|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
61.92.98.16
In Reply to: RE: Dacs powered by batteries - any successful story ? posted by beppe61 on June 11, 2015 at 01:12:01
Hi,
I helped design two of that kind. They sold (and sell) quite well.
Does that count as success?
Ciao T
At 20 bits, you are on the verge of dynamic range covering fly-farts-at-20-feet to untolerable pain. Really, what more could we need?
Follow Ups:
Hi and thanks for the interesting reply.
yes of course this is a successful story indeed.
I even wonder why batteries are not more popular.
I think that a unit could be made with a switch, for charge and listening modes.
In charge mode the unit is connected to the mains, can be used and charge the battery pack at the same time.
In listening mode the unit is disconnected from the mains and use the power stored in the battery pack.
I am working on a diy project to make something like this for a usb to spdif converter. My feeling is that the stock unit has a power supply of debatable quality. And also the my mains voltage is not clean.
The isolation from the grid that batteries can provide is very tempting.
Actually i am planning to use the type in the link.
I hope two in series would be enough.
This should give me at least complete isolation from the mains.
Thanks a lot for the valuable advice.
Kind regards,
bg
Edits: 06/13/15
Hi,> I think that a unit could be made with a switch, for
> charge and listening modes.The DAC's we make operate primarily on a USB connection and charge from USB. They can be selected to either draw power from the USB Port or not. The primary reason for this switch is not to draw down the battery of a smartphone in portable use.
Our latest upcoming USB powered DAC (no battery) has new power supply technology that moves it darn close to the battery powered performance, running directly of a generic USB Port.
And it does not use any regulators in the classic sense - we need the full 5V from USB for the DAC Chip and analogue stages, so we cannot regulate the power line down to something lower and use a regulator to reduce noise.
> I am working on a diy project to make something like this for
> a usb to spdif converter. My feeling is that the stock unit has
> a power supply of debatable quality. And also the my mains
> voltage is not clean.Batteries in themselves are not necessarily a good solution. Most batteries are noisier than a good regulator. And if the power supply inside equipment is inadequate, no matter how good the battery power, the internal PSU circuitry will compromise the results.
Power supplies are a complex subject. Batteries can help tremendously, but they need to be designed into a system inherently to deliver. Bodging then on externally mayor may not help, but for example the latest "plugtop" supply we have put into production is VERY quiet.
It is quieter than most active regulators out there, even very highly regarded ones like the latest TI Chip ones that I have seen in several recent USB power and related Gizmo's.
And this is done with a few crucial tricks to allow much lower switching noise as well as much lower audio band noise as existing designs (we are talking 0.000001V noise vs. the common 0.01 - 0.1V noise from common plugtop designs)
If we discount the "mental blockage" many audiophiles have regarding anything like this and the unjustified "worship of batteries über alles" such a simple to use plugtop supply may be as good (or better?) in most application than a battery.
Only one way to find it out, build your battery supply. I might get our people to send you one of the plug-top units when the supply situation has stabilised (we are still ramping up production and barely have enough to cover our current shipping products and a few dealer orders).
Ciao T
At 20 bits, you are on the verge of dynamic range covering fly-farts-at-20-feet to untolerable pain. Really, what more could we need?
Edits: 06/13/15
Hi and thanks a lot Mr Thorsten for the always very valuable and precious advice.
Lately i have become very interested in power supply noise.
I was even thinking to buy an instrument just to see this bloody noise out of ps.
When the noise is low even measure it becomes a very challenging task.
And devices like the Audio Precision ones (wonderful indeed !) are completely out of my reach.
One issue that matters a lot to me is isolation from mains.
Mains voltage is usually very corrupted.
For instance i do not understand why for low V and low V almost all designers use toroidals. Toroidals are quite "famous" for letting all the noise through without suppressing it.
Instead R-core or EI types having split bobbins are so much better.
I guess a very good linear ps starts with the selection of a very good transformer. This is the first choice to make.
Then after isolating from the outside noise another goal is try not to generate other noise for instance in the diodes bridge.
I think that i will just try Li Ion batteries for curiosity.
But i will end with a decent regulator. There is already one on the pcb, a lt1963.
I am just looking for decent results.
My system is quite low end ...
Thanks a lot again for the very helpful advice.
Have a nice day !
Kind regards,
bg
Hi,
> When the noise is low even measure it becomes a very
> challenging task.
I am using an EMU1616m with a "breakout box" as my second test system.
It has "120dB/192kHz" AD Converters (yes, they deliver that) and the over 60dB of gain in the microphone pre-amps are pretty useful to look at audio band noise on power supply lines. Bandwidth limit is 96kHz of course, my AP2 goes up 108k in FFT mode, more in analogue mode.
For wide-band stuff (Radio Frequencies) the analysers build into the better modern Digital Storage Oscilloscopes (DSO) are pretty useful. But you are likely still looking at several 1,000 Euro (my current 300MHz DSO retails for around 2.5k and is at the cheaper end of the range) to get something that is meaningful.
To cover things more accurately you can easily blow 50k Euro on a AP2 or Prism 'dscope and a wideband RF analyser...
> One issue that matters a lot to me is isolation from mains.
> Mains voltage is usually very corrupted.
> For instance i do not understand why for low V and low V almost all
> designers use toroidals. Toroidals are quite "famous" for letting
> all the noise through without suppressing it.
> Instead R-core or EI types having split bobbins are so much better.
Split bobbin only goes so far. I avoid split bobbins and instead use a different winding structure compared to the common one that ensures minimal leakage. But you have to have transformers custom made for that.
Something you may find paradoxical, a decent SMPS can do an excellent job of keeping corrupted mains (noise) out of the system. It must (in order to pass EMC testing) include quite substantial RF filtering.
In any SMPS the incoming voltage is first RF filtered (both directions) and then turned into DC. This DC is chopped up and applied as high frequency (67kHz and 132kHz are now common frequencies) to the mains transformer.
The actual mains transformer has very low inter-winding capacitance and is often actually shielded (needed to pass EMC) so very little noise can pass.
Suitable designs can minimise the value of the Y Capacitors or eliminate the Y capacitors entirely. Audio band noise depends on design, but can made very low, with enough effort. Of course, the voltage that comes out of the SMPS has a lot of noise at the switching frequency and up. But such noise is at predictable frequencies and of predictable nature and it should NOT present a major challenge to a competent engineer to filter this sufficiently that it does not matter.
Doing all this will cost some money, but next to custom 400Hz double C-Core transformers with special winding arrangements etc. it is downright dirt cheap and all the components needed are off the shelf commodity parts.
Filtering for the switching frequency can usually be added external to a SMPS (both input and output) for anyone doing DIY.
> But i will end with a decent regulator. There is already one
> on the pcb, a lt1963.
Hmm, it is specified as having 40uV of noise for the 3.3V part.
The old "low noise" plug-top supply which are currently replacing with a new model that is much quieter has lower levels of noise at 9V output (so much less as percentage of output voltage).
This regulator also has very poor high frequency behaviour. At 100kHz it already has less than 40dB (100 times) noise reduction capability.
I think I would probably look at this regulator as my first target.
The current DIY favourite TPS7A47 from TI (used in some linear power supplies market at quite high prices and also found in a different gizmo we recently discussed) has one tenth the noise and has ten times better high frequency rejection (at 100kHz) compare to that LT part and it retains this high noise rejection out to several MHz (so it will actually kill any noise from SMPS much more effectively).
So it may be an idea to get a "kit" and instead fix that rather awful regulator, before loosing sleep on other subjects.
Possibly adding some LC filtering on the power input (or in line with the power cable from the external PSU) could also give a major improvement, make sure the filter cutoff is at around 1kHz or lower and 2nd order. Then you will get around 80dB (10,000 times) extra noise filtering at 100khz and 120dB (1 million times) at 1MHz.
Ciao T
At 20 bits, you are on the verge of dynamic range covering fly-farts-at-20-feet to untolerable pain. Really, what more could we need?
" Hi, I am using an EMU1616m with a "breakout box" as my second test system. It has "120dB/192kHz" AD Converters (yes, they deliver that) and the over 60dB of gain in the microphone pre-amps are pretty useful to look at audio band noise on power supply lines.
Bandwidth limit is 96kHz of course, my AP2 goes up 108k in FFT mode, more in analogue mode.
For wide-band stuff (Radio Frequencies) the analysers build into the better modern Digital Storage Oscilloscopes (DSO) are pretty useful.
But you are likely still looking at several 1,000 Euro (my current 300MHz DSO retails for around 2.5k and is at the cheaper end of the range) to get something that is meaningful.
To cover things more accurately you can easily blow 50k Euro on a AP2 or Prism 'dscope and a wideband RF analyser... "
Hi again and thanks a lot indeed for the very interesting information.
So i serious investigation on noise is completely out of my budget ...
I have to live with that i guess.
" Split bobbin only goes so far. I avoid split bobbins and instead use a different winding structure compared to the common one that ensures minimal leakage. But you have to have transformers custom made for that. "
again i would like to stick with something off-the-shelf.
They are not all equal. Maybe they are not top but some still decent.
" Something you may find paradoxical, a decent SMPS can do an excellent job of keeping corrupted mains (noise) out of the system. It must (in order to pass EMC testing) include quite substantial RF filtering.
In any SMPS the incoming voltage is first RF filtered (both directions) and then turned into DC.
This DC is chopped up and applied as high frequency (67kHz and 132kHz are now common frequencies) to the mains transformer.
The actual mains transformer has very low inter-winding capacitance and is often actually shielded (needed to pass EMC) so very little noise can pass.
Suitable designs can minimise the value of the Y Capacitors or eliminate the Y capacitors entirely. Audio band noise depends on design, but can made very low, with enough effort. Of course, the voltage that comes out of the SMPS has a lot of noise at the switching frequency and up.
But such noise is at predictable frequencies and of predictable nature and it should NOT present a major challenge to a competent engineer to filter this sufficiently that it does not matter.
Doing all this will cost some money, but next to custom 400Hz double C-Core transformers with special winding arrangements etc. it is downright dirt cheap and all the components needed are off the shelf commodity parts.
Filtering for the switching frequency can usually be added external to a SMPS (both input and output) for anyone doing DIY. "
and this is another option i was thinking about. Adding a filter stage to a decent off-the-shelf SMPS.
Actually this is my best option now. I have only to add a DC panel socket on the unit and a filter inside the unit.
But i cannot find a suitable filter for such low V and low A.
" Hmm, it is specified as having 40uV of noise for the 3.3V part.
The old "low noise" plug-top supply which are currently replacing with a new model that is much quieter has lower levels of noise at 9V output (so much less as percentage of output voltage).
This regulator also has very poor high frequency behaviour.
At 100kHz it already has less than 40dB (100 times) noise reduction capability.
I think I would probably look at this regulator as my first target. "
the idea is not to mess with the pcb (the regulator is mounted on the pcb) and provide the regulator with the cleanest possible DC.
I cannot risk to ruin the pcb ...
" The current DIY favourite TPS7A47 from TI (used in some linear power supplies market at quite high prices and also found in a different gizmo we recently discussed) has one tenth the noise and has ten times better high frequency rejection (at 100kHz) compare to that LT part and it retains this high noise rejection out to several MHz (so it will actually kill any noise from SMPS much more effectively).
So it may be an idea to get a "kit" and instead fix that rather awful regulator, before loosing sleep on other subjects "
i will do that but upstream the regulator. So the regulator will not do actually the work. Clean DC in and out ...
" Possibly adding some LC filtering on the power input (or in line with the power cable from the external PSU) could also give a major improvement, make sure the filter cutoff is at around 1kHz or lower and 2nd order.
Then you will get around 80dB (10,000 times) extra noise filtering at 100khz and 120dB (1 million times) at 1MHz.
Ciao T "
thanks a lot again. Main problem is that i have no the money to set up a good measurement rig at all. I have no chance.
I have to rely on specifications of parts and kits.
But i will do some tests for sure ... this is sure.
Thanks a lot again, gino.
Kind regards,
bg
Hi,
> and this is another option i was thinking about. Adding a
> filter stage to a decent off-the-shelf SMPS.
> Actually this is my best option now. I have only to add a DC
> panel socket on the unit and a filter inside the unit.
> But i cannot find a suitable filter for such low V and low A.
You will have to calculate the filter (there are many on-line calculator options) and then purchase the correct inductors and capacitors. There are no ready-made "filter modules" I know,there is little point for any manufacturer to make such.
> the idea is not to mess with the pcb (the regulator is mounted
> on the pcb) and provide the regulator with the cleanest possible
> DC. I cannot risk to ruin the pcb ...
You give this regulator perfect DC. No noise at all. You still have 40uV noise fromthe regulator. Nomatter what you do, this is what you got.
> i will do that but upstream the regulator.
That would be pointless. It is like buying pure evian (or whatever you like) mineral water and then pee in it and insist the result is pure water because the source is pure.
> thanks a lot again. Main problem is that i have no the
> money to set up a good measurement rig at all. I have no chance.
LCfilters are quite primitive. You can probably get something towork well enough without much stuff.
Take a high ESR Electrolytic Cap as part of the filter, set the filter impedance to the ESR of this cap (say we use a 100uF/6.3V part with 1 Ohm ESR - this makes the filter impedance 1 Ohm) set frequency to 1 kHz, let the calculator work out L & C and then buy suitable L & C.
As the ESR of the cap and the filter impedance becomes lower the required capacitance goes up inductance down. All a tradeoff of size and cost.
Ciao T
At 20 bits, you are on the verge of dynamic range covering fly-farts-at-20-feet to untolerable pain. Really, what more could we need?
" Hi, You will have to calculate the filter (there are many on-line calculator options) and then purchase the correct inductors and capacitors. There are no ready-made "filter modules" I know,there is little point for any manufacturer to make such. "
Hi and thanks for the advice. I will look for them.
I am interested a lot. That solution would make things much easier.
" You give this regulator perfect DC. No noise at all.
You still have 40uV noise from the regulator. Nomatter what you do, this is what you got. "
well ... i could live with that very well.
Now i think that the actual noise is much higher than 40uV.
A lot of garbage reaches the regulator.
" That would be pointless. It is like buying pure evian (or whatever you like) mineral water and then pee in it and insist the result is pure water because the source is pure "
if i understand you mean that whatever there is upstream the regulator the outcome will be always a 40uV of noise ? really ?
This is a very fundamental point to understand for me.
If this is the case i will leave also the ps untouched.
I can bear 40uV of noise ... for me is nothing.
" LCfilters are quite primitive. You can probably get something towork well enough without much stuff.
Take a high ESR Electrolytic Cap as part of the filter, set the filter impedance to the ESR of this cap (say we use a 100uF/6.3V part with 1 Ohm ESR - this makes the filter impedance 1 Ohm) set frequency to 1 kHz, let the calculator work out L & C and then buy suitable L & C.
As the ESR of the cap and the filter impedance becomes lower the required capacitance goes up inductance down.
All a tradeoff of size and cost.
Ciao T "
thanks a lot again for the very precious advice
I have to understand well that any mod upstream the regulator will have no effect on the actual noise at its output.
If this is confirmed this changes completely my plans.
I will just solder again some caps that i took out and stop.
I should have added that after the main regulator lt1963 other local regulators follow. I cannot spot them but someone more expert than me has spotted them.
But as i said i will avoid messing with the ps.
Thanks a lot.
Kind regards,
bg
Hi,
> well ... i could live with that very well.
But this regulator makes so much noise, batteries or even our "next generation" plugtop supplies would be wasted, just as the pure water in my rather graphic example.
> Now i think that the actual noise is much higher than 40uV.
This is possible. It is almost impossible to say without suitble test ger.
> if i understand you mean that whatever there is upstream the
> regulator the outcome will be always a 40uV of noise ? really ?
Of course. All electronic circuits generate noise. The specific one you mentioned is specified as adding this much noise under ideal conditions.
> I can bear 40uV of noise ... for me is nothing.
Well, 40uV of noise might be expressed as 93dB (forgive if I am a little out) below 2V. If driving a clock, this may matter. On a single gate it will likely add around 1pS of jitter. Many modern circuits have 100's of gates. SO these 40uV can very quickly become several 100 times 1ps jitter...
> I have to understand well that any mod upstream the regulator will
> have no effect on the actual noise at its output.
It will, especially in this case. The regulator you mention does not do much to reduce noise past appx 1kHz. So giving it a passive filter in front will add the passive filters effect to the regulator.
Of course, you cannot go lower than the regulators self noise, but you can minimise the noise that gets to the regulator and past. Just there is a point where further improvements no longer help, due to build in limitations.
For fun, I can routinely make something that offers around 0.4uV of self-noise using less than one USD in parts for two channels (in 1KU), if all we need is 3.3...5V at 0.5...1A per channel and using quite generic parts. We can go as low as 0.15uV if we spend more money. Walt Jung is a good source.
> I should have added that after the main regulator lt1963
> other local regulators follow.
And if you are unlucky, they produce 400uV of noise... ;-)
Ciao T
At 20 bits, you are on the verge of dynamic range covering fly-farts-at-20-feet to untolerable pain. Really, what more could we need?
Good morning Mr. Thorsten
" But this regulator makes so much noise, batteries or even our "next generation" plugtop supplies would be wasted, just as the pure water in my rather graphic example "
actually i saw the datasheet and was surprised to see a poor performance on the higher Hz.
Clearly some kind of additional filtering is mandatory to keep noise down.
" This is possible. It is almost impossible to say without suitble test ger "
and this is my main regret because i understand that good testing rig are both complex and very very expensive. And this also normal.
" Of course. All electronic circuits generate noise. The specific one you mentioned is specified as adding this much noise under ideal conditions "
unfortunately i have no chance to test it but i trust you completely
In the end is a cheap product.
" Well, 40uV of noise might be expressed as 93dB (forgive if I am a little out) below 2V. If driving a clock, this may matter. On a single gate it will likely add around 1pS of jitter. Many modern circuits have 100's of gates. SO these 40uV can very quickly become several 100 times 1ps jitter... "
I see. However i see also very expensive master clock generators powered by what look like off-the-shelf SMPS.
Maybe there is also some kind of filterin, actually i am sure of that.
" It will, especially in this case. The regulator you mention does not do much to reduce noise past appx 1kHz. So giving it a passive filter in front will add the passive filters effect to the regulator "
and this exactly one of the reason to replace the toroidal.
I understand that a same effect of a split bobbin transformer (that i see used on high end units) can be achieved with passive filtering before the toroidal, but if space were not a constraint that toroidal will fly in the garbage bin immediately.
My approach is simple. Look what is inside the best unit and try to use it. Unfortunately schematics are not that common.
" Of course, you cannot go lower than the regulators self noise, but you can minimise the noise that gets to the regulator and past. Just there is a point where further improvements no longer help, due to build in limitations.
For fun, I can routinely make something that offers around 0.4uV of self-noise using less than one USD in parts for two channels (in 1KU), if all we need is 3.3...5V at 0.5...1A per channel and using quite generic parts. We can go as low as 0.15uV if we spend more money. Walt Jung is a good source "
wow ... that is very low indeed.
I understand that in order to suppress the noise generated by the regulator additional filtering after it would be extremely beneficial.
But i have no access to the schema and it is difficult to extract it from the pcb.
But i still could clean up a little the feed to the regulator.
Especially the HF noise that tends to pass through the regulator unsuppressed.
Even just a different transformer could be a good change.
" And if you are unlucky, they produce 400uV of noise... ;-)
Ciao T "
Yes. But as i said i cannot do anything for this.
Thinking about this a little more the idea to use an external decent SMPS or linear and put in the box just an additional filtering stage before the regulator caps is interesting.
I have just to place a new DC panel socket on the box and the filter inside.
I will look for some suitable schematic.
I see that many dvd/br player have passive mains filters.
I should be able to copy them.
Just some inductors and caps. Nothing out of this world.
I understand that the performance will be determined by the pcb design anyway.
But i can only try to feed the regulator as well as possible.
If i touch the pcb the unit i destroy everything.
And i do not like the idea to increase further the noise already generated by the regulators.
To end i have a strong feeling that the war to noise is a decisive one for sound. In the future i would like to be able to understand more on how the noise generates and how it can be suppressed.
I watched a very simple but also very interesting demo on Youtube.
The speaker compared the noise to a carpet. The more the noise the thicker the carpet.
If some objects of different sizes are then thrown on the carpet we can still see the bigger one but the smaller ones will be down in the carpet.
These small objects are like details in music.
A higher noise hides them. There is so much noise already in like.
So i think this is a very interesting field of study.
Thanks a lot again for the very kind and precious help.
Have a nice day.
Kind regards,
bg
Post a Followup:
FAQ |
Post a Message! |
Forgot Password? |
|
||||||||||||||
|
This post is made possible by the generous support of people like you and our sponsors: