Chris,
Thanks a lot for this. I really like what QuantAsylum is doing. Very clever stuff...
As far as GND on the OTOBUS™ is concerned, I was worried about that as well. If we feel that we have a real problem there, we'll have to use similar connectors with more circuits. The only drawback is that we'll have to modify our design for the rotary encoder brick, but I don't think that a single brick should drive the design of what is supposed to be a very generic interface. I'll give myself another week of background thinking about the issue before returning to the designs of the OTOBUS™ and the rotary encoder brick: my personal experience is that my judgement is often impaired when I work on the same thing for a long uninterrupted period of time. But when I come back to it after a break, I can sometimes see things through a different angle, and solve a problem that seemed intractable before.
Last edited:
Another candidate that has even lower noise and more output current capability is the diyaudio SilentSwitcher.
It achieves total isolation from the mains as it is powered from a USB charger or, for very sensitive applications, a Powerbank.
Check it out.
Jan
Hi Jan,
The Silent Switcher was explored in detail in that section of the thread along with Frex's new power supply design for the AAPSU01 (from the OSVA development thread). [FWIW I have three Silent Switchers here - 1 in the autoranger, another going to be used for PCM4222evm and another for the Bruno Putzey's preamp - they are really handy!].
I was brief in my earlier post, I should elaborate on my reasoning.
Given that the bigger context of this thread is a complex system with a bunch of ARM SoC and FPGA and hybrid FPGA+Arm CPU SoCs combined with the audio gear. I see no value in over engineering the supply to those devices (in terms of expensive low noise regulators etc).
There is the option to use 5-12V supply in to seperate regulators for that auxillary gear and use that supply in to 1/more silent switchers or AAPSU01 like designs to supply the critical/important parts. With good attention to detail on the layout this will be a really great design and my rational brain is fully satisfied.
My monkey brain can't help wonder what if more/better ideas are left on the table. It has seen lots of folks using galvanic isolation on the i2s / i2c between the DSP/processors/displays. I've seen designs that do that and still use the same upstream supply and still get objectively great results (the best kind of great results). What if we isolated the supply to the audio converter (DAC/ADC) and analogue stuff. A few ideas come to mind:
1. old school, heavy, transformers - you get extra transformers or transformers with more secondaries
2. have a separate battery supplied Silent Switcher/AAPSU01/whatever in addition to the power supply to the other components in the device. (as you point out these solve the mains isolation and ground loop problems - probably the biggest and most important issue that warrants solving)
3. have a single supply to the device and use isolated supplies where it makes sense - similar to what QuantAsylum have used.
In any of the above cases PCB layout is the most important in getting to the performance of a chip like the AKM4499. The complexity, cost, space constraints probably drive the choice between those three options.
Hopefully - that's a more coherent version of my 2c on the power supply at a synoptic level. My earlier post was just to point out that there is a 3rd option that I don't think was considered earlier.
Cheers,
Chris
Been there, done that. Ran into problems both with USB charger and a powerbank device. Both power sources produced switching noise that affected dac sound quality. Could be that CM filtering on the output of the USB charger might have attenuated the ground conducted noise enough. Similar noise from the switching regulator inside the power bank type device, even though it was completely isolated from ground. Some filtering might have helped in that case too. Not a plug and play solution in either case for my dac projects. Ultimately, settled on other power supply solutions.
Also, tried one time to see if I could get an off the shelf line powered switcher to work. Put ferrites in the input and output leads, put the switcher inside a metal box, but not luck. Then lined the inside of the box with copper foil. That helped, but still not clean enough. Wrapped the outside of the box in copper foil completely sealing it. At that point it was pretty close to good, but still not as good as a good linear supply to my ears. Didn't end up using that either.
In addition, I have tried to get LDOs such as used by TP to work with analog circuitry in dacs as well as I can do by other means. It certainly would be more convenient if simpler solutions involving less work for me could suffice. Despite trying some variations powering and loading LDOs I didn't find they could beat or equal the best linear solutions I know of for the most critical analog circuitry in dacs. Of course, still seems likely there are ways to engineer around the problems I have encountered so far, I would like to think so. However, it has bee easier so far to go with what works for me in the way of power supplies, and spend my time trying to solve other types of problems for which I have no good solution at all.
Last edited:
El numero uno was me Mark
Due to circumstances of my other commitments, work, projects etc we
discussed this via PM prior but ishizeno has graciously said OK.
On the plus side
a/ I'm happy to offer any help here with the design side where possible.
b/ I'm happy to offer help tweaking the boards for subjective performance
c/ I have lots of clients/friends with various types of recording studios who
can give valuable feedback.
One owns two Ampex ATR124 machines (this is not him)
YouTube
and Studer C37, 1" 2trk for mixdown. A nice source to test VHQ ADC on.
Some good potential users and beta testers.
Throw your hat in the ring Mark. You're experience would be highly valuable.
T
Just a quick clarification: we're not planning to use any FPGA anymore, at least not for the time being. Instead, all the DSP processing will be handled by AM5728 Sitara MCUs located in the trays and plates. This should make programming a lot easier.
Power supply - end game
Hey Ishizeno,
Thanks for all your effort and generosity to develop and share a your high performance designs for A to D and D to A.... Respect!
Re power supply:
Number 1 goal: sound quality.
Number 2 goal: reliability.
Number 3 goal: simplicity... Easy to repeat with 100% consistancy.
Number 4 goal: Safety... No accidents with a 12 volt battery!
Number 5 goal: Opens up the design to vast audience who are wary of assembling mains power supplies.... Like me!
The solution.... :
A pair of 1 Kg LiPo4 batteries, one in use and 1 on charge with auto switching via relay to disconnect from mains charging so its 100% of grid.
Hope this helps?
Cheers
Alex.
Hey Ishizeno,
Thanks for all your effort and generosity to develop and share a your high performance designs for A to D and D to A.... Respect!
Re power supply:
Number 1 goal: sound quality.
Number 2 goal: reliability.
Number 3 goal: simplicity... Easy to repeat with 100% consistancy.
Number 4 goal: Safety... No accidents with a 12 volt battery!
Number 5 goal: Opens up the design to vast audience who are wary of assembling mains power supplies.... Like me!
The solution.... :
A pair of 1 Kg LiPo4 batteries, one in use and 1 on charge with auto switching via relay to disconnect from mains charging so its 100% of grid.
Hope this helps?
Cheers
Alex.
Hi Alex,
Thanks a lot for the kind words. I've not shared much yet though, mostly dreams and vaporware at this point. But everything I do will be shared. And I really hope the ADC and DAC will be high-quality in the end, but it will take a few iterations and quite a bit of time, so be prepared for a long ride...
I totally agree and share your goals, pretty much in the same order.
Regarding batteries, you should have no problem using the ones you are planning to use. You might want to by-pass the Plate PSU board and develop your own though, because I doubt that ours will support anything other than 5V. That being said, if there is a simple way to support multiple input voltages there, we'll definitely consider it, and 12V would certainly be the way to go.
Your regulator setup is certainly very good, but I would note that the converter in the note linked is an isolated DC-DC topology, whereas the SilentSwitcher relies on whatever feeds it for isolation.
The TI push-pull controller IC and circuit that QA uses is fine, but it's low power as you mentioned.
In your opinion...
Jan has shown excellent measured results with Viktor's oscillators in another thread. I really don't think just because you alone think you hear something that this design should have to rule out DC-DC converters. This design is already not likely to satisfy your idea of audio perfection given the design goals and constraints already mentioned.
Last edited:
Sure, you're right. However, presumably the project is intended to produce better sound than could be obtained by simpler means using AK4497. I only strongly encourage the project manager here to setup the evaluation board with linear supplies, at least to begin with. He needs to have his own point of reference in order to decide for himself what performance above or below that he is willing to deem as acceptable. He wants to the final project to run from 5v. That should be doable (I think), but maybe some problems will have to be solved along the way to get it working satisfactorily. I know other people who have tried, and to my ears, failed at that with dac designs. But, I don't give up easily if I do decide to undertake to resolve a problem.
Last edited:
Mark, for the first schematic and PCB layout, I will do exactly what you recommended. Thanks!
MELF Resistors
Following Jam's advice, I will look into using Vishay MELF resistors for some parts of the design.
Following Jam's advice, I will look into using Vishay MELF resistors for some parts of the design.
I managed to find MMA 0204 and SMM 0204 MELF resistors for all the resistor values that we need for the DAC board and PSU board (all from Vishay). They're bigger than the 0603 thin film resistors that we had selected earlier, but comments from Jan Didden on the SilentSwitcher thread suggested that 1206 is as small as we should go, and 0204 and 1206 are roughly the same size. I am not sure that they will all fit on the board, but we'll give it a shot, unless someone has valid objections against it.
It should also be noted that tolerances for these resistors are not as low as what we can find with 0603 or 1206 alternatives (some do not go lower than 1%, yet this relatively-high tolerance complies to the AKM datasheet's recommendations), but the reduction of thermal noise seems to be an important factor to consider when using MELF resistors. That being said, I will let the experts voice their opinions, for I am totally ignorant on the subject.
Last edited:
These would be even better indeed, but I can't find them on Mouser. Let me look some more... Might be a case of Sunday morning fog on my part...
Thanks!
These would be even better indeed, but I can't find them on Mouser. Let me look some more... Might be a case of Sunday morning fog on my part...
Thanks!
For the UMA 0204 series, Mouser only offers these: 33Ω, 42Ω, 100Ω, 1kΩ, 10kΩ, 100kΩ.
For our application, we need the following:
5.1Ω, 10Ω, 200Ω, 270Ω, 360Ω, 510Ω, 2kΩ, 3.6kΩ, 3.83kΩ, 33kΩ.
We've managed to find the following in MMA 0204 (Precision MELF):
270Ω, 360Ω, 510Ω, 2kΩ, 3.6kΩ.
This leaves the following:
5.1Ω, 10Ω, 200Ω, 3.83kΩ, 33kΩ.
We could do 200Ω with two 100Ω Ultra Precision MELF resistors.
Newark offers these in Ultra Precision:
270Ω, 511Ω, 2kΩ.
The 511Ω model has a tolerance of 0.25%, which is great than 1Ω. Therefore this could be an acceptable part for the 510Ω (R302, R307, R322, R327).
The other resistance values are not produced by Vishay in Ultra Precision.
Bottomline: we'll go with MELF for all resistor values and use the best precision we can find for each.
Last edited:
With a bit more research, I could find Precision or Ultra Precision parts for all values but 5.1Ω, 10Ω, and 3.83kΩ.
For 3.83kΩ, there is the SMM02040E3831BB300, which is a Thin Film Mini-MELF Resistor with a 0.1% tolerance. Not sure why it's not called Precision or Ultra Precision, but it definitely fits the bill...
And for 10Ω, we have the WRM0207HPC-10RBT2 at 0.1% as well.
The only one left is 5.1Ω, for which we cannot find anything better than 1%.
Now, we should keep in mind that some of these resistor values are not stocked by a lot of distributors...
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.