Hey, i'm new to audio electronics in a sense, i'm currently studying I-V Converters for my first DAC which employs a PCM2707 and PCM1704, similar to that of many designs i've seen around here. I know I could buy someone elses PCB and simply solder the components, but I rather build it from scratch.
I've read that negative feedback op-amps are a bad choice (cant remember why), could someone enlighten me on this?
Personally, i'd for the I-V stage, i'd rather use discrete components such as a BJT, zener etc rather than opamps, just because the study of the internals can be done more comprehensively.
What are desired output characteristics of the voltage gain, snr etc?
And what nth order LPF is applicable? The higher the order (from memory), will reduce the gain, but increase the effectiveness of the LPF.
So therefore, i'm gathering the I-V stage should output a voltage gain with proportion to the order of LPF?
I know theres a few ambiguous questions here, but i'm just looking at peoples views, any help / discussion is welcome and appreciated.
Thanks,
Olmeister
I've read that negative feedback op-amps are a bad choice (cant remember why), could someone enlighten me on this?
Personally, i'd for the I-V stage, i'd rather use discrete components such as a BJT, zener etc rather than opamps, just because the study of the internals can be done more comprehensively.
What are desired output characteristics of the voltage gain, snr etc?
And what nth order LPF is applicable? The higher the order (from memory), will reduce the gain, but increase the effectiveness of the LPF.
So therefore, i'm gathering the I-V stage should output a voltage gain with proportion to the order of LPF?
I know theres a few ambiguous questions here, but i'm just looking at peoples views, any help / discussion is welcome and appreciated.
Thanks,
Olmeister
> theres a few ambiguous questions here
Yes, but basically if you want to build discrete and no feedback, you need a current conveyor with low input impedance to convey the current to a IV converting resistor. Once you get voltage, the filtering is a matter of personal taste, and depends on your source. How much voltage ? 2Vrms is standard. That you can change by simply changing the value of the R_iv.
You can have a look at the SEN circuit or the modified Hawsford IV circuit that I published earlier.
Patrick
Yes, but basically if you want to build discrete and no feedback, you need a current conveyor with low input impedance to convey the current to a IV converting resistor. Once you get voltage, the filtering is a matter of personal taste, and depends on your source. How much voltage ? 2Vrms is standard. That you can change by simply changing the value of the R_iv.
You can have a look at the SEN circuit or the modified Hawsford IV circuit that I published earlier.
Patrick
My take on this is that negative feedback op-amps aren't necessarily a bad choice just a choice to be made with care. They have a problem because their linearity depends on having lots of loop gain available - that's the essence of NFB, to trade gain for linearity.
When exposed to the signal emanating from a DAC though they're normally driven into non-linearity owing to the extremely wide-band nature of this signal. Its in effect an RF signal which is produced by a DAC and needs to be treated as such prior to LP filtering. No engineer in his/her right mind would use an audio opamp for video purposes and yet plenty of designs (even those from the DAC manufacturers themselves
) use audio opamps for I/V. Nowadays plenty of opamps are available that can handle the kind of bandwidths involved but they're not generally touted as audio parts.
When exposed to the signal emanating from a DAC though they're normally driven into non-linearity owing to the extremely wide-band nature of this signal. Its in effect an RF signal which is produced by a DAC and needs to be treated as such prior to LP filtering. No engineer in his/her right mind would use an audio opamp for video purposes and yet plenty of designs (even those from the DAC manufacturers themselves
) use audio opamps for I/V. Nowadays plenty of opamps are available that can handle the kind of bandwidths involved but they're not generally touted as audio parts.
There are any number of interesting threads here regarding the subject of I/V conversion - linked below, are just a very small sampling. Many here feel that the I/V circuit is the single most important aspect of DAC design. I think there is just too much to this subject to provide you with any simple bit of advice. Unless you are willing to immerse yourself in the study of the technology you are likely better off building a kit or an established DIY project, IMHO.
http://www.diyaudio.com/forums/digi...-u-converter-sonic-character.html#post2795940
http://www.diyaudio.com/forums/digi...-opamp-i-v-conversion-dac-31.html#post2813000
http://www.diyaudio.com/forums/digi...-u-converter-sonic-character.html#post2795940
http://www.diyaudio.com/forums/digi...-opamp-i-v-conversion-dac-31.html#post2813000
My take is that every DAC manufacturer and every serious player manufacturer are using OpAmps to do the I/V conversion. There is no substitute for virtual zero impedance that is needed for the best conversion.
Sure, some people always tend to belive that they know better than those "stupid" enginners and recomend all kind of I/V stages. Or that being different is "cool". Truth is that they never back up their claims with any measurements and alwyas hide behind the mantra "measurements can't tell how it sounds". Well, in my experience, what measured bad, sounds bad, no matter what. And if the reality was that trully a resistive I/V sounds better, the maket of hi-end audio would be full of them - because they would sell better. They are cheaper to make, and manufaturers would want to make money from that wonderfull solution. The reality is other...
For the real-life performances of the OpAmps, the newest data comes from ESS that tested a bunch of them:
http://www.esstech.com/PDF/Application_Note_Component_Selection_and_PCB_Layout.pdf
Results:
THD -114dB and DNR -132dB (differential) or -128dB (single-ended)...
That is THD=0.0002% and real DNR of 22bit.
IMO, that's plenty more than what a musical program needs.
Sure, some people always tend to belive that they know better than those "stupid" enginners and recomend all kind of I/V stages. Or that being different is "cool". Truth is that they never back up their claims with any measurements and alwyas hide behind the mantra "measurements can't tell how it sounds". Well, in my experience, what measured bad, sounds bad, no matter what. And if the reality was that trully a resistive I/V sounds better, the maket of hi-end audio would be full of them - because they would sell better. They are cheaper to make, and manufaturers would want to make money from that wonderfull solution. The reality is other...
For the real-life performances of the OpAmps, the newest data comes from ESS that tested a bunch of them:
http://www.esstech.com/PDF/Application_Note_Component_Selection_and_PCB_Layout.pdf
Results:
THD -114dB and DNR -132dB (differential) or -128dB (single-ended)...
That is THD=0.0002% and real DNR of 22bit.
IMO, that's plenty more than what a musical program needs.
Last edited:
Where the word 'serious' means 'serious about getting the best measurements' I would agree with you.
Where 'best' means 'best, measured by traditional THD+N etc.' I also agree.
Here I can't agree - no belief is necessary, just listening. You have no evidence for anyone's beliefs do you?
They have no need to back up listening claims with measurements because its obvious to them that the mantra is self-evident truth.
TI's datasheets (that's a TI DAC you're using) are full of NE5532s or the quad versions. The ESS appnote says that you can only better their DNR and THD performance by 3~4dB, which is probably inaudible (and not that easy to measure without spending real money) and which is a difference even a skilled designer might not avoid without a couple of layout iterations.
Design, draw up and build a DAC by all means, I've got a DIR9001 and a PCM1792 of my own here waiting for me to finish a few other things, but retain a sense of realism about the measurable and audible differences you can expect.
Oh, and try not to get dragged into the NE5532 haters club, it's a rare piece of modern recorded music that hasn't encountered one somewhere along the way. As you will see from the ESS appnote there are several 'audiophile' opamps that perform worse in this application. Why exactly they're hated is not clear, but it might have something to do with their total lack of exclusivity and <25 cent price. Putting them down on the basis of their 'sound' (it's never the measurements) is certainly popular with the kind of personality that likes to go in for a bit of... ...that kind of thing.
Design, draw up and build a DAC by all means, I've got a DIR9001 and a PCM1792 of my own here waiting for me to finish a few other things, but retain a sense of realism about the measurable and audible differences you can expect.
Oh, and try not to get dragged into the NE5532 haters club, it's a rare piece of modern recorded music that hasn't encountered one somewhere along the way. As you will see from the ESS appnote there are several 'audiophile' opamps that perform worse in this application. Why exactly they're hated is not clear, but it might have something to do with their total lack of exclusivity and <25 cent price. Putting them down on the basis of their 'sound' (it's never the measurements) is certainly popular with the kind of personality that likes to go in for a bit of... ...that kind of thing.
@abraxalito - beliving that all other electronics engineers (from TI, NatSemi, Wolfson, CL, Sony, Phillips, Toshiba, ESS, D&M and hundreds of other companies) are incapable to comprehend the wonders of resistors used as I/V is either just plain ignorance or megalomania.
The truth is that it was tried and failed the tests... Since the dawn of digital audio. This a billion dollar industry, if a revelation like this was real, it was used to gain economical advantage.
The truth is that it was tried and failed the tests... Since the dawn of digital audio. This a billion dollar industry, if a revelation like this was real, it was used to gain economical advantage.
Hi,
one can only marvel about SoNics knowledge about design decisions in hundreds of serious companies and his knowledge about what Truth is
With this dogmatic attitude he should maybe apply for the upcoming election of the new Pope
Just to give You something to think about:
What is the function of the resistor in the I/V-OPs feedback loop?
How does the input impedance plot of a OP-I/V look like at say around the output switching frequency?
@olemeister: You can certainly build I/V stages with impressive measurement figures using both techniques, OPs or discrete non-OP stages. The differences in figures are with a high probability just of academic interest and audibly irrelevant. Still though sonic differences exist and there´s a great number of people who prefer/like NON-OP stages sonically.
Give the simple current-conveyor circuits like the Jocko or the SEN/ZEN that EUVL already mentioned, a chance.
The same applies to the usual post filtering. With the typical upsampling DACs post filtering may be reduced or omitted with alltogether. The reduction of 2-3dB in noise-figures may be purely academic, but the absence of mainly obsolete parts within the signal path makes a difference in sonics.
jauu
Calvin
one can only marvel about SoNics knowledge about design decisions in hundreds of serious companies and his knowledge about what Truth is
With this dogmatic attitude he should maybe apply for the upcoming election of the new Pope
Just to give You something to think about:
What is the function of the resistor in the I/V-OPs feedback loop?
How does the input impedance plot of a OP-I/V look like at say around the output switching frequency?
@olemeister: You can certainly build I/V stages with impressive measurement figures using both techniques, OPs or discrete non-OP stages. The differences in figures are with a high probability just of academic interest and audibly irrelevant. Still though sonic differences exist and there´s a great number of people who prefer/like NON-OP stages sonically.
Give the simple current-conveyor circuits like the Jocko or the SEN/ZEN that EUVL already mentioned, a chance.
The same applies to the usual post filtering. With the typical upsampling DACs post filtering may be reduced or omitted with alltogether. The reduction of 2-3dB in noise-figures may be purely academic, but the absence of mainly obsolete parts within the signal path makes a difference in sonics.
jauu
Calvin
Last edited:
Current output DACs (i.e. most of them) are intended to drive into a zero resistance, so next to no voltage is generated. An inverting opamp provides this quite easily. The snag is that some opamps allegedly can't cope too well with the pulse transitions which accompany the audio. Some people claim this leads to distortion, although it is the method used in most devices.
The alternative is to use a real resistor - so-called passive I/V. The snag now is that you either need a very small resistor (less than 25 ohms?) and get some noise from the following preamp, or use a big resistor and suffer distortion (and, perhaps, clipping) from the DAC output stage (as it is not designed to produce voltage). So which do you prefer - noise or distortion? I suspect that some people who prefer passive I/V do so because they prefer mild distortion, because many published designs push the DAC well into its distortion region.
The ideal solution is probably an inverting opamp, but preceded by a diplexer to send the pulse transitions to a low value resistor. Alternatively a well-designed valve stage (grounded grid?) could cope with the pulses but may add a little second-order distortion.
PS the input impedance of an inverting op amp looks something like a low value inductor, with resistors in series and parallel.
The alternative is to use a real resistor - so-called passive I/V. The snag now is that you either need a very small resistor (less than 25 ohms?) and get some noise from the following preamp, or use a big resistor and suffer distortion (and, perhaps, clipping) from the DAC output stage (as it is not designed to produce voltage). So which do you prefer - noise or distortion? I suspect that some people who prefer passive I/V do so because they prefer mild distortion, because many published designs push the DAC well into its distortion region.
The ideal solution is probably an inverting opamp, but preceded by a diplexer to send the pulse transitions to a low value resistor. Alternatively a well-designed valve stage (grounded grid?) could cope with the pulses but may add a little second-order distortion.
PS the input impedance of an inverting op amp looks something like a low value inductor, with resistors in series and parallel.
Last edited:
Interesting to check these distortion measurements with respect to I/V resistor values:
Digi Scoop 1
Yes, such distortion measurements are interesting, but not necessarily determinative of perceived sound quality by themselves. Else, we would all build or purchase only high-feedback solid-state amplification components.
Last edited:
I believe that the 1704 compliance is limited by the output protection diodes becoming forward-biased, however, distortion starts to become objectionable before then. Even staying well within the protection diode conduction thresholds, the 1704 attempts to servo any D.C. output voltage to zero. In addition, it requires a large external capacitor to set the servo corner frequency low enough to not interfere with the A.C. signal. To my thinking, this is THE key issue surrounding the 1704 I/V implementation.
Passive I/V with the PCM1704 will best involve a low value resistor (perhaps, less than 10 ohms) and a commensurately high gain active output stage - essentially, a phono amplifier sans the RIAA equalizer.
Last edited:
I've recently tried AD603 with +/- 50mV compliance voltage (32dB gain) on the TDA1543 DAC with pleasing sonic results. It has the bonus of an in-built volume control too.
http://www.analog.com/static/imported-files/data_sheets/AD603.pdf
Yes, distortion visible to the naked eye in a time-domain plot is severe. Severe distortion is objectionable, no debate from me on that. The debate begins for THD beneath a gross level. My experience is that increasingly lower THD measurements alone often do not correlate with improvements in perceived sound quality. In fact, there often seems to be a negative correlation between perceived sound quality and vanishingly low THD. This phenomena is not only apparent with DACs, but has been apparent with amplification since the Japanese solid-state receiver THD spec. war of the late 70's early 80's.
Last edited:
I've moved on to balanced I/V stages now for which the AD605 is more suitable:
http://www.analog.com/static/imported-files/data_sheets/AD605.pdf
Owing to the limited (5V) supply voltage it can't deliver the standard CD 2VRMS in unbalanced.
http://www.analog.com/static/imported-files/data_sheets/AD605.pdf
Owing to the limited (5V) supply voltage it can't deliver the standard CD 2VRMS in unbalanced.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.