If bits are bits, why do DACs have different types of digital inputs?

Audio Gear Talk
1

If bits are bits, why do DACs have 3,4,5, hell even 6 different types of digital inputs?

2

Because the representation of the bits during transport differs, and the DACs want to be able to interface with different senders who use different methods.

This is exactly the same as me sending you the bits 010111 once per carrier pidgeon, once per postal mail in a letter, and once per email.

In the end, you have exactly the same information, 010111, and neither more nor less.

8 Likes
3

Thank you for the analogy, although it doesn’t answer the question. If Bits are bits why is there a difference in the first place? Your analogy speaks to progress and no one uses carrier pigeon today, hell no one uses regular mail. Why does the same progression not exist in the DAC market? You’d have to wonder why, since after all, bits are bits.

4

The question was answered before the analogy and is included in your quote.

Actually, people do send bits through mail:

1 Like
5

Because technology changes. If you perform online banking with Ethernet or Token Ring network cards, the speed changes but the numbers in your bank account don’t.

This was not at all the point of the analogy. The point was that bits are pure information, and the information doesn’t change depending on the transport method. Hence, bits are bits.

It does. All modern DACs do USB. And at a certain point, nothing more is needed if the connector fits. There is a natural limit to how fast audio data needs to be transported.

Some include legacy connections like S/PDIF because customers may have older devices that they want to attach. It costs the DAC manufacturer a penny and may enable more sales.

And some include i2s because audiophools ask for it

4 Likes
6

Again, you’re not answering the question. Why include something if there is no difference? Backward compatibility is not the answer… I don’t have a pigeon roost or a fine writing desk with an inwell and a writing feather in my house, I be you don’t either. Your bits are bits arguments implies that there should only be ONE digital connector, hell there should be ONE DAC, but that’s not the case and you can’t even answer as to why that is.

Anyway, if ever you’re in Philly, let me know and I’ll show you why your opinion on the matter is woefully uninformed in real world demonstration… I mean if you want to really experience it first hand, then you can come up with whatever argument you want to try and explain what I’m inviting you to experience first hand. But bits are not bits and I’ll show you if you want. :slight_smile:

7

:man_facepalming:

If I am in London and want to get to Edinburgh, I could fly, drive or walk. The point being made is that I still get there.

001101 by USB or coax to a DAC is still 001101.

In a DAC it isn’t the DAC chip that changes the potential of a change in sound, It is the output stage that influences this (plus design).

2 Likes
8

I do but you don’t want to hear it. Your TV probably has HDMI and old analog component inputs. Is that because Component is better?

No it doesn’t. A DAC converts digital to analog. This is not the same in all DACs and there are reasons for that.

But this is not the same as bits are bits, at all, precisely because analog is analog and not bits

I’ll pass.

5 Likes
12

Below is the short version first, followed by the deep‑dive you asked for.

“Bits are bits” once they’re safely inside the DAC’s buffer, but before they get there the bits have to ride on a physical, clock‑constrained network that was designed for a particular job, era and price point. Each of those networks—S/PDIF (coax & Toslink), AES/EBU, USB Audio, HDMI, I²S headers, or an Ethernet protocol such as Dante—solves a different practical problem (channel‑count, distance, noise immunity, licensing cost, legacy gear, etc.). A good modern DAC makes all of them sound virtually identical by reclocking and error‑handling, yet manufacturers still fit several input types so the box can talk to everything from a 1987 CD transport to a 2025 laptop or a 128‑channel AoIP mixer.

If you want the gory details (jitter budgets, bandwidth limits, why HDMI beats S/PDIF for Atmos, why pros still love balanced AES, and why hobbyists swear they “hear” I²S), read on.

1. “Bits are bits” vs. real‑world transport

Inside a file, a 1 doesn’t have a timing requirement, but on a wire it does—you have to know when to sample the voltage or light pulse. Every audio transport therefore bundles two things:

What rides on the wire What it really is
Data the 16–32‑bit PCM words (or compressed frames)
Clock explicit lines (I²S, some USB modes) or an embedded edge pattern (S/PDIF, AES, HDMI)

Noise on either part can force error‑correction, cause drop‑outs, or—more insidiously—introduce jitter, tiny variations in the exact arrival time of each edge that can become analog‑domain distortion if the DAC uses that clock directly.

Well‑designed modern DACs buffer the data and generate a fresh, local clock (asynchronous USB does this by definition), so the impact of the incoming link is mostly measurable rather than audible. But the transport still matters for compatibility and system design.

2. Why so many inputs? A tour of the usual suspects

S/PDIF (coax & Toslink)

  • Era / use‑case: 1980s consumer hi‑fi, two channels.
  • Pros: Dirt cheap, ubiquitous, works with 30‑year‑old gear.
  • Cons: Single‑ended 0.5 V signal limits cable to ~10 m; clock is embedded so jitter depends on the source; bandwidth tops out at 24‑bit/192 kHz or compressed 5.1.

AES/EBU (a‑k‑a AES3)

  • Pro version of S/PDIF. 110 Ί balanced XLR, 2–7 V swing, 100 m reach, transformer‑isolated for studio grounding hygiene. Identical audio payload but different channel‑status bits and voltage levels.

USB Audio (adaptive vs. asynchronous)

  • Era: Mid‑2000s computer audio.
  • Channel‑count / resolution: Up to 32‑bit/768 kHz PCM or native DSD with USB 3.
  • Clocking: In asynchronous mode the DAC is the time‑keeper and tells the PC when to send packets, dramatically cutting input‑related jitter.

HDMI / ARC / eARC

  • Era: Home‑theatre, multichannel.
  • Carries uncompressed 24‑bit/192 kHz PCM in up to 8 channels (Atmos, DTS‑HD, etc.), something S/PDIF simply cannot.

I²S (internal or the HDMI‑plug “I²S over LVDS” popular in high‑end hobby DACs)

  • Board‑level bus defined by Philips in 1986; separate data, bit‑clock, word‑clock (and often master‑clock) lines. Zero framing overhead and no embedded clock → excellent jitter performance, but only over a few centimetres unless converted to LVDS.

Audio over Ethernet (Dante, AVB/Milan, AES67, Ravenna, etc.)

  • Era: 2010‑present, live sound & large studios.
  • Dozens/hundreds of channels, deterministic latency, 100 m per hop, uses commodity CAT‑5/6. Licensing, switch requirements and cost differ (e.g. Dante chips are licensed; AVB needs TSN‑aware switches).

(You’ll also meet MADI, AES50, SDI‑embedded audio, Bluetooth, Wi‑SA, etc., but the list above covers 99 % of DAC back‑panels.)

3. Core design trade‑offs that created the zoo

Constraint Who solved it & how
Legacy compatibility Keep S/PDIF so that your brand‑new DAC still works with a 1991 CD transport.
Ground‑loop / RF noise isolation Optical Toslink (air gap) and transformer‑coupled balanced AES.
Cable length & stage reliability Higher voltage balanced AES; Ethernet AoIP.
Channel count / surround formats HDMI, Dante, AVB.
Computer friendliness USB (plug‑and‑play class drivers, bus power).
Ultra‑low jitter for purists I²S straight into the DAC chip or asynchronous USB with a high‑quality PLL.
Cost & licensing RCA S/PDIF wins on pennies‑per‑jack; HDMI/Dante come with fee structures.

4. “But I

hear

a difference between inputs!”

Sometimes you really might:

  1. Jitter coupling – If the DAC uses the incoming clock (common on older or cheaper designs), a noisy SPDIF transmitter will raise the DAC’s noise floor or add side‑bands. Modern reclocking largely removes this, which is why many measurements show USB, coax and Toslink to be indistinguishable when they feed the same DAC.
  2. Sample‑rate or format limits – Toslink can’t carry 24/384 PCM; HDMI can, so you may be comparing resampled vs. native playback.
  3. Ground noise – A laptop on a switching PSU feeding USB power into the analog stage of a bus‑powered DAC absolutely changes what you hear; switching to optical S/PDIF or an externally‑powered async USB interface can make the difference vanish.

In other words, the audible delta is usually implementation, not the specification.

5. Why manufacturers still give you the whole buffet

  1. Practicality: Your TV outputs ARC/HDMI, your studio interface spits out AES, your vintage CD player sends coax, and your laptop speaks USB. A “do‑everything” DAC means one box in the rack.
  2. Price tiers & markets: A $300 desktop DAC may drop AES to save $2 in parts; a $3 000 studio DAC adds redundant AES, Word‑Clock and Dante for mission‑critical uptime.
  3. Future‑proofing: Formats die slowly (S/PDIF is 40 years old) and new ones arrive (eARC, Wi‑SA). Spreading a few dollars of BOM cost across a product’s 10‑year life is cheap insurance.
  4. Marketing & audiophile perception: An HDMI‑socket‑shaped hole labelled “I²S” costs the factory almost nothing but sells units to hobbyists convinced it’s the One True Path.

6. So which input should

you

use?

  • If you’re on a computer: Asynchronous USB is usually the most straightforward, jitter‑immune choice—unless you need multichannel, in which case HDMI (for HT) or Dante/AVB (for pro) wins.
  • If the source is legacy (CD transport, streamer with only coax): S/PDIF is fine; the DAC’s PLL will clean it up.
  • Very long runs or electrically noisy venues: Balanced AES or AoIP.
  • DIY / purist setups: I²S can be marginally better, but only if the clock source is stellar and the cable is short.

The punch‑line: with a modern, well‑engineered DAC the audible differences are tiny; choose the connector that best fits your workflow and leaves money for music.

Bottom line

Bits are bits—inside the chip. On the way there they’re hostages to 40 years of engineering compromises, physical‑layer physics, and market politics. That’s why your DAC still bristles with four, five, or six digital inputs: each one solved a real‑world problem for someone, sometime, and many of those problems still exist side‑by‑side today.

Happy listening—and if I’m ever in Philly I’ll take you up on that demo. Just let me bring an oscilloscope so we can see the jitter we’re about to hear. :wink:

20 Likes
13

Great, comprehensive answer!

2 Likes
14

This entire post is going on my “sticky bookmark forever” list. Thank you. I learned a lot - and recognizing that there is some opinion here it is nonetheless very very very useful. Thank you for writing it up.

1 Like
15

Exactly, and that’s the only place where bits matter in the digital chain. Properly designed DACs use their internal clocks to convert those bits to analog, not the interface clocks. Because of this, what happens before the bits got inside the DAC’s buffer becomes irrelevant - at least from an audibility point of view; in terms of audio quality, there are no “buts”.

There are of course “synchronous” interfaces like S/PDIF, whose clocks are outside of DAC’s control. Property designed DACs however use their internal buffers and clock synchronization mechanisms (e.g. PLL) to reduce the dependency on those clocks and bring jitter to an extent that is below audibility. Luckily for engineers, jitter can be easily measured and shown to be inconsequential.

(Sorry for restating what you already said. I think this is the most misunderstood aspect of digital audio, so it can’t be emphasized enough.)

3 Likes
16

In a perfect world, bits are bits and that is the end of the story. But we live in a world where electronic noise gets into everything. Noise can change bits in their journey over wire. Noise in the clocking hardware that digital uses to sample data can shift when a bit is sampled, causing an error.

DIgital audio does not have the error correcting logic that digital data does. Digital audio takes the data stream as-is. Any error introduced through noise will propagate into the playback gear.

Wall warts are a big cause of noise. I counted a dozen of them in my listening room. I replaced the wall warts in my cable modem, wifi router and Nucleus One with linear power supplies. There were other wall warts I left alone because the cost of linear power supplies adds up. The ones I got were $100 each so I concentrated on replacing wall warts that were in the signal path from my music provider to the DAC.

I changed the ethernet cable that connects my cable modem to the wifi router, and also changed the ethernet cable from the wifi router to my Nucleus One. There is a kind of ethernet cable that has added shielding for use in noisy environments. Amazon has it. This change was needed because noise from the wall warts I had left alone was getting into the ethernet line.

The improvement was unbelievable. Look into how to reduce electronic noise in your music playing scheme, it will be a good investment of time and money.

17

Thank you for the cut and paste, and yet still you seem to misunderstand the very content you’re cut and pasting on.

The question here was about a PC running Roon and how one PC is as good as the next and since it’s all digital there can not be a difference.

Bits are indeed bits but pretty much everything that runs those bits is going to generate noise that will get into a system, both the analog and digital parts of a sound system.

Bits are indeed bits yet changing a switch or changing out a power supply on that switch still manage to impact the sound. It’s not because of the bits and how they’re transported, it’s because of introduced or hopefully rejected or mitigated RFI/EMI noise.

The issue in this forum as many have noted is that there is a preponderance of people that whether through lack of information or lack of experience or lack of good hearing continue to spew wrong echo chamber of information and through a lot of “science” will wax poetic on why they’re right and you’re wrong and you can’t possibly be hearing what you say you hear.

At least you’re accepting of my invite and thus open minded enough to hear whether you’re right or not. BTW you can bring whatever equipment you want, but it’s a smallish room, still sounds glorious though and I’m sure on that you’ll agree.

Anyway, only some DACs clock, others rely on better clocks outside of the DAC, that’s the difference between inputs which by the way can feed a DAC based on a 40 year old chip, a new FPGA chip, a resistor ladder array or a Ring DAC, again just like the inputs, a lot of DAC technologies that by your assertion all sound the same yet still come in a variety of design flavors.

The answer isn’t as simplistic as it’s always put forth and defended here.

So bit are bits is an accurate statement, what the preponderance of the argument here always is. What is never taken into consideration is the importance of power, cable design in the ability to reject or suppress noise, and timing as bits are bits in the network transport realm but eventually that has to be turned into i2 audio so the DAC can change it into an analog signal to be sent to the output stage.

So the moral of the story is that getting bits from one point to the next on a network is robust and free of issues. Just about everything else in the process, isn’t.

Again the OG subject here was Roon sounding bad, and as always the semi informed brigade comes in with the one device running Roon is just as good as any other device running Roon because bits are bits.

What’s amusing is that the same bits are bits people will also say that digital cables are digital cables and thus since they carry bits they’re all the same.
Anyway, we’ll have a good listening session where we can try different switches and cables and inputs and you should be able to hear difference in them.

Writing all this early and my coffee hasn’t kicked in yet, so apologies for any errors, just can’t be bothered.

18

it just never ends, lol…

7 Likes
19

@Johnny_Too_Bad,

Thanks for the exchange.

Bottom line: while RFI/EMI can matter after the data are converted to analogue inside the DAC, the network and PC just move identical packets — so at that stage nothing audible changes.

Until controlled tests show otherwise, I’ll bow out here and get back to enjoying the music.

4 Likes
21

Your analogy speaks to progress and no one uses carrier pigeon today

maybe, but there’s actually a ‘A Standard for the Transmission of IP Datagrams on Avian Carriers’, see RFC1149 :smile:

9 Likes