They feed opamp based series regulators without a pass device. As much as i share your opinion about cascading regulators there is not much leeway in this design and the input voltage needs to be held quite close to 5v.
With ES9038 and Ak4493 I never noticed that much from digital supply tweaks, and these were direct power inputs not like dam with further onboard supply distribution and processing.
always best to try yourself in your own application to be sure.
But recently Im thinking these LDOs are best avoided altogether,
batteries seem very promising, A123 LiFePo cells conveniently have 3.3V needs covered but for 5V not much you can do.
got more info ''oversized buffer'' you mention?
IME cascading TPS7A or LT304X regs has lead to an overall improvement in some cases,Are these (-)5V rails feeding any (key) part directly on the DAC?
Or do they feed another regulation stage?
Why I'm asking? On several other projects best results were achieved by
avoiding daisy chained regulation stages.
The currently best solution seems to be on oversize buffer (supercap - several Farads),
which is considered even better than a LiFePO4. I'm fully in line with this approach.
I currently run an Allos Shanti with >3F on the output feeding my current DAC and RPi .
I used to run iFi iPower with 47k Rifa on the rail (which wasn't worse then the Shanti IMO).
( I guess that answers Ricks earlier Q)
But again. I'd understand the use of Salas and 3045regs etc. if there wouldn't be any further regulation onboard. And that I don't know.
always best to try yourself in your own application to be sure.
But recently Im thinking these LDOs are best avoided altogether,
batteries seem very promising, A123 LiFePo cells conveniently have 3.3V needs covered but for 5V not much you can do.
got more info ''oversized buffer'' you mention?
IanCanada, and some of his followers are using supercaps.
Allo Shanti is using supercaps.
Supercaps are then the very final stage in your primary PS.
Watch out though! Some manufacturers call their supplies "super/ultracap-supplies" and
you'll see that they still feed a regulator as the final stage in the PS.
That's not what I'm talking about.
Since the capacities are huge these supercaps act like a unregulated, batterylike, extremely stable PSes.
Without the drawbacks that come with batteries.
Usually these supercap setups come with very low ESR. With usually 2.7V per
cap, going for higher voltages then 5.4V is not that great because you increase the ESR, complexity and price of course.
A big supercap buffer with mulitple cells needs to get balanced if you want to do it right. This can make it a bit more tricky/complex.
Also a bit tricky is the turn on/off handling and charging.
The initial charging takes a while and the discharging needs to be controlled properly.
Turn on/off thumps need to be avoided. These caps could be used for welding!
For the turn on/off handling I just recently introduced relays to the Shanti DC rails, which of course add their own impedance.
I didn't like that the Shanti supercap output stages were discharching over the load. Some devices might not like to be driven
into a greyzone.
And then you simply couldn't turn the device off by using the power switch. It could take minutes to discharge the caps.
There's no solution without compromise.
The charging current should be much higher then the load. It helps to look at your
cap buffer charging/discharging curve for proper dimensioning.
Bottom line. The Shanti with my remote controlled DIY relay stage on the output pretty much reflects my current PS preference.
I have to check if I can simply turn one rail around on that PS to get the -5V without facing issues.
Last edited:
Ian's ultra cap supply .
So the supply is like regulated supply on the surface, but really ultra cap being charged by regulator.
Cool stuff, much more convenient than batteries.
Is required maintenance what you meant by drawback of batteries or something else?
So the supply is like regulated supply on the surface, but really ultra cap being charged by regulator.
Cool stuff, much more convenient than batteries.
Is required maintenance what you meant by drawback of batteries or something else?
I got another subject.
Finding a matching amplifier.
...matching the quality level of the DAC and matching the overall gain structure.
The differential DAC delivers 2.6V. That's much too high for the typical chain setups
I'd say.
Typical setup? +/- ampgain 20+, speaker 4R and speaker sensivity 89dB/W +
Such a setup would require excessive digital VC. A NoGo.
Even at 10W output I'd just need an amp gain of around 3dB max on my setup.
What are your matching amps? Have you considered above?
Sorry for bugging you with all this. Before investing in such a device I need to make sure
that all to my critical aspects are covered.
Thx again.
Finding a matching amplifier.
...matching the quality level of the DAC and matching the overall gain structure.
The differential DAC delivers 2.6V. That's much too high for the typical chain setups
I'd say.
Typical setup? +/- ampgain 20+, speaker 4R and speaker sensivity 89dB/W +
Such a setup would require excessive digital VC. A NoGo.
Even at 10W output I'd just need an amp gain of around 3dB max on my setup.
What are your matching amps? Have you considered above?
Sorry for bugging you with all this. Before investing in such a device I need to make sure
that all to my critical aspects are covered.
Thx again.
Soundcheck, just a short remark:
In the DAM the +/-5v gets regulated down to +/-4v by the opamps in order to feed the shift registers, but the exact voltage is not particularly critical. Using Lipo batteries some users have had good results at 3.3 - 3.6v and there is no particular reason this voltage cannot be taken down to 2.7v. Obviously the 0db level will then be reduced and the required additional amplification will affect to a degree the S/N.
Should you decide to take this path, the opamp regulators on the boards can be removed/bypassed and external power applied directly to the shift registers.
In the DAM the +/-5v gets regulated down to +/-4v by the opamps in order to feed the shift registers, but the exact voltage is not particularly critical. Using Lipo batteries some users have had good results at 3.3 - 3.6v and there is no particular reason this voltage cannot be taken down to 2.7v. Obviously the 0db level will then be reduced and the required additional amplification will affect to a degree the S/N.
Should you decide to take this path, the opamp regulators on the boards can be removed/bypassed and external power applied directly to the shift registers.
Finally got my 1941 working.
I am using absurd brute force supplies for the analogue - parts I have laying about and would not expect anyone to follow. I dislike the idea of cascaded regulators. 12 volts 5 amps (per phase) power transformer - 5 henrys/26R input choke - cap - 320 mH/7.8 R choke - cap - 1 R resistor - cap and then a larger cap in place of the supplied caps and power is applied there - not at the power input header. To get 5 volts. I use a bleeder resistor to set voltages. The negative supply seems to want more current than the positive. It requires a larger value bleeder to get the same voltage. 300 R on the negative and 400 R on the positive. I think of bleeders as brute force shunt regulation. About 5.1 volts steady as can be and my FLUKE meter cannot measure the AC on the line. Just toggles between zero and .01 mV.
IFI 5 volts smps for the digital supply at the header (the header has been removed but I kept the chokes on this board this time) with an electrolytic's leads coming through the header's holes and the leads attached to that. I did not use 47mF like SOUNCHECK, only 15 mF but from the flash I get when I plug in the IFI I doubt I will try much more. Maybe it would flash without a cap? I did not try that.
I removed the USB entry and have hard wired the USB connector which is very short - about five inches. Using a SALAS BiB for USB power for the XMOS and the SD card reader.
I used to be the most anti-USB fellow in the world and figure I have to admit I am wrong. After using USB for both input and output and defeating SATA on the motherboard I am hearing excellent sound quality. I have to wonder if running two data transfer systems at once is a problem or simply that only using one is better? I had tried to use the SATA SD card readers - they work and they sound good but the being IDE instead of AHCI made for tremendous delays between tracks and forever at the beginning of a play list. The USB only sounds better to me, anyway.
I believe in slow low power computers. I am using an ASROCK 34558 board - minimum CPU speed 800 mHz. Everything is off except for USB and HPET. Graphics memory set to the lowest number. That is as far as the BIOS will let me tinker. Usually ASROCK BIOS allows memory timing changes and whole lot of other things. Less to worry with Set power mode to ECO and get the lowest amount of digital noise I have ever heard. If you want an education in digital noise set this to either normal or sport and it returns. Some people may like that sound ... The thing seems every bit as (digitally) quiet as I remember the SDTrans being - of course that is worth less than hearsay but i do think ti is true.
Player is WTF with frames set to 128 and periods to 7 with CPU usage set to 50%. I think this is an extraordinary player. Only BIG problem is that it cannot play gapless. It is the most disappointing part of the player. I have got used to it but would enjoy this being corrected.
Letting it settle in while looping files all night last night and during the working day.
Initial listening was not astonishing but then we all want nirvana and that is not possible. Certainly nothing to complain about. It does have noticeably powerful and tuneful bass.
The thing has TREMENDOUS output which leads to the question I have.
Would using the level control to lose 18 dB do harm to the sound? At the moment I have my xilica DSP boxes set to -18 dB and that doesn't seem like a good idea.
Any one using the internal attenuator for that much attenuation?
I am using absurd brute force supplies for the analogue - parts I have laying about and would not expect anyone to follow. I dislike the idea of cascaded regulators. 12 volts 5 amps (per phase) power transformer - 5 henrys/26R input choke - cap - 320 mH/7.8 R choke - cap - 1 R resistor - cap and then a larger cap in place of the supplied caps and power is applied there - not at the power input header. To get 5 volts. I use a bleeder resistor to set voltages. The negative supply seems to want more current than the positive. It requires a larger value bleeder to get the same voltage. 300 R on the negative and 400 R on the positive. I think of bleeders as brute force shunt regulation. About 5.1 volts steady as can be and my FLUKE meter cannot measure the AC on the line. Just toggles between zero and .01 mV.
IFI 5 volts smps for the digital supply at the header (the header has been removed but I kept the chokes on this board this time) with an electrolytic's leads coming through the header's holes and the leads attached to that. I did not use 47mF like SOUNCHECK, only 15 mF but from the flash I get when I plug in the IFI I doubt I will try much more. Maybe it would flash without a cap? I did not try that.
I removed the USB entry and have hard wired the USB connector which is very short - about five inches. Using a SALAS BiB for USB power for the XMOS and the SD card reader.
I used to be the most anti-USB fellow in the world and figure I have to admit I am wrong. After using USB for both input and output and defeating SATA on the motherboard I am hearing excellent sound quality. I have to wonder if running two data transfer systems at once is a problem or simply that only using one is better? I had tried to use the SATA SD card readers - they work and they sound good but the being IDE instead of AHCI made for tremendous delays between tracks and forever at the beginning of a play list. The USB only sounds better to me, anyway.
I believe in slow low power computers. I am using an ASROCK 34558 board - minimum CPU speed 800 mHz. Everything is off except for USB and HPET. Graphics memory set to the lowest number. That is as far as the BIOS will let me tinker. Usually ASROCK BIOS allows memory timing changes and whole lot of other things. Less to worry with Set power mode to ECO and get the lowest amount of digital noise I have ever heard. If you want an education in digital noise set this to either normal or sport and it returns. Some people may like that sound ... The thing seems every bit as (digitally) quiet as I remember the SDTrans being - of course that is worth less than hearsay but i do think ti is true.
Player is WTF with frames set to 128 and periods to 7 with CPU usage set to 50%. I think this is an extraordinary player. Only BIG problem is that it cannot play gapless. It is the most disappointing part of the player. I have got used to it but would enjoy this being corrected.
Letting it settle in while looping files all night last night and during the working day.
Initial listening was not astonishing but then we all want nirvana and that is not possible. Certainly nothing to complain about. It does have noticeably powerful and tuneful bass.
The thing has TREMENDOUS output which leads to the question I have.
Would using the level control to lose 18 dB do harm to the sound? At the moment I have my xilica DSP boxes set to -18 dB and that doesn't seem like a good idea.
Any one using the internal attenuator for that much attenuation?
Although the digital volume control is very good, you do sounds like you want to do it in hardware.... the outputs are a 625 ohms pure resistive, so the best way to reduce level is just a shunt resistor from output to GND, that way you also get lower output impedance.
"-18dB"
That's exactly what I'm talking about. Most amps run a gain of 20 and more.
To me the first priority therefore is to get the amp gain down.
What's missing in the discussion is that there are different recording levels
that'll add at least another -12dB to the bill.
So, we're actually talking roughly about -30dB (in above example) for certain CDs as a base-level to achieve the ""maximum"" listening level.
IMO the digital VC control - as good as it might be - should just be used to cover the differences that come with the CDs and the actual VC.
BUT. The base 0dB leveling (e.g. that -18dB on rickmcinnis system) should be done with priorities
1. the amp gain
2. the DAC output level
Dragging the DAC output voltage excessively down, either digital or analog, just to get down to your maximum listening level will never be a good idea.
However. Having a variable voltage for the ladder to cope with 2. I'd consider a brilliant approach. (analog_sas' remark goes into that direction - thx ).
That's why I think that if going for such a DAC, do not forget to think about a matching amp (se/diff/gain/impedance/coupling). That's what I currently try to figure out.
That's exactly what I'm talking about. Most amps run a gain of 20 and more.
To me the first priority therefore is to get the amp gain down.
What's missing in the discussion is that there are different recording levels
that'll add at least another -12dB to the bill.
So, we're actually talking roughly about -30dB (in above example) for certain CDs as a base-level to achieve the ""maximum"" listening level.
IMO the digital VC control - as good as it might be - should just be used to cover the differences that come with the CDs and the actual VC.
BUT. The base 0dB leveling (e.g. that -18dB on rickmcinnis system) should be done with priorities
1. the amp gain
2. the DAC output level
Dragging the DAC output voltage excessively down, either digital or analog, just to get down to your maximum listening level will never be a good idea.
However. Having a variable voltage for the ladder to cope with 2. I'd consider a brilliant approach. (analog_sas' remark goes into that direction - thx ).
That's why I think that if going for such a DAC, do not forget to think about a matching amp (se/diff/gain/impedance/coupling). That's what I currently try to figure out.
Last edited:
Doing it analogue in dac or amp doesn't really matter, does it? I suggest you do a passive attenuator after the dac with resistors as Sören suggests. So you can play your weakest recording and loud as you want with the DAM volume at 0. Then you still have 10 dB of digital gain in store if needed for a party The DAM volume is really transparent - your most certainly have other things to worrie about in your rig.
The DAM volume is really transparent - your most certainly have other things to worrie about in your rig.It does. Big time.
On the amp you'd change the gain factor. Properly dialed in, you could go 2.6V (Udiffmax) in without any issues.
What you talk about is adding a pot/voltage divider. That's attenuation. All your digital attenuated signal will on top get attenuated in the analog domain to meet
that much too high gain level of the amp (I'm pretty sure most amps (or better the owners) out there have a rather poor DAC/amp/speaker gain match in place. It's not just me!)
There are amps out there which allow to change the gain by simply swapping a single resistor btw.
I think a passive resistor attenuator is much less intrusive than twiddling with the feedback or op-amp/transistor juggling in an arbitrary amp. Sorry - I don't agree with you as it is not my experience. But you often claim outrageous differences that is not really the common understanding for a particular "problem" at hand. Good luck anyways!
//
//
Thanks for the good advice on attenuation.
I did find last night that attenuation in the DAC is better than attenuation in the DSP box - as far as sound. Now to find four matched resistors in my collection.
SOUNDCHECK - it might be that all we need is a no gain follower amp like Rothacher's MOFO and designs like that? Lots of ideas for this at TUBECAD/John Broskie and at the PASS LABS forum here. Rothacher's amp is particularly intriguing for its simplicity.
Would be great to dispense with the input stage and why not if it is not needed?
I have removed the output filter caps and wonder if that is going too far?
Going to try the PS tin foils that Michael Percy stocks - he has 1000 and 1500 pF. I think I will go with 1500.
Started off with TNTs noDC Linear filter. (That is the one you are most proud of, isn't it?) I am going to become familiar with that before trying others.
Thanks, again.
PS Looks like I will need a 100R resistor to get the gain down to where I need it.
I did find last night that attenuation in the DAC is better than attenuation in the DSP box - as far as sound. Now to find four matched resistors in my collection.
SOUNDCHECK - it might be that all we need is a no gain follower amp like Rothacher's MOFO and designs like that? Lots of ideas for this at TUBECAD/John Broskie and at the PASS LABS forum here. Rothacher's amp is particularly intriguing for its simplicity.
Would be great to dispense with the input stage and why not if it is not needed?
I have removed the output filter caps and wonder if that is going too far?
Going to try the PS tin foils that Michael Percy stocks - he has 1000 and 1500 pF. I think I will go with 1500.
Started off with TNTs noDC Linear filter. (That is the one you are most proud of, isn't it?) I am going to become familiar with that before trying others.
Thanks, again.
PS Looks like I will need a 100R resistor to get the gain down to where I need it.
Last edited:
There's nothing intrusive about changing the gain of an opamp based amp input stage,
if the amp allows it.
At aforementioned -18dB, from the 2.6V 0.33V will be left. And that's usually only to neutralize a stupid gain choice on the amp. That's wasted.
You could also take a NC400 as example. It comes with 26dB gain. In my case I'd need 6dB max. That'd make 20dB wasted.
At -30dB, IMO not a very atypcial attenuation to achieve the maximum listening level on
this or that system, we'd be down to 0.08V. Even more gets wasted.
Please. Don't tell me that changing the gain of the amp is more intrusive then
dragging the DAC output voltage down into the noise.
The problem is. Soekris and any other DAC manufacturer are pretty much tight to an
output voltage in the area of 2V. They simply can't offer more then attenuation or low output levels
for their products.
Don't ask me why this IMO stupid mismatch of amp gain vs. DAC output voltage
persists.
Last edited:
I would not be to worried about -18 dB on a Soekris DAC. It is just three bits, you would still have 24 left. Though I agree that decreasing amp gain is a better idea than reducing the DAC level, mostly from an SNR perspective. I started with a fixed resistor divider on a DAM1021, but later removed it and reduced the NC400 amp gain. This improved the sound quite a bit. I did not try a single resistor on the output though, though I would expect similar results.
Fedde
Fedde
The last lowpass filter is those 1500 pf and 625R output impedance, giving F3 at 170 Khz. If changing output impedance then the capacitor should match....
The reason us DAC manufacturers like the 2V output level is to get better S/N ratio and to be able to driver power amps directly, especially with a good digital volume control.... I believe that a overly sensitive power amp is a bad design, I like the sensitivity of pro amps....
I will attenuate at the amps.
Installed 100 uF BG Ns (that I got from Peter years ago) in place of the polymers and there is definitely a difference. Sure I expected one but ... The only ones I have so it does look a little clumsy. Had to use smaller wire between the pcb and the caps but kept them very short.
Now I am curious about trying the replacing of the ceramics which I was certain I would not do. What I heard with the above compels me to give it a try.
With all of the posts about this there was never a definitive what to do.
I can see from Peter Daniel's last picture that they must all be in parallel since I see only one leaded capacitor as the replacement. Is that a polarized BG? That is what it looks like. What value are those who have done this using?
analog_sa or Peter would make you it plain what you have done. Please.
Take care,
Installed 100 uF BG Ns (that I got from Peter years ago) in place of the polymers and there is definitely a difference. Sure I expected one but ... The only ones I have so it does look a little clumsy. Had to use smaller wire between the pcb and the caps but kept them very short.
Now I am curious about trying the replacing of the ceramics which I was certain I would not do. What I heard with the above compels me to give it a try.
With all of the posts about this there was never a definitive what to do.
I can see from Peter Daniel's last picture that they must all be in parallel since I see only one leaded capacitor as the replacement. Is that a polarized BG? That is what it looks like. What value are those who have done this using?
analog_sa or Peter would make you it plain what you have done. Please.
Take care,
Check this post: https://www.diyaudio.com/forums/ven...24-bit-384-khz-dac-module-24.html#post5768714
and this one: https://www.diyaudio.com/forums/ven...24-bit-384-khz-dac-module-30.html#post5789914
The batches of 4 ceramic caps are all in parallel, I replaced them with a single BG NX 0.1uF
and this one: https://www.diyaudio.com/forums/ven...24-bit-384-khz-dac-module-30.html#post5789914
The batches of 4 ceramic caps are all in parallel, I replaced them with a single BG NX 0.1uF
Last edited: