That's not true, a complimentary pair of MOSFET like the FD8958 turns on and off in less than 30ns when drived by CMOS logic, absolutely safe for 384kHz and also 768kHz sample rate. Moreover the slower transiction of the FDS8958 is a benefit, since it decrease ground bounce and ring.
I wonder how one can name the LVC595 "higher performance" (against the FDS8958), I don't think MSB uses them to decrease the performance of the DAC and to increase space and switching noise (just the opposite).
But there is a simple reason to use 1 x LVC595 instead of 8 x FDS8958 (+ another mosfet pair to switch to ground and the logic to drive each single bit): THE COST.
You save a bit!
You're incorrect on the FDS8958, the problem with a part like that is that they have too large dies, you get the low Rds, but to get the 30 nS switching speed you need to drive them with 1A, no FPGA or logic chip can deliver that.... And if you look at the datasheet, the Ton is lower than the Toff, meaning you get shoot through and that's bad with the low Rds, make a lot of switching noise on the power rails.... No, I would never use a part like that, and neither do MSB Tech.
But you can get more fitting dual mosfets, with Rds around 1 ohm. But it's a design question how you want to make things, some use logic chips directly, others use dual mosfet.... I'm happy with my design, although there is always room for improvement, I just haven't found anything better for my needs than the small LVC595 in qfn packaging....
You claimed "I know MSB Tech use them" (FDS8958), so I presume it's true.
Anyway, the small time difference between T-On and T-Off is not a great issue, logic gates can easily drive the mosfet pair with no more than 10-15 mA.
Although this causes current spikes on the mosfet's gates, there is no reflection to the switch output. The relatively slowness of the devices helps to mitigate the negative effect of the inrush current during the transiction due to the mismatch.
There are no ground bounce or ringing at the mosfet pair output.
Attachments
You can't both claim fast switching and then say 15 mA drive is fine, mosfet switching speed is relative with drive current, basic mosfet knowledge....
Yes, I said MSB use those small dual mosfet, not a specific p/n, and definite not the FDS8958, just take a look at the picture of a MSB board, they use 6 pins sot parts and the FDS8958 is 8 pins SO.
15 mA current spike is not reflected to the output and 40ns to get the output stable is fast switching for 192kHz sample rate.
And keep in mind that as the rds on grows the ladder precision decreases. The high output impedance of the 595 (at least 15 ohm) heavily affects the monotonicity of the DAC, you loose several bit of precision (assuming you are using 10K as 2R, the lower the resistance the worst the scenario).
Attachments
That plot of yours don't show the current shoot though, so it's worthless....
And those 40nS of yours is a looong time, errors in the time domain also creates distortion....
Actually, I measured the LVC595 to typically 14 ohms at 4V, and with an old part like that, on a mature process, you can expect they will actually typically be like that. And once you know their output resistance, it's easy to match the resistors to it, when you order enough you can get any value....
And then you can get those 0.0035% THD that I typically measure on the newer boards with 0.01% resistors.
Those 40 ns are not time domain errors, they are a constant delay at switching time, simply a smoother output wave, easier to be integrated, that's a benefit.
Instead you should worry about the time domain and frequency domain errors (phase noise) of your Si570, with its horrible phase noise performance.
How DA-8 compare with Denafrips? It use LVC574A, similar to LVC959 in @soekris DAC, but featuring parallel inputs instead of shift registers. Parallel driving is claimed producing less switching noise. It is powered with 3.3V, no extra LDO regulators close to the resistors ladder. AUDIO-GD DA-8 DAC Module R2R 24Bit / 192kHz Mono (Pair) - Audiophonics
I don't understand what they mean oversampling, there. I own R2R11, I'd like to learn more about technology.
------
Type: DAC R2R Mono X2
Oversampling: X8 384kHz 24 bit / 3072 Khz
CPLD: CMOS CPLD Xilinx XC2C64A 450Mhz
OUR Yes <---------------------------------------------- what?
Components: 48 high-precision resistors in parallel 0.1% (sorting tolerance 0.05%)
Format: 64 BCLK (Base Clock)
I don't understand what they mean oversampling, there. I own R2R11, I'd like to learn more about technology.
------
Type: DAC R2R Mono X2
Oversampling: X8 384kHz 24 bit / 3072 Khz
CPLD: CMOS CPLD Xilinx XC2C64A 450Mhz
OUR Yes <---------------------------------------------- what?
Components: 48 high-precision resistors in parallel 0.1% (sorting tolerance 0.05%)
Format: 64 BCLK (Base Clock)
Last edited:
Different rise and fall times plus noise on the mosfet drivers power supply equals time domain errors. Basic knowledge.
Stop complaining about my oscillator choice again, the Si570 is a great part. If you want something else, design your own DAC, unless that's to complicated for you.
The DA-8 is another joke by Audio-GD, they can't possible fit any useful "correction algorithm" in a XC2C64A, which have just 64 macrocells....
Just see the test of one of their finished product at ASR:
Review and Measurements of Audio-GD R8 | Audio Science Review (ASR) Forum
Please be honest and don't refer to this website. They were measuring ground loops in the gear, clearly observed in a TotalDAC tests. It was noticed by the reviewer, but instead of dealing with the problem, frequency of a power generator was adjusted to 60Hz to look uniform. On the R2R11 test a noise was so disturbing that it triggered recalibration cycle that couldn't be completed. Read a discussion following my post requesting to explain a presence of both 60Hz and 50Hz in his dashboard plot. Review and Measurements of Totaldac d1-six DAC | Audio Science Review (ASR) Forum
I was asking about design differences, but I gave you an opportunity for self advertising, shame.
Again, there are no time domain errors as you can see in the output plot, nor ground bounce (typical of CMOS tiny gate) neither ringing. It's almost a perfect square wave with rounded angle, that's a benefit.
The Si570 remains a bad oscillators as demonstrated directly by SiLabs publishing the phase noise plot. The Crystek CCHD-957, another bad oscillator, is at least 10 times better than the Si570.
And yes, as soon as I complete the working projects (oscillators with -152 dBc @10Hz phase noise, frequency doublers with -170 db noise floor, digital front-end optical isolated, battery power supply system, and so on) I start with the DAC (a simplified version is already running).
Oh... "too complicated", maybe I will ask you... I'm an ordinary diyer... building a pair of oscillators with phase noise performance in the region of the SOTA oscillators (or better) available on the market (Oscilloquartz BVA and Wenzel Uln), maybe the best oscillators for digital audio ever...
It's very curious you don't understand the point: I build audio devices as an hobby, I have no commercial interest, I have no (theoretically) budget limit, and overall I share my project with the diy community.
You have only business interest in your project, you have to sell your projects while I share my projects, so I understand perfectly your choices, I understand the reason you are using Si570 and LVC595, and so on. I also understand the reason you can't use 0.001% resistors in the ladder, because I know what I paid for them.
Finally, you are nothing to do with diy audio, yet another business man.
Yeah, I'm also not that impressed with ASR, but he do mostly fine measurements with his APx555, which in this case just confim how the DA-8 module will measure.... But the Audio-GD website is also full of incorrect statements and mixups, like the THD plots are reused from another product.... All in no circumstance can you put useful error correction inside a tiny PLD, just not possible. I don't really care if you believe the B*******, just don't complain when I point out the lies....
And I'm not pushing any of my products in this thread, just trying to get facts right.
Last edited:
Again, you can't compare a programmable oscillator with the fixed one. The Si570 IS a fine part.
Yes, just like everybody else in the "Vendor's Bazaar" I sell stuff.... The reason I started selling the dam1021 is to make them available, you need to make some qtys at at time to get a reasonable price.
I understand your reasoning, and you do fine oscillators. Just not something that I need....
A point of CPLD, may be valid, but you found a one incorrect use of a graph on the A-GD website and now (after a kindly remainder) it comes with things like "full of incorrect statements and mixups", not to mention a previous ASR referal for a pure sanity.
So I think, that your statement about CPLD is exaggerated. A fact that you are unable to find this CPLD useful doesn't mean that others didn't do better. More expensive models R1/R28 have FPGA, but still the same CPLD is used to control a ladder, it says something.
It is a known fact that R2R11 sounds great. Kingwa has managed to keep price as low as $350 and a package include very good headphone amplifier and a pre-amp for line outputs.
Last edited:
I don't think you can use "known fact" about subjective opinions.... Anyway, over at SBAF they're are also not fond of the R2R11 (or many other Audio-GD products), and I trust SBAF....
No, my statement about the CPLD is not exaggerated, it's a direct lie that the DA-8 by itself can do any processing in 64 logic cells.... Of course an external larger FPGA helps, but you started out praising the DA-8 module itself....
Just visit SBAF and enter R2R11 in their seach field and you will find many negative opinions on Audio-GD....
Also, the claim that " Parallel driving is claimed producing less switching noise." is also false.... It doesn't really matter how many bits get into the registers at a time, it's the outputs switching that create noise, and a 8 bit shift/storage register has exactly the same number of outputs switching as a 8 bits parallel register....
I could find many more false statements and lies on the Audio-GD website, but I got other things to do, any competent engineer will get a good laugh there....
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.