The KA17 from Fiio has been a blast to use. It has tons of power, and the sound quality is great. I’m just looking for a bit of advice on its proper use, primarily in terms of these filters, which I don’t quite understand:
MINI: Minimum phase (default)
Conventional minimum phase filter. No front ringing, but significant rear ringing.
FAST A: Linear phase fast roll-off apodizing
Linear phase filters incorporating special window functions to optimize transient response. Theoretical.
FAST: Linear phase fast roll-off
Traditional linear phase filter. Symmetrical ringing. Audio DACs from 20 years ago commonly used this as the default filter.
FAST R: Linear phase fast roll-off low ripple
Conventional linear phase filters that reduce some ripple.
SLOW: Linear phase slow roll-off
Conventional linear phase filters with minimal filtering.
MINI F: Minimum phase fast roll-off
Approximates a “conventional minimum phase filter,” trading some ripple for a faster roll-off.
MINI S: Minimum phase slow roll-off
This filter is recommended if the song has a high sample rate, e.g., ≥176.4kHz.
MINI S D: Minimum phase slow roll-off low dispersion
This offers the most natural sound. Note: “Natural” does not necessarily mean “good,” especially with poor quality recordings.
Also, if you have any other tips or tricks I should know about the device, I’d be glad to hear them. Thank you in advance.
For linear vs. minimum-phase the explainer I’ve latched onto is this: Linear and minimum phase – Troll Audio
The author says there’s no conclusion to draw but their own graphs tell a different story: the minimum phase filter disperses the higher frequencies by different phase amounts and produces a signal that obviously does not look like the original whereas the linear one does. Meanwhile, the supposed “unnatural pre-ringing” of the linear filter is at such low amplitudes and high frequnecy (20-22 kHz) that it can’t be heard and was never a serious scientific argument against linear filters. So I would always pick a linear filter. (The “low dispersion” min-phase is interesting if they’ve managed it, but why bother when you can have NO-dispersion with the linear?)
Slow roll-off allows ultrasonic frequencies to pass, which could cause unknown problems depending on what the amp coming after the DAC knows how to do with them or not, but most likely might cause intermodulation distortion components at audible frequencies. Sometimes slow roll-off filters start the roll-off already in the audible band and literally veil the treble like an EQ (though I don’t think FiiO are this bad at it after all this time). Better use a fast roll-off.
I would just pick the supposedly “20 years ago” linear fast, that does what a DAC’s reconstruction filter for music consumption should do. Its only bad effect is delaying the output vs. the input, which becomes a problem only if you plug a lot of effects processors into eachother because the delay adds up, but that is the stuff of complicated production/live setups. Engineers dealing with those setups are the only people who really need to worry about min-phase options and what differences there are between them.
Huh, according to GoldenSound the apodizing version of the linear filter can fix some problems introduced by the possibly bad ADC used when making the original recording. This is the first thing I’ve heard/read that makes the apodizing option attractive to me (when the term is not misused to designate just a minimum-phase filter). I might just set my KA17 to apodizing if I stick to my plan of getting one in this November in discount season.
LE:
Welp, according to the ESS datasheet for the 9069 the apodizing filter starts to roll off in the audio band around 18k and is already 5 dB down or worse by 20k. That’s a big nope for me dawg. I’d have to know for sure I’m listening to recordings that have the “ADC ringing” problem in them to use such a filter.
So it seems discount season arrived early for me as some local dude decided to re-sell his KA17 after barely any use, for about $105 shipped. I jumped immediately of course, since that kind of discount normally requires combining all the vouchers and coupons and coins and stuff on Ali, and even on 11.11 I couldn’t be sure I’d get it quite that low.
Came with firmware 0.99, didn’t quite play nice with my Android phones, I had to always toggle the EQ setting back and forth before the phones would recognize it, but after flashing v2.00 it’s fallen in line and started to behave.
Been A/B-ing it vs. my trusty Hiby FC3 and… it’s not looking nearly as good as I hoped. On the plus side, it seems to have tighter sub-bass with shorter decay, allowing relatively fast beats to be more easily perceived as separate. But in the treble it’s noticeably more veiled than the Hiby (4th device now that does this) even after I switched the reconstruction filter to Linear-Fast. At this point I have to start suspecting the Hiby for adding something it shouldn’t be adding, and I will try to PC-line-in “measure” everything I have with pink noise or some multi-tone signal, and compare all the response curves to try to get to the bottom of this. In the end there can be only one champion of treble presentation.
Stage-wise it didn’t do anything detectably better on BAL-out vs. the Hiby which is SE-only, both feeding into the ESP95X’s energizer. Figured it might be the energizer dragging them both down to its level, so I re-tested with the HE-400i, but nothing changed. I can’t hear a left-right separation difference between KA17 BAL and FC3 SE. In fact I had this funny thing happen with one song where because the KA17 doesn’t have the treble of the FC3 it wasn’t bringing forth some lateral sounds as well and it was making the stage seem narrower. Once again, after the AkLiam PD4 test, BAL-out seems to be doing nothing for me once everything is volume matched / power is sufficient for the headphones I’m using.
So far I’m not convinced this is a keeper. I’ll report back if I clarify what’s going on with that treble difference.
Yep, caught the culprit red-handed. Turns out @pkane from ASR has this free tool that does loopback Multitone-based testing and analysis, so I could very easily do exactly the measurements I was hoping for, and got the frequency response of each device in less than a minute. The KA17 has a nearly half-decibel deep V-shaped cutout centered in the 6-10 kHz region (not sure why it moves between different replots, maybe 300 tones are not accurate enough, but when I tried 400 the app just crashed) and then also rolls off early as it nears 20 kHz, even after I selected the linear-fast filter (I suspect they did this in the amp section because the ESS datasheet doesn’t show any such rolloff for the linear-fast). The half-price Hiby FC3 meanwhile is flat all the way, as a DAC should be.
Isn’t desktop mode made for using with secondary power input? I feel like you’re current starving it or something is up with attenuation. Are you running it at full power in high gain? If so, does that also happen in low gain?
It’s running off the PC’s USB 3.0 for loopback into the line-in, it’s not power starved. D.Mode is just the highest power mode, but it will take that power wherever it can find it, depending on what’s detected on both USBs. Internal volume was 48/60 for this, I turned it as high as I could without clipping, but also was plannig to not max it so it gets every chance to sound its best.
But I hear the same veil in every setup, every operating mode, BAL or SE. Only thing I still want to test is firmware 2.00 because some on YouTube say the sound changes for the worst.
Says it will always take priority from the extra input if it’s in D mode. That makes me think there’s something happening that is current limiting the output. There should be zero clipping whatsoever at full volume in low gain and if there is, it might be your source. Max volume or within a couple notches should be best performance on every amp on the planet, but that doesn’t apply if the source is clipping or if there’s a current starve issue. Windows has been known to clip for no reason unless you’re in some kind of exclusive mode, maybe that could be causing some of the issues?
Windows volume was at 90/100, the multitone app I left at 0 dB, but anyway those are clearly not the problem since the FC3 performed well when fed in the same way.
Nobody’s limiting anything, I didn’t plug in aux power and it’s capable of pushing max power if it has a good source at the data USB, even if it was a phone.
OK so I did a few more and the dip is always there. Here’s the same as above but D.Mode OFF (upped the volume to 54/60 to stay away from the noise floor but still be at least -3 dB of max):
Then I went low-gain but also switched to BAL-out to cover more ground. Unfortunately there’s something about my cabling for this scenario where using a headphone cable + 2 types of adapters at both ends and using only 1 channel into the line-in looks like it creates 30 whole-ass dB of extra noise on BAL, to the point that it’s even visible on all the FR plots as I go back up in power levels. Same happens to the AkLiam PD4 BAL-out, so I would ignore that and just focus on everything else, i.e. the dip is still there, just that the overall shape seems to change toward the brighter somehow. That could still be from the extra noise being shaped that way, not sure. (In A/B listening I still hear it as veiled + some weird sparkle above the veil area, so it’s not like it makes it seem any flatter.)
And lastly I reflashed the newer firmware v2.00 that fixes some overheating and auto-shutdown scenarios(?) and adds multiple button modes so you can have track switching on the volume buttons. The dip seems to be a tad lower this time and also the 20k rolloff now looks like another dip at 17k-ish, followed by the start of another rise. In having more hills and valleys I guess this could be called worse than v0.99, like those people on YouTube were saying.
HiBy FC3 for clean signal, but it doesn’t have much power. Still looking for a more powerful one that stays just as clean and neutral, I haven’t found it. (BTW, not the mk II of the FC3 with the screen on it. I still have doubts that it performs identically to the original because of the un-spectacular results Amir got for the mk II.)
