Still, it seems like quality 1794 dac kits are liked a lot by a lot of people, but the good kits cost more than 200 Euros. One example: https://audio-creativeshop.nl/product/dddac-1794-mk3-module/ ...only € 458,68 (€ 555,00 incl. VAT). (power supply kit available at extra cost)
Yet, for some reason there isn't so much interest Marcel's RTZ dac. Maybe some change on that when a final most recommended output stage is settled on. However, like all things people spend money on, there has to be a compelling story, and there have to be enough in the way of positive reviews.
Yet, for some reason there isn't so much interest Marcel's RTZ dac. Maybe some change on that when a final most recommended output stage is settled on. However, like all things people spend money on, there has to be a compelling story, and there have to be enough in the way of positive reviews.
Maybe usefull things ?
Here an old one with the ESS https://www.diyaudio.com/community/threads/ess-sabre-reference-dac-8-channel.117238/post-1422780
https://www.akm.com/us/en/about-us/news/2024/20241111-ak4497svq/
This thread is for discussion and advice from DAC developers.
The only change in the new IC is the addition of a programmable FIR filter.
Wouldn't this be a fun toy for digital designers?
The AK4497 is a good choice for a mini DAC build, as the current output IC and the peripheral circuitry can be simplified more than the AK4499EX.
I will continue building the AK4497, as I am interested in making thin or compact tabletop devices.
I will receive samples from AKM soon.
The board design is almost...
@analog_sa : because for the numbers that those modern dac chips exhibit I am not sure a tube is a good thing for the I/V stage layout. But with voltage dacs, tubes liked by Grunf can be added for gain, buffer... I let the specialists of transimpedance for digital output current stage to input though, it's above my head.
For 24 bits depth you need already 144 dB dynamic range, does tube I/V stage can acheive that w/o errors conversion? Human ears is limited to 21 bits depth ? And best DACs exhibits 21 to 22 bits accuracy when I read Stereophile ?!
Hi,
It is my second project of DAC, first was on AK4499EQ but after fire in AKM factory projest is useless. Now AK4499EQ chip is to expensive, over 1k USD in Ebay. Now afrer relese new chip on Munich AudioShow I downloaded Datasheet and started designing the new DAC on AK4499EXEQ.
DAC will be Dual Mono, one PCB DAC + Power Supply - one chanell. DAC will be working in Asynchronous mode.
Now it looks like this:
It is my second project of DAC, first was on AK4499EQ but after fire in AKM factory projest is useless. Now AK4499EQ chip is to expensive, over 1k USD in Ebay. Now afrer relese new chip on Munich AudioShow I downloaded Datasheet and started designing the new DAC on AK4499EXEQ.
DAC will be Dual Mono, one PCB DAC + Power Supply - one chanell. DAC will be working in Asynchronous mode.
Now it looks like this:
Here an old one with the ESS https://www.diyaudio.com/community/threads/ess-sabre-reference-dac-8-channel.117238/post-1422780
Hello,
in addition to my Headphone Amp I want to design a DAC board with SPDIF input (44,1-192kHz) with the ES9039Q2M in HW-mode. I am not an electrical engineer but was successful with some smaller projects with KiCad and I'm asking for a bit of support.
I started with the design based in the datasheet:
in addition to my Headphone Amp I want to design a DAC board with SPDIF input (44,1-192kHz) with the ES9039Q2M in HW-mode. I am not an electrical engineer but was successful with some smaller projects with KiCad and I'm asking for a bit of support.
I started with the design based in the datasheet:
- At the moment I only have +24V available, what would be the recommendation to get +5V? (LM337/LM7805..)
- What voltage is needed for the opamps (OPA1612)?
- Some values for C are not clear for me from the datasheet (marked with ???)
- In the datasheet it says...
https://www.akm.com/us/en/about-us/news/2024/20241111-ak4497svq/
This thread is for discussion and advice from DAC developers.
The only change in the new IC is the addition of a programmable FIR filter.
Wouldn't this be a fun toy for digital designers?
The AK4497 is a good choice for a mini DAC build, as the current output IC and the peripheral circuitry can be simplified more than the AK4499EX.
I will continue building the AK4497, as I am interested in making thin or compact tabletop devices.
I will receive samples from AKM soon.
The board design is almost...
In the AK4499EXEQ data sheet, they have a recommended mono external circuit summing the two hot and cold lines together after a passive LPF stage. I am using an active LPF through an FDA and need to sum the signals before the LPF. What is the ideal way to sum two hot and two cold signals? The data sheet is unfortunately very vague when it comes to mono descriptions.
This is the (v2 = improved) output stage from the ESS data sheet:
I understand that thevalues of the components around the op amp are no longer critical to avoid the hump, and no published data of 9039 DACs that I have seen exhibit the hump.
Does anyone know what the three op-amp version with slightly improved THD+N looks like? I the third op-amp maybe just for buffering the reference voltage, as a symmetrical output should have an even number of amps?
Is there a simpler way to get unbalanced output? Just using one side of the symmetrical output signal seems odd. The Chinese...
Does anyone know what the three op-amp version with slightly improved THD+N looks like? I the third op-amp maybe just for buffering the reference voltage, as a symmetrical output should have an even number of amps?
Is there a simpler way to get unbalanced output? Just using one side of the symmetrical output signal seems odd. The Chinese...
@analog_sa : because for the numbers that those modern dac chips exhibit I am not sure a tube is a good thing for the I/V stage layout. But with voltage dacs, tubes liked by Grunf can be added for gain, buffer... I let the specialists of transimpedance for digital output current stage to input though, it's above my head.
For 24 bits depth you need already 144 dB dynamic range, does tube I/V stage can acheive that w/o errors conversion? Human ears is limited to 21 bits depth ? And best DACs exhibits 21 to 22 bits accuracy when I read Stereophile ?!
Ignoring for simplicity that quantization noise is not quite the same as additive noise, that ICs are nowadays usually designed by teams rather than by individual designers and that triangular dither adds 4.77 dB to the quantization noise:
Suppose you were an IC designer and your boss would ask you to design a 120 dB dynamic range sigma-delta DAC using as little chip area and current as possible. According to Bennett's equation for quantization noise, 20 bit corresponds to 122.17 dB, so you could choose an input interface with a word length of 20 bit.
However, that would mean that you spend 10-2.17/10 • 100 % ~= 60.63 % of your noise power budget on the quantization noise at the input, leaving only 39.37 % for everything else: analogue circuit noise, jitter, requantization in the digital sigma-delta modulator and interpolation chain. You would have to design everything else for 124.05 dB dynamic range to still meet the overall spec.
Had you chosen a 24 bit interface, 120.01 dB would suffice. Analogue chip area is usually roughly proportional to the signal-to-noise power ratio, so you only need 40 % of the analogue area then.
A chip like the 90392QM needs hardly any voltage amplification and has an output impedance below 400ohms (200 in mono). Perhaps suboptimal for driving cables directly, but adding any open loop tube stage will reduce the resolution to below 16 bits... a worthwhile trade?
The often used tube - transformer IV may have the benefit of subtle out of band filtering, but again, at what cost to noise and distortion?
I am not a purist, sound manipulation is fine by me, but once done at the source it cannot be undone. It might make things easier when using crappy amps and screechy speakers, but it bothers me conceptually.
As some means of analogue VC seems mandatory imho, a line stage is still worth having. So why not have a few, with different distortion profiles, while keeping the dac electronics at ppm levels of distortion?
Then there is the added annoyance of having high voltage and high temperature sources in a box where they have no business being.
At cd resolution things are different. Tube IVs can indeed be great.
We basicly think the same so for the tube I/V aera and modern high bits dac chips with current output.
But for me I/V passive with a resistor just after the DAC current output is not a good option whatever stage you put after. Most gain with traffo are not good enough off shelves, and very good not sourcable (Cinemarg, ...) is expensive or hard to source as well as the made on spec double wires, amorphous, silver virgin smoothed silver, etc...
So the 9039QM is the best choice for Grunf. He should have already chose this. Ideal low output impedance for his Walter Jung AD811 stage & reg needeing lower impedance source than 2K ohms for the inverted input pin of that CFA.
But for me I/V passive with a resistor just after the DAC current output is not a good option whatever stage you put after. Most gain with traffo are not good enough off shelves, and very good not sourcable (Cinemarg, ...) is expensive or hard to source as well as the made on spec double wires, amorphous, silver virgin smoothed silver, etc...
So the 9039QM is the best choice for Grunf. He should have already chose this. Ideal low output impedance for his Walter Jung AD811 stage & reg needeing lower impedance source than 2K ohms for the inverted input pin of that CFA.
ESS dacs like the I/V opamps to be right next to the dac chip, with uninterrupted ground plane between the opamps and the dac chip. Same for similar dac chips such as those from AKM. IOW, they are not so good for use with modular output stages.
OTOH, here is what one user said about the versatility of Marcel's RTZ dac:
Well I've only tried the clock board with Crystek's and NDK NZ2520SDA , both of those with the clock board gave better sound to my ears than just the NDK's on the i2soverusb . I've not tried the SC-pure yet so can't comment how much gain you get with them over the cheaper alternatives. The NDK and Crysteks give a little different sound .
The thing I like about Marcels dac is it makes trying these things easy , same with the output stage so its no problem reverting back .
https://www.diyaudio.com/community/threads/general-purpose-dac-clock-board.413001/post-7982385
OTOH, here is what one user said about the versatility of Marcel's RTZ dac:
Well I've only tried the clock board with Crystek's and NDK NZ2520SDA , both of those with the clock board gave better sound to my ears than just the NDK's on the i2soverusb . I've not tried the SC-pure yet so can't comment how much gain you get with them over the cheaper alternatives. The NDK and Crysteks give a little different sound .
The thing I like about Marcels dac is it makes trying these things easy , same with the output stage so its no problem reverting back .
https://www.diyaudio.com/community/threads/general-purpose-dac-clock-board.413001/post-7982385
I'm going to take a little break this weekend and I find my topic in chaos 🤣 . I see you've already determined what I need to solder 🤣.
ES3039Q2M is currently out of stock at Mouser, but that's not a problem if I decide on it. While I draw all the PCBs and get all the parts, there is time to get two pieces.
However, I've been drawn to the AK4191/AK4499EX or maybe even an eight-channel ESS.
I also think the PCM1794A is not a bad idea, and I already have them in my drawer, which increases the chance of realization.
I don't want to limit myself with my budget. The regulators I'm using in my DAC now have four ADA4897s and four ADA4625s, all eight regulators are shunt with higher dissipation. The regulators themselves aren't cheap, but that's how it is if we want to outperform the LT3042, ADM7150,....
The I/V stage can be with an AD811, some kind of discrete current conveyor or grounded grid tube I/V (not passive). I have no limitations.
ES3039Q2M is currently out of stock at Mouser, but that's not a problem if I decide on it. While I draw all the PCBs and get all the parts, there is time to get two pieces.
However, I've been drawn to the AK4191/AK4499EX or maybe even an eight-channel ESS.
I also think the PCM1794A is not a bad idea, and I already have them in my drawer, which increases the chance of realization.
I don't want to limit myself with my budget. The regulators I'm using in my DAC now have four ADA4897s and four ADA4625s, all eight regulators are shunt with higher dissipation. The regulators themselves aren't cheap, but that's how it is if we want to outperform the LT3042, ADM7150,....
The I/V stage can be with an AD811, some kind of discrete current conveyor or grounded grid tube I/V (not passive). I have no limitations.
I know that very well, they are not 'old school' DACs ic and you can do whatever you want with them. On the same PCB there would be discrete regulators right next to the ICs themselves which would also take up a lot of space. That's the only way it would make sense.
A cheaper and less painful solution would be IC regulators but I'm not thinking about them right now.
Are they really needed ? Have those new dac chip not sota PSRR already ?
Better to list the real needed datas : bandwith and max linear low impedance of the regs where it is important for the dac chip.
How much flat bandwith length is needed for the AD811 input stage ? 50 Mhz, 100 Mhz, more ?
Is its noise level is low enough for a modern 32 bits dac chip on the needed bandwith ?
To save monney on the several layers board, maybe you can do the reg plugged vertically (with a good metal cabinet around) as you did already ?
There are more than 1700 ES9039EQ2M at Mouser seen from France.
Conceptually, you must check the solutions before drawing the board.
Better to list the real needed datas : bandwith and max linear low impedance of the regs where it is important for the dac chip.
How much flat bandwith length is needed for the AD811 input stage ? 50 Mhz, 100 Mhz, more ?
Is its noise level is low enough for a modern 32 bits dac chip on the needed bandwith ?
To save monney on the several layers board, maybe you can do the reg plugged vertically (with a good metal cabinet around) as you did already ?
There are more than 1700 ES9039EQ2M at Mouser seen from France.
Conceptually, you must check the solutions before drawing the board.
Grunf, you are right IC regulators like LT3042/3045 can not be beat easily but they are expensive. I don't know why but in my devices TPS7A4700 always is preferred when listening to devices. Its cousins TPS7A2033 and its versions are both cheap (50 Eurocent), small and quite good.
Just checked what affordable single board ES9039Q2M designs exist besides the pretty good SMSL devices. I then saw this one:
https://de.aliexpress.com/item/1005007308963856.html?src=google&pdp_npi=4@dis!EUR!53.37!26.99!!!!!@!12000040189607234!ppc!!!&src=google&albch=shopping&acnt=272-267-0231&isdl=y&slnk=&plac=&mtctp=&albbt=Google_7_shopping&aff_platform=google&aff_short_key=UneMJZVf&gclsrc=aw.ds&&albagn=888888&&ds_e_adid=726717691648&ds_e_matchtype=search&ds_e_device=c&ds_e_network=g&ds_e_product_group_id=297546631617&ds_e_product_id=de1005007308963856&ds_e_product_merchant_id=533169152&ds_e_product_country=DE&ds_e_product_language=de&ds_e_product_channel=online&ds_e_product_store_id=&ds_url_v=2&albcp=22054759359&albag=172828878419&isSmbAutoCall=false&needSmbHouyi=false&gad_source=1&gbraid=0AAAAAoukdWNUzA9mIFzB1I6IdtLWDDbLT&gclid=EAIaIQobChMIrJrh_sL7jAMVkJ9oCR1oBBgQEAQYASABEgKwA_D_BwE
Not exactly a price I was expecting knowing ES9039Q2M is 16 Euro ex taxes at distributors here!
Just checked what affordable single board ES9039Q2M designs exist besides the pretty good SMSL devices. I then saw this one:
https://de.aliexpress.com/item/1005007308963856.html?src=google&pdp_npi=4@dis!EUR!53.37!26.99!!!!!@!12000040189607234!ppc!!!&src=google&albch=shopping&acnt=272-267-0231&isdl=y&slnk=&plac=&mtctp=&albbt=Google_7_shopping&aff_platform=google&aff_short_key=UneMJZVf&gclsrc=aw.ds&&albagn=888888&&ds_e_adid=726717691648&ds_e_matchtype=search&ds_e_device=c&ds_e_network=g&ds_e_product_group_id=297546631617&ds_e_product_id=de1005007308963856&ds_e_product_merchant_id=533169152&ds_e_product_country=DE&ds_e_product_language=de&ds_e_product_channel=online&ds_e_product_store_id=&ds_url_v=2&albcp=22054759359&albag=172828878419&isSmbAutoCall=false&needSmbHouyi=false&gad_source=1&gbraid=0AAAAAoukdWNUzA9mIFzB1I6IdtLWDDbLT&gclid=EAIaIQobChMIrJrh_sL7jAMVkJ9oCR1oBBgQEAQYASABEgKwA_D_BwE
Not exactly a price I was expecting knowing ES9039Q2M is 16 Euro ex taxes at distributors here!
Last edited:
Really? Input is not sensible to stray capacitance, parasitic stray from atoo much near top ground, etc ?
TSA7A family or AD7150 is better to me than LT1963/... . LT3042 despite its noise floor never gave good result each time I tried it...
But can they all of them beat a Super few Farad caps like some Maxwell with very low ESR ??
But can they all of them beat a Super few Farad caps like some Maxwell with very low ESR ??
Last edited:
There are some very high RF frequencies on the dac analog outputs that need a low impedance return path the dac chip ground. Also, stray capacitance to ground at the input of the I/V opamp will cause frequency peaking.
... That's why I didn't understand the non interrupted gnd to the op amp input you made. the ground loop will not be much bigger if you remove few ref around and below the op amps areas needed. Anyway need to be near from the dac ic chip anyway !
The LT3042 is the worse of the 4 sound wise.
The LT3042 is the worse of the 4 sound wise.
These regulators are not best choice for ES9039Q2M because IC consuming less current.
Even VDD is internaly suplied... And specifies trough the AVDD consumption datas.
.
The best way (and previous models of ES like that way) is Vref ciruit for 3.3V with AD797 which can supply up to 50mA
way more than DAC IC need, 24 mA cca.
.
In hardware mode You can have even SPDIF option. If You want to use I2S just disconnect or stop SPDIF...
Fiter sections are locked to factory options, no dual mono mode. But everything else is way more simple to make basic try.
And to hear and measure IC without of manipulating the registers.
.Output current is not so small about 7.5mA p-p
Output impedance is 380 ohms BUT +/- 15%. Hope that it not between the channels in the same chip but just different series of bunch... 🙁
.
You can run first try with just one 3.3V worst case power supply. Just to check the sound and measure basics.
And latter work on PCB design
I am preparing to do the same
this is the PCB adapter with exposed pin to gonnect to gnd
https://www.artekit.eu/products/bre...-dip/qfn-32-to-dip-adapter-4mm-x-4mm-p0-40mm/
ES9039Q2M power (Hardware mode 2)
GND 30
GND 33 exposed pad pin at the bottom
AGND_DAC1 4 (Analog PS)
AGND_DAC2 5 (Analog PS)
name pin V mA mW
DVDD 9 1.2 (Internally Supplied. Decoup. only)
AVCC_DAC1 1 3.3 6.8 22.44 (Analog PS)
AVCC_DAC2 8 3.3 6.8 22.44 (Analog PS)
VCCA 31 3.3 0.5 1.65
AVDD 8 3.3 9.4 31.02
Total (3.3V) 23.5
Total mW 77.55
Power consumption 78
ALL PS pins decoupling 1uF
MAX MCK frequency 50 MHz
GPIO4 HW modes 16-18 S/PDIF stream input.
The ES9039Q2M has an integrated S/PDIF decoder that can be accessed in either Asynchronous Hardware or Software modes.
For Hardware mode, the S/PDIF input is on GPIO4 using HW modes 16-18.
S/PDIF input stream must be disconnected in order to use other input formats.
S/PDIF is transmitted over a single signal line using dual phase encoded data, which allows for clock extraction from the data signal line.
Output impedance
each DAC output (+ and -)
DAC1 3 Rdac 390 +-15% Ohm
DAC1B 2 Rdac 390 +-15% Ohm
DAC2 7 Rdac 390 +-15% Ohm
DAC2B 6 Rdac 390 +-15% Ohm
+/- 15% (390 Ohm) = +/- 58.5 ohm (331.5 to 448.5) !!!
AVCC_DAC = AVCC = 3.3V
Current output amplitude
Full-scale out 1000x0.886xAVCC_DAC/Rdac 7.496 mAp-p
Current output offsets
Bipolar zero out 1000x(AVCC/2-Vg)/Rdac 0 mA (for Vg=AVCC/2=1.65V)
Vg = virtual ground voltage value
4.34 mA (for Vg=0V, GND)
Last edited: