Not really. The dac chips in these boards have a very low output impedance. What it means is that the dac outputs produce the least distortion if they run into a virtual short to about +1.65v. The lowest distortion is achieved using opamp voltage to current converters offset to about +1.65v. After that, differential summing with another opamp is needed to remove common mode distortion and noise. That is just for the output stage, they also really benefit from power supply and clock improvements (some of which are pretty essential for any SQ improvement). In addition, further improvement in sound quality may be obtained by feeding them I2S from an AK4137 upsampler and DSD converter.
At the end of all that, if a little transformer coloration of the sound were desired, transformers could be connected to the outputs of the opamps.
The sort of traditional resistor and transformer dac output stage approach works much better with high output impedance dacs which don't mind a little more resistance added in series to ground. Essentially, trying to do that with one of these dacs would put it in voltage mode, or at least take it out of pure current mode. You would get sound out, but not very low distortion.
In addition, modding one of the dac boards fully for best sound quality is quite a bit of work. It would probably be a long term project for someone who really wants to do something like that. Not sure where your interest is given all the foregoing. If you are interested in best sound quality, we can give you lots of information about what is involved.
Last edited:
Most likely you could achieve current mode operation with transformers but I very much doubt with off-the-shelf transformers. Your typical line level transformer is designed to operate in voltage mode - presenting a high shunt inductance and consequently a relatively high resistance. You'd need to have a custom trafo wound which prioritized low resistance of windings and with a lower shunt inductance.
Adding even one or two ohms of resistance to the outputs measurably increases distortion. People have tried it. That makes the 'optimum' resistor size zero ohms. Shorting outputs to ground rather than a virtual 1.65v offset would also increase distortion. I can't see any sensible way to use transformers, although someone could use them just to show its possible to get somewhat distorted sound out at all.
If someone just likes the sound of transformers, or of tubes, then build a proper low distortion dac, and connect transformers and or tubes at the output. That would give the best sound quality.
If someone just likes the sound of transformers, or of tubes, then build a proper low distortion dac, and connect transformers and or tubes at the output. That would give the best sound quality.
Last edited:
Many other DACs (in the main older, multibit types) have current source outputs. Thus with them, current mode is a given, there's no other way. ESS (and more recently AKM) are the only ones using a resistor array with rather too low an output impedance for true current mode operation (current sources typically come with >100kohm Zouts). Having such a low output impedance is what precludes a universal I/V stage from working optimally.
Oops! Sorry, it was too early in the morning here. Should have said current to voltage (IV or I/V) rather than the wrong way around. Don't want to get anyone confused about how these things are supposed to work.
Speaking of confusion, as Abraxalito was explaining, current mode used in the context of dac outputs used to mean something different before ESS came along and defined voltage and current modes for their particular dacs. Unfortunately, they don't do much to help make it clear to people what they mean. They post a few old documents on the downloads page of their website and provide pretty sparse data sheets only under NDA. Far from the old days where manufacturers went to quite a bit of effort to make it easy for designers to use products by providing lots of helpful information. Maybe that was kind of a golden era that won't happen again.
Last edited:
Probably not transformers. The cheap dac boards come with a voltage mode output that works, it just doesn't sound very good to some people. Funny though, there are other people who think its pretty good. Can only guess they haven't heard better.
Don't know that trying to use transformers is worth the bother. As I keep trying to get people around diyaudio to see, there is more to making one of these dacs sound good than just changing the output stage and or adding more filter caps to AVCC. Almost everything with the cheap dacs is bad. The dac chip is good, the pcb with ground plane is good enough for modding, and the connectors are okay. Everything else is cheap junk components and mostly over simplified circuitry. Even the dac chip could have better PCM interpolation filters, and more digital overhead to handle intersample overs.
That's why making one of these dacs sound its best is so much work, and its why it is expensive to make really good dacs. In addition, not all dac designers are created equal, and many of better ones are constrained to designing to a price point in the market. In a given price range, buyers will usually go for more bells and whistles rather than more sound quality. Eye candy cases with cool looking displays and remote controls, oh boy! Lots of problems for people who just want really good sound quality at lowest possible cost. That is all IMHO, by the way.
Last edited:
TP sells output stages designed for use with higher output current ESS PRO dacs, although some could probably be modded. Our output stage is of the very accurate type, they have some that are and some that aren't. But maybe the bigger issue is that TP system costs can add up quickly. Some people may be here because of budget constraints, maybe some for other reasons.
Regarding the forthcoming transformer output, we will have to see. Some items over there may be pretty good, don't know about everything. It concerns me when I see how commercial dac makers like Benchmark Media (or Chord for that matter) can create great dacs without batteries, FIFOs or transformers and other people think they are must have items. Maybe just more than one way to skin a cat, maybe more to it than just that. Don't know.
Last edited:
I was just giving some examples of available modules, I'm not saying they will be better or even work without modification but, for someone with limited or beginner level skills (like myself) and no spare parts bin, they could be another option to try and compare.
Russ does claim the mercury should work great with other DACs as well
As for the rest, I cannot design a full DAC myself. I do have access to various parts of the system but if I cannot put the xmos chip next to the clock next to the DAC and sync the whole system for minimum jitter from those parts like the DAC-3 then I have to compensate with things like a Fifo with it's local clocks.
Sorry for the off-topic.
Russ does claim the mercury should work great with other DACs as well
As for the rest, I cannot design a full DAC myself. I do have access to various parts of the system but if I cannot put the xmos chip next to the clock next to the DAC and sync the whole system for minimum jitter from those parts like the DAC-3 then I have to compensate with things like a Fifo with it's local clocks.
Sorry for the off-topic.
Hi Mark,
Did you succeed to get good (improved?) sound using that board? Mine (768K version) arrived a few days ago. I hooked it up between Pi with Hifiberry Digi+ Pro hat and ES9038 DAC, but sound seems to get worse rather than better. Some details:
- using I2S connection (GND, Data, BCK, LRCK) from Digi+ to AK4137 board I2S input header
- from I2S Out header to ES9038 board again GND, Data, BCK, LRCK
- shorted the P/D jumper to force I2S input mode
- output modes DSD64/128/256 give sound, but it is worse than when directly connecting the Digi+. Mainly the sound stage is different and much narrower.
- output mode I2S gave highly distorted sound at first. Thinking this was due to interference I shuffled the boards around, but now these is no sound at all in this mode.
Didn't have time to measure the outputs yet. Perhaps anyone else experienced/solved this?
Hi maartn,
What are you using for input audio, 16/44? Something else?
When you say it is worse than directly connecting the Digi+, do you mean you are using SPDIF into ES9038 dac in that case?
What power supply are you using for AK4137?
Do you have one or more in-focus, close-up pictures showing the wiring interconnections that you could attach to a post here so we can see how it looks?
What software do you use on RPi to send I2S? Do you know if it is actually I2S justified, or maybe LJ or RJ? Do any of those input settings sound better on AK4137?
Haven't heard that particular complaint before. I might be inclined to try removing Digi + and just feeding AK4137 directly from RPi, maybe with a 16/44 CD rip source file. Something like that which would be a very basic test.
In my own case with the AK4137 board, I get excellent sound, but the input I2S is coming from an Amanero clone board. Some pictures of mine here: ES9038Q2M Board - Page 204 - diyAudio
What are you using for input audio, 16/44? Something else?
When you say it is worse than directly connecting the Digi+, do you mean you are using SPDIF into ES9038 dac in that case?
What power supply are you using for AK4137?
Do you have one or more in-focus, close-up pictures showing the wiring interconnections that you could attach to a post here so we can see how it looks?
What software do you use on RPi to send I2S? Do you know if it is actually I2S justified, or maybe LJ or RJ? Do any of those input settings sound better on AK4137?
Haven't heard that particular complaint before. I might be inclined to try removing Digi + and just feeding AK4137 directly from RPi, maybe with a 16/44 CD rip source file. Something like that which would be a very basic test.
In my own case with the AK4137 board, I get excellent sound, but the input I2S is coming from an Amanero clone board. Some pictures of mine here: ES9038Q2M Board - Page 204 - diyAudio
Last edited:
Master Mode PCM Works!
Just tried master mode PCM and it sounds great, very much like Katana in terms of low jitter. Cymbal details are smooth and lush. Reverb tails are clearly audible. Definitely a jump up in sound quality performance! Based on ASR measurement of Katana jitter being the lowest they ever found, even a little better than DAC-3, could be the modded dac may also have slightly lower jitter than DAC-3. It kind of sounds like that could be the case, too. It sounds really good! With jitter so low, all the other mods on the dac with output stage, AVCC, clocking, power supplies, etc., really let it shine.
Just to fill in a little of the ongoing story, master mode didn't work for me initially for some reason even though clocks where there and AK4137 was producing DATA. I must have had some register setting wrong that caused Q2M to mute. I switched control back to the MCU momentarily and it fixed whatever was the problem. AK4137 is still operating in PCM slave mode, and Q2M is generating clocks with BCLK at MCLK/16 = 6.25MHz.
Just tried master mode PCM and it sounds great, very much like Katana in terms of low jitter. Cymbal details are smooth and lush. Reverb tails are clearly audible. Definitely a jump up in sound quality performance! Based on ASR measurement of Katana jitter being the lowest they ever found, even a little better than DAC-3, could be the modded dac may also have slightly lower jitter than DAC-3. It kind of sounds like that could be the case, too. It sounds really good! With jitter so low, all the other mods on the dac with output stage, AVCC, clocking, power supplies, etc., really let it shine.
Just to fill in a little of the ongoing story, master mode didn't work for me initially for some reason even though clocks where there and AK4137 was producing DATA. I must have had some register setting wrong that caused Q2M to mute. I switched control back to the MCU momentarily and it fixed whatever was the problem. AK4137 is still operating in PCM slave mode, and Q2M is generating clocks with BCLK at MCLK/16 = 6.25MHz.
Last edited:
Master mode also works for DSD with Q2M producing 12.5MHz clocks (a little bit faster than standard DSD256). The sample rate could be made more standard if using another option for selecting Q2M clock frequency. Right now I am using the option to divide MCLK by various powers of two, typically divide by 8 or by 16, to produce 12.5MHz or 6.25MHz respectively.
At the moment DPLL is still on. Setting it to zero causes the output to mute, but didn't try it for PCM mode yet. Now that there are ways for Q2M to be clock master there are various combinations of settings to be explored.
Interestingly, even though AK4137 is set to follow Q2M clocks, there is still a little jitter that DPLL catches (at least in DSD mode). Not sure how low-jitter Q2M MCLK divider is for producing master mode clock signals. If not an issue created in Q2M, then maybe AK4137 data output signal(s) are a little jittery with respect to the clock signal(s). Or maybe just some noise being picked up with the test system out on the bench rather than in a case. I will try to see if I can narrow down what the source of remaining jitter is, particularly the jitter I see in DSD mode.
At the moment DPLL is still on. Setting it to zero causes the output to mute, but didn't try it for PCM mode yet. Now that there are ways for Q2M to be clock master there are various combinations of settings to be explored.
Interestingly, even though AK4137 is set to follow Q2M clocks, there is still a little jitter that DPLL catches (at least in DSD mode). Not sure how low-jitter Q2M MCLK divider is for producing master mode clock signals. If not an issue created in Q2M, then maybe AK4137 data output signal(s) are a little jittery with respect to the clock signal(s). Or maybe just some noise being picked up with the test system out on the bench rather than in a case. I will try to see if I can narrow down what the source of remaining jitter is, particularly the jitter I see in DSD mode.
The best and simplest way to visualize the Sabre DAC's OP is in fact a
voltage source (zero ohms OP Z) with a series resistor following.
In most of their DAC's each OP is about 195 ohms, which I believe is 50k /
number of internal parallel resistors (usually 256).
The 9038Pro is the same but with more parallel resistors = more OP
current and lower OP Z.
T
Someone asked recently if OPA1612 could be used for AVCC. I said I thought it probably could, and I had been meaning to test it sometime to verify. Well, tried one and it seems to work very nicely for AVCC, much as I find they do for the output stage. They seem free of bit of brightness that LME49720 seem to have at least around here, possibly as a result of high-ish levels of ambient RF in the area (up on a hill with lots of RF sources for maybe 90 miles to the West having line of sight to the hilltop).
Wealas said Russ said Mercury output stage should work for other dacs. Russ is the designer for TP, so he would know. The Mercury Balanced I/V Stage If you used the 9018 gain setting on one that would probably be the closest fit for an ES9038Q2M in terms of output current.
Last edited: