“The oboes are extra center-stage…” | element14

On the finish of final 12 months I accomplished a project to build a Rasbperry-Pi-based DAC streamer. Successfully it was constructing an influence provide for a Raspberry Pi zero and the related Hifiberry DAC+ pro HAT. The concept of that energy provide was to supply clear energy to the DAC to make sure the very best sound replica. Reducing a protracted story brief: I misused that energy provide a little bit for one thing else and due to that broke one of many TPS74A7 regulator modules which I selected due to their low noise capabilities. With the thought of rebuilding that channel with one thing new and with a little bit extra energy I began looking for a regulator with a very good status for audio purposes.


It did not take lengthy and I used to be studying so much about regulators and energy provides in varied DAC boards. Now right here comes the disclaimer: I do not take heed to excessive definition audio information - all my materials is 16bit/44.1kHz FLAC information. Additionally, I do not declare to have audiophile listening to capabilities. So if my FLAC information can be up-sampled to 24bit/96kHz then I'd undoubtedly not be capable of hear the distinction - in contrast to some individuals who declare that there's a clearly audible distinction. Lastly, I prefer to hear primarily to in style music - which has been mastered utterly totally different by way of the final many years: Excellent for vinyl (with all it is limitaions set by RIAA) within the 70s/80s, excellent for CD within the 90's and early 00's, and far more compressed since then to permit the music to be recognised even when my daughter performs it on the speaker of her cell phone.


After which I learn this submit from a man who claimed that with a sure energy provide to energy a particular DAC streamer he can hear the oboes clearly extra middle stage than with one other energy provide. And several other commentators agreed or supplied different audiophile energy provides to realize a fair higher or totally different sound. The streamer in query prices round 500 EUR and the facility provides in contrast have been at round 300 EUR. Now, my naive assumption can be {that a} 500 EUR DAC streamer would have it is personal energy provide part which ensures that incoming energy is regulated and filtered to an prolong that there isn't a impression on the generated audio sign - filtered as regards to EMC but in addition to ripple on the voltage sign. However clearly not...


Nicely, which parts of an influence provide do affect the output sign of a DAC streamer? The consultants within the varied boards say ripple / noise on the voltage, the impedance of the facility provide and the power of the facility provide to deal with quick load modifications. Particularly the impedance half makes full sense to me. Relating to the load modifications, nevertheless, I battle to consider that this has an excessive amount of impression - just because bottom-line a DAC streamer is a generator and never an amplifier. Due to this fact I made a decision to have a better have a look at the load modifications {that a} energy provide of a DAC streamer has to endure.

For this, my Hifiberry DAC affords a fantastic testing floor because it would not have it is personal energy part on the board (in contrast to it is extra premium successor Hifiberry DAC2 HD). Which means it even makes use of the three.3V from the converters of the Pi. Consequently, an unsuitable energy provide ought to have a direct destructive impression on the generated audio sign. The music information I streamed for these exams was despatched to the Pi/DAC by way of WiFi, so no LAN cable with a possible supply for noise was used. For the exams I powered my mixture of a Raspberry PI zero and a Hifiberry DAC+ professional with 4 totally different energy provides:

In fact I attempted to seek out variations by listening to the playback with these energy provides. For this I used a "Head Box S2" headphone amplifier from Project Audio in addition to "Amiron Home" headphones from Beyerdynamic. The music I used was "Reasonably Be" from "Clear Bandit". Why? As a result of it begins with a really distinctive violin and I discover a single violin excellent to identify "audio imperfections".

For the technical comparability I borrowed a little bit bit of drugs from 's demo pool (which is well potential for me as I work there): A MR6000 memory recorder with a 16 bit / 1 MSamples excessive decision voltage unit, and a CT6711 current sensor, which for this objective has a little bit of an "overkill bandwidth functionality" of 120MHz however on the similar time affords a superbly becoming 500mA present vary.


For every energy provide I recorded the (5V) provide voltage to the Raspberry Pi in addition to the present whereas streaming music. Within the set vary of the present sensor 500mA of measured present can be proven as an output voltage sign of 5V. The MR6000's sampling price was set to the utmost functionality of the enter unit, so each voltage in addition to present values have been recorded with 16 bits and sampled 1 million occasions per second. This ought to be greater than enough to catch even the quickest load modifications of a USB energy provide.

Above you may see the voltage in blue in addition to the measured present in purple. The motion begins with the Raspberry beginning to play the music. However even in idle mode you may see the present continually altering between 100mA and 150mA, proven on the purple scale as voltage values between 1V and 1.5V. In keeping with the present modifications the voltage reacts to this by shifting between 4.95V and 5V. As soon as the music streaming begins you may instantly see the voltage dropping extra when the present modifications. I am certain you aren't stunned after I reveal that the picture above reveals the measurement of the most affordable USB wall charger I might discover in the home. Lets zoom in a bit and have a look at a kind of load modifications:

Are you able to see how "unclean" the 5V voltage is? As you may see, the noise is unrelated to the load modifications - however once more, that is the most affordable 5V USB wall adapter I might discover in the home. Additionally, you may actually see how the voltage drops when the load will increase. However I wasn't in any respect stunned to see these outcomes. In truth, I noticed what I needed to see.


Subsequent up was the iFi iPower which prices round 70 EUR. iFi says about this energy provide: "Add the iPower into your system and also you’ll discover music come again to life. It might add intimately initially misplaced by way of standard energy provides." The ability provide used above definitely was convential - so let's hook up the iPower and play "Reasonably Be":

Sure - that is what I needed to see! As a result of I spent 70 EUR on this and I'd have actually been irritated if this screenshot would have regarded just like the one with the standard wall adapter from the primary picture. Each screenshots present the start of the identical music monitor being performed - the one factor which is totally different is the facility provide.


iFi specifies the noise of the iPower to be simply 1uV - they do not specify at which frequency these 1uV apply, however no matter that it's a actually low worth. Lets zoom in to have a look at a single load change and the cleanliness of the voltage:

Once more, the present curve appears to be like a lot cleaner and the voltage drop is decrease with the iFi than it's with a budget wall adapter. The noise on the voltage just isn't actually any totally different than with the wall adapter, although. I assume this comes from the voltage regulation on the Pi - however that is only a guess. Anyway, it appears to be like nice - however does it additionally sound nice?


Sure, it does (sound nice). However so does the music after I use a budget wall adapter. Truthfully, I am unable to hear any distinction. Two potentialities: Both my listening to capacity is so much much less audiophile than I wish to admit, or whereas the curves with the standard wall adapter look "dramatic" - perhaps they have no affect on the generated audio sign? After all of the voltage drops of the most affordable wall adapter have been 70mV in comparison with round 50mV with the iFi - so general the voltage drops the place within the vary of 1% to 1.2% for round 200 to 300 micro seconds. Is that sufficient to have an audible impression on the output sign of a DAC?


Subsequent up - my home-made energy provide primarily based on Texas Devices' TPS74A7:

"Mr. Scharrmann, do not hand over your daytime job" was the very first thing I believed when wanting on the that picture. It is in all probability not as "dramatic" because the curve of the "most cost-effective wall adapter in the home" however on the similar time it is away from the iFi. Does it get higher when zooming in?

I feel it appears to be like a lot nearer to the graphs of the iFi than to these of the most affordable wall adapter in the home. And does it sound totally different? Nope - all three energy provides end in output indicators that sound precisely the identical to my ears.


However now... the following and closing take a look at candidate is way away from the 70 EUR iFi: The subsequent candidate is a lab energy provide of the HMP collection from Rohde & Schwarz. Every of the three channels can present 5A of present so long as the full energy stays beneath 80W - load modifications within the vary we've seen above ought to be dealt with with ease.

I've to say that I used to be fairly suprised concerning the stage of voltage drop. Evaluating the graphs even the most affordable wall adapter in my home has much less of a voltage drop. However I feel there are two issues to remember:


  1. All examined energy provides aside from the HMP2030 are repair voltage energy provides with principally a lot smaller present outputs. The HMP quite the opposite has a freely selectable voltage vary between 0 and 32V and a present vary between 0 and 5A per channel. Due to this fact I'd "examine apples with pears" after I put the HMP neck on neck with the aim constructed iFi. Then again, the iFi retails for 70 EUR together with VAT whereas the HMP2030 retails for 1250 EUR excluding VAT.
  2. Additionally, I feel you will need to level that I didn't use the sense enter of the HMP2030: This measures the voltage on the DUT and permits the facility provide to counter any voltage drops or rises. Lets zoom into this image:

Once more, I'm stunned concerning the excessive stage of voltage drop. I actually did not count on this. However as talked about earlier than, I did not use the sense enter. That is why I repeated the identical measurement - however this time by utilizing the sense enter:

The above graph of the HMP2030 with the sense enter used might be essentially the most "dramatically" wanting graph. On the similar time it's an effective way to elucidate a way enter perform of an influence provide as a result of the distinction is well seen when evaluating the 2 curves (one with and one with out utilizing the sense enter). It even turns into extra clear when zooming in:

The precise voltage change might be the very best of all energy provides within the final graph. However you may already guessed it: Once more, there isn't a audible distinction when evaluating the sound. In truth I could not hear any audible distinction with whatever the energy provide used in any respect. One factor to notice is that the load modifications are under no circumstances aligned with the sound performed - so a extra highly effective half within the music doesn't end in larger present being drawn.


What do you consider this? Ought to I be stunned that the audio playback sounded the identical with each energy provide, or was that one thing you suppose was to be anticipated? Any flaws in my measurement you noticed? Any particulars lacking? Please let me find out about your ideas...

Source link

Leave a Reply

Your email address will not be published. Required fields are marked *