iFi Audio recently sent me an hip-dac2 for review and I’ve been auditioning it for a while with great pleasure.
The new version of iFi’s recently discontinued hip-dac, amongst the few low cost mobile dac-amps featured of Audioreviews’ Wall of Excellence, is marketed at a very similar price (€ 189,00) compared to its precedessor.
At the end of the day, my opinion about hip-dac2 could be condensed in a simple one-liner: as good as hip-dac, so very good for this price point, with the addition of a higher MQA reconstruction quality.
As I never published an article about original hip-dac I will take this opportunity to deliver an extended article on the “hip-dac franchise”, so to call it. I will clearly mark the differences between hip-dac2 and hip-dac within the text. Let’s go through it.
Table of contents
At-a-glance Card
PROs | CONs |
Good power delivery on medium loads | Could use better current delivery vs low sensitivity loads |
Outstanding DAC quality in this product&price category | Unimpressive stage drawing |
Commendable balanced-output dynamic range | Dull single-ended output |
No power input from USB data line | Some hissing on low impedance, high sensitivity loads |
MQA Full Decoder (hip-dac2 only) | Warm-colored (might be not a con for some) |
Spectacular design (looks, haptics, construction) |
Product analysis
Key features and general description
hip-dac2 (like its precedessor hip-dac) is a battery-equipped slim-bodied easily pocketable USB DAC-AMP.
Size-, weight- and shape-wise it’s just wonderful. The full metal shell is sturdy, greatly pocketable, and at the same time superbly stylish from the shape and finishing points of view. It “pairs” very well with an average smartphone when used in conjunction with that.
Sole audio input is the USB data port. The input connector is the “usual” iFi USB-A recessed male plug. A USB-A(f) to USB-C and a USB-A(f) to USB-A(m) short cables are supplied free. No UBS-A(f) to micro-USB nor USB-A(f) to Apple Lightning are offered in the package.
No coax, optical nor analog input available. hip-dac2 (or hip-dac) can’t be used as a pure amplifier.
Two phone outputs are available: single ended (S-balanced, actually – more on this below) 3.5mm and balanced 4.4 mm.
No line-out analog output is available, which means that hip-dac2 (or hip-dac) can’t be used as a “pure DAC”, plugged into a downstream amp device. It still can be further amplified but the internal amp section will anyhow be involved as a “pre-amp”.
The internal battery cannot be charged via the digital input USB port. A separate charge-only USB-C port is dedicated to charging (a short USB-A to USB-C cable is included in the package). This is good as it cuts on much of the source-incoming noise typically carried by an active VBUS line. On the other hand it means that even when USB-connected to (say) a laptop the hip-dac2 / hip-dac will always only take power from its internal battery, and will eventually run out of juice.
Battery autonomy as always depends on usage (highres files and high volume listening consume more of course) but you can count on some good 6-7 hours of “common spec” listening. A full recharge takes like 3 hours.
When referring to similarly priced portable DAC-AMP devices, hip-dac2 / hip-dac’s power specifications are nominally impressively high vs high impedance loads (6.2V vs 600 ohm, just wow!) and a good step above average vs mid impedance loads (400mW vs 32 ohm).
iFi doesn’t table specs vs low impedance loads (< 16ohm) though, nor hip-dac2 / hip-dac’s output impedance on either of its phone out ports is declared.
Similar to what happens for most if not all of their devices, iFi offers a selection of easily user-installable firmware alternatives for Hip Dac 2 – ultimately yielding into alternative choices in terms of digital reconstruction filters.
Lastly, the device offers a manual High Gain button (labelled “Power Match”) and an XBass+ button. More on these later.
How does it sound: DAC performance
Considering hip-dac2 / hip-dac lack a proper Line Out, DAC performances are only partially assessable as some will be influenced by the integrated amp stage.
It is nevertheless quite evident that hip-dac2 / hip-dac’s voicing is very good when looking at pretty much any other similar portable DAC-AMP on this level of budget. Auditioned from its Balanced output port (more on why later) range is very well extended both towards the bass and the highs. Bass notes are well bodied, not particularly enhanced. Treble is smooth while more than nicely airy, and mids are quite evidently the best developed section.
There’s a quite evident warm tonality – difficult if not impossible to say which section (DAC and/or AMP) contributes to that most. But it’s there. If I have to compare with my experience with other iFi devices offering Line Out options (Nano iDSD Black Label, Micro iDSD Signature) I am ready to bet this is mostly AMP-related but again… it’s a guess.
Good DAC performance doesn’t come by chance. iFi adopts high-standard components even inside their budget products like hip-dac2 / hip-dac, and this is surely one good first step – but this often happens on many chi-fi devices, which on even or very similar “internal stuff list” condition in the end sound apparently much worse. The real key is engineering competence, really – and that can’t be so easily “cloned”.
One aspect: a fundamental requisite to obtain good performances from a DAC device is avoiding interferences on the incoming digital data. Not talking about human-audible interferences, of course. You might want to read this other article of mine to get a flavour of what I’m talking about. As already mentioned above, hip-dac2 / hip-dac don’t take power from the USB data cable, this way apriori cutting a lot in terms of noise “collection”.
Another aspect: unlike the overwhelming majority of the other budget mobile devices, hip-dac2 / hip-dac offer an analog volume control, not a digital one. The reason why this is way better for DAC performances is quite technical (check here for a good, reasonable vulgarly-explained article) but putting it very simply: digital volume controls act upon the digital stream before it reaches the DAC, and deliver a “integral” digital data to the DAC only at their end-scale position (so at “100% volume” position); intermediate volume levels are realized by applying attenuatin to the digital data which de facto corresponds to reducing their digital resolution.
An analog-volume device like hip-dac2 / hip-dac always feeds its DAC chip at full digital resolution, and attenuates the analog output aposteriori only. Why not every device has this ? Quite simply because analogue volume controls are more expensive to implement and more complicated to design 🙂
Firmware options
Like most if not all other iFi DAC devices, hip-dac2 / hip-dac can run a range of firmware variants, each offering different features or optimisations. Firmware packages and the apps required to flash them are freely available on iFi’s web site, here. The flashing process is really easy and straightforward, at least on Windows platform.
The 3 significant versions to choose from for hip-dac2 are:
Supports | Does not support | |
7.3 | Full MQA Decoder, DSD up to 256 on Windows, 128 on Mac, PCM up to 384KHz | DSD 512, PCM 768 KHz |
7.3c | iFi’s proprietary GTO filter, Full MQA Decoder, DSD up to 256 on Windows, 128 on Mac, PCM up to 384KHz | DSD 512, PCM 768 KHz |
7.3b | DSD up to 512 on Windows, PCM up to 768KHz | MQA |
For the original hip-dac a very similar option is available although it may be interesting to note here that there have been two hip-dac sub-versions, one tagged with serial numbers beginning with 54010 and the other with serial numbers beginning with 54040. The latter generation accepts the same 7-generation firmware packages as hip-dac2 (labelled respectively 7.2, 7.2c and 7.2b), while the former older generation accepts older versions of the same packages ( labelled respectively 5.3, 5.3c and 5.2).
DSD is a very interesting standard but I don’t de facto currently own nor plan to own music files sampled above DSD 256, so the two options which get my attention are 7.3 and 7.3c.
Their fundamental difference is one only but a significant one at that: with 7.3c iFi’s own GTO (Gibbs Transient Optimised) filter replaces Burr Brown’s native reconstruction filters.
I strongly recommend you read iFi’s whitepaper about why and how this may be technically desireable, or not.
The paper focuses on throughly illustrating GTO’s output features while leaving another important aspect in the background: with 7.3c hip-dac2 will systematically upsample all digital input coming from the USB port up to 32 bit / 384KHz resolution prior to feeding the DAC chips. For what I seem to have understood this is fundamentally required for the GTO filter itself to work as intended.
I already experienced iFi’s GTO implementation in conjunction with Micro iDSD Signature and Nano iDSD Black Label. Simply put: on Nano iDSD BL the GTO option “sounds worse” than the native ones – for my tastes at least. Oppositely, GTO performance on Micro iDSD Signature is very significant, offering important analog reconstruction improvements on redbook-standard (16bit / 44.1KHz) tracks compared to the non-GTO firmware option.
Very similar is my experience on hip-dac2 / hip-dac, and this is one of the few notable differences between the two generations.
hip-dac2 GTO implementation (fw 7.3c) offers a very good alternative option compared to non-GTO (fw 7.3).
Oppositely, when I tested this on a first-version (ser# 54010xxxxx) original hip-dac I got a very similar result as the one I got with the Nano iDSD BL: GTO firmware is basically not worth for me. I didn’t have an opportunity to test a latter-generation hip-dac (ser# 54040xxxxx).
MQA
This is quite evidently the most important aspect about which hip-dac2 represents a significant upgrade from hip-dac: MQA reconstruction performance is evidently better.
As you may or may not already know, MQA decoding is not all equal. It depends on what sw suite (license) is present on the involved playback device(s).
Even without “any” MQA license, MQA files stay compatible with “any” sw player application which will treat them as “normal” 16 bit – 44.1 / 48 KHz files. Their sound quality won’t be much different from that of an ordinary MP3 file though, which is logical considering MQA is a compressed and – when not fully unfolded – certainly lossy format.
Many sw player applications – first and foremost Tidal’s own player app, and many others – offer a first level of MQA de-flation treatment. In MQA jargon those apps are called “MQA Core Decoders”. An MQA Core Decoder enabled player will extract (“unfold”) a part of the so-called MQA origami.
The trick happens on the sw player itself (DAP, phone or PC), and the result is an uncompressed, “standard” digital file/stream which therefore can be fed to any existing DAC, even those which are totally extraneous to the MQA project. A license fee is typically required for that to happen on the player app – often purchaseable in form of an optional “plug in”.
As mentioned, a “MQA Core Decoder” only restores a portion of the higher resolution information hidden and folded into the MQA file. The result is a higher-than-redbook (up to 24bit / 96KHz) stream which once reconstructed into analog form by the DAC will be better than the “No-Decode” case, but still not “as good as it may get”.
To go beyond that, an MQA-licensed hardware DAC device is required. When the MQA software is “inside the DAC”, in facts, all of the high res information packed inside the compressed MQA track gets unpacked (“unfolded”) by the DAC device itself and the fully extended digital high resolution information is available to the DAC to do its reconstruction work upon at the best of its abilities.
Yet, MQA makes 2 different DAC-level licensing / implementations available for their software. They are called “MQA-Renderer” and “MQA-Full Decoder”.
The most common level is “MQA-Renderer”. When a DAC device is equipped with “MQA-Renderer” software, then it can pair with a “MQA Core Decoder” source player and complete the latter’s job, i.e., the “MQA-Renderer” DAC does the second part of the unfolding job on the digital file, prior to reconstructing the analog form.
iFi hip-dac (original model), xDSD Gryphon, Pro iDSD Signature are all examples of iFi MQA-Renderer devices.
The richest and most complete MQA DAC implementation level is the “MQA-Full Decoder”, which differs from the MQA-Renderer tier on three counts.
First: the Full Decoder takes care of the entire unfolding process, all of its stages that is, on the DAC device as opposed of leaving the first unfold done at the source player app level.
Second: the actual sw code used on each different DAC device is optimised to work in conjunction with that very chip and circuitry. Alternatively said: all MQA Renderer devices use pretty much the very same MQA sw code, while every different MQA Full Decoder device runs a slightly (or not so slightly) optimised version of the code, finetuned by the hw manufacturer working together with MQA people to fully exploit the specialties of that very piece of hardware.
Thirdly: while most people often focus on the folding / unfolding aspects of MQA’s game, indeed the MQA philosophy embraces a much wider horizon. In their intents they want to work with the music makers (the artists themselves) and their producers, collect their “original” digital masters as they are officially released by their studios, and apply a sort of “genuinity seal” onto their MQA-encoded version. At the opposite end of the distribution chain an MQA Full Decoder DAC will “reveal” wether such “genuinity seal” still is unaltered on the MQA-encoded track it is working upon.
You can think of this as a sort of responsibility / transparency mechanism: if the seal is there, then the MQA Full Decoder DAC device will light a LED of a certain color, signaling it has got certified access to an “original” copy of the digital track file; it therefore takes responsibility for restituting the exact sound information as they have been approved by the artist himself in their studio (a quite sharp claim, but it’s that).
If the seal is not there instead, then the MQA Full Decoder DAC will light the LED of a different color. It will still of course do its decoding job but the listener won’t have the “device’s endorsement” on wether what they are hearing is compliant to what originally was intended by the music creator.
Hip Dac 2, Diablo, Micro iDSD Signature (with latest firmware installed), ZEN DAC v2, Neo iDSD are all examples of MQA-Full Decoders
MQA royalties and consulting fees apart, as one may easily imagine different enabling hardware makes a big difference on such a computing intensive process as MQA unfolding. Newer generation iFi models (hip-dac2, Diablo, ZEN DAC v2 etc) carry a 16 core XMOS chip with a much higher capacity and computing power (2X the clock speed, 4X the internal memory, latest USB standards compliance) so – simply put – it can “do more at the same time” than the predecessor model.
The improvement in the audible result is quite evident, and totally in line with theory. When applied to MQA-authenticated tracks hip-dac2 reconstructs a much airier, defined and detailed sound compared to the job done by hip-dac as mere Renderer on the very same tracks.
On the other hand, though, I think it’s worthwhile here to remember that – like it or not – MQA is not any sort of magical way to make a DAC sound better then it technically could when applied to a non-MQA, full resolution version of the same track.
A very easy comparison example for me is with Apogee Groove. While of course hip-dac2 will reconstruct/reproduce an MQA-master track at a higher level of audible detail and resolution compared to what Groove will do when connected as a non-MQA DAC on the same track, on the other hand Groove’s range extension, dynamic range, bass and treble control stay on a superior level even in such an “handicap-started” race. Even more evident is the DAC reconstruction quality difference of course when applying hip-dac2 to a given MQA-authenticated track, and Groove to a high-res non-MQA version of the very same track.
Long story short, I guess it all boils down to a quite trivial conclusion: MQA is no magic wand, it’s got no “hardware upgrade power”. Of course.
How does it sound: AMP performance
Based on experience I stopped expecting that low budget devices offer similar amping quality results from both their single and balanced ended outputs. It fundamentally never happens.
The fact is that in these cases balanced amping architecture is primarily adopted as an inexpensive, easy-implementable way for many manufacturers to offer a decent or above-decent output quality (cleanness, transparency, dynamic range) off of apriori difficult situations such as small / ultrasmall and low price tier pocketable devices.
Clean amping is mostly dependent on high quality power management, and in a small and/or relatively inexpensive “box” there is little “room” (physical and virtual) to fit appropriate power management circuitry. Clean power is a challange on amps of any size, and a very steep one the smaller the form factor and the budget get.
As size & cost go up it starts to be possible to encounter devices e.g. the Micro iDSD Signature whereon Single Ended and Balanced phone outs present a power difference, but negligible quality differences. Below that size and budget, I just encountered white flys. Groove, to name one, which Single Ended output is a few times over cleaner, more transparent and dynamic-extended than any other Balanced-equipped device below $300 I happened to hear. Another good case is Sony NW-A55. I have a serious hard time naming a third.
From this point of view, hip-dac2 / hip-dac follow the mainstream. Do not expect wonders from their Single Ended outputs, as in facts you won’t get any. The other way around is rather true: hip-dac2 / hip-dac’s Single Ended output is unimpressive – dull, compressed, closed-in. This, in spite of the good deeds of their S-Balanced tech.
S-Balanced is the name of some iFi’s technology, short for “Single-ended compatible Balanced”. iFi also adopts it on a number of other devices too. Refer to their own whitepaper for a nice technical description.
Also, if you are not familiar with what TRS / TRRS means, this may help.
Simply put, a cabling scheme is put in place behind both phone ports on hip-dac2 (and original hip-dac) single ended port:
- When plugging TRS plugs – the port delivers “normal” single-ended output. All single ended drivers on the market will seemlessly work in there. In addition to that, thanks to how internal cabling is designed, they will also get 50% reduced crosstalk compared to what they would get from an ordinary single-edend port – for free.
- When plugging TRRS plugs – the port delivers full “balanced-ended” output to balanced-cabled drivers, resulting in quite apparently cleaner and more dynamic sound.
In hip-dac2 and hip-dac case of course the sole “useful” application is the former: hip-dac devices offer full-blown Balanced Ended output so there’s no practical point looking for a TRRS adapter to connect a balanced-cabled IEM/HP to the S-Balanced 3.5mm port instead of the more logical 4.4 mm choice.
Different story for the Balanced Ended 4.4mm output, which comes accross evidently airier, better bilaterally extended, with a very good level of control on bass and smooth trebles, and most of all a quite nice dynamic range and good microdynamic rendering. In a word, the solid impression is that on hip-dac2 / hip-dac BE out is the sole one with enough cleanness and transparency as to offer some justice to the preceding DAC stage.
As I already mentioned above, there’s a distinct warm coloration. Is this coming from the DAC or the AMP? Difficult to determine as hip-dac2 / hip-dac don’t offer a pure Line Out option, and thus a chance to use a third-party amp like it happens on other iFi models like Nano iDSD BL or Micro iDSD Signature is precluded.
A small difference can also be identified between hip-dac2 and hip-dac’s overall output quality, namely the former being a bit more sparkly in the highs, and just a whiff less intimate as far as soundstage goes.
I didn’t mention soundstage yet, which is definitely not a shiny aspect for hip-dac2 nor hip-dac. Quite narrow, really. Is this due to scarce spatial reconstruction skills at the DAC level or due to unclean AMPing? Again, impossible to say due to the lack of a Line Out option – and after all useless to know either, as it’s not something the user can do anything about.
Lastly, I think it’s worth noting that some hiss is picked by very sensitive loads (CA Andromeda, anyone? 🙂 ). While definitely an imperfection taken per se, I guess it should be conceded to hip-dac2 / hip-dac that it’s a very common one, almost irregardlessly of the device budget.
Extra features
There are two toggle-buttons beside hip-dac2 / hip-dac’s volume knob, named Power Match and XBass.
Power Match is nice attempt at a layman-friendly naming for a Gain switch. Activating Power Match puts hip-dac2 / hip-dac in High Gain mode, which is of course recommended (only) when a low-sensitivity driver is connected. Attention though: on low-sensitivity and low-impedance devices the suggestion is flipped – Low Gain is typically a much better option.
XBass behaves like what an EQ expert would call a low shelf positive filter. By ear it pushes lows up by 2dB-ish from 100Hz down. Might occasionally turn out to be handy to help some bass-shy drivers, or as a compromise to compensate for some drivers requiring a higher level of current delivery than what hip-dac2 / hip-dac can deliver to express their best on their bass lines.
Notable pairings
You find some significant pairing impressions reported in Kazi’s article, which I already mentioned above.
I find myself totally in line with what Kazi wrote when referring to final Sonorous-III and Dunu ZEN / ZEN Pro which I also had a chance to directly test with hip-dac2 and hip-dac. Ditto for my experience with a pair of high-impedance cans, which is HD600 in my case – ultimately showing that hip-dac2 / hip-dac’s nominal 6V @ 600ohm spec is less effective than it may seem when put to the real work.
Let me just add a few other experiences here.
final E3000
Biasing-wise the pair is technically good, insofar as hip-dac2 and hip-dac both definitely deliver enough current to E3000 to open them up properly, keeping their bass transients controlled and delivering a good sense of space. The unavoidable down side is that due to E3000’s fixed cable it’s impossible to exploit hip-dac2 / hip-dac’s best amping output (the Balanced one) so the forced-single-ended pair is bound to unavoidably suffer from some dullness and lack of dynamics.
final E5000
Even when paired on the Balanced output hip-dac2 and hip-dac don’t seem to deliver enough current soon enough to brighten-up E5000’s bass line. The result is an overly thick presentation which is what very commonly one gets on E5000 from budget-tier sources.
Ikko OH-1S and Tanchjim Oxygen
Although different, the two IEMs react very similarly to hip-dac2 / hip-dac pair. Both get turned on very nicely by the balanced output, delivering much of their competence in terms of technicality. Both get “warmed up” by hip-dac2 / hip-dac’s coloration, which may be a welcome variation to many in comparison to their otherwise slightly-bright/neutral tonality. Hip-dac2 pushes both’s highmids up, luckily without passing the glare limit. Nice ones.
final Sonorous-II
Similarly to what happens on Sonorous-III, the pair has lights and shadows. Good is bass (a case where the XBass switch delivers a pleasant alternative at the user’s fingertip), and microdynamics. Less good is high-mids which get a bit too hot.
Shure SRH-1840
This is a really good pair. Power available on Low Gain is already more than enough to make SRH-1840 sing pretty well, and there’s no overdoing on the high-mids. Some treble extension is lacking. General warmth may be considered bearable in this case due to the fundamental pure neutrality of the phones taken on their own. Too bad for the narrow stage, but at this price level I’ve yet to find a better pair for SRH-1840 if I exclude Groove.
Notable comparisons
Again, some notable comparisons are already mentioned on Kazi’s article, which I once again encourage you to read. I do share his opinion about hip-dac2 vs hip-dac entirely.
vs Apogee Groove
The comparison is apriori dishomogeneous as Groove is a high-power-demanding dongle with a unique, not-general-purpose amping architecture while hip-dac2 and hip-dac are designed with full-horizontal applicability in mind. Performance differences found between the two devices should be put in the correct perspective.
That said, Groove’s DAC and AMP refinements, where applicable, are significantly better compared to hip-dac2 / hip-dac.
Hip-dac2’s DAC reconstruction prowess does challenge Groove’s resolving power exclusively when applied to MQA-authenticated tracks. On such very tracks, hip-dac2’s Full Decoder capabilities deliver superior resolution and air, while on the other hand still falling short vs Groove on range extension, bass control and treble vividness. On non-MQA material there’s no game instead.
vs Hidizs S9 Pro
Another dishomogeneous therefore “unfair” comparison, which I’m mentioning basically only due to S9 Pro’s popularity. Similarly to Groove, Hidizs S9 Pro is a battery-less dongle featuring a high host-power demand. Different from Groove, it carries a general-purpose amping architecture free from apriori pairing limitations.
Like hip-dac2 / hip-dac, S9 Pro also comes with dual phone outputs (Single Ended and Balanced Ended), and again similarly in both cases the Single Ended option, well, might also be omitted, for how underwhelming they are compared to their Balanced Ended alternatives.
That said, the sound quality difference between the two devices is nothing short of dramatic. Hip-dac2 / hip-dac are better resolving, have better estension, better dynamics and better features. Last but not least, when connected to a “noisy” host (e.g a laptop) S9 Pro degrades its cleanness and spatial reconstruction performance quite evidently, and benefits of a noise filter adoption (e.g. an iFi iSilencer or an AudioQuest Jitter Bug), while hip-dac2 / hip-dac is much more resilient off the bat.
S9 Pro costs 35% less than hip-dac2, that must be noted, too.
vs iFi Nano iDSD Black Label
Similarly priced and after all not so differently-sized, the two devices do behave similarly.
Overall, hip-dac2 comes out ahead when used as a complete (DAC+AMP) system, even more so if applied to MQA material as Nano iDSD Black label is a mere Renderer not a Full Decoder. On the flip side, Nano iDSD Black Label offers a pure Line Out option which is the big “missing bit” from hip-dac2 / hip-dac, which allows the user to “upgrade” the device with an external amp – possibly a desktop one? – and fully exploit the really nice quality of its internal DAC.
Considerations & conclusions
Hip-dac2 is an outstanding device, quite evidently the best sub-200$ battery powered pocketable DAC-AMP on the market today. It delivers very good DAC reconstruction capabilities, significant amping power, and remarkable cleannes, dynamics and air from its Balanced Ended headphone output.
Compared to its preceding version, hip-dac2 offers MQA Full Decoding which represent a solid further improvement for Tidal fans. Apart from that, its features are identical and its sound quality are so close to the preceding version that a current hip-dac owner may safely hold on to his existing investment in case Tidal Master is not his streaming service of choice.
Disclaimer
As always, a big thank you to iFi for the continued opportunity they offer me to keep assessing their products.
Hi the first thing i did after checking the ifi zen dac 2 is working good is peel off the white sticker with the qr code.\ because it looked so ugly.
I scanned it but my phonecamera said it was a nonexisting website.
I did not realise that was the serialnmr of the device.
So i cant update the firmware…
Is there an other way to see the serialnmr?
if no is there a workarround it?
grts Nico
In your case there is no way I know of to reveal your particular serial# other than asking to your vendor – possibly they do take a note of the serial numbers of the devices they sell and pair them with the invoice numbers.
That said, there is a workaround to get past the barrier and get to the download page and it’s to use add “00000” to the product family root figure, which is revealed by the ifi website.
In the case of the Hip Dac 2, the website tells you “Your Hip DAC 2 serial number should start with: 54060”
If at this point you write “5406000000” in the text field there below, it will let you through.
Ciao
Alberto