The graph that Bohrok showed mentioned in the top -4.97 dBFS and -1.97dBFs peak, which is correct when adding to signals of equal amplitude.
That in fact he offered two signals of -6.02 dB instead of -7.99 dB, means nothing but a calibration error.
To correct this, the whole graph should be shifted upwards by 1.97dB.
But for the IMD2 and IMD3, this makes no difference since all ratios remain exactly the same.
Hans
I just tried to get some more insight in differential, not SE.
About SE I'm getting more and more the feeling that more background noise could be the cause for improved sound perception.
This is extra noise is definitely the case when disabling the servo for audio and because of 3dB more uncorrelated noise.
When playing LP's the background noise is far higher as with CD's and I suspect that with Optical carts it's even higher.
But compared to CD's who sound almost unnatural clean, LP's sound more life like, at least to me.
Hans
When you say SE does that mean a simple one stage filtering network... perhaps AD811 based? My own preference is toward simplicity and am looking to implement an AD811 (not in an FIRDAC though), and do use long integration DC servo loops to avoid coupling capacitors.
It has never been that clear that differential networks ought to be sonically better in all circumstances, though the measurements often suggests it should. Yet differential XLR interconnects is something that makes sense between devices to mitigate power supply ground loop issues through the power cords.
Hans. I can understand that increased noise can create sonic improvements, though at some point there must be a negative effect. If always true why be concerned about reducing noise if this is always a sonic benefit?
I tried a few different methods. From Marcel's dac, I always went into a prototype transformer with a grounded electrostatic shield between primary and secondary. I tried with and without Marcel's output stage board, and with and without differential versus SE outputs. What I liked best was SE without any IC opamps in the output stage. However, doing that exposes other problems with the dac that IMHO and IME suggest further attention to other parts of the dac might be helpful. What makes me think that? This isn't my first time through some of this stuff with a FIRDAC in many ways much like this one. In other ways, there are some differences between this dac and my prior experience. Basically though my opinion is that, at this point giving too much focus to the output stage is probably more of a technical distraction than anything else. Its relatively easy to measure, so the light is better according to the "streetlight effect" cognitive model. Also it seems congruent with the WYSIATI cognitive model. Again, this is my personal opinion. I also understand from experience that someone else might try to parody/demean it in an effort to sabotage my credibility. Yet I refuse to be silenced by such tactics.
Noise can act in different ways to affect an audio signal. However, the only way it can expose more low level details is through some mechanism like stochastic resonance, which is not what I think is going on here in my preference for SE. Noise/distortion can sometimes also add some false sparkle/clarity in an otherwise muddy sounding system but I even more strongly feel that isn't what is going on here either. I already said what I suspect is going on in a recent post, #2,182. I will stand by that conjecture for now.
Last edited:
Your opinion is understandable. It must be frustrating when you have nothing to contribute.
So your comments about technical distraction or "streetlight effect" are not attempts to demean the efforts of others? Sorry, but it is blatantly clear why you come up with such comments. Very sad.
Absolutely not! I have great respect for the people doing serious engineering in the thread, including for trying to measure things according to engineering theory and textbook training.
Sometimes I can applaud someone's good results, sometimes I am stuck with the problem of deciding whether telling someone an opinion they may not like hearing is worth risking the trouble it may bring. However, for people who can take well intended constructive criticism as well as possible, IME its usually better to tell them the honest truth even if its an opinion. I am faced with a similar problem now with someone I am working with outside the forum. To be honest with them I had to give some criticism that what not what was wanted. I don't like it, the person I'm talking to didn't want it, but I know they understand I am just trying to be helpful. Its not intended to be hurtful.
Sometimes I can applaud someone's good results, sometimes I am stuck with the problem of deciding whether telling someone an opinion they may not like hearing is worth risking the trouble it may bring. However, for people who can take well intended constructive criticism as well as possible, IME its usually better to tell them the honest truth even if its an opinion. I am faced with a similar problem now with someone I am working with outside the forum. To be honest with them I had to give some criticism that what not what was wanted. I don't like it, the person I'm talking to didn't want it, but I know they understand I am just trying to be helpful. Its not intended to be hurtful.
Last edited:
You have posted about "streetlight effect" more than dozen times and always as an attempt to demean others for giving too much emphasis on measurements in your opinion. Instead of repeating such silly tactics why don't you try to contribute in a more positive way. If you cannot do that in this thread why don't you start a thread of your own.
The biggest problem in audio design, both diy and professionally, IMHO has to do with the mindset of engineers. I understand the mindset pretty well having been a professional engineer most of my life. Its just that as engineers we can get stuck focusing on what we know how to measure.
@soundbloke said something I through was pretty profound in another thread. It was this:
I would suggest the most important misconception is the assumption of linearity in our hearing. Our learning capabilities are (predominantly) irreversible and therefore very non-linear. Unless rigourous blind testing is carried out on a per listener arrangement (which will often be practically extremely difficult), there is little reliable information to separate the well-trained listener from a delusional one when we approach commonly accepted hearing thresholds. Regardless of our physiology, we all have a capability to train our hearing on minute details that an untrained listener would find inaudible. Likewise we all possess a substantiative capability to delude ourselves into hearing soemthing we are not. Perception and sensation are not the same thing...
https://www.diyaudio.com/community/threads/measuring-the-imaginary.409662/post-7613254
Its pretty much the same same thing about learning to hear that Howie Hoyt talked about after having submitted to years of ABX testing (it may even be that Amir has learned to listen more skillfully after going through the Harmon training). Its that people can learn how to hear little things which are inaudible other people, but the skill has to be learned methodically over time. However, there is a big problem with all this which is that people can also be quite deluded about what they are hearing. Too many EE's only believe in the "humans are quite deluded when it comes to hearing" part. They deny that anyone can be well-trained to hear what would be inaudible to others. That view is quite at odds with what soundbloke said.
Okay, so far? If so, I will try to tie it back into its relevance to this thread.
@soundbloke said something I through was pretty profound in another thread. It was this:
I would suggest the most important misconception is the assumption of linearity in our hearing. Our learning capabilities are (predominantly) irreversible and therefore very non-linear. Unless rigourous blind testing is carried out on a per listener arrangement (which will often be practically extremely difficult), there is little reliable information to separate the well-trained listener from a delusional one when we approach commonly accepted hearing thresholds. Regardless of our physiology, we all have a capability to train our hearing on minute details that an untrained listener would find inaudible. Likewise we all possess a substantiative capability to delude ourselves into hearing soemthing we are not. Perception and sensation are not the same thing...
https://www.diyaudio.com/community/threads/measuring-the-imaginary.409662/post-7613254
Its pretty much the same same thing about learning to hear that Howie Hoyt talked about after having submitted to years of ABX testing (it may even be that Amir has learned to listen more skillfully after going through the Harmon training). Its that people can learn how to hear little things which are inaudible other people, but the skill has to be learned methodically over time. However, there is a big problem with all this which is that people can also be quite deluded about what they are hearing. Too many EE's only believe in the "humans are quite deluded when it comes to hearing" part. They deny that anyone can be well-trained to hear what would be inaudible to others. That view is quite at odds with what soundbloke said.
Okay, so far? If so, I will try to tie it back into its relevance to this thread.
Last edited:
Thanks Mark. I don't know much about FIRDAC's. By the way what is the modulation frequency? I asked about the AD811 I/V because I recall you using this device, perhaps in some configuration of Walter Jung's as per below. Don't know how you found that network to work sonically.
https://cfas.waltjung.org/High_Performance_Audio_Stages_Using_TransZ_Amps.pdf
It seems the FIRDAC's can be problematic in using active I/V devices, even more so with feedback. The signals seem extremely high frequency and high level transient in nature that can permeate throughout a board (including air borne). Certainly the analysis can be highly useful, though in the end all that matters is sonically. By the way, have you ever tried an open loop AD844 feeding current into the inverting terminal and taking signals from a resistor connected to pin (5)? With low R values I believe the propagation delay can be about 5nSec.
In using the AD844 it is recommended to use a series resistance of about 100 Ohms in series with the load resistor to ground. This allows a capacitor to be applied across the load resistor for filtering, otherwise there exists some instability seemingly caused by the output network of the AD844, although it is disconnected.
Last edited:
Hierfi, I tried various amplifier ICs for I/V with IC dac chips. None of them sounded as satisfactory as a FIRDAC in SE mode with a discrete output stage design which includes a DC servo, followed by a special (very high quality by engineering standards) transformer that is good at removing any remaining RF. My guess is that the discrete output might work quite well with Marcel's dac, but I don't have a Marcel dac here anymore to try it. Regarding the discrete output stage itself, it happens to be an Andrea Mori design for FIRDACs, so it should be possible to get one to try. Also, for these FIRDACs you may find that loading all dac outputs with a resistor, including any unused outputs, may have a beneficial effect. IOW, its something that may be worth trying.
EDIT: These dacs are usually clocked at 22/24MHz or else 11/12MHz in order to be able to play DSD256. However it not the fundamental clock frequency that tends to cause problems so much as it is the fast risetime. Sometimes slowing the risetime a little can find a sweet spot where EMI/RFI is less, but the dac is still operating well as intended.
EDIT: These dacs are usually clocked at 22/24MHz or else 11/12MHz in order to be able to play DSD256. However it not the fundamental clock frequency that tends to cause problems so much as it is the fast risetime. Sometimes slowing the risetime a little can find a sweet spot where EMI/RFI is less, but the dac is still operating well as intended.
Last edited:
I asked because I am currently working on a high speed I/V converter for a "conventional" DAC. The results appear promising using a critically tuned AD811. The response below shows the input/output response for a +/- 10volt (+/- 5mA into 2K Ohm feedback) 1MHz square wave (2MHz sampling)
(100nSec/div scale). The propagation delay is about 20nSec if the scale is expanded. At the rate of rise shown a 300pF load capacitor will begin current limiting at +/- 100mA of an AD811. The second image shows well controlled current limiting with only a 1nF capacitance load.
The image is non-inverted in that the network passes through 2 inverter stages, an open loop AD844 (adding about 5nSec delay) feeding the inverting terminal of the AD811. (note that the rise and fall times are limited by my function generator)
(100nSec/div scale). The propagation delay is about 20nSec if the scale is expanded. At the rate of rise shown a 300pF load capacitor will begin current limiting at +/- 100mA of an AD811. The second image shows well controlled current limiting with only a 1nF capacitance load.
The image is non-inverted in that the network passes through 2 inverter stages, an open loop AD844 (adding about 5nSec delay) feeding the inverting terminal of the AD811. (note that the rise and fall times are limited by my function generator)
Last edited:
Actually I measured the lowest distortion (THD) when I (according to you) sabotaged the common mode loop (see post #2080). That is with stock opamp (OPA1678) in second stage.