Ideas wanted -- USB DAC product

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi all,

I'm working on a USB DAC, targeting ultra high end market, with specs comparable with Hugo 2 or ES9038-based solutions.

Like Hugo 2, I will also be using custom made DAC technology, based on DSP, FPGA and some other proprietary secret sauce.

Price will be in $2000~$3000 range, targeting Hugo 2 and Sony ZH1ES. Performance will be the top, tentative specs are:

1. THD+N: <=-120dB
2. DR: >=140dB (32-bit input)
3. Output driving: >=250mW @ 32~600 Ohms, sine wave
4. Input power: external power brick, DAC has internal low leakage isolation, eliminating ground loop and Y-capacitor leakage even without grounding.

Here comes the questions:

1. What kind of inputs are expected? USB is a must, Ethernet, IMHO, is a must for modern devices. Should I support WiFi/Bluetooth?

Keep in mind that we are a startup, and we don't have money to license proprietary technologies like aptX HD or LDAC. So if we implement BT, it will be plain SBC.

2. What sample rates are expected? 44.1/48/96/192 are baseline, should we add in support for 384 or even 768? Are 88.2/176.4 even being used?

3. What kind of outputs are expected? Currently we support XLR balanced output for line out, and TRS switchable single ended/balanced headphone output. Both share the same DAC, but the HPA also has a gain controller in between that can be bypassed if line out level doesn't have to be kept independent from HPA level. Any other ideas?

Thanks for reading the long text, and we hope to hear from your suggestions as both a DIYer and a buyer.
 
USB DAC with ethernet + wifi? That requires adding a computer with OS (in any form), way of controlling the playback software, configuration etc. That makes it a very different product. Are you sure you understand what technology is required for each of your points?
 
USB DAC with ethernet + wifi? That requires adding a computer with OS (in any form), way of controlling the playback software, configuration etc. That makes it a very different product. Are you sure you understand what technology is required for each of your points?

The feature is there, regardless you are using it or not. USB2.0 is great, but that doesn't address a few questions:

1. How to synchronize multiple devices in a studio setup? You can hook up 10 devices to a same PC, but that's it. You can't synchronize devices across different PCs. With Ethernet, you can have an IEEE1588 time server to sync all devices to exactly the same sample rate in a studio.

2. How to run long cables across a large studio or live scene? USB can't be run for too long before it generates bit error. USB audio class spec has no error correction implemented -- a bad USB packet is to be discarded, and no re-transmission is issued. With Ethernet, I can implement a protocol with re-transmission. You can argue that I can also implement a USB protocol with re-transmission, but then I will have to rewrite audio driver from scratch. Ethernet medium is intrinsically designed for long distance transmission, up to hundreds of miles, with optical fiber cable.

3. How to stream audio from multiple devices? In a home setup, how to stream audio from your phone/laptop/tablet to the DAC while not having to force them to connect to the DAC with a physical wire that has to be reconnected all the time? Also, the evil Apple has the MFi thing -- if I want to stream from Apple device, I need to be MFi certified, which I can't. With Ethernet, I can write a client app that bypasses OS audio driver, and itself handles audio file reading, decoding and Ethernet transmission to my DAC.

4. Bluetooth is a bit useless -- you don't listen SBC audio through a $3000 device. I consider it because I plan to use ESP32 as main app processor, and it has BT built in, so I don't want to waste this capability. That's it. There's no reason why I just have to support it. Actually, for $4.5, just as the Ethernet controller plus audio buffer manager, it already is worth every cent, WiFi and BT are considered free by me.

------------------------------------

My device already has USB and AES3 (coax, XLR and optical) supported, so Ethernet and WiFi/BT are additional features to be considered.

As for whether I can make it or not, I guess that's a different question. I will consider to drop features if I see some features are so hard to do and don't provide much added value, but until then, I want to have a complete wish list. After all it's Christmas season.

-------------------------------------

I know this is going to be a tough project, but I would like to take on it.

Any other suggestions?
 
Thanks for the input.

FW is long outdated, and the supporting chips are not being made anymore. Specialized chips for research and industrial purposes are too expensive, and usually require NDA to be signed.

Thunderbolt is a wonderful protocol, but again, I see no reason to use it. USB2.0 provides enough bandwidth for many, many channels of 32/768 audio, yet its latency is not too high. Sure, Thunderbolt has virtually no latency since it is just PCIe, but the extra overhead caused by USB is less than 1ms anyway.

Besides, Thunderbolt, which was promised by Intel as an open protocol, remains closed, and you need to sign an NDA and talk with an Intel representative before being able to access to their documents. Unlike USB, you can't just grab an interface chip and start working on it.
 
I must be missing something here but I can't see a price tag of $2-3k being at all credible as 'ultra high-end'. The ultra high-end price tags are at least 10X that. Ultra high-end means dCS, Aries Cerat, maybe AudioNote and certainly the top end of MSB's DACs.
In this stratospheric price range, DACs would not come with headphone amps built in, that would be too much of a compromise.
 
Last edited:
I must be missing something here but I can't see a price tag of $2-3k being at all credible as 'ultra high-end'. The ultra high-end price tags are at least 10X that. Ultra high-end means dCS, Aries Cerat, maybe AudioNote and certainly the top end of MSB's DACs.
In this stratospheric price range, DACs would not come with headphone amps built in, that would be too much of a compromise.

I'm an engineer with a PhD degree, not a dark artist. I care about specs, not how people interpret them.

For example, dcs and aries carat both have horrible specs. Only man among that list has some good figures.

I don't mind spending $20k on a top of the line audio analyzer or spending 500 hours to build one, but I only care about measured electrical data.

Bit I do like your last point -- to get rid of the headphone amp and let the user to figure out how to add their own.

I have to admit that hpa is a big compromise in my design. It causes quite some problems, and practically no one will listen to a high end dac without an amp.
 
I'm an engineer with a PhD degree, not a dark artist. I care about specs, not how people interpret them.

For example, dcs and aries carat both have horrible specs. Only man among that list has some good figures.

It seems to me that you're not quite up to speed on marketing in audio. 'Ultra high-end' is not a spec classification in the audio world, its a marketing term. If you're wanting to sell on specs alone, good luck in convincing the customer base that great specs are a guarantee of great sound. You'll need to spend more on marketing that proposition than on DAC R&D and fancy test gear I reckon.

Incidentally if you go over to MSB's website armed with your PhD you might notice their figures are fudged. The fact that this passes completely unnoticed by the consuming public might tell you something....

I don't mind spending $20k on a top of the line audio analyzer or spending 500 hours to build one, but I only care about measured electrical data.

Seems to me you'll want to open up a completely new market, one of believers in specs over and above reviewers and dealers and related hangers-on.
 
frugal-phile™
Joined 2001
Paid Member
FW is long outdated

Not at all, it is a subset of Thunderbolt.

Thunderbolt is a wonderful protocol, but again, I see no reason to use it. USB2.0 provides enough bandwidth for many, many channels of 32/768 audio, yet its latency is not too high. Sure, Thunderbolt has virtually no latency since it is just PCIe, but the extra overhead caused by USB is less than 1ms anyway.

Right, Thunderbolt, if you use the Firewire subset is designed to deal with audio, putting audio over USB is still just a convience.

Thunderbolt is not just PCIe, that is one of the subsets supported, just like FireWire, USB3, DisplayPort. Intel would really like to see the death of USB but it has far too much use for it to die anytime soon.

dave
 
I'm currently developing something similar purely as a hobby. It's a multi-channel DAC for use with PC-based FIR DSP for crossovers and perhaps room equalisation although it could equally be used as a stereo, 5.1 or 7.1 setup.

For input I will be using XMOS XUF216 for UAC2. I have already purchased a 2 channel DIYHK board to experiment with and have managed to modify the firmware for 6 channels out up to 192K/24b. At the moment it's operating as master but ultimately it will setup as slave to an FPGA. When I design the PCB for this project I'll provision for 8 channels. I believe that ethernet can also be added although I'm not interested in using it.

The setup will be {XMOS} <==> {FPGA} <==> {ISOLATION} <==> {CLOCKS, DSD_DAC}

At the moment I'm thinking that one PCB will support the XMOS, FPGA, digital power supplies, digital isolators and isolated analog supplies for the clocks and DACs. This board will provide headers for one dual clock module and four headers for stereo DACs with balanced outputs. This will give flexibility if I want to try different clock and/or DAC modules.
 
Not at all, it is a subset of Thunderbolt.



Right, Thunderbolt, if you use the Firewire subset is designed to deal with audio, putting audio over USB is still just a convience.

Thunderbolt is not just PCIe, that is one of the subsets supported, just like FireWire, USB3, DisplayPort. Intel would really like to see the death of USB but it has far too much use for it to die anytime soon.

dave

You should do some research before you post because the OP is correct and you are wrong.

Thunderbolt is multiplexed PCIe and DP put onto one connector, which it may share with USB 3.1 Gen 2. TB3 supports 4 lanes of PCI Express Gen 3 and 8 lanes of DP 1.2.

Firewire is dead. No one is making or cares to make new transceiver chips for it. The last generation of mainstream PC motherboards with FW chips were released around circa 2009. Apple will not put it on any new product. It will never appear on another PC. TB does not support FireWire. Any product using FireWire over TB is using a TB PCIe device on the other end.

Intel developed USB and drives a lot of the ongoing development of USB via its role in the USB-IF. Intel makes some of the only USB controllers on the market that actually work well and have stable drivers for all major operating systems. Intel employees actively contribute to USB controller driver development for Linux. Saying that Intel wants USB to go away shows how little of an understanding you have.
 
Minimum input interfaces for a studio DAC would be optical and coaxial SPDIF, AES/EBU, and USB. Those are the things most studios use. You do might well to take a look at the interfaces and features of the DACs which are used in many studios here: Digital to analog converters - Benchmark Media Systems, Inc.

For outputs rca unbalanced and XLR balanced are the usual.
A manual with lots of specs is here: https://cdn.shopify.com/s/files/1/0321/7609/files/DAC3_Series_Manual_Rev_B.pdf?9982830537634228604
As you can probably see, there are already some pretty good studio products on the market.
 
Last edited:
frugal-phile™
Joined 2001
Paid Member
You should do some research before you post because the OP is correct and you are wrong.

I am aways looking for FireWire (or Thunderbolt now).

Well excuse me...

I guess you missed the or Thunderbolt part. As far as i am concerned Thunderbolt is the evolution of FireWire.

I followed all the initil Thunderbolt release info, and what was said about it. Every Firewire device i plugged into my Thunderbolt ports worked fine — including my DACs. Everyting i read about Firewire interfaces plugged into Thinderbolt was that they worked. People had started to move their interfaces over to Thunderbolt.

At that point i lost the thread, spending 5 months in hospital after lifesaving surgery. And the last 8 months learning how to walk again.

I have never paid a whole lot of attention to the actual hardware implementation, but i know FireWire/Thunderbolt is harder than USB. But it can handle all the I/O without the help of the processor and is much more suitable for audio/video than USB.

dave
 
I don't mind spending $20k on a top of the line audio analyzer or spending 500 hours to build one, but I only care about measured electrical data.

You might want to be careful there. It might be worth taking a look at this: AES E-Library >> Are We Measuring the Right Things? Artefact Audibility Versus Measurement

Then again if you want to develop new kinds of instrumentation to measure things people can hear, that would be great. Also, I would point out many serious, successful equipment designers supplement measurements with listening tests. One big example would be JBL, which uses listeners it trains itself.
 
Last edited:
frugal-phile™
Joined 2001
Paid Member
input interfaces for a studio DAC

Antelope Audio, Universal Audio, Focusrite, PreSonus, Slate Digital, MOTU, Apogee, RME, Lynx, Avid, Neumann, Resident Audio, AKG, Zoom all do Thunderbolt.

And Firewire can still be had from Weiss, Universal Audio, RME, Behringer, MOTU, Metric Halo, Lexicon, Tascam, Steinberg, Echo, Focusrite, Prism Sound.

That list is probably not exhaustive.

dave
 
Thunderbolt makes perfect sense for people using a Mac, I would agree.

The main issue with having a lot of different hardware interfaces such as USB, Thunderbolt, Ethernet, and so on is that in many cases specialized hardware drivers have to be written. Writing and maintaining drivers over evolving versions of multiple operating systems can get to be a costly and time consuming undertaking.

On the other hand, SPDIF and AES have been pretty consistent over a long period of time, and SPDIF adapters for all kinds of computers are readily available. Not that great an idea for portable systems though, where too much stuff to carry around and hook up can be problematic.

Also, seems like even at a price point of $2k - $3k, one would need to focus on a particular market. The upscale (but not full blown audiophile) hi-fi home consumer market is one thing, and the studio market is another. Lots of studio features won't fit in a tiny box for one thing, a half rack space size is probably minimal. And home hi-fi doesn't need space for balanced AES interfaces.
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.