there is an optimal analog filter for DACs, a filter which should be more aggressive than usually advised. If that's so, it would be better investigated by using a voltage out DAC with single ended outputs, followed by a as simple as possible parametric LPF made of opamps in cascade.
You are welcome.
Welcome healthy skepticism. But a blanket solution? That's not an issue here.
No.
No.
YES.
That is what we are trying to do. One instance required me to travel nearly two thousand kilometers round trip. I am limited in what I can say about that now as the ramifications are still unfolding from that trip. But a DAC that omits the multi-low-bit part of the delta-sigma modulator, indicating what? Leaving the d-s modulator as the part of the chain to be affected. I have pointed this out earlier but it has not been absorbed by anyone here yet. This is a commercial product, if you have figured it out which, then please don't name here as it would not be helpful, you can PM me, but the clue is a strong one; the importer was there.
Way ahead of you.
Please don't be hasty to assume, you only have to ask. Much more is being done behind the scenes, as I have said before.
Cheers, Joe
.
Hi, Joe, I've just checked back on the goings-on of this thread and have a couple questions for you.
1) Referencing your quote above, are you saying that you've found that the Rasmussen Effect can be activated at the DAC pins while also EQ'ing the resulting frequency response droop further down the analog processing chain, thereby giving a net flat system amplitude response upon final output?
2) Have you had opportunity to observe whether the Effect manifests when decoding high sample rate content while still employing a 1dB to 2 dB 20KHz attenuation set point? I'm wondering whether the attenuation set point frequency has to be scaled proportionally with the sample rate, or whether it remains fixed.
Last edited:
I have a DIY dac with ESS 9018K2M the poor cousin of Sabre 32 and i use a clasic opamp datasheet I/V conversion. Is my circuit usable to obtain this Rasmussen effect ?
Indeed yes, and also the option of putting the EQ before the DAC.
In fact this is easy for many to try even now, if you use a USB DAC with JRiver Media Center. Apply the roll-off to the point you desire @ 20KHz and apply the data from post #461 into the 64 bit parametric EQ. This is in effect a digital pre-emphasis.
This points the way to other techniques that can be used, which I also suggested, if the correction is before the d-s DAC, then it is in the digital domain, if the correction is after the DAC, then the correction is analog - and this is easy to do in most commercial designs with opamp I/V - again sample as per post #461 repeated here:
But why would you want to do it this way? In my case, not so.
But this fact has commercial value and acceptability factor.
Of course, this is all about making the topic more palatable, for those who balk at the fact our response is no longer flat at 20KHz. In fact, I don't mind it being down, now as much as -2dB and my listeners here have not mentioned that it sounds rolled off, in fact far from it.
So Ken, this is a tactic, to point out you can still get the benefits and not necessarily be worried about frequency response or not being flat.
Bottom line: The benefit is not the result of frequency response.
I can only imagine that it is fixed, and independent from those factors. Even with pure DSD with a pure DSD processing engine. Everything done so far points to that and must be a key to do what we all want to, figure out what exactly is the mechanism we are dealing with.
-----
Let's think a bit more exotic:
Take frequency response out of the way - a thought experiment if you like. What if the d-s DAC sound better with an F3 point so low that it would be totally unacceptable? Does that mean we would never be able to listen to the DAC at the point at which is may sound best? Do you follow?
Take RIAA. Lots of pros and cons, but it is probably not far off from being a good balance. We can have a similar viewpoint here. Find the best operating point of a d-s DAC with respect to its HF response and yet still have it acceptable on the final outputs, as flat as we need it to be.
I said earlier, this is a powerful tool because the trade-off between where it sounds best can be explored by those who who wish to do so.
In that case we may find differences with different DACs as it should be able to fine-tune each one. Then some differences may arise - or simply the designer will decide what he likes best.
This remains to be seen.
Can I also point out, this is also a lot about the potential for this to be used in commercial products in the not so far future. One USB DAC manufacturer is already doing, yes, right now. This is a unit selling for $750, will be -2dB @ 20KHz and no correction. That means being brave and other manufacturers are not going to be brave enough to bring out a DAC "that measures wrong".
So that is the issue here: For those who want to manufacturer this into their product, there seems to be no need for fear of being accused "it measures wrong" as a hindrance.
Awaiting your thoughts, they are appreciated.
Cheers, Joe
.
Last edited:
Yes indeed.
This will work with Sabre DAC too, just three components per channel. Note that this is a current DAC with no DC offset and hence "+" is grounded., so it is referenced to DC ground. Your DAC is likely +1.65V, but the above still works even when you have +1.65V on "+" pins.
Another complimentary thing to do, get some 0.1F supercaps, make sure they are rated +5.5V - and put these are the two (?) power supply rails to the DAC. This cleans up the d-s modulator clipping to rail a lot and works well in tandem.
BTW, the Rasmussen Effect is not the solution perse', rather that which is being corrected.
Cheers, Joe
.
Last edited:
Pre-emphasis and de-emphasis was part of digital in the beginning. A number of early recordings had it and some DAC's had a passive low pass filter to correct for it. The first gen Crystal DAC for example worked that way.
The hazard of re-emphasis is overloading the digital and that can happen with either Analog before the input or a digital filter in the chain. Some DAC chips handle that overload really badly. It also can happen without the digital pre-emphasis as part of the internal machinations in the DAC. Most DAC's have some form of low pass filter. They are essential for correct operation (part of the reconstruction filter) and are usually set way beyond the audio band.
One challenge of evaluating a system like this is separating the the desired effect from the known effects of HF rolloff. As noted sometimes the HF rolloff sounds more like an extended response than the actual flat response. An interesting way to attemt to untagle this would be to swap the pre-emphasis and de-emphasis. Add the de-emphasis to the digital side and use an active pre-emphasis on the analog side. Adjust to get the same flat target response and see if the unpleasantness is increased significantly. If its an intrinsic effect of Delta Sigma it should get more pronounced.
The hazard of re-emphasis is overloading the digital and that can happen with either Analog before the input or a digital filter in the chain. Some DAC chips handle that overload really badly. It also can happen without the digital pre-emphasis as part of the internal machinations in the DAC. Most DAC's have some form of low pass filter. They are essential for correct operation (part of the reconstruction filter) and are usually set way beyond the audio band.
One challenge of evaluating a system like this is separating the the desired effect from the known effects of HF rolloff. As noted sometimes the HF rolloff sounds more like an extended response than the actual flat response. An interesting way to attemt to untagle this would be to swap the pre-emphasis and de-emphasis. Add the de-emphasis to the digital side and use an active pre-emphasis on the analog side. Adjust to get the same flat target response and see if the unpleasantness is increased significantly. If its an intrinsic effect of Delta Sigma it should get more pronounced.
May I also quickly add: The existing situation of bringing the DAC's output down to or close to -2dB suits me just fine and I don't need the correction as I am of the view that it works well without it. But the option discussed is about making this topic about the Rasmussen Effect more tolerable, more approachable, and commercially too, a way of removing a potential obstacle that would make them hesitate to use this.
There are of course secondary questions and those I have raised above. I am also simply trying to encourage further exploration and experimentation. Anyone of you are invited to do just that and I hope atupi does apply post #496 to his existing DAC and reports back.
Cheers, Joe
.
There are of course secondary questions and those I have raised above. I am also simply trying to encourage further exploration and experimentation. Anyone of you are invited to do just that and I hope atupi does apply post #496 to his existing DAC and reports back.
Cheers, Joe
.
Well, well, quite much happened again here. It is quite long time since I did not clicked on "last Page" of this thread.. I was thinking it is dying...
For my part I may say that I became more sceptical in the last time about this filtering method, but who knows, maybe I may try a little bit more...
Coming back soon with some of my previous done experiments/results...
For my part I may say that I became more sceptical in the last time about this filtering method, but who knows, maybe I may try a little bit more...
Coming back soon with some of my previous done experiments/results...
- Status
- Not open for further replies.