Looking for High quality DAC units

[DANGERDAN]

iv been doing some thinking and how in the world are you suppose to configure a sub with dac units, you are only left with the subs integrated lowpass filter but nothing for the mains ?????.

[Guest " "]

if using rca, run the dac out to rca y cables or y connecter. run one of the rca out on the y connecter or y cable to the amp....the other to the sub line in.

[Guest " "]

watch the sex of the connectors.....plan it out....the connector can go right to your dac...and then two new RCA outs.

[DANGERDAN]

Na i knew that but i was wondering more about the crossovers, being able to configure the crossover properly but without a filter from a receiver of some kind to filter the mains i don't see it possible. I could just use the LPF on the subwoofer to match the rolloff of the mains but wouldn't it be better cut off the mains higher with a HPF so that the sub does most of the lower frequency. ??.

[DANGERDAN]

both the peachtree Dac.it and audiolab Mdac use sabre chips (though the peachtree uses the higher end sabre chip), you are right that the same chips appears in some of the highest priced DACs in the world. I like the "play factor" of the audiolab unit allowing you to choose filters, etc. and it seems to use fairly high quality components for the price, I don´t see much about it´s output section and since it acts as a preamp it would be nice to know about how it generates and manages its analog output. I also cannot see much listed about the power supplies. I have always liked the rotel "sound" so if their DAC keeps those same family values it should be a smooth, warm sounding unit. tough choices, becuase there are a LOT of options out there. I suspect you cannot go wrong with any of the ones you listed. one last note many people classify the sound of the sabre chips as analytical while the wolfson might be considered smoother. YMMV, etc. given the flexibility of the audiolab to tailor the sound in your system I might think it owuld be your safest choice of the group. let us know how this progresses. warm regards, Tony

p.s. just to complicate things more add the rega dac to the list, really good output stage implementation there.

I seemed to have missed your post but it was helpful none the less, i had done a bit of research on the m-dac and spoke to its lead designer about a few aspects over at another forum. He gave me some insight to its technology and concerns i had which was really good, heres a few for those who are interested. I think i will possibly buy the rotel dac to see how it compares in the future.

Quote:

Originally Posted by DANOFDANGER

From what i can read with the M-DAC passive preamp being enabled it is a concern that audio quality may be attenuated at low listening levels, could you state why this is?

John: The MDAC does not have a passive pre-amplifier - its a Digital pre-amplifier so the audio signal is scaled in the digital domain - there is no change in the Analogue gain structure.

Having too much system gain requiring you to listen at high attenuation levels is a bad thing. "Gain" is never for free, each gain stage causes a reduction in sound quality - there's no such thing as a 100% perfect gain stage.

When I hear of systems where the owners are listening with level settings of -60dB or lower this suggests to me a very badly matched system.

Quote:

Originally Posted by DANOFDANGER

I would have thought since the volumes were digitally modified directly from the DAC that it would inhibit no audible distortion.

John:Simple question is how to "Reduce" something without loosing anything - you will always be reducing a signal into a "fixed" noise floor level.

You have a large bucket of water, and want to pour it into a smaller bucket - the smaller bucket will eventually overflow, and you loose the rest.... Bigger into smaller does not go.

Quote:

Originally Posted by DANOFDANGER

Or does it have something to do with the analog end like at the operational amplifier or other analog components. Is this because the DAC was designed optimally at a certain voltage level ??

John:Both optimised Analogue and Digital systems (once converted back into the analogue domain) will always face the same noise floor issue - 24bits audio is BELOW the theoretical noise floor caused by the random movement of electrons "heat" energy, unless you live near absolute zero...

Digital domain attenuation when done correctly should be no different to performing analogue domain attenuation - however practical "real world" implemention issues need to be considered.

With Digital attenuation, the Analogue system is always operating at Full gain - turn up the volume knob of any analogue amplifier with no music playing and you can hear a slight background Hum, Hiss, RF intermodulation products etc. though the speakers - this is a reality of analogue electronics.

Any Digital product by there very nature will produce RF energy. There is a practical limitation on how much you can filter this energy before you start to detrimentally impact the audio quality.

Without any form of analogue attenuation this "leakage" RF energy from the DAC is pumped directly into the Amplifiers input stage which can then be demodulated into the audio range.

Transistor inputs stages are by far the most sensitive to RF demodulation by a significant margin followed by Jfets, Tubes then MOSFET's.

Adding Analogue attenuation in front of the amplifier reduce both Audio AND RF energy - this does not happen with pure digital attenuation.

Quote:

Originally Posted by DANOFDANGER

Another thing if you disabled the passive preamp stage and say connected it with a windows computer and controlled the digital volume from there, would this have the same adverse effects of sound degradation or would this help with the distortion at low levels.

John:NEVER use windows level control - the algorithms used by the Windows sound kernel to achieve digital domain attenuation are truly horrid - unfortunately, Windows digital domain attenuator is a poster child on how digital attenuation should NOT be done.

The MDAC also has the advantage that the attenuation can be performed in a 32bit domain (in fact greater then 32 bits).

John

Originally Posted by DANOFDANGER

Hey thanks john that really helped me, one last thing as i am about to go to the atore and buy one now.. so i should leave windows volume control to full so theres no attenuation distortion coming from windows bad algerithm.

John:Yes - thankfully setting Windows level control slider to 100% can result in Bit accurate data if everything else is configured correctly

John

[sunnysal]

thanks for the follow up, interesting conversation there. too bad he didnt give much information about the output stage to see how it is implemented in that unit. regarding digital domain volume control I think you need to hear both sides of the argument, the designer doesnt really give you a very scientific explanation about why he prefers it to analog. let me try to add to the issue by quoting robert hartley "Reducing the volume in the digital domain is accomplished by multiplying each sample by a number less than one. Let's take the example of decreasing the playback volume by 6 dB. Because 6 dB represents a halving of voltage, every sample is multiplied by 0.5. The samples encode a number that represents the original analog waveform's amplitude at the time the sample was taken. By mutiplying each sample by 0.5, the amplitude of the econstructed analog signal is reduced by half--or 6 dB. But there's a price to pay for this digital slight-of-hand. Every 6 dB of attenuation reduction in volume) reduced the playback system's resolution by one bit. In other words, a 16-bit signal atenuated in the digital domain by 6 dB now has the resolution of a 15-bit signal. Lower the volume by 12 dB and you have the equivilent of a 14-bit source. Dynamic range is reduced, and the music signal gets closer to the digital noise floor. With fewer bits, low-level signals can become more coarse, particularly at high attenuation levels." interesting huh?

[sunnysal]

another comment from your friends at wolfson:

"Another drawback of digital volume control is that the signal-to-noise ratio (SNR) worsens at

low volumes. This is due to the quantisation error, which is inherent in any digital system and

may result in audible noise. Its magnitude depends on the resolution of the digital audio data

and/or the DAC. High-end audio systems are designed such that any noise is negligible

compared to the magnitude of the audio signal, and therefore inaudible. However, when

volume control is performed in the digital domain, the digital audio signals amplitude may be

decreased by several orders of magnitude while quantisation noise remains constant,

resulting in a lower SNR"

p.s. just hit 3,500 posts. great to still be learning and sharing so much with friends here on the forum after all this time!

[DANGERDAN]

I did ask him about comparison between analogue and digital attenuation and he did go on to say (in his own way) that using the digital volume control at low levels could in duce audible distortion, which i have found people claim but if you used the volume controller at high levels closest to its optimal voltage (2.1v) or 0db on the screen indicator then this should not be enough quantization to be worried about.

As i could tell there are benifits to having digital volume control but as you have seen as well there are also cons to this type of topology.

Grats on 3500 post's haha.

EDIT

The digital volume control is often seen as bad because as the volume is decreased, the bits are lost. In general, for every 6 dB attenuation, 1 bit is lost. For example, if the digital volume is set to -12 dB, then 2 bits are lost. This is a serious loss of resolution if the DAC has only 16 bits to begin with. However, this is no longer an issue with the 32-bit ES9018 DAC. With the full 32-bit data path, the digital volume control has more than enough resolution to accommodate the loss of bits. So the loss of 2 bits, as in the example, is practically inconsequential to the 32-bit DAC.

[pzannucci]

EDIT

“The digital volume control is often seen as “bad” because as the volume is decreased, the bits are lost. In general, for every 6 dB attenuation, 1 bit is lost. For example, if the digital volume is set to -12 dB, then 2 bits are lost. This is a serious loss of resolution if the DAC has only 16 bits to begin with. However, this is no longer an issue with the 32-bit ES9018 DAC. With the full 32-bit data path, the digital volume control has more than enough resolution to accommodate the loss of bits. So the loss of 2 bits, as in the example, is practically inconsequential to the 32-bit DAC.”

The key is keeping the bit width for the digital volume control higher so losses are negligible. There may be a flip side to that in the there needs to be upsampling of the data. Some people don't like that.

[sunnysal]

we agree that keeping the volume set high is key to success with digital volume controls. the problem I have in my system is that my speakers as so efficient I need to keep volume WAY down to listen. this is practically the only thing that has kept me from trying to eliminate my preamp and using a digital volume control on a DAC. I am anxious to hear about the M-DAc which seems like a cool unit. regards, tony

[Guest " "]

but wouldn't it be better cut off the mains higher with a HPF so that the sub does most of the lower frequency. ??.

my assumption was you were after a connection to sub woofer in. not too much to gain and more is loos if you do not use the sub woofer in which normally is he THX standard of 80hz. some units have varying xover points, but once you approach 120hz, you loose omnidirectional masking.

[DANGERDAN]

but wouldn't it be better cut off the mains higher with a HPF so that the sub does most of the lower frequency. ??.

my assumption was you were after a connection to sub woofer in. not too much to gain and more is loos if you do not use the sub woofer in which normally is he THX standard of 80hz. some units have varying xover points, but once you approach 120hz, you loose omnidirectional masking.

Thing is i want to be able to cut the mains off at about 80 hz with a slope of about 12-24db as this would be better than just trying to use the crossover on the sub to match the speakers, especially when the slopes on speakers are unpredictable and the frequency response my room gives off with low frequency's due to reflections and resonant distortion, adding a sub without a high pass filter would be troublesome and difficult.

[DANGERDAN]

we agree that keeping the volume set high is key to success with digital volume controls. the problem I have in my system is that my speakers as so efficient I need to keep volume WAY down to listen. this is practically the only thing that has kept me from trying to eliminate my preamp and using a digital volume control on a DAC. I am anxious to hear about the M-DAc which seems like a cool unit. regards, tony

This is the problem i face with the m-dac and i am trying to figure out the possible problems with using low attenuation with this, possible fix would be to use a -20 or -40 attenuation in the line which is a permanent fix which a lot of people have done.

[DANGERDAN]

EDIT

The digital volume control is often seen as bad because as the volume is decreased, the bits are lost. In general, for every 6 dB attenuation, 1 bit is lost. For example, if the digital volume is set to -12 dB, then 2 bits are lost. This is a serious loss of resolution if the DAC has only 16 bits to begin with. However, this is no longer an issue with the 32-bit ES9018 DAC. With the full 32-bit data path, the digital volume control has more than enough resolution to accommodate the loss of bits. So the loss of 2 bits, as in the example, is practically inconsequential to the 32-bit DAC.

The key is keeping the bit width for the digital volume control higher so losses are negligible. There may be a flip side to that in the there needs to be upsampling of the data. Some people don't like that.

From what i have been told when using digital attenuation the problem is not with bits lost but with noise shaping.

The digital volume control is often seen as bad because as the volume is decreased, the bits are lost. In general, for every 6 dB attenuation, 1 bit is lost. For example, if the digital volume is set to -12 dB, then 2 bits are lost. This is a serious loss of resolution if the DAC has only 16 bits to begin with. However, this is no longer an issue with the 32-bit ES9018 DAC. With the full 32-bit data path, the digital volume control has more than enough resolution to accommodate the loss of bits. So the loss of 2 bits, as in the example, is practically inconsequential to the 32-bit DAC.

I have been worried about having to use really low attenuation due to my speakers being really effecient, so i done some thinking on how much this should affect me and this is what i come up with.

If 1 bit is lost for every 6db then at -24db 4 bits is lost from the 32 bit dac, at -48db 8 bits is lost and finally at -78 a total of 13 bits is lost leaving 19 bits out of the 32 bit dac.

Now for people who use 24 bit content they will loose resolution but for people like me using only 16 bit will i still be fine using the digital volume below -40db ?.

With snr @ 122 db wont the noise be so low that this won't be a issue as well ??.

John: I'm afraid its not so simple - You don't loose the "bits" as you believe because of Noise shaping - the same noise shaping that allows "Bitstream" DAC's and DSD to recreate the greater then 18 Bits resolution within the audio band with just a single Bit!!!

http://en.wikipedia.org/wiki/Noise_shaping

John

[Guest " "]

as I read this I'm scratching my head wondering where did we miss the turn and how far down this road have we traveled. this whole bit discussion is inconsistent with the wow factor when using 16bit non oversampling DAC's.

[muel]

At some point you have to stop reading specs, just spend some money, and listen. Buying and selling used you aren't out too much.

[DANGERDAN]

Just so you know as I couldn't tell if you we're stating that we were talking about a dac that up sampled or down sampled, that the ess sabre 9018 doesn't do that at all.

Upsampling refers to a change in the sample frequency, not the sample size/word length which is all is done from the 32 bit dac.

[DANGERDAN]

At some point you have to stop reading specs, just spend some money, and listen. Buying and selling used you aren't out too much.

Quite true.

[LarryC]

I 've used the Audioquest DragonFly for a few days now, and think it has an excellent, musical sound quality. It's an amazingly compact thumb drive size, with a USB jack for input and a mini stereo jack for output. Perhaps some will find that limiting. Although it seemed simple to install, I did have to call Audioquest to be sure the settings were all correct (they had to be adjusted, however, so it's not necessarily easy to install).

The AQ tech support is very helpful and forthcoming.

One caution: it didn't sound that great at first, required several hours break-in before it was on the way to shaping up. $250.

[DANGERDAN]

we agree that keeping the volume set high is key to success with digital volume controls. the problem I have in my system is that my speakers as so efficient I need to keep volume WAY down to listen. this is practically the only thing that has kept me from trying to eliminate my preamp and using a digital volume control on a DAC. I am anxious to hear about the M-DAc which seems like a cool unit. regards, tony

Thought id reply to my findings on digital attenuation and your fears with converting to this type of setup, i had read through about 12 parts of threads related to the m-dac to further my education on this design, this took 2-3 days non stop reading and learnt some but wasted more hours but it had to be done haha.

Anyway what i have found is quite interesting and inspiring to digital to analog technology, adding to this my belief that preamps are becoming obsolete (voltage amplification stage) is coming clearer to me. This could be debatable but what i have found with the M-dac is there is no actual problem with quantization in regards to resolution or loss of bits, the design of the 32 bit processor is quite remarkable and is quite superior to analogue attenuation (to a point). With standard DA/C's that are usually either 16 or 24 bits, when they are being quantized they start loosing resolution or bits which reduces the dynamic range but the big problem is the actual truncation that happens because the bits are entirely removed from the signal resulting in quantization error. From what i have read this is the wrong way to do it and how it is done with most or all standard dac's, the ESS sabre on the other hand does something different, there is no upsampling or downsampling at all from what i can see, increasing the size to 32 bits is just shuffling the bit information from a smaller to a larger word length. With this increase in bit size there is more dynamic headrooom and in turn more possible range of attenuation without the effects of loss of resolution or bit loss. Truncation is also handled differently, it in fact is not removed at all but is bit relocated using a type of dithering called noise shaping. This noise shaping takes the truncated data and just moves the now distorted information into a frequency outside our hearing range, it does it in a specific way so that the signal to noise ratio is improved down the attenuation due to the bit relocation working in a specific algorithm so that the noise is transferred outside the frequency range. (moving the bottom bit).

By doing this however i think you loose less possible attenuation, like with the M-dac it is suggested that going below -40 results in possible degradation of the signal but not because of loss of resolution, because technically at -48 db 24bit source files would still have full resolution and 16 bit files would be fine right down until -78 which is only 2db from full attenuation. The problem is because of the added dithering to help reduce quantization distortion the overall attenuation range is reduced due to the artifacts of this, but this is better than not having the dithering in the first place so its acceptable.

With what you said about speaker efficiency and being forced to use heavy attenuation you can always add inline attenuators, this is what i am doing because from what i have worked out i will be hitting close to -40 on the M-dac. With these inline attenuators i can reduce the dile up to -20 or more depending on what db attenuation i decide to buy. Also matching source to destination helps with how attenuation is used, using badly matched equipment can cause problems, using a 1/2 higher sensitive input from a amp forces people to have to use more attenuation as the source is only having to work half output to reach full power from the amp. (assuming the source is abiding the average 2v output).

My M-dac will be here definitely tomorrow so i will be back to share my findings