Sorry, I chose words poorly. In my mind, no matter how much I know better, "hardware" on a chip is not actual hardware (so it I end up saying software...)
That said, on board volume control is a feature of the DAC chip. There are well understood issues with this type of thing that can be measured and are documented (so no snake oil). Loss of dynamic range and worsening SNR are typical. There is a reason the nice gear usually uses a method external to the DAC chip for volume control. The Benchmark DAC 3 for example does not use the DAC chip, nor do the PS audio DACs. Not sure about the others, but I'd be surprised if any of the ones with great reputations do.
It honestly would not surprise me if alot of the sensitivity the Katana has to power supplies is rooted in using the onboard volume control.
Either way, I will be able to comment on it in a week or so.
Benchmark doesn't use the Sabre volume control because they can't, not necessarily because they don't want to. That's because DAC-3 uses external interpolation filtering (for reasons unrelated to the volume control), and if one opts to do that, then one has to implement one's own volume control. That is because the Sabre volume control is implemented by adjusting interpolation filter coefficients to affect filter gain (which by the way is exactly the right way to do an oversampling dac digital volume control). If you read the ESS document I linked to, they agree that very well implemented analog volume controls can be even better, but usually they aren't implemented well enough to outperform the Sabre volume control. One thing they make very clear is that simple software digital volume controls are bad, and should be avoided if possible. I do recommend using software volume controls before the dac board at a fixed setting of (usuallly) about -3dBFS or -4dBFS if possible to diminish adverse distortion effects that actually can be audible due to intersample overs in digital recordings.
Regarding any audible shortcomings of the Sabre volume control, few Sabre dac designs are implemented well enough one could find audible fault with of all things the volume control. Virtually everything else in most dac board designs is far worse.
Regarding any audible shortcomings of the Sabre volume control, few Sabre dac designs are implemented well enough one could find audible fault with of all things the volume control. Virtually everything else in most dac board designs is far worse.
Last edited:
Very interesting about the Intersample Overs issue Mark. Just read up on it and I think it explains the distortion that both Greg and I found on the PCM1794 Dial DAC. Do you know how well Intersample Overs are dealt with in the Katana and its ESS DAC chip? Does it pre-attenuate?
J
Don't know about what Katana might do with regard to intersample overs. Only cdsgames would probably know.
What Benchmark does is to use pre-attenuation about as I described, except if the volume pot on the dac is turned up all the way. In that case, to get the last 3dB - 4dB of volume level, they reduce the pre-attenuation to zero. Why ever do that? If they don't, then the pre-attenuation would affect other dac specifications making them look worse in tests. Since some people judge dacs by measurements, or by some combination of measurements and listening tests, it is important for a manufacturer to have the measurements look as competitive as possible.
Another thing to be aware of with Benchmark DAC-3 is that so far as Windows 10 is concerned, the Benchmark USB driver is 24-bit or 32-bit, not 16-bit compatible. As as result, when using the Benchmark USB device as the default sound device in Windows, there is no way to prevent the OS from distorting CD audio before sending it to the dac. It should be possible to simply zero-stuff the extra bits the driver requires if playing 16-bit audio, but the Windows sound engine never does that. If an audio stream is an exact match for the default setting of the sound device, then the sound engine SRC is bypassed, otherwise the audio stream gets the full treatment which adds audible damage to sound quality. There are ways to avoid that if there is more than one sound card present and special steps are taken so that Windows leaves the Benchmark device alone. Otherwise, the safest thing is to forget the Benchmark USB driver and send the dac SPDIF somehow. At least that allows one option to avoid sound engine distortion if only a single sound card device is available.
What Benchmark does is to use pre-attenuation about as I described, except if the volume pot on the dac is turned up all the way. In that case, to get the last 3dB - 4dB of volume level, they reduce the pre-attenuation to zero. Why ever do that? If they don't, then the pre-attenuation would affect other dac specifications making them look worse in tests. Since some people judge dacs by measurements, or by some combination of measurements and listening tests, it is important for a manufacturer to have the measurements look as competitive as possible.
Another thing to be aware of with Benchmark DAC-3 is that so far as Windows 10 is concerned, the Benchmark USB driver is 24-bit or 32-bit, not 16-bit compatible. As as result, when using the Benchmark USB device as the default sound device in Windows, there is no way to prevent the OS from distorting CD audio before sending it to the dac. It should be possible to simply zero-stuff the extra bits the driver requires if playing 16-bit audio, but the Windows sound engine never does that. If an audio stream is an exact match for the default setting of the sound device, then the sound engine SRC is bypassed, otherwise the audio stream gets the full treatment which adds audible damage to sound quality. There are ways to avoid that if there is more than one sound card present and special steps are taken so that Windows leaves the Benchmark device alone. Otherwise, the safest thing is to forget the Benchmark USB driver and send the dac SPDIF somehow. At least that allows one option to avoid sound engine distortion if only a single sound card device is available.
Last edited:
Hi Guys,
I just finished my Katana setup with good power supplies.
And now, the katana THD achieves a really great sound.
I used MPaudio power supplies as Greg does partially, and they are really great to my ears.
one 2x5v for Katana dac and micro controler
one 2x15v for the output stage
a last one 5v with 5A for the rpi and the isolator from Ian.
I added the film cap bank that Mark help me to craft.
All usb connector were bypassed on katana and RPI.
All these weren't cheap. It was really expensive compared to my chinese 9038q2m setup. But results are there.
Timber and basses are beautiful and soundstage is really wide.
Every intrument is really detailed and well separated.
Music is filling the room really better that before and trebles aren't harsh anymore.
The only problem I have is that it looks that I can't play any music above 32/192khz.
So I can't even play dsd64 using DoP. It produces a lot of noise.
I was thinking that katana was able to play dsd64 and dsd128 using DoP.
I was thinking that it was because of Ian isolator use, but it is rated for DSD512/768khz master & slave mode and I use it on my chinese 9038q2m playing DSD256 DoP without any problem.
The katana is connected to Isolator/rpi using pin 3/5/12/35/40 as it is written in katana documentation.
The rpi is using Moode 4.4 with MPD 0.21.3
Tell me if you have an idea of what may be the problem and if you can play DSD64 &128 Dop without problem.
Thank you again Greg & Mark for all your help achieving this.
I just finished my Katana setup with good power supplies.
And now, the katana THD achieves a really great sound.
I used MPaudio power supplies as Greg does partially, and they are really great to my ears.
one 2x5v for Katana dac and micro controler
one 2x15v for the output stage
a last one 5v with 5A for the rpi and the isolator from Ian.
I added the film cap bank that Mark help me to craft.
All usb connector were bypassed on katana and RPI.
All these weren't cheap. It was really expensive compared to my chinese 9038q2m setup. But results are there.
Timber and basses are beautiful and soundstage is really wide.
Every intrument is really detailed and well separated.
Music is filling the room really better that before and trebles aren't harsh anymore.
The only problem I have is that it looks that I can't play any music above 32/192khz.
So I can't even play dsd64 using DoP. It produces a lot of noise.
I was thinking that katana was able to play dsd64 and dsd128 using DoP.
I was thinking that it was because of Ian isolator use, but it is rated for DSD512/768khz master & slave mode and I use it on my chinese 9038q2m playing DSD256 DoP without any problem.
The katana is connected to Isolator/rpi using pin 3/5/12/35/40 as it is written in katana documentation.
The rpi is using Moode 4.4 with MPD 0.21.3
Tell me if you have an idea of what may be the problem and if you can play DSD64 &128 Dop without problem.
Thank you again Greg & Mark for all your help achieving this.
Attachments
I have the same Moode and MPD versions as you, and I can play DSD64:
Encoded at: DSD64, 1 bit, 2.822 mbps Stereo
Decoded to: PCM, 32 bit, 352.8 kHz, Stereo, 22.579 mbps
No idea what the problem could be though.
I'm not using an isolator.
Try using Moode 4.4 with MPD version 20.20.
I tried with MPD version 21.4 and I was unable to play DSD64 files without noise. Tim from Moode was unable to reproduce my issue using the same configuration. Switching back to MPD version 20.20 solved my problem.
Hi,
thank you.
If you are talking about j12/j13 jumpers, my isolator is on the Master settings.
No I didn't listened without the bank for now.
I setup everything at one time
As I have these little problems, I will reopen the box.
I will listen to the system without the film cap bank when I'll check if the problem is the isolator or mpd.
@terry22, that is great news about your results with the MPAudio supplies. While I didn't have any reservations that the dual 5V MPAudio supplies would work well, I haven't yet had the chance to build-up and try them for a +-15V supply. Good to know they are working well and I'll add that as another confirmed option in the first post to this thread.
Also a VERY nice build. What case are you using? It looks like one of the higher-end HTPC cases, right?
AND are you using the single-ended or balanced outputs?
Curious what you'll find when you listen without the cap bank AND solve your DSD64 issue. Sorry I can't help you with that one, the bulk of my source material is ripped CDs.
At some point I'll get my Stammheim boards built (I got them from when MPAudio still sold raw boards... Michael doesn't anymore due to the level of support required for people not experienced in SMD builds) and compare to my current +-15V supplies, but have been busy these last few weeks with work, organizing my hobby work areas (both audio & model aviation), and a recent unexpected effort of recovering from a small flood when a pipe came loose in our kitchen. The latter wasn't too bad and the only significant damage is to some wood flooring, but at least we now have our living areas functional again.
@Jonathan P, on the Katana balanced outputs, my advice is to go ahead and try them. The circuitry and filtering used by Allo is very similar to that used by 90%-95% of the commercial ESS-chip-based DACs out in the world. Once Allo said that the balanced outputs filtering was equivalent to the single-ended outputs, I have no reservations suggesting people try them. I understand what @Markw4 was discussing around balanced output circuitry and while it has merit, I worry that too much theoretical discussions outside of what can be done with the stock Allo Katana hardware with only minor mods (like bypassing the USB3 for 5V power to the Microprocessor board or all of the powering-options that everyone explored, including @Markw4's cap bank) will only muddy the waters and confuse many. Perhaps it is time for an 'Allo Katana SERIOUS Mods' or 'Improvements for the Katana 2.0' thread?
BTW, if you look closely at the SMD components right at each set of balanced output pads, you'll see the passive filtering components there.
@Jonathan P, please let us know how the balanced outputs work for you if you try them. Doing that is also on my list of things to try, but see my comments above.
Also, over the years since the early ESS 9008 chips, there has been a lot of discussion back-and-forth on whether the on-chip volume processing produces better results than very good hardware volume. The general consensus from those using them in this forums was that the ESS processing should better any HW volume setup. I personally have 'evolved' several good passive volume control options over the years (since 1987 or so) and only recently dipped my toe into this realm with the completion of my 2 Twisted Pear Audio Buffalo IIIPro setups earlier this summer. Without having done much back and forth, my casual impressions are that I can't discern much difference between my passive setups and the on-chip volume processing, at least when using only low amounts of attenuation. BUT I only find using the on-chip volume processing convenient with a setup like Twisted Pear's where their firmware lets you use a simple pot to set the on-chip volume OR Ian Canada's ESS DAC controller with both a physical knob AND remote control. Having to access the driver means that with the current Katana setup, I'll stick with my passives. OR is there a way to physically access the on-chip control for the Katana that I've missed?
Greg in Mississippi
Also a VERY nice build. What case are you using? It looks like one of the higher-end HTPC cases, right?
AND are you using the single-ended or balanced outputs?
Curious what you'll find when you listen without the cap bank AND solve your DSD64 issue. Sorry I can't help you with that one, the bulk of my source material is ripped CDs.
At some point I'll get my Stammheim boards built (I got them from when MPAudio still sold raw boards... Michael doesn't anymore due to the level of support required for people not experienced in SMD builds) and compare to my current +-15V supplies, but have been busy these last few weeks with work, organizing my hobby work areas (both audio & model aviation), and a recent unexpected effort of recovering from a small flood when a pipe came loose in our kitchen. The latter wasn't too bad and the only significant damage is to some wood flooring, but at least we now have our living areas functional again.
@Jonathan P, on the Katana balanced outputs, my advice is to go ahead and try them. The circuitry and filtering used by Allo is very similar to that used by 90%-95% of the commercial ESS-chip-based DACs out in the world. Once Allo said that the balanced outputs filtering was equivalent to the single-ended outputs, I have no reservations suggesting people try them. I understand what @Markw4 was discussing around balanced output circuitry and while it has merit, I worry that too much theoretical discussions outside of what can be done with the stock Allo Katana hardware with only minor mods (like bypassing the USB3 for 5V power to the Microprocessor board or all of the powering-options that everyone explored, including @Markw4's cap bank) will only muddy the waters and confuse many. Perhaps it is time for an 'Allo Katana SERIOUS Mods' or 'Improvements for the Katana 2.0' thread?
BTW, if you look closely at the SMD components right at each set of balanced output pads, you'll see the passive filtering components there.
@Jonathan P, please let us know how the balanced outputs work for you if you try them. Doing that is also on my list of things to try, but see my comments above.
Also, over the years since the early ESS 9008 chips, there has been a lot of discussion back-and-forth on whether the on-chip volume processing produces better results than very good hardware volume. The general consensus from those using them in this forums was that the ESS processing should better any HW volume setup. I personally have 'evolved' several good passive volume control options over the years (since 1987 or so) and only recently dipped my toe into this realm with the completion of my 2 Twisted Pear Audio Buffalo IIIPro setups earlier this summer. Without having done much back and forth, my casual impressions are that I can't discern much difference between my passive setups and the on-chip volume processing, at least when using only low amounts of attenuation. BUT I only find using the on-chip volume processing convenient with a setup like Twisted Pear's where their firmware lets you use a simple pot to set the on-chip volume OR Ian Canada's ESS DAC controller with both a physical knob AND remote control. Having to access the driver means that with the current Katana setup, I'll stick with my passives. OR is there a way to physically access the on-chip control for the Katana that I've missed?
Greg in Mississippi
It works! I'm now able to play DoP DSD64-128-256
But, it's a bit annoying to resample all my pcm music to 176.4khHz, and DSD128 must be 352.8kHz.
Do you have an idea of why it works this way?
Hi Greg,
The case is a streacom fc9. REALLY not cheap but good build.
I use single-ended outputs as my Amp accept only single end cables.
I hope I have time this week to test setup with another mpd version, with and without isolator, with and without film cap bank.
Very curious to ear the differences too.
Isolator is behaving differently in slave and master modes. This has something to do with the fact that on master mode isolator introduces 2 delays...one on BCK going to RPI and one on DATA coming from RPI (while on slave mode , all are delayed the same way so its canceled)
This is why we advise 24/192Khz as maximum if using isolator By resampling you are in fact lowering the delays and your music is playing correctly