Hi abraxalito,
Pretty much agree with you.
Hi SoNic_real_one,
I have worked with the TDA1541A enough. It's just okay, nothing to write home about. The best sounding D/A chips I have heard so far are the "co-linear" ones from BB. I have a Denon DCD-S10 using 4x PCM1702 chips and the 20 bit digital filter. That's the nicest sounding digital source I have heard so far. Having said that, I also have a Nakamichi OMS-7, ancient machine. It sounds very good using a 14 bit Philips chip set (and NEC everything else, incl laser head), the TDA1540 x 2. The secret in getting the best performance seems to be in supplying a clean eye pattern to the demod and error correction circuits, and in having nothing short of excellence in the audio path outside the D/A. Of course, we are assuming a nice clean power supply for everything too.
With a current output D/A, the game has shifted as abraxalito pointed out. You now have to pay attention to the voltage compliance of the D/A chip, and also where it begins to become unhappy. The current type output helps with stray capacitance, so start thinking about current flow. A virtual earth configuration (inverting stage) would be my first choice. You'll need to pay attention to the current noise of your op amp as well as it's slew rate. CMRR plays a minor role here since the inputs are at ground potentials. There shouldn't be any type large signal on either input pin. Another consideration would be how clean a potential op amp sounds, low distortion, low noise and very good step response with minimal ringing. Circuit layout and component types may overshadow your choice of op amp. Be ready to figure out what might be at fault if you are unhappy with the sound quality.
-Chris
Pretty much agree with you.
Hi SoNic_real_one,
I have worked with the TDA1541A enough. It's just okay, nothing to write home about. The best sounding D/A chips I have heard so far are the "co-linear" ones from BB. I have a Denon DCD-S10 using 4x PCM1702 chips and the 20 bit digital filter. That's the nicest sounding digital source I have heard so far. Having said that, I also have a Nakamichi OMS-7, ancient machine. It sounds very good using a 14 bit Philips chip set (and NEC everything else, incl laser head), the TDA1540 x 2. The secret in getting the best performance seems to be in supplying a clean eye pattern to the demod and error correction circuits, and in having nothing short of excellence in the audio path outside the D/A. Of course, we are assuming a nice clean power supply for everything too.
With a current output D/A, the game has shifted as abraxalito pointed out. You now have to pay attention to the voltage compliance of the D/A chip, and also where it begins to become unhappy. The current type output helps with stray capacitance, so start thinking about current flow. A virtual earth configuration (inverting stage) would be my first choice. You'll need to pay attention to the current noise of your op amp as well as it's slew rate. CMRR plays a minor role here since the inputs are at ground potentials. There shouldn't be any type large signal on either input pin. Another consideration would be how clean a potential op amp sounds, low distortion, low noise and very good step response with minimal ringing. Circuit layout and component types may overshadow your choice of op amp. Be ready to figure out what might be at fault if you are unhappy with the sound quality.
-Chris
Oh, I would like to hear that DCD-S10 
I hope you got rid of those lame NJM2041 - the "Asia" model came with OP275 as I/V!
I see that it has the same Alpha filter - SM5845 - like my DCD-360. The only issue is that in my schematics is shown as either 5845 or 5848 and on board is SM5845. None of them have public datasheets (Denon proprietary special order).
I didn't care for the sound of my TDA1541 and I did assume that, even the voltage-out delta-sigma PCM1791 does a better job than the TDA, none of the multibit will be better. Sure, it's not an -A version, but my gut feeling is that even that cannot make up for the difference.
Now I saw that the R-2R is a different animal compared with the current sources and passive dividers used in Philips. Or maybe the Denon Alpha filters are way better than the SAA7220?
Also agree, the demodulation and error circuits are VERY important too - in their job to eliminate the mechanical jitter produced by the PLL loop (that includes the spindle motor). Spindle motor quality (and wear) might have a say into that too... 16kB buffer might be enough for some, 32kB (used by some Sony circuits) might make a big difference for others.
I hope you got rid of those lame NJM2041 - the "Asia" model came with OP275 as I/V!I see that it has the same Alpha filter - SM5845 - like my DCD-360. The only issue is that in my schematics is shown as either 5845 or 5848 and on board is SM5845. None of them have public datasheets (Denon proprietary special order).
I didn't care for the sound of my TDA1541 and I did assume that, even the voltage-out delta-sigma PCM1791 does a better job than the TDA, none of the multibit will be better. Sure, it's not an -A version, but my gut feeling is that even that cannot make up for the difference.
Now I saw that the R-2R is a different animal compared with the current sources and passive dividers used in Philips. Or maybe the Denon Alpha filters are way better than the SAA7220?
Also agree, the demodulation and error circuits are VERY important too - in their job to eliminate the mechanical jitter produced by the PLL loop (that includes the spindle motor). Spindle motor quality (and wear) might have a say into that too... 16kB buffer might be enough for some, 32kB (used by some Sony circuits) might make a big difference for others.
Attachments
Last edited:
Hi SoNic_real_one,
Denon has produced other models that are essentially the same as the DCD-S10. I fell in love with the cosmetics and Denon had some in stock. I got it as "service stock". It pays to be a nice person sometimes.
This model has a digital input and output, plus real balanced outputs in addition to the normal RCA type you would expect. As a straight D/A converter, it kills other outboard units. Interestingly, Teac has some models that used the PCM-1702 also in their upper units. Nakamichi used the PCM-1704 in a later top of the line model too.
You should take a careful look at how this chip operates. The term co-linear designates a real cool design that reduces non-linearity quite effectively. Couple that with the matching digital filter and you are close to having magic happen. All this great stuff can be ruined by the analog stage though, watch out.
The slew rate of an op amp is not the most important specification for this application. Not only that, but once you have a high enough slew rate, going for more may only bring trouble. The design is audio engineering, not pick-the-fastest (or best) op amp in the catalogue. Whenever a "designer" starts talking about using the "best" parts, you can be pretty sure they do not know what they are doing. The best parts for what? The most expensive, or the current darlings in audio press? A good audio engineer will choose the part that has characteristics that most closely match the requirements with an eye to cost. A more expensive part may not actually bring anything to the table that a less expensive part has. It would be pretty stupid to go with the more expensive part in that case. Of course, the marketing dept. may force the issue.
-Chris
Denon has produced other models that are essentially the same as the DCD-S10. I fell in love with the cosmetics and Denon had some in stock. I got it as "service stock".
It pays to be a nice person sometimes.This model has a digital input and output, plus real balanced outputs in addition to the normal RCA type you would expect. As a straight D/A converter, it kills other outboard units. Interestingly, Teac has some models that used the PCM-1702 also in their upper units. Nakamichi used the PCM-1704 in a later top of the line model too.
You should take a careful look at how this chip operates. The term co-linear designates a real cool design that reduces non-linearity quite effectively. Couple that with the matching digital filter and you are close to having magic happen. All this great stuff can be ruined by the analog stage though, watch out.
The slew rate of an op amp is not the most important specification for this application. Not only that, but once you have a high enough slew rate, going for more may only bring trouble. The design is audio engineering, not pick-the-fastest (or best) op amp in the catalogue. Whenever a "designer" starts talking about using the "best" parts, you can be pretty sure they do not know what they are doing. The best parts for what? The most expensive, or the current darlings in audio press? A good audio engineer will choose the part that has characteristics that most closely match the requirements with an eye to cost. A more expensive part may not actually bring anything to the table that a less expensive part has. It would be pretty stupid to go with the more expensive part in that case. Of course, the marketing dept. may force the issue.
-Chris
Well, I do think with my mind not marketing dept. To faitfully reproduce the encoded signal, I would think that the I/V stage needs to be alble to follow the settling time of the DAC itself. Some 200ns for PCM1702... That inherently means slew rate at the voltage output. Output of 2V needs 2/0.200us=8V/us. Faster DAC's mean faster SR at the output.
Last edited:
If you've found an opamp that settles to 20bits in under 200nS (I recall settling times in series tend to add as sum-of-squares so overall you'd get 280nS with a 200nS opamp) then I'd love to see the datasheet.
Hi SoNic_real_one,
Please understand no insult is ever intended from me - unless it is clear from the context from the thread. In your case, none intended. I just like to try and be clear of my meaning. Sometimes that doesn't help either.
The I/V stage can also perform some low pass filtering at the same time. All that really matters is that the circuit, op amp or otherwise, doesn't clip or misbehave in various ways. This is one of those things that make some op amps better than others when slew rate specs would seem to favour the faster device. What is important to remember is what the expected output is. If you wanted to preserve the individual pulses at whatever sampling rate the circuit is running at, then you would be concerned about very high slew rates. Then your calculation would be concerned with the sample rate, not the maximum speed of the D/A converter.
-Chris
Please understand no insult is ever intended from me - unless it is clear from the context from the thread. In your case, none intended. I just like to try and be clear of my meaning. Sometimes that doesn't help either.
The I/V stage can also perform some low pass filtering at the same time. All that really matters is that the circuit, op amp or otherwise, doesn't clip or misbehave in various ways. This is one of those things that make some op amps better than others when slew rate specs would seem to favour the faster device. What is important to remember is what the expected output is. If you wanted to preserve the individual pulses at whatever sampling rate the circuit is running at, then you would be concerned about very high slew rates. Then your calculation would be concerned with the sample rate, not the maximum speed of the D/A converter.
-Chris
we should want Audio - not "perfect" steps out of the I/V
technically a "high" output slew rate is not at all needed if you plan on filtering the DAC output - as you should at some fraction of Nyquist to cut image frequencies
as I pointed out on the 1st page a feedback C in parallel with the feedback R can be incorporated into the anti-imaging/reconstruction filter response
assuming adequate upsampling ratio I would put a 3rd order or higher Butterworth filter with fc ~ 40 KHz after the DAC
the real pole of that filter should then be the 1st order feedback C corner of ~ 40 KHz in the I/V which gives only 0.5 V/uS required slew rate at the I/V op amp output
the upsampling interpolation/digital filtering also reduces DAC output max step size - again cutting I/V slew rate requirements to close to that required by the audio signal itself
audio frequency accuracy requires excess loop gain - it is GBW that "needs" to be maximized, slew rate isn't a direct concern except as a proxy for high input linearity
with the I/V feedback C the transient diff V at the I/V op amp input is determined by the fc/GBW and the "feedthrough Zero" of the current step edge flowing in the op amp output impedance through the high frequency "short" of the feedback C
the really low Vnoise 1 nV/rtHz op amps can be used with an input RC shunt to gnd right at the DAC output – this “Zobel” soaks up some of the DAC current step edge before the op amp responds
the technique is better used with the low noise input I/V op amp because it creates a “noise gain” corner frequency above which the op amp input noise gets multiplied by a 1st order rising response
can be seen in the AD797 datasheet Figure 53. A Professional Audio DAC Buffer
technically a "high" output slew rate is not at all needed if you plan on filtering the DAC output - as you should at some fraction of Nyquist to cut image frequencies
as I pointed out on the 1st page a feedback C in parallel with the feedback R can be incorporated into the anti-imaging/reconstruction filter response
assuming adequate upsampling ratio I would put a 3rd order or higher Butterworth filter with fc ~ 40 KHz after the DAC
the real pole of that filter should then be the 1st order feedback C corner of ~ 40 KHz in the I/V which gives only 0.5 V/uS required slew rate at the I/V op amp output
the upsampling interpolation/digital filtering also reduces DAC output max step size - again cutting I/V slew rate requirements to close to that required by the audio signal itself
audio frequency accuracy requires excess loop gain - it is GBW that "needs" to be maximized, slew rate isn't a direct concern except as a proxy for high input linearity
with the I/V feedback C the transient diff V at the I/V op amp input is determined by the fc/GBW and the "feedthrough Zero" of the current step edge flowing in the op amp output impedance through the high frequency "short" of the feedback C
the really low Vnoise 1 nV/rtHz op amps can be used with an input RC shunt to gnd right at the DAC output – this “Zobel” soaks up some of the DAC current step edge before the op amp responds
the technique is better used with the low noise input I/V op amp because it creates a “noise gain” corner frequency above which the op amp input noise gets multiplied by a 1st order rising response
can be seen in the AD797 datasheet Figure 53. A Professional Audio DAC Buffer
Some of the newer delta sigma implemenations that I saw, have (somehow) faster OpAms and a place holder for the input capacitor but nothing is installed there.
The older multibit implemenations that I saw have very slow OpAmps (what was avail at that time) with that capacitor installed.
I assume that filtering in the same stage as I/V conversion (input capacitor and feedback one) will have some drawbacks - related to the final THD+N and that's why was abandoned (at the input) or minimized (feedback) in newer implementations.
Since now newer OpAmps apeared with low noise and higher SR/GWB, I was assuming that some improvement can be gained from that. I am using now as I/V the dual LM4562 that has decent performance in all those intrinsec aspects (noise, SR, GWB, THD) - close of single AD797. I am planning to experiment with reducing/eliminating the input and feedback capacitors.
Faster than that I listed earlier in this post, but I am not sure if anything is to gain over 20V/us (I use this just as quick metric in the topic, because the GWB and settling time are somehow related to this).
Unfortunatelly, I don't have access anymore to a good distortion meter and I don't trust SPICE models too much.
The older multibit implemenations that I saw have very slow OpAmps (what was avail at that time) with that capacitor installed.
I assume that filtering in the same stage as I/V conversion (input capacitor and feedback one) will have some drawbacks - related to the final THD+N and that's why was abandoned (at the input) or minimized (feedback) in newer implementations.
Since now newer OpAmps apeared with low noise and higher SR/GWB, I was assuming that some improvement can be gained from that. I am using now as I/V the dual LM4562 that has decent performance in all those intrinsec aspects (noise, SR, GWB, THD) - close of single AD797. I am planning to experiment with reducing/eliminating the input and feedback capacitors.
Faster than that I listed earlier in this post, but I am not sure if anything is to gain over 20V/us (I use this just as quick metric in the topic, because the GWB and settling time are somehow related to this).
Unfortunatelly, I don't have access anymore to a good distortion meter and I don't trust SPICE models too much.
Last edited:
Hi SoNic_real_one,
You could learn a lot from paying attention to jcx and abraxalito (among others). The points these two people brought up are important. They go a long way towards explaining why the op amp in an I/V converter is not a speed demon - that's apart from the fact that optimizing a part for one spec generally degrades other behaviour.
The one and only thing that matters in an op amp used for I/V converter use is that the input signal doesn't drive it to being non-linear. So as long as all the other specifications for high quality audio are met, you're good to go.
-Chris
You could learn a lot from paying attention to jcx and abraxalito (among others). The points these two people brought up are important. They go a long way towards explaining why the op amp in an I/V converter is not a speed demon - that's apart from the fact that optimizing a part for one spec generally degrades other behaviour.
Technology improved. The way any particular chip is implemented is up to the designer, so they often change the design from the app. notes. Always remember to, a design for a great DAC may not have had personnel on staff that designed great op amps. Even something as simple as running out of silicon can force compromise for an option (the on-board op amp).
Yes they do. Generally not done for audio applications since almost everything available has sufficient slew rate and settling times.
That has no bearing on anything. Asking for low power and slew rate + settling time is pushing things. You can't have everything you ask for. A part used for video signals would not be excluded from your requirements just because it's intended for a video application. The part is either capable of the required performance, or not. It is that simple.
The one and only thing that matters in an op amp used for I/V converter use is that the input signal doesn't drive it to being non-linear. So as long as all the other specifications for high quality audio are met, you're good to go.
That is a problem for you. The going price for an HP 339A distortion measuring set is in the $400 range, or less. Many years ago, a Leader THD meter cost me about $2.5K. I was forced to buy that one new, and the HP 339A blows it away. If you are serious about what you are doing, $400 is cheap. I have seen them sell for as little as $150 from time to time. In addition, make sure you look for the sample frequency at the output from your filter. Sync to the clock and watch the audio for these nasty spikes. Sometimes it is a slow op amp that protects equipment down the line from this HF garbage. A non-oversampling DAC without it's brick wall filter will output a strong 44.1 KHz signal. Amplifiers and preamps will pass this frequency just fine, but you could burn out the zobel in the amp, plus the tweeters. If the amp goes unstable, you could lose the entire lot (save the preamplifier). Sometimes there are reasons why certain parts are used.
-Chris
@Sonic_real_one: I use a smd dac board from muse-audio that's equipped with a 250MHz 130V/us quad-op
and 'my brain says "yes" to the sound'.
Another device, also equipped with the same dac-ic but with NE5532 i/v is much noisier and sounds smeary.
@anatech: Men try to rule their technology, woman try to rule their sexuality, or,
men try to rule womens sexuality and women try to rule mens technology.
and 'my brain says "yes" to the sound'.
Another device, also equipped with the same dac-ic but with NE5532 i/v is much noisier and sounds smeary.
@anatech: Men try to rule their technology, woman try to rule their sexuality, or,
men try to rule womens sexuality and women try to rule mens technology.
Hi!
So, summarizing, which specs do really matter for an I/V converter?
Sorry, I know the answer may be implicit in your previous replies there're so many information that I couldn't filter them.
So, summarizing, which specs do really matter for an I/V converter?
Sorry, I know the answer may be implicit in your previous replies there're so many information that I couldn't filter them.
Last edited:
My gut tells me that main figures are: input noise voltage, THD+N, slew rate (that's connected with GWB). Now, I need to figure a compromise: where lowering the noise stops to bring any substatial improvements. What is the value that rising the SR over, brings better results than a lower audio-band THD+N. How low can i get with the SR reducing capacitors.
Now, I have another gut feeling - all those are important only for multibit DAC's. The D-S seems to have lower settling times that multibit (even if not published and the manufacturer won't answer to that simple question directly).
PS: Looking at PCM1794 - for fs=96kHz, max F=768fs=73.728MHz. 1/F=13nS. In order to get a full swing, requires 24 bit to be "turned". That, due to "advanced segment" architecture will take at least 12 periods, no? That makes some 162ns... Comparable with the 200ns given in the PCM1704 datasheet.
Now, I have another gut feeling - all those are important only for multibit DAC's. The D-S seems to have lower settling times that multibit (even if not published and the manufacturer won't answer to that simple question directly).
PS: Looking at PCM1794 - for fs=96kHz, max F=768fs=73.728MHz. 1/F=13nS. In order to get a full swing, requires 24 bit to be "turned". That, due to "advanced segment" architecture will take at least 12 periods, no? That makes some 162ns... Comparable with the 200ns given in the PCM1704 datasheet.
Last edited:
In my view : immunity of input stage to overload, GBW, settling time, cleanliness of settling transient, linearity. Noise could only be a significant issue for >20bits and I don't know of any recordings that have a better than 20bit noise floor in practice. Slew rate isn't a major issue in that high GBW and good linearity pretty much guarantee adequate slew rate.
Well... GWB practically was "created" to express SR, open loop amplification and settling time with one number. So it is kind of redundant.
Linearity is related to feedback loop, therefore again with GWB.
Noise floor IMO is important. Even if recording doesn't have that dynamic, I don't like that noise flowing thru my system. Mudds up the sound...
PS: Haha, you are concerned about the noise shaping in a D-S DAC (at some -120..130dB), but you are OK with higher level noise if is in the I/V stage?
Linearity is related to feedback loop, therefore again with GWB.
Noise floor IMO is important. Even if recording doesn't have that dynamic, I don't like that noise flowing thru my system. Mudds up the sound...
PS: Haha, you are concerned about the noise shaping in a D-S DAC (at some -120..130dB), but you are OK with higher level noise if is in the I/V stage?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.