Yes, maybe it is worth to set independently for every bit and put some thermal common sink over the bodies. And to use devices from the same butch?
.
I dont knowww... Not talking about some "perfect" design, but some realistic, optimised, educational and relatively simple. For the starting point and device to comparing others?
.
I am not hunging-up for ultrahigh prec. Rs. And expensive too... For my opinion for the start is not good advice to spend a lot of money into Rs? Without examining the behaviour and sound first with economy but appropriate types?
As I said I used for my first diskrete R2R dac 1% worst case Rs. And it is not possible to trim, to measure for days with 0000 resistance meter... 🙁
...
Before %tol options to choos, people should know which format is neded as input digital stream, what is the typa of discrete DAC (voltage current), value of the reistors, etc etc.
All these things which are not stated in almost every R2R, non-fpga, dac in the net.
Does anyone have the figures referenced in that article. They might be somewhere on archive.org but I don't know offhand how to find them.
I've seen this (missing figures and diagrams in tech articles) many times on older EDN artickes and possibly other tech magazines online, and it's frustrating. I'm sure I'd have a much better understanding of the article with the figures.
I built a rather pedestrian 16 bit R-2R DAC from scratch a few years ago using a Cypress Semiconductor PSOC4 chip to drive the ladder. I quickly found that the difference in the on resistance of the port pin pull up driver something like 27 ohms on resistance vs the pull down driver, maybe 7 ohms swamped out the error in my hand matched resistor network. If you want really well matched resistors, just buy a few hundred 0.1% and sort them. The DAC worked fine and the highest harmonic for a mid band sinewave was 70 dB down from the fundamental, so maybe 12 bit performance. Many people said it sounded great. Fun project. If I were to do it again I could see using two ports in parallel, one configured for pull up only and one configured for pull down only, with a 33 ohm trim pot in series with the lower resistance pull down port to balance it with the pull up port resistance.
Last edited:
Its not the identical article but covers much of the same material and includes figures : https://www.analog.com/media/en/technical-documentation/application-notes/an86f.pdf
Suppose you build a DAC that, due to inaccuracies, makes a 20 LSB instead of a 1 LSB step when the MSB switches. You then use a feedback loop with an accurate ADC to find the DAC code that produces an output voltage that fits best with the intended code, and store it in memory or so. You then still have a 20 LSB step somewhere in the centre of the range.
I think you need a coarse and a fine DAC, where the range of the fine DAC is big enough to cover gaps in the output of the coarse DAC. The input code can then go straight to the coarse DAC and to the address input of a memory that looks up the corresponding fine DAC correction code.
You would need 16777216 fairly short words of memory for a 24 bit DAC, and you would need to calibrate all 16777216 steps. Maybe you could leave some of the lower bits uncorrected to reduce this.
Edit: I see that's essentially what is done in the article abraxalito linked to, except for the memory look up. I had only seen their figure 1, which is somewhat misleading because it doesn't show that there is actually overlap between two DACs.
I think you need a coarse and a fine DAC, where the range of the fine DAC is big enough to cover gaps in the output of the coarse DAC. The input code can then go straight to the coarse DAC and to the address input of a memory that looks up the corresponding fine DAC correction code.
You would need 16777216 fairly short words of memory for a 24 bit DAC, and you would need to calibrate all 16777216 steps. Maybe you could leave some of the lower bits uncorrected to reduce this.
Edit: I see that's essentially what is done in the article abraxalito linked to, except for the memory look up. I had only seen their figure 1, which is somewhat misleading because it doesn't show that there is actually overlap between two DACs.
Last edited:
I have a Soekris related question, but it may or may not apply to R2R DACs in general. I just got a DAM 1021 0.05 board working with a Wave IO input USB board. In comparison I have a basic Chinese AK4490 DAC with op-amp output. The big differences I hear are:
Soekris - very smooth, detailed and instrumental tonalities are "natural" but with one exception - dynamics. The dynamics are rather "flat" meaning I'm not getting a proper attack on the leading edge of notes. Like if you enter a flat tone to notes in Logic/Garageband rather than using a touch sensitive keyboard which gives you the proper attack and decay. This is most obvious with a Steinway piano which sounds flat rather than percussive and steely as it should, and does on the AK4490.
AK4490 - detailed and proper attack and decay. Not as refined as the Soekris but more involving because of the dynamics.
I don't know if this is a characteristic of R2R DACs in general, or just my implementation of the Soekris. The PSU is 3.3V and 5V DC from bench supplies and +-12V DC from a regulated supply with LM317/337.
The other salient feature is that the guy who set it up for me downloaded a different filter which was popular in threads, will have to ask which filter it was.
Input is a Mac Pro. Output is a 2a3 amp configured Hammond 1:4 SUT 1140-LN-C > 10Y > 2a3. These both work fine for dynamics with the AK4490. I believe the output of the Soekris is 1.4V straight off the board.
I need to solve this problem to be happy using the Soekris. There are some very clever guys on this thread, so could I please have some idea of where to look for answers to this question?
1. I can make a different +-12V supply if you think that's the issue. e.g. CLC
2. If you think the filter is the issue I can look at downloading another filter
3. Or could it be something else?
4. Or might it be a characteristic of R2R DACs?
I'm a musician and this is really bugging me.
Soekris - very smooth, detailed and instrumental tonalities are "natural" but with one exception - dynamics. The dynamics are rather "flat" meaning I'm not getting a proper attack on the leading edge of notes. Like if you enter a flat tone to notes in Logic/Garageband rather than using a touch sensitive keyboard which gives you the proper attack and decay. This is most obvious with a Steinway piano which sounds flat rather than percussive and steely as it should, and does on the AK4490.
AK4490 - detailed and proper attack and decay. Not as refined as the Soekris but more involving because of the dynamics.
I don't know if this is a characteristic of R2R DACs in general, or just my implementation of the Soekris. The PSU is 3.3V and 5V DC from bench supplies and +-12V DC from a regulated supply with LM317/337.
The other salient feature is that the guy who set it up for me downloaded a different filter which was popular in threads, will have to ask which filter it was.
Input is a Mac Pro. Output is a 2a3 amp configured Hammond 1:4 SUT 1140-LN-C > 10Y > 2a3. These both work fine for dynamics with the AK4490. I believe the output of the Soekris is 1.4V straight off the board.
I need to solve this problem to be happy using the Soekris. There are some very clever guys on this thread, so could I please have some idea of where to look for answers to this question?
1. I can make a different +-12V supply if you think that's the issue. e.g. CLC
2. If you think the filter is the issue I can look at downloading another filter
3. Or could it be something else?
4. Or might it be a characteristic of R2R DACs?
I'm a musician and this is really bugging me.
You are already answered your own question: "I believe the output of the Soekris is 1.4V straight off the board."
At the moment we have a prototype DSD only discrete resistor dac running here. It is an Andrea Mori design. With a top notch transformer used for DC removal and differential to single ended output, it sounds remarkably good at DSD256. Unfortunately the dac is better than the hardware PCM to DSD converters that can be found in forum threads. Using something like HQ Player or a particular offline PCM to DSD software can give excellent results. Imaging in the best case is rather spectacular. Of course, with a dac of its caliber it matters if the rest of the system at a similar level to get the full benefit. The dacs should be available for people that want them before too long.
Pic shows dual shunt regulator to power dac, the dac, and the output board. The output transformer is located externally at the moment.
Last edited:
Andrea Mori designed a DSD DAC? That's remarkable. DSD is a sigma-delta modulate and he used to dislike anything sigma-delta.
He changed his mind after he visited here once while on holiday. He said he never heard an oversampling dac that sounded like my AK4499 dac running in DSD256 mode. At one point he and his friends said they thought his discrete resistor PCM dac sounded about the same as my AK4499 dac when were doing A/B comparisons. Also, we suggested he try making a DSD dac. Then at some point he posted he was working on one in a thread on his TheWellAudio website. Both of his dacs are very good, BTW, but we think the DSD dac is a tad more to our liking. At some point we are expecting firmware to run it in RTZ mode too.
Last edited:
Andrea did not design a DSD converter. It is a DSD only DAC. Up to to the user to choose a converter.
Regarding an FFT, don't have one. Don't know how to measure imaging with one anyway. IOW, what's important to me are not necessarily the same things that are easy to measure with only a sound card. IMHO focusing on what is easy to measure is like the old story of the drunk looking for his car keys under the street light: https://en.wikipedia.org/wiki/Streetlight_effect
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7829028/
Regarding an FFT, don't have one. Don't know how to measure imaging with one anyway. IOW, what's important to me are not necessarily the same things that are easy to measure with only a sound card. IMHO focusing on what is easy to measure is like the old story of the drunk looking for his car keys under the street light: https://en.wikipedia.org/wiki/Streetlight_effect
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7829028/
Hi Mark ... Thanks for your feedback. As it is I don't as such disagree with you, however, the FFT IMHO gives some insight into the "structural" performance of a circuitry, i.e. its static level of distortion etc. IME many other factors influence the actual sound - but, well, I tend to find some information in the FFT as to the possible attainable level of performance of a design. Anyway ... I didn't mean to open a discussion about this - I reckon it is/has been addressed elsewhere here on diyaudio. Just was curious about distortions, noise etc., if there were some information about this.
In any case - thanks for considering 😉
Jesper
In any case - thanks for considering 😉
Jesper
""The “streetlight effect” refers to the parable of a drunk man searching underneath a streetlamp for a key he dropped on the other side of the street because, as the drunkard reasons, “This is where the light is.”1 A version of this parable is found in 13th-century Sufi literature,2 suggesting an enduring human quest to find solutions not necessarily where the problem lies, but wherever the search seems easiest.""
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7829028/
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7829028/
There should be a parable about someone who refuses to use a flashlight when looking for keys because it reminds him too much of a streetlight.
Well, you could always coin such a parable now. Maybe it will persist over the centuries if sufficiently apt.
Regarding the original parable, of course all analogies fail at some point if picked apart. Usually their purpose is simply to introduce a new idea by superficial comparison to something already familiar, or by else by reference to something easily understood by common intuition.
Regarding the original parable, of course all analogies fail at some point if picked apart. Usually their purpose is simply to introduce a new idea by superficial comparison to something already familiar, or by else by reference to something easily understood by common intuition.
Last edited:
Hi Markw4 😉
It's as if you eat a dish in total darkness that you will find excellent, then as soon as the light comes on, the presentation not being up to the taste, this same dish then becomes tasteless... 🤔
The 'numbers race' is well understood by resellers, at least they use it very well 💲💲💲
Cordially.
It's as if you eat a dish in total darkness that you will find excellent, then as soon as the light comes on, the presentation not being up to the taste, this same dish then becomes tasteless... 🤔
The 'numbers race' is well understood by resellers, at least they use it very well 💲💲💲
Cordially.
