Hello Mark,
I realized on one of your pictures about VCCA powering (we discussed before) you placed the jumper wire between both VCCA straight over the DAC chip - probably it catches some RF here? I have routed it below the chip, as all external power lines run below the GND plane.. just a guess...
I realized on one of your pictures about VCCA powering (we discussed before) you placed the jumper wire between both VCCA straight over the DAC chip - probably it catches some RF here? I have routed it below the chip, as all external power lines run below the GND plane.. just a guess...
Hi freezebox,
I shortened that wire some and moved its connection points to cap solder pads closer to the dac chip, but no clear change from that. In both cases, the old position of the wire and the new position, there was enough extra wire length to push it around with a plastic stick to see if any sensitivity with respect to its position. So far as I can tell, if there is any effect at all it would appear to be very small, and much too small to account for what I am seeing.
The RF around the dac as it can affect jitter is so sensitive that sometimes my standing or leaning over near the dac can affect DPLL tuning stability. With the bag of ferrites in place it is much more stable in terms of more distant influences. It just shows how far EM fields exist in space around a 100MHz clock source with edge harmonic frequencies possibly extending up into GHz. Then there is the 11.2MHz DSD with traces exposed on one side going into the dac chip. Some of the wavelengths especially from the clock may be pretty small. The effects are well beyond the capabilities of my scope to help with, both in terms of frequency and in terms of small signal levels.
Another thing is that while we have replaced a lot of the analog electronics used in dac circuitry with high quality parts, we are still using whatever cheap decoupling caps they used everywhere on the dac board. Of course, they probably used something less than the best quality RF parts for that, and instead used parts they figured were good enough for a dac of the quality they were selling. That would be my guess. We might have to do some online research and some testing to figure out what would be the best course of action.
I shortened that wire some and moved its connection points to cap solder pads closer to the dac chip, but no clear change from that. In both cases, the old position of the wire and the new position, there was enough extra wire length to push it around with a plastic stick to see if any sensitivity with respect to its position. So far as I can tell, if there is any effect at all it would appear to be very small, and much too small to account for what I am seeing.
The RF around the dac as it can affect jitter is so sensitive that sometimes my standing or leaning over near the dac can affect DPLL tuning stability. With the bag of ferrites in place it is much more stable in terms of more distant influences. It just shows how far EM fields exist in space around a 100MHz clock source with edge harmonic frequencies possibly extending up into GHz. Then there is the 11.2MHz DSD with traces exposed on one side going into the dac chip. Some of the wavelengths especially from the clock may be pretty small. The effects are well beyond the capabilities of my scope to help with, both in terms of frequency and in terms of small signal levels.
Another thing is that while we have replaced a lot of the analog electronics used in dac circuitry with high quality parts, we are still using whatever cheap decoupling caps they used everywhere on the dac board. Of course, they probably used something less than the best quality RF parts for that, and instead used parts they figured were good enough for a dac of the quality they were selling. That would be my guess. We might have to do some online research and some testing to figure out what would be the best course of action.
Switching back to talking about Katana for a bit, just tried Katana 1.2 for the first time. I am starting out as the manual recommends with the two power supply configuration. Also, right now the 'sound quality' output stage board is still in the stack as that was the configuration at shipped. The sound quality this way is like something we surpassed a long time ago with our modded dacs, kind of rough, grainy, forward, and clearly a bit distorted. However, it is listenable, and may well be better than other dacs some people have been using. For the RPi, I used the SMPS that came with it; for the other 5v supply I used a $50 ifi brand claimed-low-noise type. I don't have a reason to think its too bad for an SMPS, in fact it may be pretty good, but probably not compared to a very clean linear supply.
I then decided to do something that according to the manual is clearly not recommended and is 'advanced,' which is to use three 5v supplies. In this case the Katana dac board is powered by the additional 5v supply. I used a 7805 type linear supply for that. The linear supply does have ultra-soft recovery rectifier diodes but is otherwise pretty standard. Katana sounds much better this way. Even if Allo doesn't recommend it, I would definitely recommend giving it a try. All the bad kind of words I used to describe the sound with two power supplies are still somewhat applicable, but at much lower levels (and likely lower yet with the 'thd' output stage). However, while clearly better, it is still pretty far from where we are with our modded dacs at this point (our fully modded and register tweaked dacs, that is).
For the next step rather than switching output stage boards, I would like to try the isolator. One slight problem though, I just ran out of suitable 5v supplies. Another one will be here tomorrow, for another $50. Will pick up with isolator and 'thd' output stage listening comparisons when able. Also would like to try external +-15v supply soon.
While waiting for a power supply, started thinking about the following: Taking a brief look at prospective costs so far, we might have for a dac with isolator something like around $250 for Katana, +$40 for isolator, +(cost of 4 each, 5v power supplies dedicated to dac use, could be $200 if all ifi powered, for example). That's okay though. I want sound quality above all else, so the real question for me is first how to get it, then secondarily, how to pay for it. Hopefully, tomorrow will get back on the path to finding better sound quality through Katana dac.
I then decided to do something that according to the manual is clearly not recommended and is 'advanced,' which is to use three 5v supplies. In this case the Katana dac board is powered by the additional 5v supply. I used a 7805 type linear supply for that. The linear supply does have ultra-soft recovery rectifier diodes but is otherwise pretty standard. Katana sounds much better this way. Even if Allo doesn't recommend it, I would definitely recommend giving it a try. All the bad kind of words I used to describe the sound with two power supplies are still somewhat applicable, but at much lower levels (and likely lower yet with the 'thd' output stage). However, while clearly better, it is still pretty far from where we are with our modded dacs at this point (our fully modded and register tweaked dacs, that is).
For the next step rather than switching output stage boards, I would like to try the isolator. One slight problem though, I just ran out of suitable 5v supplies. Another one will be here tomorrow, for another $50. Will pick up with isolator and 'thd' output stage listening comparisons when able. Also would like to try external +-15v supply soon.
While waiting for a power supply, started thinking about the following: Taking a brief look at prospective costs so far, we might have for a dac with isolator something like around $250 for Katana, +$40 for isolator, +(cost of 4 each, 5v power supplies dedicated to dac use, could be $200 if all ifi powered, for example). That's okay though. I want sound quality above all else, so the real question for me is first how to get it, then secondarily, how to pay for it. Hopefully, tomorrow will get back on the path to finding better sound quality through Katana dac.
Last edited:
Hi Mark
We recommend 1-2 days minimum for the burn in . Second if you add the isolator you need the same number of PSUs, isolator will provide power to RPI. I am not sure what PSUs you are using on your modded dac but I suspect that some of the differences you are hearing might be due to the fact that you used different PSUs (linear maybe) on your setup.
A good jump in sound quality is switching the DC/DC convertor off and using linear dual rail for the opamp board (I think thats what you are doing on your modded dac) .
At last , SQ stage has the Rasmussen mod included , maybe you can hear more distortion , try the THD+N stage and it might be better suited .
Looking forward to your findings.
We recommend 1-2 days minimum for the burn in . Second if you add the isolator you need the same number of PSUs, isolator will provide power to RPI. I am not sure what PSUs you are using on your modded dac but I suspect that some of the differences you are hearing might be due to the fact that you used different PSUs (linear maybe) on your setup.
A good jump in sound quality is switching the DC/DC convertor off and using linear dual rail for the opamp board (I think thats what you are doing on your modded dac) .
At last , SQ stage has the Rasmussen mod included , maybe you can hear more distortion , try the THD+N stage and it might be better suited .
Looking forward to your findings.
Last edited:
Hi cdgames,
Thanks for clarification on the isolator. No doubt you are right that things will get better with better power, and with the 'thd' output stage.
I just wanted to start at the bottom at work my way up, so to speak. Since the manual say things that could lead the average customer to think what I did on the first go around is fine, I wanted to see what they would get. Why the manual is like that I have no idea.
EDIT: Also, I will give some time for burn in, as you suggest. However, I am not much of a believer in it personally. My own opinion is that burn in is mostly something that happens in the listener's brain as they get used to a new sound device. Nevertheless, I respect that other's may see it differently.
Thanks for clarification on the isolator. No doubt you are right that things will get better with better power, and with the 'thd' output stage.
I just wanted to start at the bottom at work my way up, so to speak. Since the manual say things that could lead the average customer to think what I did on the first go around is fine, I wanted to see what they would get. Why the manual is like that I have no idea.
EDIT: Also, I will give some time for burn in, as you suggest. However, I am not much of a believer in it personally. My own opinion is that burn in is mostly something that happens in the listener's brain as they get used to a new sound device. Nevertheless, I respect that other's may see it differently.
Last edited:
Hi Mark
I think that most people will use the Katana with 2 SMPS. (or one with dual output like we are developing). If someone is not afraid of getting dirt under fingernails , then multiple PSU will elevate the SQ
In hi end dacs , power is very important . A fair comparison will be running both DACs will clean power (and frankly you only need 2 x 5V one for RPI/isolator and one for Ess9038 board) and one dual rail 15V (that you probably already have on the modded dac)
On each successive PSU , SQ is getting better. How good ?
I am biased of course , but in my opinion not many commercial DACs will beat Katana
About burn in , it might be psychological , in our testing we cannot see any change. Still noise reduction in uV might change the way we perceive the sound , but its all theory .
At last , I will repeat . Katana is a good DAC with 2 SMPS , with clean power 2x5V or 3x5V and dual rails 15V , its on a different level.
Greg Stewart and one more team is testing Katana with good PSUs against pretty hi end gear and I think everyone is happy so far but lets wait for the end conclusion .
Happy testing Mark , lets see what you find.
I think that most people will use the Katana with 2 SMPS. (or one with dual output like we are developing). If someone is not afraid of getting dirt under fingernails , then multiple PSU will elevate the SQ
In hi end dacs , power is very important . A fair comparison will be running both DACs will clean power (and frankly you only need 2 x 5V one for RPI/isolator and one for Ess9038 board) and one dual rail 15V (that you probably already have on the modded dac)
On each successive PSU , SQ is getting better. How good ?
I am biased of course , but in my opinion not many commercial DACs will beat Katana
About burn in , it might be psychological , in our testing we cannot see any change. Still noise reduction in uV might change the way we perceive the sound , but its all theory .
At last , I will repeat . Katana is a good DAC with 2 SMPS , with clean power 2x5V or 3x5V and dual rails 15V , its on a different level.
Greg Stewart and one more team is testing Katana with good PSUs against pretty hi end gear and I think everyone is happy so far but lets wait for the end conclusion .
Happy testing Mark , lets see what you find.
Hi cdsgames,
Actually, I think you probably right about that. I don't listen to most dacs, but based on the few I listen to most are quite bad, IMHO. That probably says more about my calibration for what I think dacs should be than it says about the dacs. Or, just maybe most dacs are quite bad by the metric of how close they come to just playing everything on a CD without adding a sound of their own.
The fact is, it is very hard to make a truly good dac, as measured by the metric I suggested. Dacs that come quite close are very expensive. That is why I started trying to help people make better dacs. Somehow, we need make good dacs more affordable so people can get more accustomed to what good sound is.
In that regard, I applaud Allo's introduction of Katana. I appreciate what you are trying to do, and I have a pretty good idea how very difficult it must be to do it in an RPi stack form factor.
I guess I am never satisfied. Likely, that keeps me working on my own dac project. It also means that while I am very happy to see Katana come to the point it is now, I wish that more work had been done to figure out recommended power supplies, even if only a diy schematic. That is why I suggested a possible topology to try in the another thread.
In the world of audio technology we have today, it seems a dac designer's work is never done. There is never the perfect dac, only good, which is the best, and less than good, which is almost all dacs.
Why do I say a the best dac is only good? Because the best dac doesn't make music sound good, or sound better. It does nothing to music other than accurately reproduce what is on the record. If the recording is bad, the best dac will sound bad because that is the truth of what is on the record.
So, when I describe the audible deficiencies of Katana, I am describing how it differs from perfection. I don't think that should be omitted from the description of a dac intended to be a serious dac, one that competes on what we consider to be the high end.
However, if you would like you could send me on loan another dac to compare with. There are some that many believe to be quite good, that measure well, and that may not be at the level of Katana (at least when suitably powered). That might provide some perspective for people who don't have a near perfect dac (a good dac) as point of reference to compare with.
The thing is the better the DAC gets the better good recordings can sound. The odd thing is that lesser DACs can also mask the deficiencies of bad recordings as well and actually make them more listenable. Does a better DAC always give better sound then? I am leaning towards the thinking that it depends on the recordings you have or prefer. Doesn't it? remember the trick is to find serendipity or let it come to you by trying. On that basis it is best to have a palette of DACs then?
Yes, I am in the process of building either a tube buffer as a loop or a totally separate tube preamp.
I changed the op amp in the AVCC supply to an AD817 and there is more improvement there. Previously it appears not to be stability issues but the rail was too low for the LME49710. With a little more heatsinking on the pass transistor I think I can bump up the input rail voltage to something the LME49710 likes and try that sometime. Meanwhile the AD817 is fully specd to operate with single rails as low as 5V so it is happy and sounding very refined. Higher bandwidth and performance op amps/ gain blocks are audible in the AVCC. The tonality does not change but the transparency and sense of space, immediacy and microdynamics are where it is heard. Of course a high resolution system is required, hence why Mark insists on anyone not already having a high res system to try and use an HPA and a set of phones.
I changed the op amp in the AVCC supply to an AD817 and there is more improvement there. Previously it appears not to be stability issues but the rail was too low for the LME49710. With a little more heatsinking on the pass transistor I think I can bump up the input rail voltage to something the LME49710 likes and try that sometime. Meanwhile the AD817 is fully specd to operate with single rails as low as 5V so it is happy and sounding very refined. Higher bandwidth and performance op amps/ gain blocks are audible in the AVCC. The tonality does not change but the transparency and sense of space, immediacy and microdynamics are where it is heard. Of course a high resolution system is required, hence why Mark insists on anyone not already having a high res system to try and use an HPA and a set of phones.
Last edited:
I agree. In another thread , I got a complaint about the fact that Katana is "too detailed" ...hmmm
""In that regard, I applaud Allo's introduction of Katana. I appreciate what you are trying to do, and I have a pretty good idea how very difficult it must be to do it in an RPi stack form factor.""
Yeap not easy . Still I am satisfied with how things turned .
""It also means that while I am very happy to see Katana come to the point it is now, I wish that more work had been done to figure out recommended power supplies, even if only a diy schematic. That is why I suggested a possible topology to try in the another thread.""
Will do so very soon .
""However, if you would like you could send me on loan another dac to compare with. There are some that many believe to be quite good, that measure well, and that may not be at the level of Katana (at least when suitably powered). That might provide some perspective for people who don't have a near perfect dac (a good dac) as point of reference to compare with.""
I will think on it , but pretty much any DAC that its normally powered by SMPS (unless its a uber engineered SMPS like the benchmark) will have a hard time against Katana powered by SMPSs
Did some more with Katana earlier today. It is equipped with Isolator and 'thd' output stage. Hooked it up to linear 5v and +-15v power supplies, which are in turn behind a very good power conditioner. It actually sounds very good this way. The power supply itself is not expensive. One transformer with multiple windings, and 7805 for the 5v, LT1086 for +15v, and LM337 for -15v. Sounds like jitter is very low. There is some slight graininess, a little distortion, to audio output. Sounds like maybe the output stage without the so-called 'Rasmussen' filter (actually the history on that idea goes back much further, at least to Hawksford, IIRC).
How does it compare to my modded dac? First, the differences are pretty small. I can hear them plainly, but today my son is having trouble. I'll let him try again tomorrow with fresh ears (he has been playing video games with headphones). Here is how it sounds to me: With no register tweaking of DPLL in the modded dac, Katana sounds better, primarily due to what sounds like much less jitter. If I adjust the modded dac for stability with DPLL at 2 out of 10, then set it back to 5 out of 10, I still slightly prefer Katana. If I adjust DPLL for stability at 1 out of 10, and then set it back to 4, it clearly sounds cleaner than Katana. Jitter performance sounds pretty close, the biggest difference that kind of is bigger than any jitter difference is some graininess of Katana vs extreme smoothness (or lack of any audible distortion) of the modded dac. Drums and instruments sound more real with the modded dac in that adjustment state. Its much more like drums are real and right there vs sound like very good reproduction of drums. To my ears in that state the modded dac is much more closing the distance on DAC-3.
However, my present methods for adjusting and maintaining adjustment of DPLL stability down at 1 out of 10 needs more work. I have more ferrites coming, and some other ideas to try going forward. For now I will assume that the technical issues can be solved satisfactorily for the mod builders.
Later, I want to give both dacs a little more of a chance to compete better. For the modded dac, I will try to work on improving DPLL tuning stability. For Katana, I will try adding some very low .053 ohm ESR, 10,000uf filter caps to the linear power supplies to see if I can get some more improvement out of Katana SQ that way. I will keep you informed as things progress.
How does it compare to my modded dac? First, the differences are pretty small. I can hear them plainly, but today my son is having trouble. I'll let him try again tomorrow with fresh ears (he has been playing video games with headphones). Here is how it sounds to me: With no register tweaking of DPLL in the modded dac, Katana sounds better, primarily due to what sounds like much less jitter. If I adjust the modded dac for stability with DPLL at 2 out of 10, then set it back to 5 out of 10, I still slightly prefer Katana. If I adjust DPLL for stability at 1 out of 10, and then set it back to 4, it clearly sounds cleaner than Katana. Jitter performance sounds pretty close, the biggest difference that kind of is bigger than any jitter difference is some graininess of Katana vs extreme smoothness (or lack of any audible distortion) of the modded dac. Drums and instruments sound more real with the modded dac in that adjustment state. Its much more like drums are real and right there vs sound like very good reproduction of drums. To my ears in that state the modded dac is much more closing the distance on DAC-3.
However, my present methods for adjusting and maintaining adjustment of DPLL stability down at 1 out of 10 needs more work. I have more ferrites coming, and some other ideas to try going forward. For now I will assume that the technical issues can be solved satisfactorily for the mod builders.
Later, I want to give both dacs a little more of a chance to compete better. For the modded dac, I will try to work on improving DPLL tuning stability. For Katana, I will try adding some very low .053 ohm ESR, 10,000uf filter caps to the linear power supplies to see if I can get some more improvement out of Katana SQ that way. I will keep you informed as things progress.
Last edited:
Hi Mark
thx you for the testing and am glad that both DACs are so close to each other. I have been watching this thread your efforts and approach have been ..groundbreaking . ( I dont say this lightly)
Your approach of feeding DSD direct to the ess9038q2m might show the direction for the new designs.
Back to your findings. Katana does not use any DPLL (its turned off=0) , our jitter is close to jitter of clock + buffer that are feed individually with PI filters and LDOs (at audiosciencereview they measure jitter in analog domain of Katana and its lower than any DAC on the market including Benchmark 3)
I suspect that what you hear, its still noise in the opamp rails . Yes 10.000 uF on the power will probably result in another jump in SQ (thats what Greg was saying in a different thread). I am wondering if more capacitance is added after the LDOs (where we have the 1000uF big cap)..how it will change the sound..
One last idea.
So far you have been trying to improve a poorly designed ess9039 board. Except for the the fact that Katana wont accept the DSD (and I know its a big thing) , I am wondering what type of sound you can get by tweaking and modifying our board. You dont have to worry about clocks , dplls , just clean power and output stage.
My 2 cents.
thx you for the testing and am glad that both DACs are so close to each other. I have been watching this thread your efforts and approach have been ..groundbreaking . ( I dont say this lightly)
Your approach of feeding DSD direct to the ess9038q2m might show the direction for the new designs.
Back to your findings. Katana does not use any DPLL (its turned off=0) , our jitter is close to jitter of clock + buffer that are feed individually with PI filters and LDOs (at audiosciencereview they measure jitter in analog domain of Katana and its lower than any DAC on the market including Benchmark 3)
I suspect that what you hear, its still noise in the opamp rails . Yes 10.000 uF on the power will probably result in another jump in SQ (thats what Greg was saying in a different thread). I am wondering if more capacitance is added after the LDOs (where we have the 1000uF big cap)..how it will change the sound..
One last idea.
So far you have been trying to improve a poorly designed ess9039 board. Except for the the fact that Katana wont accept the DSD (and I know its a big thing) , I am wondering what type of sound you can get by tweaking and modifying our board. You dont have to worry about clocks , dplls , just clean power and output stage.
My 2 cents.
Last edited:
I wonder about this too. In fact, I mentioned the possibility in a PM to Greg. Maybe we can talk about it a little more offline.
Ah, ha! That explains why reverb tails are so easily audible with Katana. Same thing I noticed with the modded dac when tuned for best DPLL stability.
Some posts back before this, I recall mentioning that I thought the modded dac seemed maybe even slightly better than DAC-3 in that one regard (reverb tail audibility), but otherwise still sounds 'less sophisticated.'
The audiosciencereview measurement of Katana might be some evidence to help confirm what I was hearing in the reverb tails was associated with very low jitter and not some other aspect of dac performance.
It is all very interesting, only one problem. Right now keeping DPLL tuning stability in the modded dac where I want it is not good enough for most practical purposes. It has to be tweaked a tiny bit too frequently to be more than a laboratory curiosity. Don't know what it would take to make it really solid and reliable. Probably have to try various things. Could require a lower frequency dac clock (I have a 49MHz one I bought from Allo in stock here somewhere!). It might need to be shielded, more power supply regulation, etc., who knows what. However, the benefits of solving the puzzle would be well worth some effort being spent on it.
Also, I still don't really understand the clock quality for what came installed in the AK4137 board. Trying other supposedly low jitter clocks made it sound worse. I kind of assume their jitter while presumably smaller may be more deterministic (or maybe not really smaller?), whereas the clock that came with the AK4137 board may have more random type of jitter. Don't know, just speculating. Sure would like to measure it, but I have to think its got to be smaller than the jitter meter on loan here from Richard Marsh. That one only goes down to a pS (I think) and we need to go 10 or 100 times lower, and do so in a way that lets us look at offset frequency and jitter composition. If the AK4137 was directly driving a dac, then maybe J-test would show something, but a Sabre dac uses a separate clock that would mostly determine J-test results. Also, the Arta J-test only works for at one frequency (if I understand correctly), and more detailed info might be useful.
Oh well, we have gotten this far without much test equipment. Just have to do a little more that way I guess. We'll see.
Last edited:
It turns out sound quality with Q2M can be very similar to PRO. In terms of distortion, ESS says Q2M can do -120dB and PRO can do -122dB. PRO has eight channels, but Q2M has enough for stereo. PRO chips cost more, and take more and higher current power supplies, and also need higher current output stages. The benefit of those last few things is potentially lower noise. However, Q2M noise is down at -122dB, probably low enough for many purposes. So, cost and complexity go up a quite a bit with PRO, with not necessarily all that much better sound quality. It seems like one of goals with Katana has been to provide excellent sound quality at very reasonable cost. Q2M was probably the right choice in that case, IMHO.
Another thing to maybe keep in mind is that designing a good dac with either Q2M or PRO is not trivial. There are a lot of rather compromised designs out there. Opting to use a PRO chip in a dac that is to sell for $250 almost certainly means that a lot of corners will have to be cut elsewhere in the design. Overall sound quality would most likely suffer vs going with a well-optimized Q2M design, again, IMHO.
Thank you Mark, agreed, that makes sence.
I don't like to compromise, so I will be plowing on modding the PRO, for me it's a bit of fun while I learn about the DAC... and I don't care if it ends up consuming more power than my class a amps Maybe one day I'll end up with a AudioGD D77 
?志?响
Hi Mikett,
Don't know if you have seen...
ESS Technology :: Voltage Regulators
I don't like to compromise, so I will be plowing on modding the PRO, for me it's a bit of fun while I learn about the DAC... and I don't care if it ends up consuming more power than my class a amps
Maybe one day I'll end up with a AudioGD D77 ?志?响
Hi Mikett,
Don't know if you have seen...
ESS Technology :: Voltage Regulators
Thank you Mark, agreed, that makes sence.
I don't like to compromise, so I will be plowing on modding the PRO, for me it's a bit of fun while I learn about the DAC... and I don't care if it ends up consuming more power than my class a amps
?志?响
Hi Mikett,
Don't know if you have seen...
ESS Technology :: Voltage Regulators
About 9 months ago actually. I think I was the first top point this thing out.