Not sure what you mean by that?
Regarding a very good 20-year old regulator design, the superreg, no doubt it is likely better than many common regulators in use today. That doesn't mean it is the only way to get the best results, or good enough results, in today's world. A good point about it is that it is fully described in the public domain. A less good set of points is that it is complex, bulky, and more expensive than some potentially viable alternatives. Boards are not currently available either, a further downside meaning availability is unreliable for many people.
Some people still firmly believe in the old ways of designing power amps, maybe because it is the only good solution to the problem that they know how to do. Because they understand it reasonably well, they may believe it is the only way possible.
For me, I found a good power amp in Benchmark AHB2. It uses a switch mode power supply that supplies power near-instantly only as it is needed. It also has a class A output that eliminates cross over distortion. It measures great, sounds great, and is the most truthful, honest, and detailed power amp available. The only one that can fully reproduce everything that distinguishes the best dacs from one another. All, IMHO, of course.
Other people only believe in the old ways with huge banks of caps and huge heat sinks. Just like some only believe in the best 20-year old regulators. I don't want to assume that is the only way to go. I know it isn't the only way to go because compelling evidence exists to the contrary. However, it may be a very sure way to get extremely good results, and that makes it still very valuable to have available. Again, all IMHO.
Last edited:
If your board is like mine, then it is not so easy to do. The board has 4 layers and a single AVCC power plane is inside for both channels. One may try to separate the surface track by drilling out connection or lifting DAC pin, but I do not think its worth the effort. A single AVCC power is fine providing the reg can cope with noise caused by the load. In reality, AVCC still gets polluted by DAC and onboard Vref (AVCC/2) filters do not manage to clean it enough. As I understood that was one of the DAC3 know-how regarding Vref filtering. Maybe Mark can share the info? I simply separated Vref from AVCC and put it on a separate reg, what is not the exactly correct way, but the sound became better than ever.
Last edited:
I was thinking that if Vref is tapped off the AVCC, would that not mean that disturbances in AVCC will be repeated at the Vref and it would track and possibly null it out? I am new at this but I have to think this aspect through carefully.
A dip in AVCC will result in less current output at the DAC output pin. That dip would be carried across in Vref as well but it would be reduced by the smoothing cap at the voltage divider.
What would the result of this be at the transimpedance stage? Well the lower current will result in a dip in the output voltage leaving the transimpedance stage. However the voltage dip at the Vref what how does this shift the voltage from that aspect?
Is it theoretically better to have an independent Vref? Or one that tracks?
I don't know.
Interesting point raised there. Don't know to what extent it might be a problem, but I guess we could find out by using a high gain preamp to listen to Vref while the dac is playing at full volume without a power amp to make any sound. Instead we can listen to whatever signal is on Vref. Anybody want to try it? (EDIT: Actually, one could probably just disconnect the dac outputs from the I/V stages. Then the only input signal to the output stage would be Vref. Maybe try it with and without the dac outputs tied to ground or just left floating open. Neither one is exactly the same as virtual grounding to the Vref offset, but probably close enough to see if there any audible effects.)
If it is a problem, then filtering and or buffering are the only ways I can think of off hand we might use to fix it. We could buffer AVCC rather than Vref. Vref could be taken from something like LTC6655 with or without its own buffer. Offset voltage might be an issue, but so might resistive voltage divider tolerances for making Vref in the first place. We could keep any issues to a minimum.
All that being said, I do not recall seeing an issue raised about Vref quality. One easy thing to do would be what I recently suggested regarding questions of where to best put the Vref divider, on the AVCC board, or on the output stage board. I said better to leave it on the AVCC board, but add another filter on the output stage board, if desired. I would say the same for this. A two-pole filter will roll off faster than a one-pole filter, is what it amounts too. We can increase the time constants some too. At worst it would add some warm up time if we modestly increased the time constants. If we increase it too much then we might have to worry about secondary issues, but probably not if we keep it reasonable.
Last edited:
Ahh...but think about this. I would cut the trace leading to AVCC L&R. I would also cut the traces leading to Vref. That should isolate the connections from within the board. I would then connect directly to AVCC L&R right at the DAC chip and take the sense lines back to the Super Regs from there. If indeed it is superior to connect the Vref to the DAC AVCC, then I could use some copper foil and route a flat plane across from the DAC connection to Vref. Or I would simply use the 3.3V from my LT3042 to Vref if that is better. I guess I could try it both ways. The sense lines from the Super Reg allows longer than optimal runs from the power supply and is a big part of the reg.
Connecting at the dac chip would require lifting all the AVCC pins. There is no telling which pins are shorted together by traces running under the dac chip. Left and right might be shorted together in multiple places.
Take a look at the board and trace AVCC path. I hesitated to demolish the board at the extent required to separate the power for both channels.
It might be easier to buy an unpopulated ES9028Q2M board (with dac chip) from diyinhk for $80 or so, and patch it into your output stage on the existing dac board, and to as many of the power supplies as you want to still use. 
https://www.diyinhk.com/shop/audio-....html#/pcb_option-pcb_with_es9028pro_dac_chip
AVCC is separated and there is a place to put whatever new clock you want. You could patch in existing I2C firmware control lines to the new board.
https://www.diyinhk.com/shop/audio-....html#/pcb_option-pcb_with_es9028pro_dac_chip
AVCC is separated and there is a place to put whatever new clock you want. You could patch in existing I2C firmware control lines to the new board.
Last edited:
I'm thinking of cutting the traces that connect to the AVCCs right by at the pins. That way back to the reg nothing is connected. Same for the Vref lines at the transimpedance amps, I totally isolate the board traces.
But that goes against one of the criteria I had, that the board should always be restorable to stock in the event I move to something else.
Understood. But what I am trying to explain is that you might not be able to cut the traces if they run from the back of one pin pad, under the dac chip, to the back of another pin pad. If you wanted to cut that trace, first you would have to remove the dac chip from the board, or cut a hole through the dac chip to get out under the back side of it, or tunnel though the back side of the board underneath the dac chip. There is no rule saying traces have to approach pin pads from the outside where they could be cut, and sometimes they don't. Hope I was more clear this time.
Its best to stick with recommended AD825, HF decoupling upsets the AD797, - look at the output using ARTA
We need 3.3V. For 5V I could have used AD817 for which I have 6 of them.
At 3.3V I need to get some AD8031
Well I know the AD797 is very unstable on this board and hence why 20 years ago, i stayed away from it. The LME49710 is stable and gives just about the same performance as the AD797 as reported by others using the identical board.
the PRO and Q2M boards are very different, have different bottlenecks. i would range the mods according to their impact on sound:
1. 205R FB in I/V and high ohmic LPF
2. low impedance opamp power
3. low noise clock power
4. low jitter clock
5. low noise Vref
6. low noise and impedance AVCC
7. low noise digital power
you may have AVCC mod leading putting the regs directly on pins though. an unexpected thing was very prominent effect of the FB cap type. silver mica took over PP.
I saw Jan's recomendation, did you have a look at the datasheet for LTC6252 - attached in the post below.
Super Regulator
I still think your doing it the hard way - Moving from 9038Q2M to a 9028Pro board Use the LTC6655 if you don't like the NiCd idea.
Last edited:
Before putting stuff into the 9028pro. I now put it on a test rig to see what happens there first before I put it onto the ESS9038PRO rig. If there is a correlation then I know I am likely getting the effect of the mod. If not, then I have to soul search.
For the LDO, it feeds off the IV analog supply stage +-15v and the LDO feeds the 1794 DAC pretty much at the pins. A very nice layout. I am starting to get positive correlation between changing the op amps on the reg feeding this rig. You can hear each incremental change in going to a higher performance op amp so far the order in performance is NE5534, AD817 and LME40710. Each change affects the imaging and transparency and immediacy. It is now plainly obvious much more so than I ever expected. This is the same thing that ESS had outlined and it can be heard but it is simply feeding the LDO. As a comparison the Sulzer supply also used the 5534 but in a different topology. The super reg using the same op amp leaves the Sulzer in the dust. Not in the same league at all. So why is that? I attribute it to the topology that the op amp regulates the supply to itself. On the sulzer the power to the op amp is taking off the incoming line and it does not use its speed for regulating itself.
I now think from this that the performance of a reg is largely affected by the incoming voltage feed in normal regs. Why? the results between the Sulzer and Super Reg and the results of putting on a better reg in front of the LDO on the 1794.
One could then conjure up putting identical regs in series to get the same performance but in that case, the output impedance increases and this is documented when others put in a prereg in front of the super reg. And what contributes to the performance..the noise aspect or the low output impedance. I will draw my own conclusion when I put on the different power supplies on the 9028pro. The problem here is that I have to make sure my mind does not slip and I let the smoke out AGAIN.
Last edited: