[00940]

Quote:

Originally Posted by cetoole /img/forum/go_quote.gif Yeah, up to about 2MHz. Havnt been able to keep it from going inductive above that, except by adding capacitance to the input, which will increase distortion. Maybe with different transistors than I have tried it would work nicely.

You're right. This cfp thing goes completly crazy at high frequencies. I've tried a few transistors in ltspice and nothing changes much.

This being said... The digital filter inside the pcm1798 achieves a 120dB attenuation at 4fs. For 44.1KHz, wouldn't it mean that there isn't too much junk at 2MHz ?

Here's what I'm having right now in ltspice:

 

[Cauhtemoc]

Quote:

Originally Posted by 00940 You're right. This cfp thing goes completly crazy at high frequencies. I've tried a few transistors in ltspice and nothing changes much. This being said... The digital filter inside the pcm1798 achieves a 120dB attenuation at 4fs. For 44.1KHz, wouldn't it mean that there isn't too much junk at 2MHz ?

Sigma-delta DACs rely on extreme noise shaping and so have a lot of noise past 100 kHz, this can cause problems if the input impedance change.

R2R DACs don't have this problem, but they have very fast settling times, in the order of 0.1 us, which again causes problems. Feed it with a pulse with a rise time of 0.1 us and watch the input and you will see what I mean.

Quote:

Originally Posted by 00940 Here's what I'm having right now in ltspice:

You should use a non-ideal current source to simulate a real DAC, a voltage source with a 1k resistor is a good start.

 

[Russ White]

Quote:

Originally Posted by Cauhtemoc /img/forum/go_quote.gif Alright, so it's a super symmetric balanced opamp. The common mode control loop, I noticed. I also noticed that you do not have a current mirror here, you should seriously consider adding one. The common mode control loop plays a crucial role in minimizing THD when using only one output (referenced to ground instead of differentially between the two outputs). Ah, but then distortion isn't everything. The design posted by cetoole above has an input impedance of 2.5 ohms, and meaning that the output of a PCM1794 swings some 10 mV. This is hardly enough to generate noticable distortion.

I think you are looking at my first design, my second can be found here:

Notice there are definitely current mirrors for the corrector.

diyAudio Forums - Discrete Super Symmetric(I think) Opamp for I/V Etc...

Now here is the main point. With my I/V stage the voltage at the input is only... 300uv at full scale output from DC to 1mhz. thats one third an mv.

and yes I think that can make a difference.

To me, while not everything distortion is pretty critical.

I will experiment with using a resistor in series with a voltage as a current source. I have just never done it that way.

Cheers!
Russ

 

[Russ White]

Ok I changed my simulation to module the current in as a voltage in series with a resistor. I used the PCM1794A DS to calculate a series resistor of about 362 ohms.

Simulated as such I actually got better numbers then I had before.

And this into a 300ohm load with 400pf of capacitance in parallel.

Really I wanted my amp to be able to drive headphones etc directly.

Cheers!

Fourier components of V(out+,out-)
DC component:5.79172e-008

HarmonicFrequency Fourier Normalized Phase Normalized
Number [Hz] Component Component[degree]Phase [deg]
1 1.000e+033.687e+001.000e+00 -0.02° 0.00°
2 2.000e+031.150e-073.118e-08 90.06° 90.08°
3 3.000e+031.328e-073.602e-08 73.27° 73.29°
4 4.000e+031.144e-073.102e-08 89.92° 89.94°
5 5.000e+031.165e-073.160e-08 96.83° 96.85°
6 6.000e+031.163e-073.155e-08 90.29° 90.31°
7 7.000e+031.156e-073.135e-08 84.15° 84.17°
8 8.000e+031.147e-073.110e-08 89.76° 89.78°
9 9.000e+031.147e-073.110e-08 88.52° 88.54°
Total Harmonic Distortion: 0.000009%


.step f=10000
Fourier components of V(out+,out-)
DC component:1.31338e-009

HarmonicFrequency Fourier Normalized Phase Normalized
Number [Hz] Component Component[degree]Phase [deg]
1 1.000e+043.687e+001.000e+00 -0.21° 0.00°
2 2.000e+043.152e-098.548e-10 -56.24° -56.04°
3 3.000e+041.294e-073.509e-08 84.26° 84.46°
4 4.000e+044.319e-101.171e-10 115.13° 115.34°
5 5.000e+041.128e-083.060e-09 42.66° 42.86°
6 6.000e+043.642e-099.879e-10 10.91° 11.12°
7 7.000e+041.724e-094.676e-10 -115.59° -115.39°
8 8.000e+043.552e-099.634e-10 26.33° 26.54°
9 9.000e+042.603e-097.059e-10 119.28° 119.48°
Total Harmonic Distortion: 0.000004%

.step f=20000
Fourier components of V(out+,out-)
DC component:-7.28242e-010

HarmonicFrequency Fourier Normalized Phase Normalized
Number [Hz] Component Component[degree]Phase [deg]
1 2.000e+043.687e+001.000e+00 -0.41° 0.00°
2 4.000e+041.454e-083.943e-09 -20.90° -20.49°
3 6.000e+042.551e-076.920e-08 96.55° 96.96°
4 8.000e+042.279e-086.182e-09 -4.74° -4.33°
5 1.000e+052.984e-088.093e-09 -5.85° -5.44°
6 1.200e+053.353e-089.094e-09 -2.14° -1.72°
7 1.400e+054.591e-081.245e-08 2.55° 2.96°
8 1.600e+054.810e-081.305e-08 -6.98° -6.57°
9 1.800e+055.488e-081.489e-08 -2.66° -2.25°
Total Harmonic Distortion: 0.000007%

 

[00940]

Input impedance down to 1.75 ohm. But the 2sa970 is hitting the limit of its linear range. There is 16ma and 9ma iddle accross the 2sc2240/2sa970.

 

[00940]

I just saw this... amazingly low input impedance (less than 0.5R). Have a look at the last schematic.

homecinema-fr.com : Voir le sujet - Etage IV sans condo de liaison , sommation de sortie de dac

 

[cetoole]

Quote:

Originally Posted by 00940 /img/forum/go_quote.gif I just saw this... amazingly low input impedance (less than 0.5R). Have a look at the last schematic. homecinema-fr.com : Voir le sujet - Etage IV sans condo de liaison , sommation de sortie de dac

Yeah, via a feedback mechanism, a la Hawksford. Have you guys seen his paper Current Steering Transimpedance Amplifiers for High-Resolution Digital-to-Analogue Converters? It uses the same current mirror feedback to the bias transistor as shown in that schematic. Good, but also has the problematic rising impedance at HF.

Here is a simulation showing the input impedance for my I/V stage using CFPs. Hard to tell, but the impedance is 260mohm, with the first rising step at 2.37MHz. Also, Pars was correct, I am using a voltage source followed by a series 2.1kohm resistor to simulate the nonideal nature of the DAC current output. I am designing for use with the AD1862N-J DAC, of which I have four.

 

[Pars]

Quote:

Originally Posted by 00940 /img/forum/go_quote.gif Input impedance down to 1.75 ohm. But the 2sa970 is hitting the limit of its linear range. There is 16ma and 9ma iddle accross the 2sc2240/2sa970.

Yeah, when building these, when you get up around 10mA the SA970/SC2240s get touchy. I had to reduce the current in both rbroer's and cetoole's stages to get them stable at all with these transistors. Small air currents, etc. would cause alot of drift.

This is what I wound up with for the rbroer stage:



Quote:

Originally Posted by 00940 /img/forum/go_quote.gif I just saw this... amazingly low input impedance (less than 0.5R). Have a look at the last schematic. homecinema-fr.com : Voir le sujet - Etage IV sans condo de liaison , sommation de sortie de dac

Interesting... I assume you mean the last one on the page?

 

[00940]

Pars : yes, the last one. Combine it with this one :

And you get something quite clever perfectly suited to balanced out DAC (in mono mode). The only thing I don't like is that the simulation shows dissimilar input impedances with the 2sa970-2sc2240. 0.02R (!) vs 0.1R @20KHz. At 2Mhz, we're back to 3R for both inputs.

Another good thing in that design is that you get offset nulling at both inputs and output with only three trimmers. Regal will be happy.

No need for balanced to unbalanced conversion.


Cetoole: yes, impedance rises with frequency too.

If I might ask, what command do you use in ltspice to plot the input impedance vs frequency ?

 

[cetoole]

00940, as shown in my simulations, though difficult to see, I use ".ac oct 100 1 100000000", and then to get it to show impedance instead of voltage, just divide the voltage at the input by the current through the resistor.

 

[Russ White]

Ok if I measure input impedance at each input, this is what I get:

Yes, that's 38 mOhms at 20Khz. And it stays pretty flat up through 5Mhz.

Cheers!
Russ

 

[Rescue Toaster]

Quote:

Originally Posted by cetoole /img/forum/go_quote.gif 00940, as shown in my simulations, though difficult to see, I use ".ac oct 100 1 100000000", and then to get it to show impedance instead of voltage, just divide the voltage at the input by the current through the resistor.

cteoole, what is the limiting factor for max impedance at high freqs? Hitting 5 ohms at ~105 Mhz isn't bad or anything I'm just wondering what is the dominant effect above 20Mhz or so. And how much AC current do you experience in the supply at that range?

 

[Cauhtemoc]

Quote:

Originally Posted by Russ White I think you are looking at my first design, my second can be found here: Notice there are definitely current mirrors for the corrector. diyAudio Forums - Discrete Super Symmetric(I think) Opamp for I/V Etc...

Yes, my apologies.

Quote:

Originally Posted by Russ White Ok I changed my simulation to module the current in as a voltage in series with a resistor. I used the PCM1794A DS to calculate a series resistor of about 362 ohms. Simulated as such I actually got better numbers then I had before. And this into a 300ohm load with 400pf of capacitance in parallel. Really I wanted my amp to be able to drive headphones etc directly. Cheers!

Just to prove that I also can play the feedback game, I took the famous JE-990 discrete opamp and put it to I/V conversion duty. The result was some 0.000001% THD.

The JE-990 uses some 10 transistors. How many do you use?

 

[Russ White]

Cauhtemoc. No apologies required.

I am not playing any games at all, feedback or otherwise. I am Only trying to explain why I like I/V stages that are low distortion.

And yes if I omit the filter caps in two spots (which does not effect the amp's stability, just its linearity and bandwidth) I can achieve a zero distortion number at 20K too. Thats really not the point.

I have a very similar single ended version of this opamp which also uses only 10Qs and gets the same THD numbers.

The only difference between you and I is I have shown my schematic for critical review.

Cheers!
Russ

 

[Russ White]

For example, here is 2V RMS at 20Khz into same load with no filter caps.

Fourier components of V(out+,out-)
DC component:-7.26524e-014

HarmonicFrequency Fourier Normalized Phase Normalized
Number [Hz] Component Component[degree]Phase [deg]
1 2.000e+049.945e-011.000e+00 -0.05° 0.00°
2 4.000e+043.141e-123.159e-12 -2.68° -2.63°
3 6.000e+043.296e-093.314e-09 74.97° 75.02°
4 8.000e+046.279e-126.314e-12 -1.26° -1.21°
5 1.000e+055.622e-105.653e-10 105.90° 105.95°
6 1.200e+059.416e-129.469e-12 -0.75° -0.70°
7 1.400e+053.094e-103.111e-10 -151.45° -151.40°
8 1.600e+051.255e-111.262e-11 -0.48° -0.43°
9 1.800e+053.362e-103.380e-10 22.37° 22.42°
Total Harmonic Distortion: 0.000000%