Skip to main content

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: No 24-bit recording with generic Realtek HDA Windows drivers? (Read 10941 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

No 24-bit recording with generic Realtek HDA Windows drivers?

I finally got my new Soundblaster Audigy FX to work today (hint: restart Windows). Funnily enough, it looks like the CA0113 is a fairly standard HD Audio controller IC, which they are then using to operate a Realtek ALC898. The latter was picked up just fine by the generic Realtek HDA driver already installed (R2.71, so fairly current still). Oddly enough, the recording side showed sample rate choices up to 192 kHz (it's a Vista machine), but all 16 bit! Huh?

Now I do know that this card does 24-bit recording, as the Creative driver installed thereafter offers 24/44 to 24/96. (BTW, 24/44 actually seems to give 16/44 only, while the playback side works fine. Smells like a driver bug.) And not too badly, actually, as I can get almost 103 dB(A) of dynamic range in an RMAA loopback test. (*)

Another case: A notebook with ALC269. Granted, its ADC is not quite as fancy at a nominal 98 dB(A) SNR, but still the product page for the chip claims "4-channel ADC supports 16/20/24-bit PCM format for independent two stereo channel audio inputs". Again, the latest R2.73 driver directly from Realtek (Win7 x64) does not offer 24-bit recording, and all I ever get in a loopback test is 87-ish dB(A) (up to 90 with a headphone amp in between and at 96 kHz). I think this is quite odd. ADC readout is not something that would be expected to vary by OEM (it's given by the chip), so I'd say 24-bit recording could be implemented by the generic driver quite safely assuming the chip can do it.

I have looked at a few of the massive INF files, hoping to possibly be able to hack them a bit, but unfortunately this stuff is way over my head (I'm not a Windows driver guru). Apparently they're using one or more WAVEFORMATEXTENSIBLE type structs in a binary representation, or something along these lines. Ouch.

Any thoughts?

BTW, for 24/44 playback to work properly, you need a halfway recent version of the Realtek drivers. Build 6263 from late 2011 definitely is too old, build 6873 from R2.71 is fine.

*) THD is OK though it goes a bit higher than I'd like to see at low frequencies and high levels (2nd/3rd are hitting the mid-80s at 60 Hz). Periodic passband ripple in loopback is about 0.05 dB p-p, OK but worse than the ALC898 spec (0.02 dB rec, 0.005 dB or somesuch pb). Looks like the front out does about 2 Vrms and the others 3-6 dB less, but I'll have to ask the multimeter. R_out,front = 22 ohm, both at rear and front panel. (Apparently uses MAX97220A. NJM4558s on rear/center/sub.)

No 24-bit recording with generic Realtek HDA Windows drivers?

Reply #1
Periodic passband ripple in loopback is about 0.05 dB p-p, OK but worse than the ALC898 spec (0.02 dB rec, 0.005 dB or somesuch pb).


Perhaps there is sample rate conversion in the drivers (that is, assuming that the ripple should really be 0.02 dB for a loopback, I did not find it in the datasheet, and it is often worse for recording than for playback) ? It is sometimes difficult to avoid on WIndows.

And not too badly, actually, as I can get almost 103 dB(A) of dynamic range in an RMAA loopback test. (*)


Which in reality is less by 2-2.5 dB, because RMAA does not calculate the A-weighting correctly (I guess it does not normalize the response of the filter to 0 dB at 1 kHz).

*) THD is OK though it goes a bit higher than I'd like to see at low frequencies and high levels (2nd/3rd are hitting the mid-80s at 60 Hz).


That is probably caused by too small and/or low quality DC blocking capacitors. Is there also some bass roll-off ?

Looks like the front out does about 2 Vrms and the others 3-6 dB less, but I'll have to ask the multimeter.


It is most likely the 1.2 Vrms output level of the Realtek codec on the other channels, with the 4558 buffers having unity gain.

No 24-bit recording with generic Realtek HDA Windows drivers?

Reply #2
Greetings all. First post - I hope is of some use.

Have recently been able to test the audio on a Gigabyte motherboard (cannot remember the exact model but a current 8 series chipset / socket 1150 / Haswell) with a Realtek ALC887/888 audio chip. The two stereo ADC's are specced for 16/24bit 44.1/48/96kHz. Using the Gigabyte/Realtek driver I could only see 16bit recording as an option but reverting to the generic windows driver allows 24bit recording - but on one of the two stereo ADC's only. Whether the other ADC could still do 16bit alongside 24bit on the other I did not test nor did I confirm that the 24bit input was actually 24bit. I was not impressed.

An email to Gigabyte had them initially avoiding my question and treating me like an idiot (they linked to a paragraph on wikipedia explaining what an ADC is  - thanks for that!) When I pressed them further they confirmed that playback is 24bit and recording is 16bit (which I guess is their way of confirming that recording is not possible at 24bit despite what the advertised specs say). I had also asked whether a different model of board using a different Realtek chip could do 24bit on both ADCs and unfortunately it seems to be a problem with all Realtek chips and/or drivers (though they would not admit it). Maybe it's a driver problem that will get fixed but it seems to be quite a poor situation as it stands. I realise that traditionally consumers have been led to not expect much from onboard sound, but can anyone imagine a manfacturer getting away with a graphics solution that offered 1920x1200 but actually only does 1024x768!

Another Gigabyte 7 series board I have used with a VIA VT2021 audio chip does not have this problem and can indeed do 24bit recording on both stereo ADCs simultaneously (4 mono channels). Unfortunately Gigabyte do not offer any 8 series boards with VIA chips and I could not find another company that does either. Anyone interested in this capability may wish to note that there is an offset of about 8 samples (at 96kHz) between the two ADCs on the VIA chip (around 50 samples on the Realtek chip) but this does fortunately remain consistent so long as the audio driver is not interupted (for example by focusing windows from one audio program to another) and can therefore be manually compensated for in your recording software should you require sample accurate allignment across all channels. It's still a shame that the ADCs are not alligned on the same sample since all five DACs are alligned.

I also think it's a real shame that a built-into-motherboard solution for 4 channel recording at 24/96 cannot be sorted out properly. The quality may not compare with a pro soundcard but many home recordists may not notice or care anyway. Consider that the cheapest dedicated 4 channel recording unit costs about the same as a decent PC motherboard and you have to remind yourself of the value for money aspect that PC component buyers often take for granted. I realise it's the same with dedicated graphics cards and all about economics of scale and consumer expectations, but still - get your act together Realtek!

No 24-bit recording with generic Realtek HDA Windows drivers?

Reply #3
I had also asked whether a different model of board using a different Realtek chip could do 24bit on both ADCs and unfortunately it seems to be a problem with all Realtek chips and/or drivers (though they would not admit it).


It is probably a limitation of the Windows drivers. I could successfully record in 32-bit PCM format from an ALC887 on Linux (using the ALSA 'hw' interface to bypass software processing), and the resulting WAV file indeed seems to contain audio data in the 24 most significant bits of each sample. The 24-bit recording is of limited practical usefulness, though, since I measured an A-weighted noise floor of -92 and -90 dBFS on the left and right channel, respectively; it is only a few tenths of a dB improvement over recording in 16-bit format.

No 24-bit recording with generic Realtek HDA Windows drivers?

Reply #4
True that onboard 24bit recording is likely to offer only a marginal increase in useful dynamic range. I think the real advantage is practical rather than sonic in that you can just relax a bit when setting recording levels - if you don't mind trading noise for resolution.

Thinking about this now I may have read somewhere it was a Windows 7 problem (should have mentioned I was using W7). But it's actually the windows driver that allows 24bit on one of the two ADC and the Realtek driver that doesn't allow on either. I'm afraid I'm no expert on the mysteries of hardware/OS interactions.

It may be interesting if anyone is able to confirm 24bit input on both stereo ADC simultaneously for any Realtek chip on either a 7 series motherboard (or earlier) using Windows XP or Windows 8, or on an 8 series board with Windows 8 (no support for good ol' XP on the latest boards!). ASIO4All may be useful in picking up separate audio devices if the two ADCs have to be handled in that way - though I don't see why that should be the case.

No 24-bit recording with generic Realtek HDA Windows drivers?

Reply #5
I would be happy if I could do 24-bit on one ADC at a time. It bumps up loopback DR by about 13 dB in this case. (It'll be kinda pointless with the likes of an ALC892 and its 90 dB(A) ADC, that's true, but I prefer to be limited by analog rather than quantization noise.)

Anyway, measurements time on the Audigy FX:
Rear front output: 2.2 Vrms
Front headphone output: 1.0 Vrms
That's using WDM KS (Audacity daily build), and 3 dB less with WASAPI / DSound / MME.
The difference between Rear front --> line-in to Front headphone --> mic in was about 2.5 dB, vs. 6+ dB of output level difference, so I guess the mic-ins are more sensitive (or rather the line-in has some negative input gain in order to be able to take >2 Vrms).

All of the outputs in this card should be DC-coupled, as in previous Creative Labs cards. They don't like spending money on big coupling caps, so they're rather using output buffers (which are needed anyway) preceded by smallish caps.

Oh, and I looked at ALC898 digital filter specs again. Passband ripple is given as +/-0.03 dB for the ADC and +/-0.0005 dB for the DAC. +/-0.03 dB is 0.06 dB p-p, so what I'm seeing is all ADC filter ripple, which then would be within spec.

 

No 24-bit recording with generic Realtek HDA Windows drivers?

Reply #6
PS: Interesting. Switched speaker config from 5.1 to stereo, and now I'm getting 2.0 Vrms on the front headphone out (Audacity WDM-KS / Foobar WASAPI exclusive). Nonetheless, hiss into rather sensitive 16 ohm in-ears remains well-controlled (the 22 ohm output resistance helps, though it rules this card out for truly finicky cans). Subjective SQ is great btw, my HD580s aren't complaining.

No 24-bit recording with generic Realtek HDA Windows drivers?

Reply #7
PPS: Audigy FX rec level handling at the onset of clipping.
Line-in: Approx 2.0 Vrms
Front Mic: Approx 1.2 Vrms (presumably wired through to ALC898)
Each @~-0.9 dBFS.

Mic Noise:
Input-referred white noise approx. 110 dB below FS (~4-5 µV @20k) @+30 dB preamp, plus a fair bit of 1/f. Bias voltage (3.2V/2k2) may contribute up to 20 dB more up to several kHz @input open, practical results depending on mic impedance. Even the high-frequency white noise part is about 3 dB quieter in L-R when compared to L+R, presumably power supply noise.
Practical results with a cheap old dynamic mic (Vivanco DM22, -54 dB/Pa, 600 ohms, TS plug) are decent, I can get up to 60-ish dB of SNR when talking/singing directly into mic @+20 dB. For a noise floor lower than 55-60 dB SPL, get a mic preamp or a condenser mic. Telephony ought to be just fine as-is.
ALC269 (@Lifebook S761 vPro) is about 6 dB noisier, which subjectively makes the difference between a voice recording that is virtually noise-free and one with audible hiss.

Re: No 24-bit recording with generic Realtek HDA Windows drivers?

Reply #8
Have same problem with ALC892 and looks like we have solution, hope someone help this.
From datasheet for ALC892:
Quote
Two stereo ADCs support 16/20/24-bit PCM format, multiple stereo recording
All ADCs supports 44.1k/48k/96k/192kHz sample rate

1. Update drivers, latest R282 have some issues, like not working manager in Win7.
Check tenforums\mydigitallife\station-drivers for latest [modded\Dolby-enabled] drivers. here some links:
Spoiler (click to show/hide)
2. Enable 24-bit recording with Realtek internal tool, thx for alanfox2000
Download AudioDevice_sky.rar(Recomend)\RTKDeviceConfig.rar here -
https://www.mediafire.com/folder/a0tmggmy9j864/Realtek_Tool
https://www.techpowerup.com/forums/threads/realtek-modded-audio-driver-for-windows-7-8-1-10.250915/post-4126739
enable and save rtkhdaud.dat with this parametrs:
DrvCtrl20> Vista24bitRecord;
DrvCtrl44> Support24bitRecOnIntMic;
After reboot in Windows7 Sound config now possible choose 24bit for line-in and front-panel mic.

Re: No 24-bit recording with generic Realtek HDA Windows drivers?

Reply #9
Realtek Driver Mods - I have an ALC892 board and Asus stopped supporting it long ago so most of  the native features were not functional or crippled.  Alan has made major progress and most if not all Realtek hardware and the branded suites are rolled into a signed driver package now. Its pretty awesome. I can play music/movies out SPDIF to my receiver in DTS and out game sounds to a headset via front panel. I don't use the suites or do recording but it seems to have added all available range.  Just a heads up for people with older motherboards who want great audio.