... and of course if you use discount code DIYA you get 10% off!
Some article details here: http://www.diyaudio.com/forums/everything-else/171767-linear-audio-volume-2-out-32.html#post2688723
jan didden
Some article details here: http://www.diyaudio.com/forums/everything-else/171767-linear-audio-volume-2-out-32.html#post2688723
jan didden
Patrick and Jan, thank you both so much for contributing! This IV converter certainly looks to be very, very exciting! Can't wait to sign up for the evaluation kit 🙂
Let's give it 2 weeks, so that the magazine could get to some of you who has already subscribed.
Patrick
Patrick
Really interesting stuff patrick - can't wait to read more about it in the mag.
I noticed on the measurements that the noise floor looks slightly higher on the CEN compared to SEN - is that component specific or a typology based difference?
I noticed on the measurements that the noise floor looks slightly higher on the CEN compared to SEN - is that component specific or a typology based difference?
I asked this question on the Zen I/V thread too: is the linearity of this circuit at least as good as that of a high end 24 bits DAC ?
If it isn't then the linearity of the system DAC+converter will be limited by the weakest link, i.e. the converter.
If it isn't then the linearity of the system DAC+converter will be limited by the weakest link, i.e. the converter.
Good point. (I assume that you mean the I/V converter would be the weakest link, not the Digital-to-Analog Converter - there are two converters!)I asked this question on the Zen I/V thread too: is the linearity of this circuit at least as good as that of a high end 24 bits DAC ?
If it isn't then the linearity of the system DAC+converter will be limited by the weakest link, i.e. the converter.
If you look at linearity, I think you'll find that the power amp and loudspeakers are the weakest links.
Good point. (I assume that you mean the I/V converter would be the weakest link, not the Digital-to-Analog Converter - there are two converters!)
That's what I meant.
If you look at linearity, I think you'll find that the power amp and loudspeakers are the weakest links.
Ok, but let's assume that I want to use a high performance DAC for something else than feeding a power amp, let's say I want to use it to build a high performance signal generator.
In this case I don't want my I/V converter to degrade the linearity of the DAC.
Is the circuit presented here good enough for this purpose ?
Take the specs of ES9018 and PCM1704K, which are about as high end as you get these days for Sigma delta and for R2R DACs.
ES9018 specs says -120dB THD, and PCM1704K says 0.001%, or -100dB.
Even at a transimpedance of 2700R, which is factors higher than anything this simple in discrete circuits that I know of, the simulated distortion for SEN IV is 0.0012% (-99dB), and is largely 2nd harmonics. In balanced mode, it will go well below -100dB, i.e. matching that of a PCM1704 K grade.
There are tricks described in the article to reduce this by a further factor of 3.
And so far I have not seen any discussion of how the circuit functions. The key to its performance is that it functions totally different from say a ZEN IV. Only when you understand how it works would you aprreciate its elegance.
If you only want to chase after THD numbers, you should not make anything discrete and just use high open-loop gain opamps.
I build discrete because they sound wonderful.
Patrick
.
ES9018 specs says -120dB THD, and PCM1704K says 0.001%, or -100dB.
Even at a transimpedance of 2700R, which is factors higher than anything this simple in discrete circuits that I know of, the simulated distortion for SEN IV is 0.0012% (-99dB), and is largely 2nd harmonics. In balanced mode, it will go well below -100dB, i.e. matching that of a PCM1704 K grade.
There are tricks described in the article to reduce this by a further factor of 3.
And so far I have not seen any discussion of how the circuit functions. The key to its performance is that it functions totally different from say a ZEN IV. Only when you understand how it works would you aprreciate its elegance.
If you only want to chase after THD numbers, you should not make anything discrete and just use high open-loop gain opamps.
I build discrete because they sound wonderful.
Patrick
.
Last edited:
If you want to use a DAC to build a functions generator, the only way you can get decent performance (below -120dB) digitally is IMHO to notch filter the DAC output.
You can of course always use voltage out DACs with built-in opamps (on top of the notch filtering).
It does not need to sound nice, just low distortion.
Patrick
.
You can of course always use voltage out DACs with built-in opamps (on top of the notch filtering).
It does not need to sound nice, just low distortion.
Patrick
.
Last edited:
The noise floor on the SEN is lower by about 3dB because there I had 2 JFETs in parallel.
Well known trick to lower JFET noise.
Patrick
Well known trick to lower JFET noise.
Patrick
In balanced mode, it will go well below -100dB, i.e. matching that of a PCM1704 K grade.
...
If you only want to chase after THD numbers, you should not make anything discrete and just use high open-loop gain opamps.
I build discrete because they sound wonderful.
Patrick
.
So that means that you would be able to distinguish between the sound of your I/V circuit and the sound of an op. amp. based I/V converter that has the same or better THD performance (i.e. better than -100db) ?
Obviously the rest of the audio chain would have to be identical (the same digital source, DAC, amp. and speakers).
Last edited:
Here are the schematics as published in Linear Audio, courtesy of Jan.
CEN for complementary, SEN for single ended.
Patrick
what is the input impedance as seen by the DAC ??
doesn't it have to be ultra low to be effective as an IV converter ??
I can't see how source impedance of two JFETS could be very low unless they were biased very high.
regards
Trev
15R as per schematics published.
It can be halved by changing the 2SK170BLs to 2SK369BLs in SEN IV, at a slight increase in distortion.
Again, all discussed in detail in the article.
Patrick
.
It can be halved by changing the 2SK170BLs to 2SK369BLs in SEN IV, at a slight increase in distortion.
Again, all discussed in detail in the article.
Patrick
.
Last edited:
The DAC chips that I've worked with want 25 Ω or less.doesn't it have to be ultra low to be effective as an IV converter ??
My experience with the ESS9018 DAC, which has lowest THD with 0R load, is that sonics is not only dependant on the load (no surprise I guess). I have used 10R load with the ESS9018 and it can sound fantastic with that load. I have alsu used transformers with the ESS9018 (and PCM1792) and sonics can be just amazing with loads up to several k ohms.
/IR
/IR
The DAC chips that I've worked with want 25 Ω or less.
With discrete designs you can make such elegant solutions and using chip OPmaps is not much of a challenge/fun.
One can make a discrete OPamp with THD as low as you wish (almost), but they tend to be complex, and using top notch components then make it rather costly. Thus, a simple discrete solution with the best components is a great solution.
EUVL, you have made an outstanding elegant IV converter - hats off for you!

I am not eagerly awaiting my copy of the magazine
/IR
One can make a discrete OPamp with THD as low as you wish (almost), but they tend to be complex, and using top notch components then make it rather costly. Thus, a simple discrete solution with the best components is a great solution.
EUVL, you have made an outstanding elegant IV converter - hats off for you!

I am not eagerly awaiting my copy of the magazine
/IR
If you only want to chase after THD numbers, you should not make anything discrete and just use high open-loop gain opamps.
I build discrete because they sound wonderful.
Patrick
.
A few of you asked about details of the subscription of the evaluation kit.
You guys are really impatient.
Firstly the cost.
4 quad JEFT heatsinks plus 4 dual JFET heatsinks alone cost 30 Euros.
For the small quantities we are making, there are no alternative to wire cut EDM.
Extrusion might have been cheaper, but we need to do 1000kg min !!
Plus 4 PCBs, we shall set the cost at 50 Euros per kit including registered post world wide.
Other courier at extra cost, on request.
Delivery is as usual FOB. We are not responsible for postal losses.
Matched 2SJ74BL's (Idss match only) is limited to 40 pairs on first come first serve basis.
And limited to one pair per person.
See also post #7 for alternative source.
Secondly subscription
I shall open a GB thread around mid September.
By that time you should have had a chance to read the article and decide whether it is worth trying.
As long as we go over 60 kits, I shall start collecting money and then order the PCBs and the heatsinks.
Expect 3-4 weeks for the sinks. The PCBs are usually much quicker.
JFETs are all ready matched.
And we hope you like the results as much as we do. 🙂
Patrick
You guys are really impatient.
Firstly the cost.
4 quad JEFT heatsinks plus 4 dual JFET heatsinks alone cost 30 Euros.
For the small quantities we are making, there are no alternative to wire cut EDM.
Extrusion might have been cheaper, but we need to do 1000kg min !!
Plus 4 PCBs, we shall set the cost at 50 Euros per kit including registered post world wide.
Other courier at extra cost, on request.
Delivery is as usual FOB. We are not responsible for postal losses.
Matched 2SJ74BL's (Idss match only) is limited to 40 pairs on first come first serve basis.
And limited to one pair per person.
See also post #7 for alternative source.
Secondly subscription
I shall open a GB thread around mid September.
By that time you should have had a chance to read the article and decide whether it is worth trying.
As long as we go over 60 kits, I shall start collecting money and then order the PCBs and the heatsinks.
Expect 3-4 weeks for the sinks. The PCBs are usually much quicker.
JFETs are all ready matched.
And we hope you like the results as much as we do. 🙂
Patrick
- Home
- Source & Line
- Digital Line Level
- Zen -> Cen -> Sen, evolution of a minimalistic IV Converter