I see what you are talking about. It doesn't show in my measurements. I'll duplicate the measurements they are doing with the demo board and see if I get similar. It won't be until the weekend. Its possible that is an artifact of the AP ADC or something else. In the quick check I did of old plots of the AK4430 I see possibly 5 dB of noise floor shift but that's from -145dB to -140dB. While its not ideal that doesn't strike me as an issue that would ever be detectable on the other side of an acoustic transducer.
I'll try several tricks to get a really good valid look into the issue. I know Keith Johnson bent my ear for a few hours about it.
I'll try several tricks to get a really good valid look into the issue. I know Keith Johnson bent my ear for a few hours about it.
This is from memory so the figures might be a little off, but from what I recall the THD+N is around -105dB at full scale, whereas at -60dB you get -60dB. This for their top of the range (or near-top). The residual is clearly not HD at the highest output, that's obvious from the FFT plot they provide - so it can only be noise. Bear in mind that noise modulation in S-D modulators is an instantaneous level sensitive thing, the FFT will only show the averaged noise and so with a sinewave as stimulus (spending very little of its time close to zero) it won't show up well.
Last edited:
That's very interesting . Looking at the cheap 24 bt chip I gave a link for it is about - 30 dB distortion typical at -60 db . I dare say that isn't so awful if other factors are good . I am tempted to buy one to test my theory . I remember years ago some some un-dithered - 90 db distortion to be about 30% . This might improve to 3 % using dither . Thinking of bits available that seems ambitious . The 3 % at - 60 dB seems easy to understand . What might be better is using a high grade CD burner and test the whole system , so easy to do and should show differences .
This might be a better explanation of the nanosecond timing requirements that seem now to be accepted . With minimal bits low distortion is dependent on integration taking place at the correct time . Even in the crude RC integrator produces a curve good enough to substitute the original as long as the start point is correct . Interestingly this cheap chip has the advantage of Delta Sigma D to A to do the integration . This is not a lightly taken choice . It has an internal I to V converter one assumes as it has usable output . So strange it has such modest spec . I am willing to bet is sounds better than it measures . As you say how it is used is the big deal . I have Naim CDi with a dying DAC . It would make an ideal testbed .
Thanks for the data . Any good links on building DAC clocks . Simple would be best .
This might be a better explanation of the nanosecond timing requirements that seem now to be accepted . With minimal bits low distortion is dependent on integration taking place at the correct time . Even in the crude RC integrator produces a curve good enough to substitute the original as long as the start point is correct . Interestingly this cheap chip has the advantage of Delta Sigma D to A to do the integration . This is not a lightly taken choice . It has an internal I to V converter one assumes as it has usable output . So strange it has such modest spec . I am willing to bet is sounds better than it measures . As you say how it is used is the big deal . I have Naim CDi with a dying DAC . It would make an ideal testbed .
Thanks for the data . Any good links on building DAC clocks . Simple would be best .
If cheap chips interest you, try TDA1545AT (here its $0.30, new) or TDA1387 ($0.08, recycled from Soundblaster cards). Both these easily beat off all the S-D offerings I've heard for instrumental timbre (not heard ES9018 mind). But then they're not one-stop solutions, to get the best out of them you need to add a filter and amp stage, so overall not so cheap as S-D DACs. They don't seem to be particularly clock sensitive either - I've not heard differences between feeding the DAC from my QA550 wav player I2S (crystal controlled) and from an S/PDIF receiver chip.
I was shocked how the cheaper chip was better explained in it's data . Honest to a fault . I will look at your recommendations . I'm far from happy with how CD works . I now have a first class machine that I always liked that needs a DAC . Why not learn as much as I can if doing this .
Anyone put a non 16 bit DAC into a Naim CDi ? To my mind remove the digital input pin from the Naim DAC and use it as the new DAC input , job done ? Care with ground and PSU using spectrum analyzer to be certain . Best I to V converter ? Valve even .
Anyone put a non 16 bit DAC into a Naim CDi ? To my mind remove the digital input pin from the Naim DAC and use it as the new DAC input , job done ? Care with ground and PSU using spectrum analyzer to be certain . Best I to V converter ? Valve even .
http://www.hqew.net/files/pdf/PHILIPS/TDA1545AT/TDA1545AT_www.hqew.com.pdf?key=TDA1545AT
Very similar specs and again honest . I dare say very close to the Naim original . Naim told me a bit of Bluetak under the chip helps the LSB error ! Philips told Naim about this as a possible tweak . Originally Philips were concerned about overheating . They learned that engineers did it to help them solder SMD and nothing to do with sound . No overheating and less errors . They found Blutak standard chips out performing selected grade known as crown . Philips concluded that the DAC works as a strain gauge at the 30 uV level representing bit 1 . If Delta sigma I would guess it less important . For 16 bit worth trying .
Very similar specs and again honest . I dare say very close to the Naim original . Naim told me a bit of Bluetak under the chip helps the LSB error ! Philips told Naim about this as a possible tweak . Originally Philips were concerned about overheating . They learned that engineers did it to help them solder SMD and nothing to do with sound . No overheating and less errors . They found Blutak standard chips out performing selected grade known as crown . Philips concluded that the DAC works as a strain gauge at the 30 uV level representing bit 1 . If Delta sigma I would guess it less important . For 16 bit worth trying .
An interesting dilemma. As Abraxalito quite rightly points out, the alternatives are not a one stop solution.
While there is a lot of good to be said for high degrees of integration in case of one stop solutions, it has its drawbaks as well, namely that it is what it is, and there's little you can do about it.
I can't help being reminded of the old dilemma, integrated op amps vs. discrete solutions. This seems to be just about the same thing.
Integrated gives you everything in one chip - easy, cheap, compatibility guaranteed, but locked up, no intervention possible.
Older chips require a periphery, which allows you to choose and select the peripherals to your liking, including sources. For example, I'd probably go for AD op amps because by and large I am happy with the sound they afford me, Nige would almost certainly use some NE variant, and so forth. More expensive, more extensive on the PCB, but relatively felxible and adjustable to suit various tastes.
So, gentlemen, start your engines!
While there is a lot of good to be said for high degrees of integration in case of one stop solutions, it has its drawbaks as well, namely that it is what it is, and there's little you can do about it.
I can't help being reminded of the old dilemma, integrated op amps vs. discrete solutions. This seems to be just about the same thing.
Integrated gives you everything in one chip - easy, cheap, compatibility guaranteed, but locked up, no intervention possible.
Older chips require a periphery, which allows you to choose and select the peripherals to your liking, including sources. For example, I'd probably go for AD op amps because by and large I am happy with the sound they afford me, Nige would almost certainly use some NE variant, and so forth. More expensive, more extensive on the PCB, but relatively felxible and adjustable to suit various tastes.
So, gentlemen, start your engines!
Personally, by far the best solution I have ever come across so far, meaning a better that usual sounding solution not costing an arm and a leg, is one of those "real time DACs", which use 8 parallelled Phuilips 1543 DACs, and because it does it all in one pass, uses no brickwall filters.
I purchased my version from DIY Australia some3 or 4 years ago, for Australian $ 250. They asked what to put in the declaration to make it easy for me at my local customs office.
Anyway, sonically, there's a freedom and appearent dynamics iz offers which I am hard pressed to remember from any big name company at anything less tha say US$ 5,000 or so.
These days, I am told, the Chinese are offering the same or similar for way less, like US$ 50 or so. This is the whole deal, DAC and full output stage, often using AD 847, with separate rectifiers and power supplies for the digital and analog stages.
I use my Yamaha CDX-993 player (app. 23 lbs of uir, separate transformers for digital and analog, PSUs using 4 6,800 uF/50V caps, etc) as the transport, and the final result is a sound with much detail, yet not sounding "digital" at all, no harshness, no fizz, no spitting, and truly reminscent of an LP source. The only giveaway detail is the total absence of crackle and pop, a deathly background silence.
I purchased my version from DIY Australia some3 or 4 years ago, for Australian $ 250. They asked what to put in the declaration to make it easy for me at my local customs office.
Anyway, sonically, there's a freedom and appearent dynamics iz offers which I am hard pressed to remember from any big name company at anything less tha say US$ 5,000 or so.
These days, I am told, the Chinese are offering the same or similar for way less, like US$ 50 or so. This is the whole deal, DAC and full output stage, often using AD 847, with separate rectifiers and power supplies for the digital and analog stages.
I use my Yamaha CDX-993 player (app. 23 lbs of uir, separate transformers for digital and analog, PSUs using 4 6,800 uF/50V caps, etc) as the transport, and the final result is a sound with much detail, yet not sounding "digital" at all, no harshness, no fizz, no spitting, and truly reminscent of an LP source. The only giveaway detail is the total absence of crackle and pop, a deathly background silence.
http://www.lampizator.eu/lampizator/LINKS AND DOWNLOADS/DATAMINING/meridian cs4329.pdf
This was the DAC I liked most in the past .
This was the DAC I liked most in the past .
It sounds to me very much like you're describig the Lite DAC-AH, which here goes for about 800rmb, under $150 delivered. I have a fair amount of familiarity with this design as I've bought several with the intention of modding them. They make an excellent base for such activities because the PCB is well-made, robust and doesn't have those nasty opaque solder masks which are sometimes found on Chinese DAC PCBs. This means you can easily see what connects where, particularly in respect of the ground plane.
Indeed the grround plane is the no.1 thing to modify because they use the same one for power and signal. This is time consuming - cutting through the thermal reliefs and then instituting star-earthing. Sonically though well worth one's time. I also cut out all the TDA1543s on mine and plugged in TDA1545s (in one case TDA1387s) - the DIL package of the 1545 is pin compatible, you do though need to twiddle some pins on the CS8414 receiver to get the correct digital format for it - its not I2S.
Changing from a bipolar to a CMOS DAC introduces harshness if you don't filter the output, I reckon this comes about from the inherent glitchiness of CMOS. I began with just long rows of ferrite beads (at least 20) forming a simple LR filter. There is one thing I hate about this DAC - on all the ones I've bought, they've used hot melt to secure the S/PDIF input connector, whih itself is of very shoddy quality. In order to disassemble the DAC, you'll probably have to sacrifice that connector and drill out the back panel for a superior one.
All in all though, an unsung bargain in terms of sound satisfaction. The AD847 despite being an opamp, does mitigate one of the major weaknesses of opamps used in high RF environments - the LTP input is nicely degenerated.
I rather agree with Thorsten - no one solution to those requirements. Optimize for 16/44 with a multibit DAC, then go for something like an ES9018 for the rest.
Why not to use something like pcm1704 with filters in DSP? Even DSD input can be converted to 24/176. It will be really universal solution and allow selecting best filter for each record (useful feature in many cases).
Because in my experience (i.e. from listening) that series of chips introduces colourations of instrumental timbre, in particular on brass. I very much doubt its caused by the FIR filters, rather its much more likely noise modulation from the S-D modulators.
Oops, sorry - my bad - my eyesight is poor, I misread PCM1704 as PCM1794. The PCM1704 is a fine chip, just rather too expensive to recommend in this instance. Although I've not listened to it myself, I'm dubious it beats out the much cheaper TDA1545A on RBCD owing to being R2R architecture.
Oops, sorry - my bad - my eyesight is poor, I misread PCM1704 as PCM1794. The PCM1704 is a fine chip, just rather too expensive to recommend in this instance. Although I've not listened to it myself, I'm dubious it beats out the much cheaper TDA1545A on RBCD owing to being R2R architecture.
Last edited:
@Abraxalito
Many thanks for the pointers.
I would just point out that my version preceeded the Chinese slew of similar products by two or three years, and is, AFAIK, an Australian product. At least, it seems to be that. And they were far from being stingy - it's full of Dale resitors, which I like to see and hear, the filter caps are high grade Sanyo products, and the whole deal included a ratrher good 50VA toroidal transformer as well. Everything but the case, in fact. And it was a real pleasure doing business with that crew - courteous, very prompt and accurate to the minutest of points. Thumbs up.
As for the sound, obviously, the thing worket straight off as adverstised, and in fact, their own advertisement doesn't do them justice. It's way better than advertised. My own acid test is playing three CDs:
1. The New Seekers - The best of (outstanding rendition of vocals, male and female)
2. Enigma - 1991 (the first CD, unbelievable dynamic range, mine is some special edition)
3. The Blue Man Group - the first CD, from extremely low volume to literally thunder (which is what you have when a man strukes full force a bass drum with a diameter of 16 feet, leaves most amps and speakers gasping for air (1)).
The sound was everything I ever wanted from the CD format - too bad modern mastering techniques are so focused on volume, in fact killing most of the dynamic range. But where it truly shined was with realatively low volume, subtle reordnings; for example, its rendition of Incredible String Band's "Air" was truly stunning (THE junkie song of all junkie songs, in my view), no problem in visualising the band right there, in the room, with you. It also shines with very complex, rich sounds of people like The Steeleye Span, or Pete Seeger's concert performance of say "Wimoweh".
In short, it makes listening to music a true pleasure, be it civilized music like Gordon Lightfoot or Joni Mitshell, or hell raising stuff like Alice Cooper. Not perfect, of course, but it somes awfully near to portraying live music in your room.
It also makes you wonder why some jackass ruined a fine format for the sake of sheer brutish volume.
BUT, you have infected me with a bug. I'll have to take a closer look at how it's made, as you say, there may well be room for improvement. I am not above making a new board if I feel it may bring some benefit.
(1) With my apologies to Wayne, neither my amp, nor the speakers, are left gasping for air, just my wife, who barges in with blood on her mind. Admittedly, I refer to H/K Citation 24, in "high voltage" setting (as opposed to the "High Current" setting), PSU rails at +/-56V, nominally 100W/8 Ohms, and I do admit to having refreshed its original four 6,800 uF caps with 10,000 uF new ones from Fischer&Tauche (German made, my favorites).
Incidentally, in case of the Blue Man Group, I would agree with you Wayne that in this rare instance, you would very likely need hundreds of Watts and something like 20 Amps of current or more for a faithful rendition of that drum even at normal household SPL levels.
Many thanks for the pointers.
I would just point out that my version preceeded the Chinese slew of similar products by two or three years, and is, AFAIK, an Australian product. At least, it seems to be that. And they were far from being stingy - it's full of Dale resitors, which I like to see and hear, the filter caps are high grade Sanyo products, and the whole deal included a ratrher good 50VA toroidal transformer as well. Everything but the case, in fact. And it was a real pleasure doing business with that crew - courteous, very prompt and accurate to the minutest of points. Thumbs up.
As for the sound, obviously, the thing worket straight off as adverstised, and in fact, their own advertisement doesn't do them justice. It's way better than advertised. My own acid test is playing three CDs:
1. The New Seekers - The best of (outstanding rendition of vocals, male and female)
2. Enigma - 1991 (the first CD, unbelievable dynamic range, mine is some special edition)
3. The Blue Man Group - the first CD, from extremely low volume to literally thunder (which is what you have when a man strukes full force a bass drum with a diameter of 16 feet, leaves most amps and speakers gasping for air (1)).
The sound was everything I ever wanted from the CD format - too bad modern mastering techniques are so focused on volume, in fact killing most of the dynamic range. But where it truly shined was with realatively low volume, subtle reordnings; for example, its rendition of Incredible String Band's "Air" was truly stunning (THE junkie song of all junkie songs, in my view), no problem in visualising the band right there, in the room, with you. It also shines with very complex, rich sounds of people like The Steeleye Span, or Pete Seeger's concert performance of say "Wimoweh".
In short, it makes listening to music a true pleasure, be it civilized music like Gordon Lightfoot or Joni Mitshell, or hell raising stuff like Alice Cooper. Not perfect, of course, but it somes awfully near to portraying live music in your room.
It also makes you wonder why some jackass ruined a fine format for the sake of sheer brutish volume.
BUT, you have infected me with a bug. I'll have to take a closer look at how it's made, as you say, there may well be room for improvement. I am not above making a new board if I feel it may bring some benefit.
(1) With my apologies to Wayne, neither my amp, nor the speakers, are left gasping for air, just my wife, who barges in with blood on her mind. Admittedly, I refer to H/K Citation 24, in "high voltage" setting (as opposed to the "High Current" setting), PSU rails at +/-56V, nominally 100W/8 Ohms, and I do admit to having refreshed its original four 6,800 uF caps with 10,000 uF new ones from Fischer&Tauche (German made, my favorites).
Incidentally, in case of the Blue Man Group, I would agree with you Wayne that in this rare instance, you would very likely need hundreds of Watts and something like 20 Amps of current or more for a faithful rendition of that drum even at normal household SPL levels.
Last edited:
Really helpful
Does anyone else think LTP degeneration in op amps useful when I to V conversion ? It is slightly counter-intuitive as it is an integrator . However I can clearly see what the feedback loop does is no indication of RF happiness . PDF below is interesting . It is on my tool bar .
http://www.analog.com/static/imported-files/tutorials/MT-096.pdf
As it happens, at the end of last year I kicked off this thread with the intention of explaining how to hot-rod the DAC-AH. My version very much retained the quasi-boutique components - Dale resistors, Sanyo capacitors and even went one better on the trafo with a 3 secondary R-core. However there was very little interest and then inexplicably I got banned from there I discovered in the process that modding and tweaking was far more satisfying than talking about modding and tweaking to a relatively disinterested audience.
I discovered in the process that modding and tweaking was far more satisfying than talking about modding and tweaking to a relatively disinterested audience.Amen to that .
I did a bit on how capacitors work and how to save a fortune . Deathly silence form the tweakers . Basically do not overlook non polar electrolytic's and keep them at a low voltage if you can ( 0.4 V , ideal for feedback arm ) . Do not deliberately polarize caps in a false belief it is better if you have a choice .
Look at resistor choices . Sometimes carbon comp is OK . Measure before you say no .
I discovered in this brief look at DAC's that about 2 % THD - 60 dB is acceptable for a DAC . I had imagined that with all the trickery of a modern DAC the waveform would be better . A number of questions come out of this . What would a Dolby Pro Revox do at - 60 dB ( it's dynamic range is like 16 bit , done to extend the life of the Revox etc , Fancy Dolby A ) ? Also what is the frequency distribution of that distortion in both cases . Would dither although altering dynamic range change the THD to be better /nicer . Is putting in a triangle wave all I need do . What triangle wave ? Is white noise better despite people saying not ? Pink ( ish ) noise looks a choice also. Reading up on quantization noise . Early commentators who knew of triangle-wave dither said white was better as it exactly replicates the distortion types of a tape recorder at the chosen level . Logically it would as bias is almost a digital process . The more remarkable thing is if compared with tape recording nothing has been traded to do this ( remember a microphone is setting some parameters ) . Logically speaking that would be true if bias is analogous to digital .The beauty of analogue tape recording is that the playback needs no clock apart for the obvious speed stability .
I did a bit on how capacitors work and how to save a fortune . Deathly silence form the tweakers . Basically do not overlook non polar electrolytic's and keep them at a low voltage if you can ( 0.4 V , ideal for feedback arm ) . Do not deliberately polarize caps in a false belief it is better if you have a choice .
Look at resistor choices . Sometimes carbon comp is OK . Measure before you say no .
I discovered in this brief look at DAC's that about 2 % THD - 60 dB is acceptable for a DAC . I had imagined that with all the trickery of a modern DAC the waveform would be better . A number of questions come out of this . What would a Dolby Pro Revox do at - 60 dB ( it's dynamic range is like 16 bit , done to extend the life of the Revox etc , Fancy Dolby A ) ? Also what is the frequency distribution of that distortion in both cases . Would dither although altering dynamic range change the THD to be better /nicer . Is putting in a triangle wave all I need do . What triangle wave ? Is white noise better despite people saying not ? Pink ( ish ) noise looks a choice also. Reading up on quantization noise . Early commentators who knew of triangle-wave dither said white was better as it exactly replicates the distortion types of a tape recorder at the chosen level . Logically it would as bias is almost a digital process . The more remarkable thing is if compared with tape recording nothing has been traded to do this ( remember a microphone is setting some parameters ) . Logically speaking that would be true if bias is analogous to digital .The beauty of analogue tape recording is that the playback needs no clock apart for the obvious speed stability .
Last edited:
- Status
- Not open for further replies.