diyAudio

diyAudio (http://www.diyaudio.com/forums/)
-   Digital Source (http://www.diyaudio.com/forums/digital-source/)
-   -   The pcm1794a datasheet I/V converter and how to improve it. (http://www.diyaudio.com/forums/digital-source/218244-pcm1794a-datasheet-i-v-converter-how-improve.html)

smms73 20th August 2012 12:47 PM

The pcm1794a datasheet I/V converter and how to improve it.
 
1 Attachment(s)
The I/V converter in pcm1794a datasheet is one good example of how you should not use op ampīs, and is because circuits like this that op ampīs have such a bad reputation.
Op amps normally have low quiescent current, and should use large load impedances if one want that they stay in "class a" operation (low distortion).
The circuit in pcm1794a datasheet use very low output impedances in the load of op ampīs and because of that, they work in "class b" (high distortion).

This circuit has 0.003% distortion at 10Khz

Ken Newton 20th August 2012 12:55 PM

I concur. It seems they recommend such low resistances to minimize Johnson noise. Such noise is likely the limiting factor in approaching 24-bit static resolution from the converter. Personally, I'd choose higher noise over higher distortion.

smms73 20th August 2012 01:04 PM

1 Attachment(s)
One way to remedy this is using a buffer at each op amp.
below is the same circuit but with the addition of 3 buffers, now each op amp sees a much higher load impedance and the buffers work in class a.
The distortion is much lower. Now we have a distortion of only 0.00007% at 10Khz, consisting only of 2š and 3š harmonics, the distortion is even lower at 1Khz (0.000005%).
This is a big improvement over the original circuit.

This simple buffers have been shown to me by Xslavic at post 20 of
http://www.diyaudio.com/forums/digit...istortion.html

smms73 20th August 2012 01:16 PM

Quote:

Originally Posted by Ken Newton (Post 3133417)
I concur. It seems they recommend such low resistances to minimize Johnson noise. Such noise is likely the limiting factor in approaching 24-bit static resolution from the converter. Personally, I'd choose higher noise over higher distortion.

I think that with the buffers we can have both low distortion and low noise.
what you guys think? and what is your experience.

dirkwright 20th August 2012 01:51 PM

I learned something. thanks. Is the noise the stuff @ -200dB in your FFT for this circuit?

I suppose an IC buffer would be OK also?

smms73 20th August 2012 02:37 PM

Quote:

Originally Posted by dirkwright (Post 3133466)
I learned something. thanks. Is the noise the stuff @ -200dB in your FFT for this circuit?

I suppose an IC buffer would be OK also?

No Dirk, unfortunately the noise floor will be higher. LTspice donīt simulate noise in transient analysis. You need to run a Noise simulation for that, so just ignore this -200db

the U1 op amp has a 21ma pp swing, so even a ic buffer may enter class b, and is more expensive.

Ken Newton 20th August 2012 03:06 PM

Quote:

Originally Posted by smms73 (Post 3133437)
I think that with the buffers we can have both low distortion and low noise.
what you guys think? and what is your experience.

This is an interesting topic. From a overall systems perspective, and from my own emprical experience, I've come to conclude that Johnson noise is not much of a listening enjoyment impediment for any hifi system with audiophile pretensions. Consider vinyl LP. Many LPs have only 70dB to 80dB of SNR, yet few audiophiles seem to find that this significantly interferes with their listening enjoyment.

I feel that the Johnson noise (not the quantization noise, or dither noise) of a digital audio system, what with it's potential for well over 90dB SNR, is even less of an listening impediment. Truly random noise seems to recede from our concious awareness, so long as more important subjective factors are are being properly handled. For example, I don't feel that quantized resolution is anywhere near the most pressing problem for digital audio sound quality which the faults of the CD medium have led people to conclude. I feel that the main subjective problems of digital audio are time-domain related.

Those low value resistors in the datasheet application examples can make a difference in a laboratory measurement of static resolution for what is marketed as a 24-bit converter. However, for the reproduction of dynamic A.C. signals such as music, which were recorded in a natural ambient noise floor environment, and using microphones and other fromt-end electronics, taking pains to objectively approach 24-bit noise levels might be counter-productive to more important subjective factors.

dirkwright 20th August 2012 03:07 PM

Quote:

Originally Posted by Ken Newton (Post 3133526)
This is an interesting topic. From a overall systems perspective, and from my own emprical experience, I've come to conclude that Johnson noise is not much of a listening enjoyment impediment for any hifi system with audiophile pretensions. Consider vinyl LP. Many LPs have only 70dB to 80dB of SNR, yet few audiophiles seem to find that this significantly interferes with their listening enjoyment.

I feel that the Johnson noise (not the quantization noise, or dither noise) of a digital audio system, what with it's potential for well over 90dB SNR, is even less of an listening impediment. Truly random noise seems to recede from our concious awareness as long as more important subjective factors are are being properly handled. I don't feel that quantized resolution is anywhere near the most pressing problem for digital audio sound quality. I feel it's main subjective problems are time-domain related.

Well, I suffer from tinnitus so I know all about ignoring noise.

dirkwright 20th August 2012 03:13 PM

Quote:

Originally Posted by smms73 (Post 3133503)
No Dirk, unfortunately the noise floor will be higher. LTspice donīt simulate noise in transient analysis. You need to run a Noise simulation for that, so just ignore this -200db

the U1 op amp has a 21ma pp swing, so even a ic buffer may enter class b, and is more expensive.

How about a trio of LME49600? :eek:

There's always a bigger hammer.

Cost shouldn't be a consideration in my opinion because a power supply and a nice box will be far more expensive than a few chips.

Ken Newton 20th August 2012 03:29 PM

Quote:

Originally Posted by dirkwright (Post 3133535)
How about a trio of LME49600? :eek:

There's always a bigger hammer.

Cost shouldn't be a consideration in my opinion because a power supply and a nice box will be far more expensive than a few chips.

Three op-amps in parallel could reduce Johnson noise, but by no more than 9.5dB relative to a single such op-amp. However, you would need to isolate the low impedance output of each op-amp from that of the others with a resistor, so not the have them buck each other, which would increase distortion. Not what we want. Output isolation resistors will contribute their own Johnson noise, reducing that theoretical 9.5dB figure.


All times are GMT. The time now is 06:02 AM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright Đ1999-2014 diyAudio


Content Relevant URLs by vBSEO 3.3.2