I have a personal experience here. A friend was testing a line preamp prototype using different transformers in the power supply.
Switching on the lights in the room was sending a big bang in the speakers with the toroidal. Much much less with the EI. Just a little click.
I understand that also toroidal can work as isolation transformer, still I am more confident on EI.
For such low VA efficiency is not an issue.
The toroidal type transformer it have the worst AC noise isolation. A slightly worse than the EI type. This isolation issue is somehow improved by a galvanic shield between windings in both cases. Before these type transformers, it should be used AC main filtering for sensitive devices powered so.
Actually all the power transformers (no matter their types) it may be called also isolation transformers, as indeed, it isolate the user device from the main AC rail.
This is a fundamental point. How large the cap following the regulator can be ?
I have at hand some 16V/2200uF Nichicon Gold. Will 2200uF be too much ? does this depend on the actual regulator used ?
I like this idea a lot indeed. In this way there will be a lot of power reserve just close to the dac.
You can very safe use after (at least 10A, Schottky diodes bridge) around 100 000µF capacity. There is more about the place available you may have for these caps... After the regulator device, you can go so high as 1F... The modern regulators have no problem to deal with such capacities. The problem when using so important capacities in a PSU is its quite slow start up. So, one may chose a good functioning compromise between the max acceptable capacity after the regulator device, so to not have a very slow power on sequence. There is here also about the place of a such PSU in the device`s system. A too slow start up it can disturb another regulator stages into the rest of the system, or may disturb a designed/implemented power up sequence. A general appreciation may be taken before increase too much the filtering capacities. In your case, as a primary PSU, a slower start up it may not be an important issue. As I can see... However, you should decide by yourself, and by experimenting. The slow start up issue of a linear PSU, using large filtering capacities, it can be somehow solved by having the PSU always connected to the main AC, and then connect (disconnect) the device to (from) the regulated power provided by that PSU...
Exactly my thinking ! when i look at the ripple figures i see about 1% of Vout for the good smps and from 20 to 50 times less for good linear psu.
Even high quality digital units like master clock generators use smps.
Actually the dac i am using has dc generation on board.
But again i would like very much that the power supply is not in any way a limiting factor. As i am going to stick with this dac for long i will take some time about this issue.
Thanks a lot again for directing me towards R-core transformers.
I will check immediately. I guess a 12VAC unit should be fine for a 12VDC power supply.
In the designed systems today it is used on large scale the SMPS, not only as main PSU, but as local PSUs. This it have its advantages, but important disadvantages in my opinion. However, there is not possible to use linear PSUs everywhere on a complex board of a whatsoever device.
I experienced this myself: there is a huge improving in the quality of the outputted signals (no matter there are audio, video, or measurements signals, digital or analogue) when the main PSU is a linear one but not a SMPS. A main SMPS it do the bad even worse in a powered so digital/analogue complex system. A linear PSU it attenuate the overall noise level in a powered so system. Why is so, it may be another discussion...
The big problem with linear PSUs is the important heat dissipation, and quite low energetically efficiency. While a SMPS is very high efficient, it dissipate very low heat, for huge power capabilities. All these advantages are only spoiled by a very high HF noise on its outputs, which is spread into the all powered system. To filter out such very large spectre noises is an enough challenge for designers, and therefore it may increase too the final price. Because this expensive filtering, the most manufacturers accept in the end a quality/price compromise. And the final user in general accept this too... for the price they get. Those few, who do not accept such compromises, it tweak the systems for a linear PSU, and it get so the quality back, for a higher overall price... So it works...🙂
Switching on the lights in the room was sending a big bang in the speakers with the toroidal. Much much less with the EI. Just a little click.
I understand that also toroidal can work as isolation transformer, still I am more confident on EI.
For such low VA efficiency is not an issue.
The toroidal type transformer it have the worst AC noise isolation. A slightly worse than the EI type. This isolation issue is somehow improved by a galvanic shield between windings in both cases. Before these type transformers, it should be used AC main filtering for sensitive devices powered so.
Actually all the power transformers (no matter their types) it may be called also isolation transformers, as indeed, it isolate the user device from the main AC rail.
This is a fundamental point. How large the cap following the regulator can be ?
I have at hand some 16V/2200uF Nichicon Gold. Will 2200uF be too much ? does this depend on the actual regulator used ?
I like this idea a lot indeed. In this way there will be a lot of power reserve just close to the dac.
You can very safe use after (at least 10A, Schottky diodes bridge) around 100 000µF capacity. There is more about the place available you may have for these caps... After the regulator device, you can go so high as 1F... The modern regulators have no problem to deal with such capacities. The problem when using so important capacities in a PSU is its quite slow start up. So, one may chose a good functioning compromise between the max acceptable capacity after the regulator device, so to not have a very slow power on sequence. There is here also about the place of a such PSU in the device`s system. A too slow start up it can disturb another regulator stages into the rest of the system, or may disturb a designed/implemented power up sequence. A general appreciation may be taken before increase too much the filtering capacities. In your case, as a primary PSU, a slower start up it may not be an important issue. As I can see... However, you should decide by yourself, and by experimenting. The slow start up issue of a linear PSU, using large filtering capacities, it can be somehow solved by having the PSU always connected to the main AC, and then connect (disconnect) the device to (from) the regulated power provided by that PSU...
Exactly my thinking ! when i look at the ripple figures i see about 1% of Vout for the good smps and from 20 to 50 times less for good linear psu.
Even high quality digital units like master clock generators use smps.
Actually the dac i am using has dc generation on board.
But again i would like very much that the power supply is not in any way a limiting factor. As i am going to stick with this dac for long i will take some time about this issue.
Thanks a lot again for directing me towards R-core transformers.
I will check immediately. I guess a 12VAC unit should be fine for a 12VDC power supply.
In the designed systems today it is used on large scale the SMPS, not only as main PSU, but as local PSUs. This it have its advantages, but important disadvantages in my opinion. However, there is not possible to use linear PSUs everywhere on a complex board of a whatsoever device.
I experienced this myself: there is a huge improving in the quality of the outputted signals (no matter there are audio, video, or measurements signals, digital or analogue) when the main PSU is a linear one but not a SMPS. A main SMPS it do the bad even worse in a powered so digital/analogue complex system. A linear PSU it attenuate the overall noise level in a powered so system. Why is so, it may be another discussion...
The big problem with linear PSUs is the important heat dissipation, and quite low energetically efficiency. While a SMPS is very high efficient, it dissipate very low heat, for huge power capabilities. All these advantages are only spoiled by a very high HF noise on its outputs, which is spread into the all powered system. To filter out such very large spectre noises is an enough challenge for designers, and therefore it may increase too the final price. Because this expensive filtering, the most manufacturers accept in the end a quality/price compromise. And the final user in general accept this too... for the price they get. Those few, who do not accept such compromises, it tweak the systems for a linear PSU, and it get so the quality back, for a higher overall price... So it works...🙂
Last edited:
Perfect ! message received and thanks a lot for confirming my impressions.
I will look seriously to R-core type anyway. I have seen them in very high quality units. And the primary and secondary windings are completely separated.
I see. I will stat low then. After all the smps do not have big output uF
I will stay around 1000uF or so.
From what i understand this type of units have dc to dc converters on board that provide all the needed supply.
As i have already mentioned if the problem is on the board no external ps can solve it, like a very big noise peak on one of them (see attachment).
The peak persists in the same position with the same level no matter which psu i use 🙁
My question is .. is there an instrumental way to assess this ?
i am only able to perform these noise floor graph, for now
I think that they can be useful to spot some issues.
Honestly i have to say that some bad cases aside the results are quite similar. No big variation at least for the noise floor.
Mine is more a question of principle. I see many graph of noise floors and they are usually very flat. That is what i like. A floor must be flat.
Yes i think i have got it. The better is always the bigger and the more expensive. I am using now a 12V/3A smps quite smaller than a pack of sigarettes 😱
A similar but linear would be much bigger and i guess expensive.
But probably much better for audio.
For casual listening also a smps could do fine.
But for more deep listening i do not think so.
Thanks a lot again, gino
Attachments
However a linear PSU do not have a such spread noise spectre as a SMPS... This is a solid fact.
There are of course many aspects/sides of this endless discussion subject, about noises and power supplies types, and the use of such in audio field.
Here is about using a linear PSU as primary power supply instead of a SMPS. In my above considerations (based on my experience) I mainly refer to the use of a primary PSU to power a complex digital/analogue domain of the/a device. There is a very solid fact that the simple replacement: a linear PSU instead a SMPS (as primary/main power supply), it bring an overall and dramatic improvement for the outputted signals of that device. And this it refer to both digital and analogue domains of that device. Why is like this, it may be a enough comprehensive, interesting, and maybe endless discussion.
A main/primary power supply is meant to provide the row power rails to a device/system. This row power is to be further improved, filtered, regulated and so on, for the use in local and sensitive areas, stages, circuits.
There is one thing to start a power quality processing/improvement in a system/device, starting from a quite low noise, and good enough quality linear PSU, and a very different thing to do the same processing/improvement of the power rails, based on raw power provided by a switching PSU. The overall lower noise levels, as the different noise spectre of a linear PSU, comparing to a SMPS it may very well be one main explanation for the overall improvements of performances of the targeted device/system. Or in other words, a raw explanation of this phenomenon...🙂
I would like to show here (I`m sure you know also these things very well...), how it looks the outputs of two power rails of a linear PSU and a SMPS one (as primary/main PSUs). This power rails are to provide power further to another small SMPS for digital stage, as power for linear regulators into the system... The conclusion is enough clear, I suppose...
It not matter at all in case of a primary PSU if its outputted noise it is in the audio domain, or not. It is mainly about the amount of overall noise level, and the huge spread spectre of such noise in case of SMPS. And is about a main PSU, to power a complex (digital/analogue) system.
And another exemple: Here is the original SMPS used by Agilent in the scope I own. Can you imagine? A measurement device, powered by a such power supply, with 1,4Vpp noise level... That SMPS is loaded under measurement, for the same current as the device it used. After this scope it got a linear power supply, its overall noise level it decreased dramatically...
Attachments
Last edited:
A SMPS it can not have large capacities on its outputs. It can not function so. And it not need, as the high frequency ripple it can be well filtered by low capacities. This is another advantage of SMPSs: smaler overall dimensions.
But another advantage of a linear PSU: it can tolerate very large capacities, and this it improve a lot its capabilities to deliver a clean/cleaner power. The cons are bigger overall dimensions...
Thanks a lot sincerely again for all the very valuable explanations.
I am attaching a picture of the unit. The idea is to place a proper dc socket on the back panel to free to try any 12VDC around without having to worry with sizes.
Actually this is the most tricky part for me.
I will report asa i will have something to report.
During the weekend i should have all the pieces to build something. Linear of course 😱
Kind regards, gino
Attachments
Hi ! thanks a lot again.
This is what makes me drooling 🙂
But let me ask you a trivial question.
I can see on my pc the resulting noise floor spectrum of the soundcard up to 96kHz.
This card has DC to DC converters on board, a very common solution I understand for this type of devices.
I have to test still linear psu. I have tried only smps for now.
But if I will see the same noise floor swapping between a smps and a linear one can I be confident about the good sound from both ?
or maybe the noise above 96kHz, that I cannot see, it is what makes the difference in favor of the linear psu ?
Or there is something that the instruments cannot see ?
Very very interesting ! thanks a lot !
I really think I will be experimenting a lot with linear psu.
I am in the parts gathering phase now.
I will post some pictures asa I will have the pieces assembled.
In particular I am now interested in regulator kits.
I need a kit to start. Assembled or not. 12V/1A.
Unfortunately the range above 96kHZ will be out of my reach.
Thanks a lot again for the very kind and interesting advice.
Kind regards, gino
Last edited:
My experience differs greatly from yours regarding power supplies, again this is all tailored towards SMPSs being bad which is not the case....
Hi !
I think that a very important factor is also the PSRR of the circuit that the power supply powers.
For instance some op-amps have PSRR of even more than 100 dB 😱
This means that almost any psu noise will be suppressed to minimal level.
The problem maybe is how this PSRR changes with Hz ?
I am attaching the values for the AD797.
So even if a power supply is more noisy than another the end result can be very similar because the higher noise will be suppressed by the PSRR of the circuit ?
I think that a very important factor is also the PSRR of the circuit that the power supply powers.
For instance some op-amps have PSRR of even more than 100 dB 😱
This means that almost any psu noise will be suppressed to minimal level.
The problem maybe is how this PSRR changes with Hz ?
I am attaching the values for the AD797.
So even if a power supply is more noisy than another the end result can be very similar because the higher noise will be suppressed by the PSRR of the circuit ?
Attachments
Well, what it is really happen when using a serial/linear PSU on a digital stage of a device is quite interesting, and for sure it have to be analysed more in deep. I didn`t do it myself yet...
What I observed so far is that the quiet linear PSU (with a very low noise on its output resistive loaded), when is connected (as a primary/raw power source) to a digital stage, it become enough noisy... This is quite normal, as the digital system is a very noisy one by nature, and a large spectre HF noises are spread everywhere through ground, power rails, and as EMi emissions. Even thought, the improvements for that digital stage performances, and for the final results produced by that system, are dramatic and obvious at once. My explanation so far is that the linear PSU it may lower somehow the overall noise level into the system, or change its spectre domain, so to produce improvements on either processing or the signals quality. While using a SMPS to power the same system, the increasing of overall noise level into the system, as enlarger its HF spectre and complexity, it looks enough normal to me. The SMPS it contribute with its huge high level noise domain to that of the digital system itself. Maybe someone else may explain better what really happen, but the fact is: a linear PSU powering a digital stage it produce improvements.
Indeed, there is not possible to see just everything on measurement instruments, and to see enough much it mean huge research efforts, lot of dedicated time, and so on. Maybe years of investigations... Personally i prefer more immediate results, and if I get what I expect, then my interest to explain more about one or another, it may decrease quite much, and fast...
I was myself very sceptical to use a linear PSU to power a digital system, as I knew (and it was also confirmed to me) that the noisy digital system it will spoil the quietness of a linear PSU, and such approach is very low effective, it produce lot of heat, and so on. After testing it myself, I experienced the improvements, and so I changed my scepticisme...
Such powering approach it is not possible always, and at least SMPS it have its functional well defined functional domain, as a powering method.
There is quite old the idea to power a entire computer from a serial PSU, to improve its overall performances or for some stages of it. It are also some trials in this directions. In my opinion, this is just not possible (when about usual computer systems), but it may works fine for small (low energy) computer systems. I haven`t get so far yet in my experiments, but I did it for my sound card installed in a SMPS powered computer. I changed the whole power system for my sound card (to a linear one), improving it, and isolate it completely form the rest of the computer. The result is was/is above all expectations...
So, my conclusion and advice is to use such linear powering approach rationally, making a good compromise between system efficiency, available space, the application to be fit it to, and so on...
Well, I do not doubt at all your experience, but you may admit that these two powering concepts are radically different, when about the quality of their outputs (especially when about noise generation), as their use in hifi fields. We may agree that the linear PSU come much better out of such comparative...
However, this subject it seems to me to be an enough endless discussion...🙂
If you may show a so clear scope line on an SMSP output, as we can see on a linear PSU, then I will not say anything anymore...
Hi and thanks again for the very helpful advice.
From now on I will be experimenting different linear power solutions for sure.
As I said changing smps has not has a big of an impact, stock one aside (why they did not care about that issue I do not understand really).
I will test first the psu outside the box and then I will try to place the regulation stage inside and keep the transformer outside.
I have the feeling that this card as a huge potential indeed as AD and DA converter.
Actually I bought it especially to use it as sub to AES interface as I have another AD/DA converter of which I like the sound.
If the result will be positive I could use a linear psu also with that converter
An externally hosted image should be here but it was not working when we last tested it.
Thanks a lot again, gino
Hi ! very impressive indeed ! great project.
I guess that there is DC going through the umbilical ?
I tend to prefer the AC option, placing reservoir caps as close as possible to the final stages to avoid any lag in the current delivery ?
I read about that concept somewhere but I do not remember where.
Smaller and fast caps close to the output devices ... I like the idea.
Transformers are bad animals. Heavy, vibrating, EMI generators ... better to keep them away from circuits. I would do that with my amp if I had the time. Just extract the transformer and put it in a separate cage.
In this way there will be more space inside the amp for better and bigger caps.
Thanks a lot, gino
Would suggest you learn more about digital system noise, linear or SMPS if its a good supply wont make a difference, same with analogue.
I would start with learning about SSN (simultaneous switching noise).... your started for 10, then we bcan look at the effect layout has on digital noise, we'll start with ground plane slots here.....
Last edited:
SMPS are not bad at all ! It is just that linear PSUs don't switch and thus chances are likely that they don't generate lots of noise in various bands. Heat is not a real problem when LDO regs are used and transformer voltage is adjusted accordingly. In audio devices not much power is needed, it would be a different case for high power devices.
Some observations regarding SMPS in audio devices::
- Manufacturers use SMPS adapters/wall warts that are CE/UL tested and approved so they don't need to do that process with self designed PSUs for their devices.
- Newer devices with SMPS chips have an enable pin which can be controlled by a processor which can be convenient and it is cheaper to produce.
- A real 230V AC power switch simply costs too much. An SMPS with soft on/soft off might be cheaper than a real power switch.
- a transformer costs too much (copper!) and using one creates the need for a sturdier case. The weight of the device will be way higher leading to higher shipping costs.
- for a linear PSU a heatsink is often needed which adds to the total cost.
- Euro government involvement in "green/eco technology" demands very low standby power. Well, a power switch was already taken out of the equation and a transformer can not beat an SMPS here. A transformer is better not left switched on because of copper/iron losses.
- when SMPS are chosen it seems the cheapest electrolytic caps are the best option. I can not count the times I had to replace leaking caps of dubious quality in suddenly failing young equipment. This, together with the heaps of ripple and RF creates a prejudice against SMPS...
So economical or ecological reasons lead to choosing SMPS rather than technical merits. For consumer stuff produced in the millions this of course has a smaller carbon footprint. Those few thousands of audio hobbyists won't have a very large influence on the whole picture. What can be simpler than a well decoupled linear PSU with a good old transformer when small power devices need a good low noise PSU ? Linear PSUs also don't break down as much as many an SMPS tends to do. Personally I use well decoupled linear PSUs without any doubt when power is below 20 W max.
Some observations regarding SMPS in audio devices::
- Manufacturers use SMPS adapters/wall warts that are CE/UL tested and approved so they don't need to do that process with self designed PSUs for their devices.
- Newer devices with SMPS chips have an enable pin which can be controlled by a processor which can be convenient and it is cheaper to produce.
- A real 230V AC power switch simply costs too much. An SMPS with soft on/soft off might be cheaper than a real power switch.
- a transformer costs too much (copper!) and using one creates the need for a sturdier case. The weight of the device will be way higher leading to higher shipping costs.
- for a linear PSU a heatsink is often needed which adds to the total cost.
- Euro government involvement in "green/eco technology" demands very low standby power. Well, a power switch was already taken out of the equation and a transformer can not beat an SMPS here. A transformer is better not left switched on because of copper/iron losses.
- when SMPS are chosen it seems the cheapest electrolytic caps are the best option. I can not count the times I had to replace leaking caps of dubious quality in suddenly failing young equipment. This, together with the heaps of ripple and RF creates a prejudice against SMPS...
So economical or ecological reasons lead to choosing SMPS rather than technical merits. For consumer stuff produced in the millions this of course has a smaller carbon footprint. Those few thousands of audio hobbyists won't have a very large influence on the whole picture. What can be simpler than a well decoupled linear PSU with a good old transformer when small power devices need a good low noise PSU ? Linear PSUs also don't break down as much as many an SMPS tends to do. Personally I use well decoupled linear PSUs without any doubt when power is below 20 W max.
Last edited:
thanks, no way i could have done it on a single chassis...
there are two umbilicals one for dc and the other for ac...
Hi ! thanks a lot for the valuable advice.
Actually i read that the only noise difficult to suppress maybe is the one coming from the diodes bridge ?
And also i read that while single chip regulators are quite effective at low Hz they are less effective at higher Hz.
So maybe with linear psu it is just to select the right diodes ? and use a good isolation transformer ?
Thanks again for the very helpful explanations.
May i just ask you what well decoupled means ?
is it about the caps type and value used in it ?
Kind regards, gino
I never ever have seen (and I`ve seen a lot along teens of years) a SMPS which it can present a decent clean line on a scope, as a linear PSU it do.
I will be very glad if you can show a such exemple.
Else there is lot of to learn about, all the time... I agree. Only the time available may be a serious issue... sometimes... anytime...
- Status
- Not open for further replies.