Thanks Terry have you time to built and test a circuit for 3.3v?
You may need a 2.7V zener + vbe =3.3V if using walt circuit.
Morning coffee helped me to find the detailed measurements it that very long thread.
Posts from #461 to #463.
D-Noizator: a magic active noise canceller to retrofit & upgrade any 317-based V.Reg.
Actually for a PS with TL431 filtered ref and NE5532 err amp noise is about 1.5uV RMS in 20Hz-20kHz band. Measured ones. What about yours then?
LT3042 is $5.60 at mouser. Here's a whole finished board with a buffer for under $15.
Low Noise LT3042 Linear Regulator Power Supply Board DC Converter Overvoltage | eBay
What makes the cost such an insurmountable burden?
Low Noise LT3042 Linear Regulator Power Supply Board DC Converter Overvoltage | eBay
What makes the cost such an insurmountable burden?
Uhm.
I have seen those results in the past, and did not like them. My first hand experience, the data sheet results and other peoples results are completely different what regards the lt3042 curves. But strongly.
Then I had a bit more read into that thread and I see that Gerhard brought up exactly my concerns..
https://www.diyaudio.com/forums/pow...fit-upgrade-317-based-reg-50.html#post6045099
What I don't like about these things that also You had to know about these, but had chosen to fall back to those somehow skewed measurements, instead of listening to the debate around them...
I dont like cults developed. I do like that denoiser thread, that is the circuit suggested aeons ago by Wenzel.
A nice trick, can bring improvements, but never forget that in the end it must follow the original regulator behaviour. A limited, 10kHz bandwith regulator remains just that.
If progress had resulted in just fundamentally better solutions, why is it so hard to accept that..
Again, what Gerhard says too:
https://www.diyaudio.com/forums/pow...fit-upgrade-317-based-reg-51.html#post6046409
Ciao, George
I have seen those results in the past, and did not like them. My first hand experience, the data sheet results and other peoples results are completely different what regards the lt3042 curves. But strongly.
Then I had a bit more read into that thread and I see that Gerhard brought up exactly my concerns..
https://www.diyaudio.com/forums/pow...fit-upgrade-317-based-reg-50.html#post6045099
What I don't like about these things that also You had to know about these, but had chosen to fall back to those somehow skewed measurements, instead of listening to the debate around them...
I dont like cults developed. I do like that denoiser thread, that is the circuit suggested aeons ago by Wenzel.
A nice trick, can bring improvements, but never forget that in the end it must follow the original regulator behaviour. A limited, 10kHz bandwith regulator remains just that.
If progress had resulted in just fundamentally better solutions, why is it so hard to accept that..
Again, what Gerhard says too:
https://www.diyaudio.com/forums/pow...fit-upgrade-317-based-reg-51.html#post6046409
Ciao, George
so for what load the above is relevant? I believe if we speak about DACs AVCC/Vref, then a ref/err amp or a shunt is a fine choice. discrete or monolite. there are also pretty good 317 based ones like the Nazars one. for the LPF if it is op amp based one may be better off skipping the reg at all. fine elcaps and well designed filter sounds so sweet 
IMHO of course.
IMHO of course.
Last edited:
My naive original question was: is regulator with over 120 dB PSRR, 0,8 µV RMS noise in the whole audio bandwidth, extremely low output impedance and 10 mV voltage drift, good enough to be used as voltage reference for the DAC. Fact, that such results are provided by LM317 is another “issue” altogether.
I should have received short answer: No, it is not.
I don’t like cults and devoted believers versus facts myself. There is no cult around denoiser. It is just very nice circuit and denoiser principle can be applied to some of the LT regulators as well, LT1963 for instance, pushing their already good results further.
As for the noise level, Gerhard agreed on it with Elvee in the post #530.
Anyway, this thread was so many times hijacked and derailed that I complained, and now I’m doing the same.
So it’s time to cut the cr.. um, I mean the denoiser discussion. I got my answer.
I would agree that the specs above are fine for the AVCC of the DAC discussed here. But.. the thing is... there is more than that... there is some "magic" in mating particular DAC with particular reg. most likely this "magic" is measurable, but it may take some effort, and then again the question remains what kind of sonic outcome one prefers. I would suggest instead of seeking definite solutions in advance just build several regs and choose the one you like. I take it as a freedom to select my own preferences despite whatever "gurus" may insist. it is diyaudio after all
Last edited:
I am planning this. I have a pi3 in the streamer already but I have seen that most info online follows the ardiuno, I have one spare and may use that. That is the next step and plan to figure it out soon, if there is a good place to look for instructions it would be a great help. Logic analyzer still to arrive.
Steps i have completed are all power lines isolated, X5 LDO from individual windings, clock is late to arrive but will be swapped. i2s isolation added. Battery power supply has been finished and sounds awful, now removed. Working on op amp stage now for comparison, currently has tube output.
power line isolation is only by way of individual windings to each required supply.
My goal in this project is to improve on the last DAC build but also to allow parts to swap out for testing, measuring and overall learning. for this reason i have fitted a board on top of the dac with terminal connections for all items that require power as well as dac outputs. This is the reason for each VR board being different. I fully appreciate this has been explored and confirmed elsewhere i just want to measure it for myself and understand clearer as i go mistakes and all. obviously the extra wiring needed will bring its own issues, and learning from it.
x6 secondaries 5.2V AC
LDO boards being used following data sheet builds:
DiodesZetex AP7312-1233W6-7, Dual LDO Regulator, 150mA, 1.2 V, 3.3 V, ±2% 6-Pin, SOT-26
ON Semiconductor NCP612SQ50T2G, LDO Regulator, 200mA, 5 V, 2% 5-Pin, SC-70 to ON Semiconductor NCP51460SN33T1G, Fixed Series Voltage Reference 3.3V, ±1.0 % 3-Pin, SOT-23
TL431 with a 317 as a CCS as well as TL431 without CCS
1117 adj
op amp AVCCL+R to be created following ESS guide doc
My goal in this project is to improve on the last DAC build but also to allow parts to swap out for testing, measuring and overall learning. for this reason i have fitted a board on top of the dac with terminal connections for all items that require power as well as dac outputs. This is the reason for each VR board being different. I fully appreciate this has been explored and confirmed elsewhere i just want to measure it for myself and understand clearer as i go mistakes and all. obviously the extra wiring needed will bring its own issues, and learning from it.
x6 secondaries 5.2V AC
LDO boards being used following data sheet builds:
DiodesZetex AP7312-1233W6-7, Dual LDO Regulator, 150mA, 1.2 V, 3.3 V, ±2% 6-Pin, SOT-26
ON Semiconductor NCP612SQ50T2G, LDO Regulator, 200mA, 5 V, 2% 5-Pin, SC-70 to ON Semiconductor NCP51460SN33T1G, Fixed Series Voltage Reference 3.3V, ±1.0 % 3-Pin, SOT-23
TL431 with a 317 as a CCS as well as TL431 without CCS
1117 adj
op amp AVCCL+R to be created following ESS guide doc
Maybe more issues than you will be able to learn about, given the plan as described. Allo used 5-pin regulators, either LP5907 or XC6219 judging from the SMD code on them, and found out how much input and output filtering it took for each regulator before no further FFT improvements could be seen. For the two clocks, each clock got its own regulator and there was a long string of filter components before and after the regulators, partly to maintain lowest measured noise and lowest impedance across the band at each load power input. Although clocks had longer filters than other loads, all the regulators had input and output pi-filters with parallel caps in the legs to ground, and some series legs were frequency compensated. It took them many months of measurements to optimize measured performance, but in the end the dac sounded very good if powered off a clean linear external supply, then letting the local LDOs do the final dropping to 3.3v. All opamps connected to clean +-15 external power if best sound quality was desired. They took an ES9038Q2M and made a dac out of it that (only with external linear supplies) sounded like > $1k dac. Pretty good!
Despite all that it still didn't sound as good as a Benchmark DAC-3, which itself lost to another dac in listening tests. The best sound dac of the three was my modded ES9038Q2M, but I used my own methods that were not intended to minimize cost. I just wanted to see how good I could get the dac to sound, and so there was some overkill involved in the design, no question. It came out preferred over DAC-3 by listeners, but didn't sound better in every single way. I figured it was probably possible to do better and it is, Topping D90 is better than DAC-3 in all ways.
Last edited:
Here is my opinion about modding DACs at this point. It might not be worth it anymore if saving money is the end game. If it is for learning and fun, then OK.
Here is the basis for my conclusion. As Mark described, the intricacies of design now depends on a lot of small factors to get the desired performance to a high level. These iterations and measurements and attempts are largely nor able to be done by the casual modder.
This was painfully obvious when I got a Khadas Tone Board. the size and the close proximity was really good for the asking price of $99, one could do no better at that price point. The same designer went onto forming his own company and then makes a higher performing one for $199 with a CNC machined case... all finished with superior IV circuits and he eliminated the ESS hump. LA-QXD1 Also look at the nose levels this is achieving at $199. Not saying it sound good as I have not heard it but the numbers look like some serious homework was done. Nothing that a DIYer can likely achieve at the kitchen table.
Remember the Khadas has no enclosure. Both mentioned are only a sample of what is available at a price point and the performance levels keep changing to the better in time.
Some companies in China now are doing first class work like in the D90. If I counted up the dollars spent on modding, parts and power supplies etc. I would have possibly spent more than the cost of Topping D90. However, I did learn a few things and caught up some on the current DAC world and what is happening. That was the fun factor and learning. However, I am pretty sure for the asking price of a D90, I will be unable to match it sound. The circuit layout at the crazy noise and distortion levels these new DACS are at are unachievable by the DIYer starting with suboptimal circuits.
So if modding low cost is for fun, then Ok. Otherwise just purchase a well designed machine from somewhere.
Now some other DACs might flavor the sound but I am leaving that territory. I think a DAC should be as neutral possible being the source. You can flavor the sound downstream via speakers, amps, preamps headphones etc. That way a more predictable outcome is likely to be had.
Just got a D90 MQA on its way. I want to sample what MQA has to offer on Tidal in my main rig. My 9028pro comes to my office. My KTB, moves to the bedroom.
Here is the basis for my conclusion. As Mark described, the intricacies of design now depends on a lot of small factors to get the desired performance to a high level. These iterations and measurements and attempts are largely nor able to be done by the casual modder.
This was painfully obvious when I got a Khadas Tone Board. the size and the close proximity was really good for the asking price of $99, one could do no better at that price point. The same designer went onto forming his own company and then makes a higher performing one for $199 with a CNC machined case... all finished with superior IV circuits and he eliminated the ESS hump. LA-QXD1 Also look at the nose levels this is achieving at $199. Not saying it sound good as I have not heard it but the numbers look like some serious homework was done. Nothing that a DIYer can likely achieve at the kitchen table.
Remember the Khadas has no enclosure. Both mentioned are only a sample of what is available at a price point and the performance levels keep changing to the better in time.
Some companies in China now are doing first class work like in the D90. If I counted up the dollars spent on modding, parts and power supplies etc. I would have possibly spent more than the cost of Topping D90. However, I did learn a few things and caught up some on the current DAC world and what is happening. That was the fun factor and learning. However, I am pretty sure for the asking price of a D90, I will be unable to match it sound. The circuit layout at the crazy noise and distortion levels these new DACS are at are unachievable by the DIYer starting with suboptimal circuits.
So if modding low cost is for fun, then Ok. Otherwise just purchase a well designed machine from somewhere.
Now some other DACs might flavor the sound but I am leaving that territory. I think a DAC should be as neutral possible being the source. You can flavor the sound downstream via speakers, amps, preamps headphones etc. That way a more predictable outcome is likely to be had.
Just got a D90 MQA on its way. I want to sample what MQA has to offer on Tidal in my main rig. My 9028pro comes to my office. My KTB, moves to the bedroom.
Last edited:
Mike,
I don't know if things are that bad for modder's. We now know a great deal about how to mod ES9038Q2M quite well on a first try. Starting from scratch just isn't the best way to go. Its not a simple nut to crack. Allo has been working on ES9038Q2M dacs from since about when I started modding dac boards. They are now coming out with their latest version, called Revolution, which features the latest improvements they learned how to make in ES9038Q2M implementation. While I don't know their exact most recent solutions, I do know the problems that are trying to solve and how to go about solving them myself. Its just that I'm more interested in AK4499, and other things right now.
I would suggest to prospective dac modder's that they learn the basics of what we already known before deciding on what experiments should be run. Otherwise, they could spend years and still not produce a dac that comes close to competing with well designed $200 dacs.
Also, I think the approach one takes when the goal is to build one of something is very different than the approach one takes when the goal is to build 100's or 1000's. In particular, its not worth spending a lot of time figuring out how to optimize cost verses performance if only building one item. Allo's solution with low cost regulators and lots of filtering is designed to be low cost to manufacture while still delivering high performance. Not only does it take a lot of time and effort, but it takes a large investment in test equipment. The total cost to develop is actually quite high unless they can make it up selling many dacs to distribute the costs around. So, when modding cheap Chinese ES9038Q2M boards I would take a different approach. It can be fun and a great learning experience, but I would agree if all one wants is a dac that sounds good there are some very good and price competitive commercial designs to choose from.
I don't know if things are that bad for modder's. We now know a great deal about how to mod ES9038Q2M quite well on a first try. Starting from scratch just isn't the best way to go. Its not a simple nut to crack. Allo has been working on ES9038Q2M dacs from since about when I started modding dac boards. They are now coming out with their latest version, called Revolution, which features the latest improvements they learned how to make in ES9038Q2M implementation. While I don't know their exact most recent solutions, I do know the problems that are trying to solve and how to go about solving them myself. Its just that I'm more interested in AK4499, and other things right now.
I would suggest to prospective dac modder's that they learn the basics of what we already known before deciding on what experiments should be run. Otherwise, they could spend years and still not produce a dac that comes close to competing with well designed $200 dacs.
Also, I think the approach one takes when the goal is to build one of something is very different than the approach one takes when the goal is to build 100's or 1000's. In particular, its not worth spending a lot of time figuring out how to optimize cost verses performance if only building one item. Allo's solution with low cost regulators and lots of filtering is designed to be low cost to manufacture while still delivering high performance. Not only does it take a lot of time and effort, but it takes a large investment in test equipment. The total cost to develop is actually quite high unless they can make it up selling many dacs to distribute the costs around. So, when modding cheap Chinese ES9038Q2M boards I would take a different approach. It can be fun and a great learning experience, but I would agree if all one wants is a dac that sounds good there are some very good and price competitive commercial designs to choose from.
Last edited:
For some time now, I have been listening to the separate power supply for AVCC-L resp. AVCC-R with some AD797. My first mod back then was a decent power supply, +/- 15V and a separate supply for the reclocking part of the I2Sover USB.
As of immediate instant, it is obvious what a difference the separate 2 x 3.3V AVCC supply makes which has already been mentioned various times here before. The dac is now really making music. There is no point in summarizing again all the relevant aspects of what makes it more musical: you are encouraged to just try for yourself. A keeper: one is more than zero, is my motto. Even, or especially, in the digital domain zeros and ones are not all the same. This was also illustrated firmly by my musings with the iFi iGalvanic and Micro3.0 from the laptop to the I2SoverUSB from JLSounds.
As for the implementation, I have kept it simple as possible. A separate board fitting under the ES9038q2m board. On the ES9038q2m board only the smd resistors were removed. As I only use solid core wire (thanks, late Allen Wright), the wires to the AD797 board are secured with shrink tubing across the existing elcos.
Next up is a separate supply for the clock. It looks like only one sms resistor is to be removed from the trace that used to serve the AVCC too. Then an external 3.3V supply can be added. While it is not necessary to constantly supply the whole dac, keeping a constant power on the clock definitely has an advantage. In my view, the clock needs the most time to warm up (please no flames here). In order to obtain a constant supply, a rechargeable 9V battery comes to mind. But then a regulator should preferably eat as little current as possible. As the LM431 requires a Iz min of 0.4mA (400uA), the TLV431 with only 30uA of Iz (in the datasheet called Ik) looks like an interesting candidate.
Should you have another suggestion, please share.
In order to calculate the total current though Rs, the current of the clock should be included. The crystal on the rather old version of the ES9038q2m board says CBN6ZSK (E100.00), and not much is to be found about it.
Can you give an estimate of the probably very low current?
As of immediate instant, it is obvious what a difference the separate 2 x 3.3V AVCC supply makes which has already been mentioned various times here before. The dac is now really making music. There is no point in summarizing again all the relevant aspects of what makes it more musical: you are encouraged to just try for yourself. A keeper: one is more than zero, is my motto. Even, or especially, in the digital domain zeros and ones are not all the same. This was also illustrated firmly by my musings with the iFi iGalvanic and Micro3.0 from the laptop to the I2SoverUSB from JLSounds.
As for the implementation, I have kept it simple as possible. A separate board fitting under the ES9038q2m board. On the ES9038q2m board only the smd resistors were removed. As I only use solid core wire (thanks, late Allen Wright), the wires to the AD797 board are secured with shrink tubing across the existing elcos.
Next up is a separate supply for the clock. It looks like only one sms resistor is to be removed from the trace that used to serve the AVCC too. Then an external 3.3V supply can be added. While it is not necessary to constantly supply the whole dac, keeping a constant power on the clock definitely has an advantage. In my view, the clock needs the most time to warm up (please no flames here). In order to obtain a constant supply, a rechargeable 9V battery comes to mind. But then a regulator should preferably eat as little current as possible. As the LM431 requires a Iz min of 0.4mA (400uA), the TLV431 with only 30uA of Iz (in the datasheet called Ik) looks like an interesting candidate.
Should you have another suggestion, please share.
In order to calculate the total current though Rs, the current of the clock should be included. The crystal on the rather old version of the ES9038q2m board says CBN6ZSK (E100.00), and not much is to be found about it.
Can you give an estimate of the probably very low current?
Last edited: