The Lavry DA11: For your ears only

[synfreak]

Originally Posted by Dan Lavry

The typical noise of the DA11 is very low, typically -114dB unweighetd. I chose to specify it at -112dB typical and -110dB minimum, giving myself some margin. The DA11 is about 2-3dB better then DA10 (which is also very good DA).

Note: almost all the converter makers specified converter noise with "A weighted" curve which makes things look better by a few dB, but I prefer to specify with flat response 20-22KHz. I just do not like it when the same folks that sell you 96KHz (48KHz audio bandwidth) telling you that you need all that extra bandwidth, and at the same time they specify the noise floor with "A weighting" which virtually eliminate the noise measurement at frequencies over 10KHz. This is what I call "marketing"...

Quite true! Its very anoying to see measurements treated that way.


Dan, could you provide some (FFT-based) measurements of the signal-to-noise performance of the DA 11 (and - maybe - as comparison from the DA 10)? Would be highly appreciated ...

The specs of your new baby just look very promising for a solution that could fit my needs - and at a reasonable margin too.

Thanks
Harald

[Dan Lavry]

Originally Posted by Wodgy

Dan, you seem to be missing the core of what people are saying to you. They're not saying to ditch your system, and they're not saying to add a stepped attenuator. They're suggesting you use a rotary encoder, a digital rotary encoder rather than an up/down toggle switch. That's all they're suggesting. A rotary encoder would be more usable.

Here is some of my thinking:

Is a 56 position rotary is practical? Imagine 56 positions at about 6 degree per position... I do not know if such a device is available, but is so; I would not wish to paint 56 numbers on a front panel. That would call for a display such as I use (My 2 digit display).

Most rotaries based device use 2 devices, such as one device for 1dB step and the other for a 10dB step. So every 10dB, you need to reach the 10dB step device, and then reach for the 1dB step device and make a 9 dB "correction". Say you want to go from 19 to 20dB:

If first you change the 1 to 2, which gets the volume up to 29, an increase of 10dB (blowing your ears) then you need to change the 9 to 0, to 20dB.

A better way is to first lower the gain from 19dB to 10dB, and then go for 10dB device and change it from 1 to 2, for a 20dB.

In other words, you need to think about the order - which rotary switch to use first. If you want to reduce from say 40 to 39, you have to reverse the order you access the switch, in order to prevent a sudden intermediary 10dB boost...

Of course there are other considerations here that I did not mention. In fact, designing calls for a lot of considerations, not all are obvious and some are best kept proprietary.

I do my best to listen to inputs, and act on it. The DA11 is the third device with a volume switch and 2 digit display (the other are DA10 and MP10 micpre). The vast majority of feedback was very favorable.

I find the idea of up for more volume and down for less volume very user friendly and intuitive. Also note that for the >PIC< (tm) I used horizontal switches (activate sideways). If you want a wider image on the left, push the left image switch to the left. A narrower image on the left calls of pushing the left switch to the rights (of course the right channel switch is opposite). That is very intuitive beacuse the hand motion goes along with the desired action. You can see that I am very interested in user friendly and intuitive controls.

You said that a rotary will be more usable. Why? I like feedback but I would like to have it coupled with reason. What is it about an up down switch with a clear digital display that is less usable? What is it about turning a knob that makes it better then pushing a switch up for more volume, or down for less volume? Why is it that a rotation to the right is better for more volume? Why is it that some tiny numbers printed on a front panel (for attenuation) are better then my large lit numbers? The lit numbers are certainly clearer in a dark room...

Regards
Dan Lavry

[Dan Lavry]

Originally Posted by synfreak

Quite true! Its very anoying to see measurements treated that way.


Dan, could you provide some (FFT-based) measurements of the signal-to-noise performance of the DA 11 (and - maybe - as comparison from the DA 10)? Would be highly appreciated ...

The specs of your new baby just look very promising for a solution that could fit my needs - and at a reasonable margin too.

Thanks
Harald

Hello Harald

I am working at home (design is best done with minimum interference), and also I have not learned how to post plots and diagrams on this forum. Your request is an additional task, but I will comply but only if you really insist. My specifications are properly stated, better then most gear out there.

For now here are a few comment relating to what you said:

The output of DA’s are analog, therefore the direct measurement of the noise is not FFT based but an analog noise measurement. Many test systems do indeed utilize a some quality AD converter as an intermediary built in measurement tool , but one way or the other, for DA, the noise energy over the measurement bandwidth is combined into a single number that expresses the noise (dynamic range).

One can use an AD converter and do an FFT, which will show the combined AD and DA noise floor. That will show detail, and is can be useful to a designer. One has to keep in mind that, an FFT shows a noise at a level far below the “combined number”. This is a great way to “confuse the innocent”. For a 0-20Khz bandwidth, an FFT noise plot is around 40dB “better” then the dynamic range ”single number “. That is so because the FFT breaks the noise into many narrow frequency bins, each bin has only some of the overall noise.
For example, a CD format (16 bits) has a theoretical FFT noise level at around -139dB, but the specification is around 96dB (around 6 dB per bit).

That too has been a “marketing thing”. When one talks about noise, they should refer to the overall noise, which is a single combined number of the energy in all the bins (the whole audio range). But if one is looking at an FFT, the “noise floor” is at a much lower level. So some people that tell you that they have some dither letting you hear below the noise, are really talking about the FFT noise plot, not about the accepted definition of noise (such as dynamic range).

One can not hear music below the noise level of an FFT plot. But one can certainly hear below the dynamic range (single number ) given a good system and low noise environment.

Regards
Dan Lavry

[thomaspf]

Originally Posted by Dan Lavry

Here is some of my thinking:

Is a 56 position rotary is practical? Imagine 56 positions at about 6 degree per position... I do not know if such a device is available, but is so; I would not wish to paint 56 numbers on a front panel. That would call for a display such as I use (My 2 digit display).

Most rotaries based device use 2 devices, such as one device for 1dB step and the other for a 10dB step. So every 10dB, you need to reach the 10dB step device, and then reach for the 1dB step device and make a 9 dB "correction". Say you want to go from 19 to 20dB:

If first you change the 1 to 2, which gets the volume up to 29, an increase of 10dB (blowing your ears) then you need to change the 9 to 0, to 20dB.

A better way is to first lower the gain from 19dB to 10dB, and then go for 10dB device and change it from 1 to 2, for a 20dB.

In other words, you need to think about the order - which rotary switch to use first. If you want to reduce from say 40 to 39, you have to reverse the order you access the switch, in order to prevent a sudden intermediary 10dB boost...

Of course there are other considerations here that I did not mention. In fact, designing calls for a lot of considerations, not all are obvious and some are best kept proprietary.

I do my best to listen to inputs, and act on it. The DA11 is the third device with a volume switch and 2 digit display (the other are DA10 and MP10 micpre). The vast majority of feedback was very favorable.

I find the idea of up for more volume and down for less volume very user friendly and intuitive. Also note that for the >PIC< (tm) I used horizontal switches (activate sideways). If you want a wider image on the left, push the left image switch to the left. A narrower image on the left calls of pushing the left switch to the rights (of course the right channel switch is opposite). That is very intuitive beacuse the hand motion goes along with the desired action. You can see that I am very interested in user friendly and intuitive controls.

You said that a rotary will be more usable. Why? I like feedback but I would like to have it coupled with reason. What is it about an up down switch with a clear digital display that is less usable? What is it about turning a knob that makes it better then pushing a switch up for more volume, or down for less volume? Why is it that a rotation to the right is better for more volume? Why is it that some tiny numbers printed on a front panel (for attenuation) are better then my large lit numbers? The lit numbers are certainly clearer in a dark room...

Regards
Dan Lavry

Hi Dan,

Keep the display and the numbers just use a rotary encoder to control the up and down. No painting numbers, no super fine control and as many steps as you like.

Interesting question about the UI aspect. My wife just can't get the hang of it. Probably has to do with past experiences. However she is not alone, so far I had to explain the operation of the volume control to everyone who I showed your DA10 to.

Cheers

Thomas

[thomaspf]

A single 10KOhm resistor at room temperature measured over 20KHz bandwidth is about 1.8uV noise source. That alone is more noise then many OP amps circuits. For comparison, many semiconductor switches yield .15uV to .5uV.

Wow, it looks like a lot of progress has been made with semiconductor switches if the self noise of the attenuation element is an order of magnitude higher than the switch.

Cheers

Thomas

[Dan Lavry]

Originally Posted by thomaspf

Wow, it looks like a lot of progress has been made with semiconductor switches if the self noise of the attenuation element is an order of magnitude higher than the switch.

Cheers

Thomas

Hello Thomas,

The "attenuation element" is the resistor itself, and the higher the resistance value, the higher the noise. Similarly, The basic paramenter responsible to noise is the resistance of the silicon when the FET is on (FET On resistance). A lot of the older FET switches have low resistance thus yield low noise. Some video oriented devices, as well as some power devices have incredibly low on resistance yielding very low noise, as low as .05uV over the audio range, or less!. Of course most such devices are not at all good for audio. There are a lot of other factors that matter, on resistance is just one of the many parameters and considerations. At the end of the day, it is the combination of circuit and devices that count. A great device in a poor design is no better then a poor device in a great design...

But yes, technology does improve, and I am all for taking advantage of it whenever possible.

Regards
Dan Lavry

[synfreak]

Originally Posted by Dan Lavry

I am working at home (design is best done with minimum interference), and also I have not learned how to post plots and diagrams on this forum. Your request is an additional task, but I will comply but only if you really insist. My specifications are properly stated, better then most gear out there.

Hello Dan!
I really would like to see some FFT-based mesurements. You could put them on your website too, and just bring in the links.

I´m aware of the facts you mentioned in your answer, but I find it much more evident to look at a FFT (in addition to standard "dynamic-range" measurements). As you already noted, some manufacturers insist on "A-weighted" figures, which - at least for me - aren´t very helpful at all.

Thanks for your time + sharing your opinions with us!

Greetings
Harald

[Dan Lavry]

Originally Posted by synfreak

Hello Dan!
I really would like to see some FFT-based mesurements. You could put them on your website too, and just bring in the links.

I´m aware of the facts you mentioned in your answer, but I find it much more evident to look at a FFT (in addition to standard "dynamic-range" measurements). As you already noted, some manufacturers insist on "A-weighted" figures, which - at least for me - aren´t very helpful at all.

Thanks for your time + sharing your opinions with us!

Greetings
Harald

OK, give me a few days, I will find some time to do that additional task (I assume that I can post a plot here), if not I will post a link to a pdf...

Regards
Dan Lavry

[synfreak]

Originally Posted by Dan Lavry

OK, give me a few days, I will find some time to do that additional task (I assume that I can post a plot here), if not I will post a link to a pdf...

[Dan Lavry]

Originally Posted by synfreak

Here are the plots:

Look at the blue lines, they are for volume setting = 56, which is the max level (24dBu, 12.28V, 34,72V peak to peak). That is reference level for that condition thus a full scale signal would hit 0dB. Of course I "stretched the plot" so that the top is at -100dB, to show better detail.

The other lines (green, yellow and red) are added just to show the noise distribution over the frequency range at different gain settings (50, 40 and 30), and in those cases the amplitudes are of course lower.

Note that the DA11 has lower noise at the hearing sensitive range (1-4KHz). The DA10 is clean and flat at around -140dB. The DA11 is at least 6dB lower noise at around 1-4Khz, where the ear is most sensitive, and that is one more effective bit. Both DA10 and DA11 are about the same noise at 20KHz and that is why the overall (single number) dynamic range difference is around 3dB...

Also note that while the DA11 dynamic range is around 114dB, the FFT plots are much lower, as expected (I explained it in a previous post). Again, the overall dynamic number represents the "accumulated" noise of the whole curve into a single number.

Computers drive me crazy. It took minutes to measure and a lot of time to figure how to convert to gif, jpg and bit map... I have higher resolution files, but they may be too large . Let me know if the plots are clear enough...

Regards
Dan Lavry

[USAudio]

Thanks for taking the time to put together those plots Dan!

[AlanY]

Those are very impressive measurements. Thanks for posting that. Looks like the DA11 is a nice improvement over the DA10.

[shamu144]

Very nice indeed. The DA11 definitely looks like a significant improvement (at least technically) over the DA10. I should receive my DA11 now in a few days.

Dan, when using the DA11 at fixed volume as a DAC, is there any way to determine the optimal volume setting to get overall lower noise floor.

Thanks

[synfreak]

Hi S
Dan!

Thanks for the time you´ve spent for the measurements!

And they do look very good indeed! To say the least

Seems like I have to get my fingers on a DA 11 as fast as possible. With the USB input and the remote controlable volume it should perfectly suit my needs.

Greetings
Harald

[USAudio]

Here: Lavry Engineering
I see in the features list of the DA11 it lists "High power discrete headphone output".
Can someone more technically minded than myself explain exactly what that means?

By "discrete" I assume that means it's a discrete audio circuit and not an IC?
By "High power" I guess that means it will easily drive headphones from anywhere between say, 30 and 700 Ohms?

Does anyone know how, power wise, the DA11 compares to other similar products such as the Benchmark DAC1?