Getting the best out of Allo.com's new Katana DAC...

Member
Joined 2003
Paid Member
Some quick responses...

@terry22, on testing the IsolatorPi on the Katana, I have not. Actually, I was not planning to do that test myself, just find out if the IsolatorPi should work ok under a Katana and post that info here.

Ian did tell me his I2C signals are bi-directional, but I have yet to find out exactly what mods Allo made to the Isolator 1.2 and what lines need to be bi-directional to make it work. Should have that later this week. IF you have one of Ian's IsolatorPi's, give it a try and let us know... as long as you have things powered correctly, it should not hurt anything. EXCEPT I would not have it hooked up to an amplifer and speakers (or expensive headphones) to test. In situations such as this, I plug in a cheap RCA cable at the DAC output and use a cheap cellphone headset across the output contacts to verify if things are working correctly before I risk my amps and speakers.


@alexkosha, I have to correct what I said above... I have a couple of Salas Mini-Reflektor boards almost built for 5V each, that's what I'll try at some point, but again likely not soon. BUT I suspect the latest Salas UltraBiB will be near the top of the performance curve for all of the rails on the Katana and very likely above what I'm using and have tried so far.


@Markw$, I also have some other large (100uF/200V & 1,500uF/700V) polyprops I can try across various lines... sort of like with the Salas above, someday.


@stellarelephant, thanks for the review... I'll send you a check later this week ;). Actually, Kevin's tests with the modified Jamecos are why I started considering them for things like Katanas. I originally modified several for use replacing the SMPSs on networking gear, with the mods mostly designed to lower noise fed back into the AC line by the Jamecos (adding an AC filter cap, a John Swenson transformer secondary ringing snubber, and good Schottky diodes). I never intended to use them as low-cost alternatives power supplies for DACs and boards like Isolators and Kalis. Darned surprised that they work as well as they do!


Then a general note... I want to be clear I DO NOT intend to try to test ALL available options under consideration or discussion here. I do not have interest, time, money, bench space, bench time, or listening time to do so. I've got a lot of other projects in the works... now including improving my other 'best' digital sources. I'm hopeful that others will continue to chime in with their suggestions, ideas, and results. AND thanks for those who already have!!!!

One area I do want to test further that I haven't mentioned before is output stages. I strongly suspect a LOT of the SQ I'm hearing out of the Katana is due to the Sparkos discrete opamps used in the output stages. I think it's worthwhile to confirm that and how they stack up to some other options. BUT, using a different output stage on the Katana would be tricky, given how that board is sandwiched between the DAC board and Microprocessor board. My thought is to use Ian's ES9028Q2M board as a test bed for those comparisons, as his controller is a seperate unit and his output stage is the top board of the stack. I've given some thought on how to adapt the Katana output board onto Ian's boards (heck, I've given some thought on how to fit one on my ES9028 BIIIPRO setup too!) and think it is do-able. I'll get Markw4's OPA1622 I/V stage info from his ES9038Q2M board thread. I've got Ian's OPA1632 board here (need to populate it) AND am getting info on his OPA861 I/V setup too. Another one that is pretty easy to implement is one based on the AD844 chip, there's a long thread on that in another forum here.

AN interesting aside on this, there's a recent thread on Audio Asylum discussing alternative opamps to replace LM4562s for DACs and other purposes here:

LM4562's - bobwire - Tweakers' Asylum

While it goes off the rails for a bit, it has some interesting discussion of various options including the OPA1622 (and the FET-input OPA1642, which also now goes onto my list) AND several posted experiences with using discrete opamps too, including the Sparkos. Their comments mirror what I've seen in other similar threads, that the discretes generally beat the monolithics, but it will be useful to try I think.


Finally, I did some additional listening tests on alternative supplies last night with some VERY positive results... should make @randytsuch happy... more later today or tomorrow.


Greg in Mississippi
 
I'll get Markw4's OPA1622 I/V stage info from his ES9038Q2M board thread.

Greg,
It's OPA1612, just to be clear. Post #3003 contains that schematic, although I am not using nor do I recommend the balanced outputs as MrSlim showed them. I replied in a subsequent post about a preferred approach if balanced out were wanted. Output stage assembly instructions I was requested to create can be found here: Dropbox - Output Stage Instructions.zip Also, after writing up the instructions, I went back and put in heavier gauge +-15v wires for each opamp to help keep power supply impedance low back to the bigger filter caps. In addition, Katana locates the AVCC dividers to produce Vref on the output stage boards, whereas the modded dacs do that closer to the AVCC power supply. As a result, the dividers are not shown on the output stage schematic.

Regarding opamps, I tend to go for low distortion types, and try to avoid picking opamps to color the sound. Philosophically, I don't want to color the sound and most low distortion opamps used for audio should sound almost exactly the same unless there is still a problem with the circuitry somewhere. For LME49720/LME49860, RF such as from DECT phones can also cause problems. Also, when discrete opamps are preferred over IC, I always have to wonder if a little low level euphonic distortion is behind the preference. It can sound nice in some ways, but there may be more IMD too for people who notice that. It may be more audible on musically dense recordings.
 
Last edited:
Member
Joined 2003
Paid Member
@markw4,

Thanks for the correction. Yes, it was the OPA1612 that was mentioned in that Audio Asylum thread as a very good monolithic IC opamp for I/V purposes. Apologies for the confusion.

Also thanks for the guideance on where to pickup the details on your OPA1612-based I/V stage. I've downloaded the instructions and will look at those posts.

AND I'm getting the impression that we have somewhat different perspectives on what constitutes good audio design. I tend to align with those who suggest that after a certain level of measured distortion lowering, further improvements are not audible. AND that some of the techniques used to produce extremely low distortion measurement in circuits and filters will produce poorer sounding results.

Of course, given that each of us have different and personal 'hearing mechanisms' (physical and our internal processing systems), different sets of criterial by which we judge how audio systems should be designed, built, and sound, and that the filtering of our former through our latter will always produce somewhat different results person-to-person, I do think it's good we have plenty of space in our hobby for a variety of views. AND I appreciate and respect yours!

Greg in Mississippi
 
I tend to align with those who suggest that after a certain level of measured distortion lowering, further improvements are not audible. AND that some of the techniques used to produce extremely low distortion measurement in circuits and filters will produce poorer sounding results.

No, that's not it I don't think. I can hear very low distortion levels here, and you probably could too if you were here. For one thing, my power amp is ultra low distortion. Also, I don't know about "techniques that produce low distortion produce poor sound." The only times I see something like that happen is when the low distortion "lowers the water to expose the rocks." In other words, it makes audible that there is still some problem with the circuit. What some designers do is use other low level distortion to mask the remaining distortion they don't know how to get rid of. That is exactly what I was talking about when I said something to you (by PM?) about a particular brand of headphone amp that sounded better at first, until the remaining problems where fixed, then it sounded worse.
 
Greg,
You might be surprised to know that Benchmark DAC-3 uses plain old LME49860 IC opamps, and it is on the Stereophile A+ recommended list. Another dac on that same list is Chord Dave, which according to its maker uses composite IC opamps. A composite opamp is two opamps combined together to effectively make one higher performance opamp. They do that so that feedback loop gain can be very high and stay that way up to pretty high frequencies. It makes the distortion so low that they don't have to worry about it.

I agree with the philosophy of Benchmark and Chord that one should strive for low distortion and if it seems to sound worse when one source of distortion is lowered, then find the source of the now audible distortion that you just unmasked and fix it too. Then, if customers want to follow a low distortion dac with tube preamp or something to add a little color back in, fine. Or people who just want clean, elegant, low distortion that sounds great can have that too.
 
Last edited:
Member
Joined 2003
Paid Member
Report on listening to Ultracap-based supplies on the Katana...

I'm going to start this report off with DIRE warnings... DON'T EVEN THINK OF trying this at home! ESPECIALLY IF WHEN YOU LOOK AT IT, you have to ask a LOT of questions about how to implement it and do it safely.

Please understand the amount if energy stored in each of the Ultracap sets I'm using is STAGGERING! Back in the 1979, I built my first full audio gear kit, a Hafler DH-200. This 100wt/8ohms rated amplifer had a +-60V supply, with 10,000uF/80V capacitors as stock. That calculates out to 18 Joules of energy stored in each cap. The supply I'm using has 2 325F/2.7V Ultracaps, with each charged to 2.5V, which comes out to 1015 Joules or 56x the energy stored in each Hafler amplifier cap!!!!! AND these caps will release that energy in metal-welding hundreds-of-amp rates. That won't do anything good for your Katana (or anything else connected) if a fault happens in the gear that shorts the output.

Also someone who works with these did tell me they will kill. I didn't think so before because of the low-voltages involved, but that is too danged serious of a warning to just ignore!

Also, this setup has NO safeties built-in. I also use a number of Uptone Audio LPS-1.2's to power various pieces of gear in various configurations in my systems. These 'off-the-grid' devices use 2 continually-swapped banks of 7 10F Ultracaps to provide the output voltages (Let's estimate about 220 Joules per bank when used at the max output voltage). The designer, John Swenson, spent a LOT of effort designing, coding, and testing the processes to keep the units ALWAYS safe, including:

- power-on Ultracap discharge cycle

- controlled charge rates with no output until both banks are charged

- active charge level balancing

- Overtemp and overcurrent cutoffs

- power-off Ultracap discharge cycle


These are the safety features I've encountered or heard of others encountering using the LPS-1.2. I suspect there are others that I don't know of.

The setups I used HAVE NONE OF THESE! ONE USES AT YOUR OWN RISK!

You can find info on these Ultracap setups in these 3 threads:

ES9018K2M, ES9028Q2M, 9038Q2M DSD/I2S DAC HATs for Raspberry Pi (I have some comments in this thread in posts 622 & 629

Develop ultra capacitor power supply and LiFePO4 battery power supply

Ian's Ultracapacitor Explore PLUS - build thread

Again, if you read all of the relevant comments and the schematic (that is posted in one of the threads) AND you have to start asking questions, please don't pursue this route. Just wait for Ian's ready to use-version that hopefully will have less of a possibility of severely damaging your equipment or you.

Ok, having said all that, what did I test and how did they sound?

The Ultracap supplies use 2 of the 325F/2.7V Ultracaps in series with a simple resistor-balancing scheme. A person who is MUCH more knowledgeable about Ultracaps than I said that will work at low charge/discharge rates (which is what we see here), but not has you rise into the .5A range. They have an on-board 5V LT3042 regulator which limits the charge to about 300mA (Ian says 200mA is MUCH safer). Charging them takes roughly an hour (I've not timed it!). AND I did not use them without a charging device connected, so I don't know how long they'll last on the Katana without being float-charged. BUT based on my testing (and Ian's) of these in other setups, it will sound slightly to somewhat better without the charger connected, depending on the float charging device (and possibly the driven circuit, but I don't have sufficient experience to say that yet).

As with the tests of the modified Jamecos above, I only replaced the 2 5V supplies feeding the Isolator output / DAC board and Microprocessor board in this test. AND I tested 3 configurations using the Ultracaps against my current 'best' AC-connected power supplies.

The 3 configuration varied by the type of 'float' charging supplies:

- The same 2 modified Jameco 12V/1A units from above

- 2 Uptone Audio LPS-1.2s set to 7V

- 2 Uptone AUdio LPS-1s set to 7V

AND for good measure, I also listened to the LPS-1.2's powering those same 2 Katana power inputs directly without the Ultracaps in place.

First, the Ultracaps charged by the modified 12V Jameco. Remember, the same units feeding the Katana through both halves of the 5V OPC dual 4||LT3042 boards and these feeding had me rate the sound as 'a hint of greatness'. With the Ultracap supplies, SQ pushed a bit closer to my 'best' AC-connected setup, perhaps I could say 'a bit of greatness'. Much of the difference was from the earlier 12V Jameco-based setup was in slightly greater definition in the midrange and up. Interesting, in all of the trials with the Ultracap supplies, the bass was not as dynamic, defined, or impactive as the AC-connected setup. I don't know why... burn-in perhaps? Then there was another change I noted that could be an issue... a slight shelve upwards from the middle midrange throgh the mid-treble. I observed this in other applications, but there the Ultracaps were charged to the final circuit voltage (3.3V) and directly connected to the powered chips. The shelve effect was less AND the benefits of the Ultracap power were greater. Again, it might be break-in, it might be the sonic signature of these caps, it might be something else entirely. I'll update as I put more time on them. The good thing is that the cost of the parts for this setup not a lot higher than that for the OPC (and lower than the already assembled MPAudio units), so conditionally a slightly better setup for a little less $$$. @Randytsuch, you'll be happy!

Next, the Ultracaps charged by the LPS-1s & LPS-1.2s. These were similar in most ways, but I prefer the older LPS-1s in this application, as that 'shelve' mentioned above is slightly diminished. Possibly output regulators (TPS7A4700's in the LPS-1 versus paralleled LT3045 in the LPS-1.2). I'd rate both of these as roughly equivalent to the 'best' AC-connected setup... both of these Ultracap setups had slightly quieter backgrounds, similar dynamics, again less bass power and definition. Because of the 'shelve' I'd put this setup powered by the LPS-1.2s as a bit below the 'best' AC-connected setup, the other a bit above. Ah, but for a cost... at $435 per LPS-1.2 (the earlier LPS-1 was $395), AND I put together linear supplies to replace the provided SMPS's as I'm just keeping them out of my systems, figure almost $500/$460 per unit. SO this adds significantly to the power supply costs. Depending on what's important to a listener, that delta might be worth it. For me, I could listen to either, but marginally prefer the AC-connected setup.

Bottom line.. the Ultracap setup powered by the Jameco showed great promise, though that 'shelve' could be an issue in some setups and to some ears. Curious if any of the other brands/lines of Ultracaps exhibit this. The LPS-1/1.2 powered setups were better, but the incremental cost would not be worth it for most.

Finally, a pair of LPS-1.2s alone providing the 2 5V feeds? Above the 12V Jameco-charged Ultracap setup in that quieter background and lack of the shelve, lower than the LPS-1/1.2 charged setups with slightly lower dynamics and bass power/definition compared to all 3 of them. Not a win given the increased cost.

Note that grounding proved to be important for both the previous 2 Jameco powered setups, the Jameco-charged Ultracap setup, and the LPS-1/1.2 setups. This system has 4 isolated zones:

- Music server -> networking gear -> FMC (transformer's frames grounded, but none of the gear)

- FMC / RPi -> Isolator input side (grounded to AC ground via the RPi USB connectors)

- Isolator output side -> Katana -> input side of transformer attenuator (not grounded to AC ground using the AC supply, required grounding for these supplies)

- output side of transformer attenuator -> amplifier -> speakers (grounded to AC ground)

SOME of the time I was listening to the Jameco powered and LPS-1/1.2 powered setups, it was cold-ish here (lower 30s F to upper 20s F) and I had static discharge noises IF I did not connect the Katana ground to AC ground. This has not been an issue with my 'best' AC-connected setup. I suspect that is because the transformers' frames ARE connected to the AC grounds AND they have no electrostatic shield, so they may sufficiently ground the static. This is not an issue, just an observation that attention to grounding MAY be needed for different power supply setups, especially if you are using an isolator before the Katana... AND also if you have an optical network link.

Another few days, another set of tests. Whew!

Later y'all!

Greg in Mississippi
 
Last edited:
Member
Joined 2003
Paid Member
Mark,

Good question... sorry I wasn't more clear. From from the middle midrange through the mid-treble, the apparent volume is slightly higher than the rest of the frequency ranges. Very much like a shelving EQ, it is turned up a bit. I can hear it in my system, I wouldn't call it a fatal flaw, but it is noticeable. IF a system has any other emphasis in that range OR a person is sensitive to it, it might eliminate this solution for them. I've heard recordings with similar types of slight shifts in limited frequency ranges... a lot like that.

It is only there using the 325F Ultracap supplies. The Uptone Audio LPS-1/1.2s don't exhibit this, even though they DO have the same manufacturer's Ultracaps in some production runs.

Greg in Mississippi
 
@alexkosha, I have to correct what I said above... I have a couple of Salas Mini-Reflektor boards almost built for 5V each, that's what I'll try at some point, but again likely not soon. BUT I suspect the latest Salas UltraBiB will be near the top of the performance curve for all of the rails on the Katana and very likely above what I'm using and have tried so far.




Greg in Mississippi



Understood. I’ll get my UltraBiB boards soon and I’ll start the setup. I’ll report my audition results and general findings, if any.
It would save my time if you will give me measured current consumption for all of 4 PSUs. I saw some info in your first post, but just need to confirm if all is still correct there.
I need to set Salas boards with some margins and don’t want to guess.
 
Easy enough to ring out ground on the balanced output connections with a DVM, but maybe not such a good idea to use balanced rather than the unbalanced outputs. Usually, the way balanced output are derived from a 6-opamp output stage like the one in Katana means that not all output stage filter processing has been done at that output. Unless the user knows how to finish that, probably better to either use the unbalanced outputs directly, or rebalance them with a two opamp circuit commonly used for that purpose (such as the one shown below).
 

Attachments

  • Balanced Out.gif
    Balanced Out.gif
    3.4 KB · Views: 178
Member
Joined 2003
Paid Member
@atca,

Hmmm... the pinout for the XLR's IS in the original Katana manual, but I didn't see it in the later one.

Pinout on each is + || Grnd || - as viewed from above, with '-' at the '0' printed on the right side of each XLR pads.

@Markw4, I'm curious what Allo will have to say about the output filtering on the XLR outputs.

Greg in Mississippi
 
Member
Joined 2003
Paid Member
A Katana SQ report with different supplies from a friend...

@Terry22, very curious how Ian's IsolatorPi works with the Katana... thanks for taking this on, sorry I haven't gotten an answer back to you yet.


I recently received a Katana report from a friend with lots of experience in the audio industry and as an audio enthusiast. He's been testing his Katana without the Isolator 1.2 at first, trying a number of power supply configurations and alternatives.

His base configuration now is:

- RPI3B+ <- 3A SMPS
- Katana DAC board <- LT3042 board with pass transistor
- Microprocessor board <- LT3042 board with pass transistor
- Opamp board <- Twisted Pear Placid HD with Antek tranformer

He's been very methodical in his testing...

- Started with the 3A SMPS powering the entire RPi/Katana stack ('Ok, modded Boss was probably better')

- Moved to separately powering the DAC board ('Better')

- Added a lab supply for the Microprocessor board and Opamp board via the DC-DC converters ('Better again')

- Added the +-15V via the Placid HD ('Wow. Almost mid-fi versus hi-fi. Blacker, clearer, more defined, yet musical.')


He then tested several different options on the DAC and Microprocessor board supplies.

On the DAC board, he tried a TPS7A4700 regulator board, then an LT3042/45 board with a pass transistor. His comments were:

'On the Boss with the isolator, I prefer the TPS7A7400 implementation. The difference is tough to describe. On the Katana, powering the DAC board With the LT3042 with a series pass transistor in a low noise configuration, things just seem… right. With the TPS7A7400 regulator, nothing really seems wrong, it is just kind of dry, and, mostly, un-involving - you almost instantly become uninterested in listening to the music. Funny.'


Then on the Microprocessor board, he heard something surprising:

"I tried a bench supply, LM317, LT3042, and TPS7A7400 implementations. Frankly, powering just the micro controller separately, I did not expect much difference, having already separately powered the I/V and buffer and the DAC. But, no! A bench supply or a generic regulator are just sort of awfully limiting. Limiting in terms of soundstage width and a depth particularly. With the LT3042 with a series pass transistor in a low noise configuration, the width of the presentation simply leaps outside and beyond the width of the speakers in the horizontal plane compared to using a garden variety supply. Depth goes much deeper too. My preference again runs to a LT3042/LT3045 implementation.'

I found this surprising too. Since receiving his report, I tried a sightly better regulator powering the Microprocessor board in my setup (moving to a single 3||LT3045 regulator board from MPAudio versus the dual paralleled output 4||LT3042 regulator board from former DIYAudio poster OPC). I heard improvements too. I would not count them as large as his, as in my experience the differences between those two configurations is not large, but I have consistently preferred the MPAudio board to the OPC one where I've compared them.

He also experimented with the shunt current on the Placid HD, setting it for 300mA, 200mA, and 100mA. He didn't offer an absolute assessment on them, saying '... you could adjust how much dynamic slam, and to some extent the amount of bass. As to which setting is “best,” well that’s personal preference/system dependent.'

What I found most interesting was his finding on powering the Microprocessor board. Before his report, I would not have expected much if any difference between the supplies he tried.

Also found interesting that his overall experiences in going to his 'better' supplies mirrored mine. I especially found his comments comparing powering the DAC board with the TPS7A4700 regulator versus the LT3042 / pass transistor setup similar to mine on using the DC-DC converter versus a good +-15V supply... with the latter, I just wanted to listen to the music. With the former, I kept wanting to go back to the separate +-15V supplies.

Anyone else have any new reports on the Katana's SQ with good to great supplies?

Greg in Mississippi
 
@Greg Stewart
I love you man for sharing all that knowledge!
So I have latest Katana 1.2, Isolator 1.3 and 2 iPower 5V PSU.
The sound's pretty good but after reading your posts, I want to use 5 PSU!
I apologize in advance if I missed it in the posts but I was wondering if you could point me to detailed instructions and jumper settings. I could find info in allo's doc but it's not clear to me in case of Isolator.
 
Jonathan P,

The biggest difference when running the Katana stack with separate power supplies, and then adding the isolator into the mix is that you need the jumper on J30 on the DAC board. This allows power from the DAC board to feed the clean side of the isolator. Without the jumper I mentioned (J30), you'll get an ALSA/MPD error in MoOde Audio, at least, because the Pi can't "see" the DAC.