Review and Measurements of Okto DAC8 8Ch DAC & Amp

[JohnYang1997]

So you say you can hear a difference between the Okto one and the Schiit one? What measurement would point to that? What measurement determines which give a worse sound?

You want to see the power supply noise at or 60hz and their harmonics(120 180 240 etc)(or 50hz for many other countries). If this is excessive it will alter the perception of bass, body of instrument, thicknesses etc.
Harmonic distortion, in my experience distortion under 0.01% is small but smaller distortion even in 0.001% range can change the tonality of sound. It may not be the harmonics itself but indication of something else.
Nosie, well engineered dacs have normally noise of around 2uV or lower. But some dacs have much higher noise than can alter the perception of sound. Again low frequency noise can produce body to the sound which many people want but not originally there.
The rest can be found in another thread
https://www.audiosciencereview.com/...-thresholds-of-amp-and-dac-measurements.5734/

 

[g29]

Thanks, I have read the entire threat indeed. They mention that shorting the hot lead to the ground overloads the output stage causing distortion. I would hardly call this elaborate. They recommend a ~300 dollar Jensen solution that is unavailable in EU. Considering the serious limitations of transformer based isolators, and the mere recommendation by Oktoresearch, I felt it could be worthwhile to explore alternatives. I figured that more people would be interested in this.

@1234VICE ,

FWIW, here is a USB/single ended 8 channel DAC that may be of interest if unbalanced outputs are a hard and fast requirement. Also 1/4th the cost and available in Europe. It has been put on the measurement wish-list, but I don't think Amir has tested one yet.

miniDSP U-DAC8

 

[1234VICE]

What is steady state, you have to ask yourself. How a complicated alternating signal has anything to do with "steady state". None dacs can predict such signal. So thinking it's steady state is not accurate.

Very fundamentally, I am considering the issue that mathematically there is no one-to-one projection between transfer function and linear operator. For instance, let "P" be some time shift operator, such that

P exp( i w t ) == exp( i w (t - 2pi/w) ) = exp ( i w t ) ("steady state")

I.e. this introduces a time delay equal to one cycle

Then, consider some system for which the output voltage V_out is described as

V_out = P V_in

Clearly, this system's transfer function is perfect, but it's output is not. More importantly, we no longer can compute the time domain response by using its transfer function as a kernel inside the FFT.

Now, this operator is likely not physical, and the crux is that we need to prove that for all physical systems, there is a one-to-one projection from transfer function to time-domain response, i.e. for the subset of differential equations describing physical systems to transfer functions.

You are mentioning that dac's cannot predict the signal, I interpret this as a suggesting the previous statement. However, this is not very convincing IMO, as feedback can be viewed as a form of predicting. I already mentioned room acoustics as a physical system where things go wrong. Another example would be driver integration of a speaker with a multiple cycle offset. For instance, a woofer that starts playing in phase with a tweeter after "n" cycles because of some time delay circuit. Real example; the KEF ls50 has woofer and tweeter wired in opposite polarity to account for the effective time delay between both drivers, this is revealed by the impulse response only.

We have thus already proven that there is no one-to-one projection between transfer function and differential equation of physical systems, and assessing the performance of a piece a equipment based on the transfer function alone might not be justified.

This null test sounds appealing indeed. For this is direct proof that the system is fully described by its transfer function. Indeed, effort should still be taken to resolve and characterize the linear and non-linear distortion. This null test purpose is to act as a sanity check. Perhaps even better is not to compare to the signal itself, but by the signal transformed to the predicted output based on the measured transfer function. It seems hard to define a metric based on this test though.

 

[1234VICE]

@1234VICE ,

FWIW, here is a USB/single ended 8 channel DAC that may be of interest if unbalanced outputs are a hard and fast requirement. Also 1/4th the cost and available in Europe. It has been put on the measurement wish-list, but I don't think Amir has tested one yet.

Thank you for mentioning this. Indeed, I am aware of this device. The miniDSP SHD is also an interesting offering, considering I only need 4 Ch and I do require a preamp. The DAC8 with miniSHARC was basically a SHD on steroids + much cheaper.

There are a couple of reasons why I will go for the DAC8 though
1. I admire the engineering
2. The company profile / philosophy strongly resonates with me
3. The price is right
4. Not unimportant: it's a damn sexy box IMO

Maybe I should just buy subwoofers with xlr inputs instead .

 

[Olli]

Thank you for mentioning this. Indeed, I am aware of this device. The miniDSP SHD is also an interesting offering, considering I only need 4 Ch and I do require a preamp.

The MiniDSP SHD is a 2 CH DAC, not 4 CH. You can apply different XO settings to the 4 outputs however.

 

[JohnYang1997]

Very fundamentally, I am considering the issue that mathematically there is no one-to-one projection between transfer function and linear operator. For instance, let "P" be some time shift operator, such that

P exp( i w t ) == exp( i w (t - 2pi/w) ) = exp ( i w t ) ("steady state")

I.e. this introduces a time delay equal to one cycle

Then, consider some system for which the output voltage V_out is described as

V_out = P V_in

Clearly, this system's transfer function is perfect, but it's output is not. More importantly, we no longer can compute the time domain response by using its transfer function as a kernel inside the FFT.

Now, this operator is likely not physical, and the crux is that we need to prove that for all physical systems, there is a one-to-one projection from transfer function to time-domain response, i.e. for the subset of differential equations describing physical systems to transfer functions.

You are mentioning that dac's cannot predict the signal, I interpret this as a suggesting the previous statement. However, this is not very convincing IMO, as feedback can be viewed as a form of predicting. I already mentioned room acoustics as a physical system where things go wrong. Another example would be driver integration of a speaker with a multiple cycle offset. For instance, a woofer that starts playing in phase with a tweeter after "n" cycles because of some time delay circuit. Real example; the KEF ls50 has woofer and tweeter wired in opposite polarity to account for the effective time delay between both drivers, this is revealed by the impulse response only.

We have thus already proven that there is no one-to-one projection between transfer function and differential equation of physical systems, and assessing the performance of a piece a equipment based on the transfer function alone might not be justified.

This null test sounds appealing indeed. For this is direct proof that the system is fully described by its transfer function. Indeed, effort should still be taken to resolve and characterize the linear and non-linear distortion. This null test purpose is to act as a sanity check. Perhaps even better is not to compare to the signal itself, but by the signal transformed to the predicted output based on the measured transfer function. It seems hard to define a metric based on this test though.

1 prediction Feedback is some sort of prediction but it doesn't predict the whole cycle. So periodic signal is not different from actual music signal.
2 About the shift operator. The shift operator shifts the signal depending on the signal (period). So it's not a shift/time invariant filter. If that happens everything will fall apart, which is not possible in audio equipment (if that happens in a little bit that's jitter which does exist).
3 Many people have done different measurements along with null tests and nothing is proving be contradictory. So all previous measurements are valid and is substitute to null test in most part. The most similar test to null test is multitone td+n measurements. And it reveals much much more information in one graph. In null test you only know how much is left but you hardly know what's left.

 

[rickyhgarcia]

Is there a published picture of the back panel of the DAC8 Stereo?

 

[rickyhgarcia]

The OKTO DAC8 Stereo does not do RCA and they advise against XLR to RCA cable usage.

So the DAC8 Stereo is not a good choice for a legacy mid 80’s tube (Audio Research) preamp? Would the RME be a better choice?

 

[g29]

So the DAC8 Stereo is not a good choice for a legacy mid 80’s tube (Audio Research) preamp? Would the RME be a better choice?

As far as I know, the DAC8 Stereo, IS THE PREAMP (read: shorter signal chain, less distortion summing). If you want to add the tube preamp after the DAC8 Stereo as a tube buffer, you can use one of their recommended RCA-2-XLR recommended solutions, but the tube buffer will be degrading/coloring the DAC8's signal by doing so.

Little information has been published about it and neither the DAC8 PRO or Stereo production versions have been released/materialized yet. Maybe @Okto Research can give us an update on their respective release dates ?

 

[rickyhgarcia]

As far as I know, the DAC8 Stereo, IS THE PREAMP (read: shorter signal chain, less distortion summing). If you want to add the tube preamp after the DAC8 Stereo as a tube buffer, you can use one of their recommended RCA-2-XLR recommended solutions, but the tube buffer will be degrading/coloring the DAC8's signal by doing so.

Little information has been published about it and neither the DAC8 PRO or Stereo production versions have been released/materialized yet. Maybe @Okto Research can give us an update on their respective release dates ?

The reason I need the preamp is for the phono stage. The preamp would input the turntable and the DAC. To the DAC I plan to connect a PC (USB) and a CD Player (either optical or coax). The preamp will output to a tube amplifier, then to speakers. The DAC sole purpose is to incorporate a PC and CD player to a tube analog system.

 

[dc655321]

We have thus already proven that there is no one-to-one projection between transfer function and differential equation of physical systems, and assessing the performance of a piece a equipment based on the transfer function alone might not be justified.

Please take some time to study LTI/LSI systems.

 

[1234VICE]

Please take some time to study LTI/LSI systems.

This wikipedia page is exceptionally well written, probably confusion on this topic has been persistent . Thank you for linking this.

I like how they conceptualize the transfer function based on the convolution with the impulse response / Greens function, this seems like a good way to think about this. It illustrates well that the timing information / impulse behavior must be encoded somewhere in the transfer function.

This then actually means that locally a room response must be fully invertible with a filter of sufficient resolution; quite mind boggling. I still do not understand then why humans do not prefer a flat in-room response in terms of timbre? Do we enjoy reflections? Do we measure the transfer function wrongly? Is the effect too localized?

It must be clear that I should indeed study this further, and write down the math sometime. I am definitely not smart enough to grasp the full extend intuitively. For now, it is sufficient for me to realize that I was wrong. Thank you again.

 

[dc655321]

For now, it is sufficient for me to realize that I was wrong. Thank you again.

No! Thank you!!
Do you know how rare admitting to a lack of/incomplete knowledge is???

This then actually means that locally a room response must be fully invertible with a filter of sufficient resolution;

Not quite "fully invertible".
Room responses, in general, are not minimum phase (I'll let you look that one up) over the entire bandwidth of interest, hence not entirely invertible.

 

[josh358]

As far as I know, the DAC8 Stereo, IS THE PREAMP (read: shorter signal chain, less distortion summing). If you want to add the tube preamp after the DAC8 Stereo as a tube buffer, you can use one of their recommended RCA-2-XLR recommended solutions, but the tube buffer will be degrading/coloring the DAC8's signal by doing so.

Little information has been published about it and neither the DAC8 PRO or Stereo production versions have been released/materialized yet. Maybe @Okto Research can give us an update on their respective release dates ?

Pavel told me that the DAC8 PRO will be shipping in a week, but because of preorders, new orders won't be shipped for 1-2 months. The DAC8 Stereo will be launched in about a month.

 

[josh358]

This wikipedia page is exceptionally well written, probably confusion on this topic has been persistent . Thank you for linking this.

I like how they conceptualize the transfer function based on the convolution with the impulse response / Greens function, this seems like a good way to think about this. It illustrates well that the timing information / impulse behavior must be encoded somewhere in the transfer function.

This then actually means that locally a room response must be fully invertible with a filter of sufficient resolution; quite mind boggling. I still do not understand then why humans do not prefer a flat in-room response in terms of timbre? Do we enjoy reflections? Do we measure the transfer function wrongly? Is the effect too localized?

It must be clear that I should indeed study this further, and write down the math sometime. I am definitely not smart enough to grasp the full extend intuitively. For now, it is sufficient for me to realize that I was wrong. Thank you again.

Well, first off, we do pretty much prefer a flat response. But two channel stereo requires a declining "house curve," e.g., 1 dB/octave, to sound flat. This has been attributed to the interaction of the standard stereo angle with the head-related transfer function. Indeed, one of the challenges in acoustic design is keeping room response and reverberation relatively flat; often, rooms end up sounding too dead because thin absorbers like carpets are more effective at high frequencies.

Reflections -- well, that really depends on the *type* of reflection, e.g., when it arrives. A loud reflection that arrives too early will color the sound, whereas if it arrives later, it will lend a pleasing sense of ambiance or reverb and space. So in general we try to keep reflections earlier than about 20 ms 20 dB down. At the opposite extreme, a reflection that arrives too late will be interpreted as an echo, which is even more undesirable. The best reverberation time for a room (generally specified as RT60) depends on the application (music vs. theater, etc.), the radiation pattern of the speakers, the size of the room, and to some extent, listener preference and the function of the room (a studio will generally have a "drier" or less reverberant acoustic than is desirable in a home environment, because the recording engineer has to hear every detail).

Two channel stereo just doesn't work right without room reflections, because we're used to hearing room reverberation and stereo can't provide that. Multichannel systems can benefit from a less reverberant acoustic.

Another area in which room reflections are problematic is in the bass and midbass region below the Schroeder frequency where room modes (resonances) create position-dependent frequency response irregularities. Above the Schroeder frequency, the brain is less sensitive to nulls and peaks caused by reflections. And the brain is quite good too at using room reflections to gauge the size of the room, proximity to surfaces, etc. For example, if an instrument is 5' from a wall, the first reflection off that wall will be delayed by roughly 10 ms; if 10', 20 ms. So the brain can use the timing of the first reflection to estimate the distance to the wall.

 

[Okto Research]

Hi everyone,

hopefully I'll be able to shed some light on the previously discussed XLR-RCA interfacing.

The main reason why two wires aren't enough for a high-performance audio interconnect is that no matter what you do, there will be current caused by a different ground potentials and/or a real-world interference flowing through the shield conductor. Since the shield conductor has a non-zero impedance, this current will produce a voltage and in case the shield conductor also doubles as a signal conductor, this voltage will be mixed straight into your audio.

To quote Bruno Putzeys:

The whole idea of having one of the actual signal wires also do the dirty work of shunting equalizing currents away is barmy. To then try and solve the problem by eliminating those currents is bone-headed. To try and mollycoddle the sonic defects this causes by making outlandish cables is madness. The RCA connector and all it stands for should be banned by law.

If the receiver is made differential, i.e. it only cares about the difference between the two signal conductors (XLR pins 2 and 3 or hot and cold), the "dirty work" is offloaded to shield (or XLR pin1) conductor. As long as those are not connected to the PCB on either transmitter or receiver side (which causes so called Pin1 problem) but rather diverted to the Protective Earth mains conductor, this makes for a hum-free connection.

When interfacing XLR outputs to RCA inputs, you have these options:

1) The best one from a technical standpoint: retrofitting a XLR differential input to a legacy equipment
It might be either built inside or made as an outside box with separate power. The XLR shield and pin1 would need to be grounded to the chassis on the receiving side or the whole thing might be bolted to it. A company called THAT Corporation makes very good receiver chips that would fit this task and a module using those has even been a commercially available product at some point .
https://www.neurochrome.com/that-receiver-rev-1-0/

2) A drop-in solution if you don't mind some distortion: tranformers
The transformer-based approach (like with Neutrik NA2F-DOB-TX or much more expensive Jensen transformers) will work and avoid ground loop issues. Transformers also have higher common-mode rejection ratio than what can be achieved with semiconductors. The distortion might not matter for use with tube amplifiers that already have a transformer in the signal path or with subwoofers.

3) Off the shelf XLR-RCA cables
As mentioned, these short XLR pin 3 (cold) to pin 1 (ground). That makes them work both with direct-coupled and transformer-coupled XLR inputs/outputs, but the direct-coupled ones might not be happy about the short. While the outputs of DAC8 PRO are short-circuit proof, operational amplifiers do not like huge loading and will respond with increased distortion (by several orders of magnitude).

4) Cable with floating pin3 (cold)
Using just one of the signals from the XLR side to interface with RCA is a dirty and cheap solution. We cannot guarantee this will be hum-free (depends on circumstances) or that DAC8's output will be pop-free on turn on/off in this configuration, but it is certainly a better way than the previous one. There will still be an increased distortion compared to interfacing a regular XLR input since you will be picking up the common-mode distortion products that would otherwise be rejected. Because you are only taking half of the signal, you will lose 6dB. If you still want to go that way and don't have pliers, Benchmark already makes cables like that: https://benchmarkmedia.com/collecti...dapter-cable-pin-3-floating?variant=547761981

Is there any problems with "numbers" pairing Octo DACstereo with Denafrips Hestia pre amp ? They would be paired to tube monoblocks & speakers. Hestia has XLR in and RCA (and XLR) out for monoblocks that have just RCA in.

Denafrips Hestia specs:

https://www.denafrips.com/specs-hestia-hyperion

Thanks & cheers

They don't specify the input voltage nor the gain, but given the fact that their output works with the same voltage as DAC8 (4V RMS), I don't expect any issues.

@Okto Research
What is the latest ETA on the DAC8 PRO launch ?
When will production units be shipping ?

DAC8 PRO launch coming up in the following days! Shipping will start shortly after that. Stay tuned for the news.

Pavel, Okto Research

 

[maty]

For now there are three Okto Research products.

Okto Dac8 Pro

https://www.oktoresearch.com/dac8pro.htm

Okto Dac8 with DSP

without page link

Okto Dac8 Stereo, with Raspberry Pi running Volumio

https://www.oktoresearch.com/dac8stereo.htm

We can see a RF/EMI inlet filter with switch, from Schaffner?

BTW, I like most of what I can see inside the box and the aluminum box. And:

 

[Okto Research]

For now there are three Okto Research products.

None of those is our current or upcoming product. The first and third one are a custom builds we made last year, the second one is the DAC8 DSP which was discontinued.

 

[maty]

I would change, if until is possible, the placement of the ventilation grill away from the audio electronics. And already put, better that they were small round holes.

 

[Snafu]

I would change, if until is possible, the placement of the ventilation grill away from the audio electronics. And already put, better that they were small round holes.

could you please stop spamming / spreading disinformation like this ?