Testing the pcm1794

[smms73]

Since the beginning of my quest to design a good dac around the pcm1794, i had some doubts about some characteristics of this chip.
And it seems that is a lot of contradictory info in this forum about them. some of my doubts are the output impedance, the distortion and noise of using the dac in single end VS differential, and the influence of the input impedance of the I/V converter.

hope that with your help we can shed some light on the subject.

This tests were made with pcm1794 and dir9001 in a breadboard.
I use the line input and spdif out of the sound card (emu1212m), for distortion measurements.

[smms73]

this first test is about the distortion level caused by the impedance at the output of the pcm1794. There is the widespread idea that the lower the output impedance seen by the dac the better. but how low is enough?

this test was made by connecting the pin IoutL+ to a classical ampop I/V converter (opa2134) with a resistor of 50 , 100, and 150 ohms in serie.

the pin IoutL- was connected to ground.

the first picture is with no resistence in serie, the 2º is with 100 ohms (r1) , the 3º is with 150 ohms. the picture with 50 ohms is ommited because there is no visible difference between the one that has no serie resistence.

[smms73]

this test was made by connecting the pin IoutL+ with a resistor of 50 ohms to ground and measure the distortion at the pin IoutL+.

The pin IoutL- was connected to ground.

The distortion and noise are real low. (0,0005%) the output level of the dac was -1 dbFS.

[smms73]

One thing that is important in the case you opt by using this type of I/V converter is to choose a high quality capacitor for C1 (polipropilene or C0G ceramic). this pictures are the distortion of using a polipropilene (MKP) capacitor or a bad quality capacitor.

this test use a differencial I/V converter.

[smms73]

summarizing:

There is no problem using a I/V converter with a input impedance of 50 ohms or lower.
There is not a great difference between differencial or single end.
Ween using a passive resistence as a I/V converter connect the IoutL- and IoutR- pins to the ground.

I will let you digest this information for now.
If you have any doubts or questions please feel free to ask.

[Ken Newton]

Sergio, thanks for providing this battery of very interesting test results. Your 150R I/V THD figure of 0.0022% exactly matches my own test result with 150R , except that my DAC's output is completely passive. I don't utilize any gain stage for normalizing signal levels to 2VRMS prior to sending them to the linestage unit for any needed gain.

What I/V resistance value was utilized in the differential mode test?

[smms73]

Thank Ken, there is even more.

completely passive !!! how ? do you use a transformer ?

In post #4 i use 1k ohm , this post is just to warn about the use of bad quality capacitor in the feedback path, that can have a huge impact in the linearity.

the best results in balanced mode was 0.00022%, that is also the limit of the emu1212m.
But breadboard is not the most perfect environment for sensible audio equipment ( a good ground is very important to keep the distortion and noise values low).

In this test i use 510 ohms and no capacitor.

[smms73]

One important value is the output impedance of the PCM1794.

Some say it is 2k ohms, others say is less than 1K ohms, Ti says is 2 Mega ohms.
By my experience it is more than 100K ohms .

I have use the following circuit to come up to this result.

first i have use a 10k in "Rx", and measure the output with the switch close and then with the switch open, then i repeat the same procedure with 20k , 100k and at last without Rx. the results are this:

switch - close - open

10k - 21.16db - 21.33db
20k - 21.16db - 21.24db
100k - 21.16db - 21.18db
Zero - 21.16db - 21.16db

As there is no diference in output voltage when there is no resistence in parallel with the output of the dac, i assume the impedance is greater than 100k , because when we use 100k, we can sense a difference betwen the switch is close and when is open.

I have try this with -1db , -21db , -43db ,always with the same result.

It is very difficult to measure better than this, because of the temperature and noise. the 2 Mega ohms that Texas presents is plausible but i can not measure that.

[smms73]

This is the output voltage of the PCM1794 without filter and with a first-order filter.
The cutoff frequency is 72Khz, its a litle bit low , but this is only for demostration.

[smms73]

Forgot to say that all the distortion test until now, are made with -1db of PCM1794 full scale and 44.1Khz , 24bit.