Yes, the Jung has crazy low impedance. Salas too (I am a Salas fanboy).
OTOH, I did come across this post:
Thing is, Silent Switcher is reputedly good although it switches. Even running it from clean power creates hash. The linear supply I am using now I think is fine. My concern is primarily for people who might not have any way of knowing for sure whether they are getting hash into their supplies that either is not easily regulated, or that tends to find its way around filters, contaminate ground planes, etc., by sneaky means (parasitic coupling, ground coupling, radiated EMI, etc.) that are not shown on a schematic. Only way to really know is to have suitable measurement equipment or ways to substitute and or compare. One cannot a know priori by design that nothing unexpected can happen. You know. This is what worries me. I want other builders besides you and me to have some way of knowing how their device is actually performing, and if not performing well, how to track down and fix the cause.
Quote: Batteries give better soundstage and some comfort while salas shunt sounds more "polite".
So, how does one figure out which is right, or if either one is? That is the problem. You shouldn't hear batteries and shunt regulators. Something is wrong.
Of course, you shouldn't hear reconstruction filters either but that is a different type of problem. It is a deep down design problem, something not fully fixable with Q2M.
So, how does one figure out which is right, or if either one is? That is the problem. You shouldn't hear batteries and shunt regulators. Something is wrong.
Of course, you shouldn't hear reconstruction filters either but that is a different type of problem. It is a deep down design problem, something not fully fixable with Q2M.
Batteries do solve a lot of problems on boards such as these and may level the playing field among users, where reliable power isn't a given. But, unsurprisingly, I'm dubious of the personal impressions front due to its unreliability.
Given the option between the older superregulators and a LT3045/LT3042 or similar brethren, sign me up for the monolithic, if for no reason but layout.
Given the option between the older superregulators and a LT3045/LT3042 or similar brethren, sign me up for the monolithic, if for no reason but layout.
No doubt, it should be possible to measure some of the reported sound effects involved with AVCC power, especially with bass. LF harmonic distortion, IMD (depending, maybe), seem like things to look at. 1kHz might not show much though.
HF hash may just look like a raised noise floor and generally increased harmonic distortion. Depends. How would one know the cause of those symptoms? Or, what if it doesn't sound very good, but no measurement equipment? What if it doesn't sound nearly as good as it could, but one doesn't know how good it should sound if working properly?
As far as people fooling themselves with listening tests, a lot can be helped in that regard by twisting the volume control up and down and listening for things about the sound that don't change with volume. Try comparing those things when trying other power supplies, etc. Already loudness, possibly the biggest contributor to being fooled is at least somewhat compensated for. Unfortunately, I don't have any published data to show on that though, in full disclosure.
HF hash may just look like a raised noise floor and generally increased harmonic distortion. Depends. How would one know the cause of those symptoms? Or, what if it doesn't sound very good, but no measurement equipment? What if it doesn't sound nearly as good as it could, but one doesn't know how good it should sound if working properly?
As far as people fooling themselves with listening tests, a lot can be helped in that regard by twisting the volume control up and down and listening for things about the sound that don't change with volume. Try comparing those things when trying other power supplies, etc. Already loudness, possibly the biggest contributor to being fooled is at least somewhat compensated for. Unfortunately, I don't have any published data to show on that though, in full disclosure.
Last edited:
Mark, there's always going to be a "you must be this tall to ride this ride" effect with technical pursuits. The more you can understand cause and effect (requiring *some* sort of measurement) the better you can engineer a solution. For those without such means, then a hierarchy of improvements off basic principles and common knowledge is the best we can do.
Which, as you and I have both pointed out, is getting a decent I/V circuit bolted in and supplying the board with decent +/- 15V. Everything else gets muddy quickly.
Which, as you and I have both pointed out, is getting a decent I/V circuit bolted in and supplying the board with decent +/- 15V. Everything else gets muddy quickly.
Funny you mention that. I just swapped out the 45mhz stock clock on the hifime usb receiver (SA9227) for a Crystek 957. I really wasn't expecting much especially as it isn't the dac clock. But it sounds excellent. Too afraid to swap out the main clock though. The can looks like a PITA to desolder.
I've only recently (yesterday!) gotten myself an EMU 0404 usb for testing/measurement (and keeping at my work desk to have something infinitely better than a laptop out for headphones), so give me some time to characterize what I have before we loop back to clocks. I do appreciate your offer, that's very generous.
I'm working on a program to do that and more, functionally that is, I'm not so interested in the fancy display part. Presently it just seizes the I2C but to do some things the existing microcontroller can't do. When the bus is released the microcontroller picks up again without any problem.
Right now the only thing my program does is allow reading and writing any registers, including the 2-byte signed int harmonic distortion compensation adjustments for H2 and H3. If I want to adjust the volume or change filters I just release the bus for that and then seize it again.
I am hoping to make a closed version available if people want to try to adjust distortion. Just found a problem though in a difference between the Arduino serial monitor and terminal emulators such as 'putty.' Will work on it.
Anyway, I believe it is possible to use avrdude to extract an image file from an Arduino which by default is called 'mystery.' It is a binary file that I think avrdude can upload to any similar Arduino. This would work with the Arduino called Pro Trinket 3.3v (FTDI). They sell for about $10, and maybe that much again for the USB to FTDI adapter.
Once the mystery file is uploaded (assuming I get the terminal problem fixed) then people should be able use it without knowing how the code works. I could even remove the general register access and leave only the distortion compensation part available. And, I could obscure the details of how that works too, just say enter a number and see what you get. If you don't like it try another number. It wouldn't take very many passes to get it adjusted as long as it isn't made to appear chaotic (which it would not be, at all).
By the way, one can figure out quite a bit by reading the Hi-Fi Duino code. The .ino files are actually text files and can be opened in any text editor. Notepad++ is good free one. Turn on line numbers and start looking around. If you get that far and can't find anything let me know. There is a search term that I don't remember offhand that gets to the interesting part. I read all that before I went NDA. Apparently, it's okay with ESS.
EDIT: Just remembered I don't need a mystery file, I know where find the .hex file for uploading. Have to find out the avrdude command for doing that over USB using the Arduino software uploader. (There are various ways to load Arduinos and the commands can be a bit arcane.)
Last edited:
Well I took off the Cset 33uF chinesium electrolytic caps off the LT3042 and in place put on a hybrid set of caps I had lying around. It consisted of 3x4.7Uf polypropylene with a small bypass. The effects of the 33uF is now gone and it actually improved the midrange. In theory, there really should be no discernible difference from a technical standpoint but it was clear.
Power supplies do make a serious difference for inexplicable reasons. Just like when I left the world of passive preamps and went to the 744/811 preamps with Superregs. There was no going back. I was a convert to preamps. It should not make as large a difference but it does. When I looked at the APOC circuit that B&W hung onto the speakers, the effect should be inaudible but it was like hearing bad cables.
Well I can't wait to hang on my set of IVs and op amp AVCC and see what that brings. I will try and lay out the board so that I can use the opamp or LT3042 and switch between.
Wires, component quality , layout and grounding all play at this point. Can someone replicate what one has concocted I doubt it. I might get clues to try different things but the conclusions we reach may indeed differ. Isn't that part of the fun?
It's only such a slight improvement in the presence of LT3045 (instead AMS)? Did you do the measurements using the unmodded board using 1-2 LDO, and did you powered the controller from the same AMS? I wonder if this level still remained after full modding?!
Last edited:
Superreg is my favourite regulator. Due to the very low noise floor music becomes very real (you are there) and it makes huge dynamic as well. Highly recommended to power the opa-s. As far as know, the lowest voltage is about 5 V.
https://linearaudio.nl/sites/linearaudio.net/files/superreg V2.3.pdf
https://linearaudio.nl/sites/linearaudio.net/files/superreg V2.3.pdf
Last edited:
Just a bit to add to the superreg thing. I have two Silent Switchers from Linear Audio.
First thing I noticed with them is that if they are run off a wall wart then some switching hash from the wall wart makes it through to the +-15 output.
Even if SS is run off a very clean 5v linear power supply they still produce a small amount of some kind of noise at the output as compared to a very clean linear supply. In fact, I could hear the difference with the exact Chines DAC we are working on in this project.
(EDIT: However, SS running from a clean 5v linear supply is probably better than anything other than a very clean linear +-15volt supply.)
In addition, if even if no switcher in front of an LT3045 module, depending of the value of Cset they might be not suitable for directly powering AVCC. In the ebay ads for the modules some show the picture I will post below of the output noise spectrum. Probably fine for powering opamp rails since they have good PSRR at LF. AVCC on the other hand has no PSRR at all. And adding larger output filter caps to LT3045 doesn't help much according to the data sheet because it reduces regulation bandwidth.
Therefore, if you can replace Cset with 22uf it would be better for powering AVCC. Probably even better, use the LT3045 to power an opamp and use well filtered reference voltage and the opamp to power AVCC.
Attachments
Last edited:
If those are the distortion levels being discussed, then he's still using the as-supplied buffer, rather than building up an i/v circuit. The voltage out mode severely hampers the overall performance of the dac, and I'd expect the improvement in ldo to be mostly a lower noise floor, which gets lost in a single quoted thd+n number.