[Cauhtemoc]

Quote:

Originally Posted by Russ White I am not playing any games at all, feedback or otherwise.

Of course not. My point was that designs with feedback will always achieve very good THD figures, but as I said before, THD isn't everything. There is a reason for why Wadia uses a non feedback I/V converter in their $10'000 581i CD player.

 

[00940]

Of course 0.00000000000000000000000 sounds more impressive than :

HarmonicFrequency Fourier Normalized Phase Normalized
Number [Hz] Component Component[degree]Phase [deg]
1 2.000e+041.451e+001.000e+00 161.09° 0.00°
2 4.000e+041.861e-041.282e-04 59.57° -101.53°
3 6.000e+045.511e-063.797e-06 135.82° -25.27°
4 8.000e+041.955e-071.347e-07 -142.05° -303.14°
5 1.000e+051.474e-081.016e-08 -76.23° -237.32°
6 1.200e+051.418e-099.772e-10 -73.68° -234.78°
7 1.400e+052.510e-091.729e-09 29.32° -131.78°
8 1.600e+058.599e-105.924e-10 -168.82° -329.91°
9 1.800e+052.275e-091.567e-09 -156.77° -317.87°

Total Harmonic Distortion: 0.012827% (it's for the I/V I posted above)

or

HarmonicFrequency Fourier Normalized Phase Normalized
Number [Hz] Component Component[degree]Phase [deg]
1 2.000e+042.951e+001.000e+00 -7.27° 0.00°
2 4.000e+046.286e-052.131e-05 -167.13° -159.86°
3 6.000e+047.141e-052.420e-05 0.03° 7.30°
4 8.000e+041.243e-064.214e-07 8.25° 15.52°
5 1.000e+055.357e-071.816e-07 -141.65° -134.38°
6 1.200e+052.703e-089.162e-09 148.32° 155.59°
7 1.400e+058.763e-092.970e-09 25.50° 32.77°
8 1.600e+056.814e-102.310e-10 -60.95° -53.68°
9 1.800e+053.674e-091.245e-09 102.98° 110.25°

Total Harmonic Distortion: 0.003225% (for the stuff with current mirror found on homecinema-fr.com : Source de Passion)

Both at 2Vrms, 20khz of course

 

[Russ White]

The bottom line is, how good does it sound? And this is always between any two decent designs 100% subjective

I have not built my circuit yet so I don't know. But it seems interesting.

I have nothing against no feedback designs. If you think they sound nice, then use them. I have and do!!

Certainly there is a niche there, so perhaps I could make one just to please the masses.

I just happen to believe that a very linear feedback design can sound better! That's just my personal observation.

I guess part of me thinks that if your going to use a DAC capable of .0004% THD, you should at least "try" to create an output stage that can match it.

I know in reality it is not likely that a simulated circuit will live up to its promises when you get to building it. But hey its all good fun.

Also I have known for quite a long time now that the cost of a component sometimes has nothing to do with its performance. Its all about perception.

There is no good or bad in that. Things go in and out of fashon all the time.

Cheers!
Russ

 

[Cauhtemoc]

Try this on for size:

THD is 0.0006% mainly second harmonic, input impedance is 0.003 ohm.

Long story short, I was playing with current conveyors and came up with the above. It has low enough THD and input impedance to sway even the most stuborn feedback user (like Russ above!

).

I mentioned Wadia earlier, this is something along the lines of what they use.

 

[00940]

This is very impressive. Out of curiosity, how does the input impedance change wrt frequency ?

With this more common stage, I'm now at around 0.008% (real life won't be as good of course...).

 

[Russ White]

Quote:

Originally Posted by Cauhtemoc /img/forum/go_quote.gif Try this on for size: THD is 0.0006% mainly second harmonic, input impedance is 0.003 ohm. Long story short, I was playing with current conveyors and came up with the above. It has low enough THD and input impedance to sway even the most stuborn feedback user (like Russ above! ). I mentioned Wadia earlier, this is something along the lines of what they use.

Now that is very interesting!

I actually have looked at the current conveyor before too. But not for a while. It does seem promising. Why the choice of Q types?

Now just to clear the air, its not that I would only ever choose to use global feedback excluding all else, its just that I am not adverse to using it either rhetorically or electrically.

I am learning quite a lot. So thanks for the healthy and respectful conversation.

Cheers!
Russ

 

[Cauhtemoc]

Quote:

Originally Posted by 00940 This is very impressive. Out of curiosity, how does the input impedance change wrt frequency ?

The input impedance does rise with frequency, up to about 4.5 ohm at 100 MHz before falling of again. It uses feedback, albeit not global feedback, so this is to be expected.

 

[Cauhtemoc]

Quote:

Originally Posted by Russ White Now that is very interesting! I actually have looked at the current conveyor before too. But not for a while. It does seem promising.

Feel free to build it and compare to your balanced opamp converter.

Quote:

Originally Posted by Russ White Why the choice of Q types?

The 2SC2240/2SA970 is a very low noise general purpose transistor with a very linear hfe in the region of 1 mA to 10 mA. I use it for many of my designs, as do a lot of people. You can use other small signal transistors here if you want, performance just won't be as good. Some examples would be BC550/BC560 or 2N5087/2N5089.

 

[regal]

Quote:

Originally Posted by Russ White /img/forum/go_quote.gif Proper.... hmmm what exactly is "proper" My design also has orders of magnitude lower distortion than simple passive I/V (resistor or transformer) + buffer approach (like the zapfilter). Cheers! Russ

I looked at the zapfilter and the best I can tell is it is active I/V. Unless they tell you after you buy it that you have to add a resistor to ground? Or am I missing someting?

Yours should be much better since it is configurable to the PCM1794. But how is the performance single ended? Does it have the common rejection that a balanced to unbalanced IC opamp conversion would have?

 

[Russ White]

I was not really trying to say it was not active. I was simply pointing out the difference between it and a feedback type approach.

I actually think its a clever design.

Cheers!
Russ

 

[regal]

Quote:

Originally Posted by Russ White /img/forum/go_quote.gif I was not really trying to say it was not active. I was simply pointing out the difference between it and a feedback type approach. I actually think its a clever design. Cheers! Russ

Jeff,

Have you considered adding something like this to a balance to single transfer (minus the filter) ?


http://home.quicknet.nl/qn/prive/ra....tstage_PCB.pdf

 

[00940]

Quote:

Originally Posted by Cauhtemoc /img/forum/go_quote.gif The input impedance does rise with frequency, up to about 4.5 ohm at 100 MHz before falling of again. It uses feedback, albeit not global feedback, so this is to be expected.

It's still a very smooth curve. it's not worse that a cfp or a folded cascode with high current transistor.

To balance the possible input offset of a dac, I however had to play with the value of R6, which reduces the current going through the stage and slightly worsens the distortion figures.

 

[Cauhtemoc]

R6 is set to allow about 10 mA through the Q9. If you go below this the distortion increases, above this and the 2SC2240/2SA970 will start to loose their linearity.

You should adjust R5 to get 10 mA through Q6 including the DAC offset. For example, if you are using the TDA1541, which has 2 mA offset, you should set R5 to provide 8 mA. This gives a total of 10 mA through Q6 (8 mA from the current source plus 2 mA from the DAC). This is the same as the current provided by R6, so no current goes through the output resistor R7, and thus there is output voltage.

Adjusting this way means that there will always be 10 mA through the stage, which will help to keep distortion down.

 

[Pars]

Quote:

Originally Posted by regal /img/forum/go_quote.gif I looked at the zapfilter and the best I can tell is it is active I/V. Unless they tell you after you buy it that you have to add a resistor to ground? Or am I missing someting?

I would say the 4.02K resistors and the 470pf caps are the I/V resistor(s) and low pass filter in this design. I'd have to look at their instructions (also online, IIRC), but the solder bubbles at the input (1-6) are how you configure this for current out or voltage out DACs and single ended or balanced input. I'm guessing that solder bubble 3 is the one to bridge for current in (I/V) single-ended stage, and 3 and 5 for balanced I-in.

 

[cetoole]

Quote:

Originally Posted by Pars /img/forum/go_quote.gif I would say the 4.02K resistors and the 470pf caps are the I/V resistor(s) and low pass filter in this design. I'd have to look at their instructions (also online, IIRC), but the solder bubbles at the input (1-6) are how you configure this for current out or voltage out DACs and single ended or balanced input. I'm guessing that solder bubble 3 is the one to bridge for current in (I/V) single-ended stage, and 3 and 5 for balanced I-in.

I would think I/V is actually performed by the two 300ohm resistors going to v- through that zener diode.