Interesting effects and analyses. Every time soundstage is improved, it is usually a question of better transient recording and reproduction for me: so, attack times, slew rates, settling times and digital resolutions are what I would look for in simulations and testing.
Hi Barbara, I took my time and also had a number of things on my plate and I did not respond to quickly without considering your DAB paper.
So, feel free to correct me, this is about taking a differential output (I or V) and forming a Balun (making bal/differential into a single-ended unbalanced) and while doing so, you observed:
"The central point of good sound is the Cap (C1) in the middle of the bridge configuration. It is not intended to build a filter, but it may be understood as a kind of integrator. It will change its charge-value, each time the time-slice ratio of the switches behind each current output changes. It makes sound big deal more transparent and 'lively'. Music sounds more fluently and leaves more air between instruments and voices. Only good foil type caps..."
A few point:
1. The observation re sound, you wrote this before this thread or the earlier mention of the topic on the Oppo 105 thread/topic. And what you describe accurately is what I have heard, and so has Coris and Ken. I have known about if for a while, but you may have heard it even before me.
But it is powerful, one way or the other.
2. The word that jumps at me is "integrator" - and I understand what you say about time slices (differentially) - but help me here, what we see coming out of the DAC hasn't been completely reconstituted (reconstruction filter) and that can only be done post-DAC? Or at least completed? They are still 'slices of current'? I am an analogue guy, so I am quite willing to learn.
Understanding something is understanding how to look at it.
Some years ago, Sony made CD players, and something not commonly known was that the output of the Digital Filter was pure DSD (up sampled or whatever you might call it, some sort of algorithm at work?) even though this was only a CD player. Later they did also with their 1st generation SACD player, the SCD-1 and onward. When playing CDs through SCD-1, the output of the Digital Filter (DF) was 'upsampled' - but was SACDs, the output was the full resolution DSD signal, separate Left and Right (differential) and 2.8MHz.
The late Allen Wright, around 200, asked Ed Meitner if a DSD signal fed through an analog filter (bandpass hat obviously include LPF) and buffered as an output (the output would be 0.6V or -11dB) and a grumpy Ed said 'of course' and we went ahead and did that.
We were also told by a friend who at that time worked for Philips, that in no way could we take that 'digital' DSD signal out of the box and the solution we came up with had to be in box or else... very heavy stuff. Only a final analogue signal was allowed to come out via RCA or XLRs.
So that DSD signal, even in separate Left and Right, were still to be regarded regarded as 'digital' - but as we are talking about 2.83 Million DSD(density/width pulses representing analog - PWM or PDM) needed to become true analog. That is where the filter comes in, it was a reconstruction filter, an integrator, and beyond that point it was not regarded by the heavies at Philips and Sony as digital anymore.
That now brings us to the outputs of delta-sigma DACs. In light of above, although the "Sword of Damocles" is not hanging over us, in what light should we look at it. Like DSD is not yet fully reconstituted analogue and that filter (cap) is actually where it actually is reconstructed as such?
I find that fascinating - and does it help to explain what we, and that includes you, myself and two others on this thread, but also a dozen others who have not seen sight of here, they most certainly have heard it.
One thing for sure - a lot of other people should be playing with this - and here is an opportunity for the DIY guys to lead - and maybe one day we will see actual products for sale that will be using this - and we can then look back with some satisfaction.
Cheers, Joe
PS: Consider the second order reconstruction filter, on the output of Class D Amplifiers - got me thinking, tweaking that inductor/cap, so that it also intrudes at 20KHz and not flat. That ought to be tried - while listening to changes.
.
Last edited:
The thing with DSD is that its digital form is similar to what an analog signal does because the width density can equate to sound pressure. In addition, you can have DSD coming out of a line in its pure digital form, and a simple normal analogue filter will allow you to hear the sound properly.
Since I'm interested in both DSD and DACs, it would be great to know if the improvement using this capacitor or DAB arrangement also helps with some of the existing (or future) DSD-capable DACs. Also, the interplay if any with non-oversampling DACs and using R2R ladders rather than chip DACs.
Since I'm interested in both DSD and DACs, it would be great to know if the improvement using this capacitor or DAB arrangement also helps with some of the existing (or future) DSD-capable DACs. Also, the interplay if any with non-oversampling DACs and using R2R ladders rather than chip DACs.
To know the possible answers about what is in your interest, it could be nice to just try yourself to use this filter, or experiment with it... And if you want, then share here your conclusions... or findings...🙂
That's the point.
Would the ES9018 Sabre DAC not include in the 'existing' DACs you mention. Certainly it does. Also been tried with Burr-Brown PCM1794 current DACs (and similar in the PCM-DSD179x range) as well as Cirrus Logic voltage DAC. All DSD capable.
I don't think it will be a factor with those DACs, but if I am proven wrong about that, I shall not be displeased.
Cheers, Joe
.
I meant DAC as in a complete construction rather than the chip itself.
Does anyone know if using extension leads for easy capacitor-swapping is detrimental to the effect?
The only DAC I could possibly experiment with for now is a bit problematic: all SMD, uses an Asahi Kesei (which is good) with on-chip SCF, so I'm not sure to what extent I can hack it.
I think we haven`t discussed yet a particular side of this filtering approach:
It does this (cap between DAC chip differential output) filter be used in conjunction (or not) with the classical filtering between I/V converter and final opamp?
It may be fortunate or not to heave so many filters in a signal path?
I know the theoretical purists I may want so many and sophisticated filters as needed, to filter out all what it may not be in audio spectre, but is that the best way to be done? I really do not think so...
It seems to me that Joe have experimented this filtering method heaving the another filters in his existent device/configuration. Nice if Joe can confirm or not this.
From my part, I have used/experimented so far with this filter/cap in a completely non filtered system. This cap being the only one filter in my configuration (post DAC analogue processing).
So, maybe we may have a "talk" about these aspects?
Referring to YashN last post. I will suggest using of a film cap with the smallest physical dimensions (best SMD - 50/100v) one. Long legs is not at all fortunate for a cap and the circuit it is to be placed in. My opinion...
I`m a strong (reasonable...) supporter of the SMD components... and of the shortest possible wires, traces, and so on...
It does this (cap between DAC chip differential output) filter be used in conjunction (or not) with the classical filtering between I/V converter and final opamp?
It may be fortunate or not to heave so many filters in a signal path?
I know the theoretical purists I may want so many and sophisticated filters as needed, to filter out all what it may not be in audio spectre, but is that the best way to be done? I really do not think so...
It seems to me that Joe have experimented this filtering method heaving the another filters in his existent device/configuration. Nice if Joe can confirm or not this.
From my part, I have used/experimented so far with this filter/cap in a completely non filtered system. This cap being the only one filter in my configuration (post DAC analogue processing).
So, maybe we may have a "talk" about these aspects?
Referring to YashN last post. I will suggest using of a film cap with the smallest physical dimensions (best SMD - 50/100v) one. Long legs is not at all fortunate for a cap and the circuit it is to be placed in. My opinion...
I`m a strong (reasonable...) supporter of the SMD components... and of the shortest possible wires, traces, and so on...
Last edited:
I was interested here because I'm dealing with the LP filter for DAC's right now. Below is a simulation of it in the ESS case (using really simplified mode;l for the ESS output) It really is nothing more than a single pole low pass filter set to a lower frequency that usual practice. (not too different from FM tuner practice). its a good idea to keep the HF stuff out of the opamp. The second plot shows what the opamp needs to contend with if there is no filter. Even with feedback caps etc the input differential is squeeling to keep up.
The second plot is the real HF noise from a number of different DAC chips playing a 1 KHz tone. This is real and if at full amplitude will fry many power amps. Fortunately most preamps don't have the bandwidth to pass this stuff. I won't pass on which chips but they are from the major vendors. Your favorite is there most likely.
I think the issue is the subjective response to the low pass filter. Floyd Toole did a lot of work on this and showed that listeners prefer the attenuated high frequencies over flat response. I am also coming to the feeling that the DSD 30-50 KHz (5 pole) low pass filter contributes more to the preference for DSD than the technology, especially since most modern DAC's use variations on DSD (Delta Sigma) for their conversion.
The second plot is the real HF noise from a number of different DAC chips playing a 1 KHz tone. This is real and if at full amplitude will fry many power amps. Fortunately most preamps don't have the bandwidth to pass this stuff. I won't pass on which chips but they are from the major vendors. Your favorite is there most likely.
I think the issue is the subjective response to the low pass filter. Floyd Toole did a lot of work on this and showed that listeners prefer the attenuated high frequencies over flat response. I am also coming to the feeling that the DSD 30-50 KHz (5 pole) low pass filter contributes more to the preference for DSD than the technology, especially since most modern DAC's use variations on DSD (Delta Sigma) for their conversion.
Attachments
This was shared with me as confidential so I don't want to reveal too much. I may make my own measurements if I can get my hands on some of these and I'll post them.
The simulation also shows that too much cap or too small resistors will produce oscillation. I did not look at the impact on noise. It may not be good.
The simulation also shows that too much cap or too small resistors will produce oscillation. I did not look at the impact on noise. It may not be good.
I have also seen myself too, that higher capacity leads to HF oscillations increasing... It looks like an increasing of the HF noises, but it may well be just oscillations from the opamp... I did not insisted too much on these measurements so far. I think to repeated it later on.
Maybe so, but the issue here is not a frequency response thing. Keep in mind that several people have said "the effect is not subtle" and in fact "very obvious." A little bit down at 20KHz in FR terms would be subjectively subtle, this is not.
There is something else going on that cannot be just explained by FR. Trust me (I know trust is becoming a rarity), once you have heard it, then your reaction at that point will be "what is going on?"
I did this to an Oppo that was brought to me this week, it was -1.4dB after it left there. He took it home and connected it up and something unexpected happened. His partner, who I understand she is not an audiophile in any sense of the word, within sixty seconds realised something was different (she was part of a blind test and had no idea) and said so. "I don't know, I can't put my finger on it, but it is different.." and then added shortly afterwards she added "and I like it."
He not only heard it just as she did, but he was at least expecting something and she wasn't. He was amazed at her reaction and also knew what he was hearing was indeed "better". He then called me and told me what happened and added "you are going to be busy doing this." He also remarked that also the low frequencies were tonally also improved. And yet this is a HF filter.
I would like everybody to be busy doing this.
So please, by all means measure the filter, but also give it a listen.
So what is the mechanism at work? Whatever it is, there is a rational reason for it.
Again, this has only been observed with delta-sigma DACs. I think that has to be considered more than just a simple clue, but the essence of it.
I think that it has to do with what Barbera suggested, that this cap fits across the + and - phases, it is improving in the way of being an Integrator of packed density pulses (whether voltage or current) and hence we are getting a cleaner analog reconstituted.
I am not implying it is about reducing slew rate problems (it may) because I use it with a system that is as immune as things can get.
So if you have implemented the filter to a delta-sigma DAC, then by all means measure it, but do also take a listen. You may be pleasantly surprised as has several dozen people so far (and that number is growing).
Cheers, Joe
.
Last edited:
All,
I think that it would be interesting at this point to know who on this thread has attempted listening for the described effect for themselves, and whether they did, or didn't feel that they heard anything unexpected. If you choose to respond, please specify the DAC chip, the I/V circuit input resistance, and the value of the filter cap., thanks.
Separately, and lending support to what Joe has said about the effect not likely being simply due to the output filter's rolling off of the treble, I made the following observations regarding how modifying the 1st order output filter response by only a few tenths of a dB @ 20KHz (relative to D.C.) subjectively activated the effect or not:
-0.4dB = the effect is not perceptible
-0.5dB = the effect becomes just perceptible
-1.0dB = the effect is clearly perceptible
-1.3db = the effect seems close to optimum
-1.6dB = the high treble begins to sound soft
The key point of the above SUBJECTIVE observations is that seemingly small changes in filter response produced subjectively large changes in the sound quality, going far beyond changes in perceived tonality. To objectively assess whether what we are hearing is likely due to the greater suppression of out-of-band noise, let's compute what difference an 1st order filter is making to OOB rejection with respect to not perceiving the effect, and optimally perceiving it. The following figures are for a 1st order filter comprised of a single capacitor in parallel with an 75 ohm resistor passive I/V circuit.
@20KHz
-0.3dB
-1.3dB
@40KHz
-0.7dB
-2.4dB
@80KHz
-1.4dB
-4.3dB
At 80KHz, the difference in filter rejection is only 2.9dB. Hardly seems enough to be responsible for activating the effect or not. In addition, any 2nd order or greater output filter, which was flat in-band to 20KHz, would far better reject ultrasonic noise than can a 1st order filter set to -1.3dB @ 20KHz. Therefore, 'the effect' should have been heard via every cheap CD player made since sigma-delta D/A chips took over the low end of the market.
Last edited:
No wish to reignite the rancor here earlier but really Joe, how do you know this? Yes its not an audio band FR response thing because as Ken points out the FR changes around 20kHz are fairly subtle but the perceived effect isn't. However the cap is affecting FR way into the MHz region and the effects there aren't subtle. So why isn't what's being heard an ultrasonic/RF FR effect on the OOB noise produced by S-D DAC chips? Do you have observations to share which tend to rule this hypothesis out?
The cheaper CD players I've looked at, and the appnotes for the DACs used in them normally only implement 2nd order (sometimes 3rd order) active filters. Not 2nd order passive ones - which would require an inductor. The RF rejection of those active filters is going to be poor because it depends on the GBW of opamps. NE5532 is a typical choice and it doesn't have much loop gain available at all at 1MHz - so its output impedance is going to be pretty high. Meaning not much rejection.
Better rejection than from a passive single pole filter at RF, perhaps. Better at ultrasonic frequencies, I doubt. But as you observe, we really don't understand the mechanism behind what we are hearing. The search for the mechanism is one of the purposes of this thread.
Last edited:
I don't disagree that traditional opamp stages will be better below a couple of hundred kHz. My hypothesis for this effect is its the RF energy (beyond 200kHz) being reduced that accounts for the SQ improvements.
The key is, based on my experience, to get rid of the RF before any active circuitry sees it. If that happens, there's fold down into the audio band and the damage cannot be undone.
The key is, based on my experience, to get rid of the RF before any active circuitry sees it. If that happens, there's fold down into the audio band and the damage cannot be undone.
Last edited:
Mine too. In fact I have spent now decades eliminated slew rate distortion.
But in this case, there are enough indications to show this is not because the post-DAC is prone distortion - in fact, using a transformer coupled (200KHz bandwidth) into purely non-feedback tube amp that totally is immune to slew rate distortion (feedback opamps having poor overload margins is very common) and yet for some reason, getting the response down by approx 1.3dB @ 20KHz is proving remarkably consistent, from DAC to DAC, irrespective of the nature of the post-DAC circuit.
There has to be a different common denominator.
We are simply searching for a mechanism (which I guess that you have not yet sampled), that explains it. In my humble view, this cap has an effect on the DAC rather than limiting HF and VHF problems. And that is what I mean by saying this it is not an FR thing - because it doesn't explain it.
Cheers, Joe
That's the kind of extra datapoint I was hoping to get (about the tube amp).
Firstly distortion isn't necessarily 'slew rate' distortion. That's one possible mechanism but isn't the only one.
Saying 'the cap has an effect on the DAC' doesn't count as an explanation because you're not positing a mechanism by which it has the alleged effect. The 'effect on the DAC' mechanism would have to be something remarkably consistent between different vendors of DACs. That doesn't of course rule it out - but the OOB filtering hypothesis equally doesn't get ruled out by the tube amp example because that's an example amp not immune to all possible distortion mechanisms by which RF can fold down into the audio band.
Firstly distortion isn't necessarily 'slew rate' distortion. That's one possible mechanism but isn't the only one.
Saying 'the cap has an effect on the DAC' doesn't count as an explanation because you're not positing a mechanism by which it has the alleged effect. The 'effect on the DAC' mechanism would have to be something remarkably consistent between different vendors of DACs. That doesn't of course rule it out - but the OOB filtering hypothesis equally doesn't get ruled out by the tube amp example because that's an example amp not immune to all possible distortion mechanisms by which RF can fold down into the audio band.
- Status
- Not open for further replies.