Russ

Could someone who is familair with kX perform some simple tests for me?

I think there my be an issue with my card, and I would like to verify how things work for other people to be sure. (I want to make sure I am not giving people bad instructions when helping them (although noone has said anything), based on how it is working with my card (if there is a bug with my card, etc)).

I will posts the steps using Line In, but any AC97 source should work the same.

1:
Clear the DSP, and load epilog and then prolog (prolog should not be connected to epilog in any way), and a peak plugin (connected to the AC97 pins of prolog).
Select Line In as the AC97 source on the AC97 page of kxmixer.
Unmute the 'AC97 Rec Level' slider (Recording page) and set the level to 0 dB.
The 'Master Recording Level' slider (Recording page) should be at 100%, and the 'AC97 In' slider (Ins and Outs page) should be muted, as well as the 'Line In' slider (AC97 page). The 'Analog Recording Gain' slider (AC97 page) should not be muted (the level does not matter).
Play some audio through Line In, and see if you are able to record it using MME (i.e. using Sound Recorder or similair. Obviously you will not be able to hear the recording, as there are no connections to the speakers, but you shoud see something (other than a flat line) on Sound Recorder's GUI while recording and during playback, if it is recording the signal).

2:
Same settings, except this time mute the 'AC97 Rec Level' slider (Recording page), and unmute the 'AC97 In' slider (Ins and Outs page), and set level to 100%.
Play audio though Line In and see if you can see the signal on the peak plugin.

3:
Same settings as 2 (if you could see the signal, otherwise skip this test).
Play some audio through Line In (should again be able to see the signal on the peak meter), and then mute (while the audio is still playing) the 'Analog Recording Gain' slider (AC97 page), and see if you still see anything on the peak plugin after doing so.

Post your results.

With my card, when I mute the 'Analog Recording Gain' slider, I still see a signal on the peak meter (a DC signal (i.e. peak meter looks frozen, but does show a signal)). I am guessing that other people do not get this behavior. This possible bug, in addition to the steps outlined in the guide for setting up the MIC (which for my card would not be right (i.e. I cannot think of any reason why I would have have the 'AC97 In', 'AC97 Recording Level', and 'Line In' sliders all unmuted at the same time, based on the observed behavior with my card), are making me think that maybe there are some more issues with my card then I was aware of, but I would like to be sure.

Thanks,
-Russ

Lex Nahumury

I don't think your card is broken perse. Some thoughts:
It's might be a matter of (incorrectly) asserting your AC97 codec's registers.
Do you have it's spec sheet?

Also IIRC, there is some ac97 input buffer on 10kx.
(review 8010.h)

btw; is the DC level always the same, or kind of 'last sample'.

Russ

It seems to be the last sample.
I do have the spec sheet for my AC97 codec, and muting Analog Recoding Gain, does mute the Record Gain (AC97) register like it is supposed to do. I am just not sure what else it is supposed to do, in order to clear the last value of the AC97 inputs (and I am not sure if whatever is causing this, is affecting anything else).

I will take a look at 8010.h, and do some tests with those registers as well.


BTW: Regarding the tests:

Test1: I can record with those settings.
Test2: I can see the signal on the peak meter.
Test3: I do see a signal on the peak meter (DC signal (see above)).

Please let me know if anyone gets different results (especially with Test 1 or 2).

Maddogg6

Test 1: No level on peak meter - and NO recording.

test 2: No level on peak meter and No recording

Test 3: No level on peak meter

Same results on both my cards.

Russ

Thanks for testing.
I was afraid of that. If that is how it is for most people, then it means that I have been giving the wrong instructions to people for some recording setups, becuase it works differently for me.

BTW: You did not have the Analog Recording Gain slider muted for Tests 1 and 2 right? (I forget to mention that it should not be muted for those tests (the level of that slider does not matter, as long as it is not muted)).

It would also be good if some other SB022x users (with STAC9758), could let me know how it works for them, and/or anyone that gets different results from Maddogg6.

Maddogg6

No problem, I played with the settings abit to try to get something show on the peak - the ONLY thing that made a difference was un-muting linein & adjust slider in the AC97 screen. But NEVER seen any signal going into winrecorder untill I connect something to the WinMM record input of Epilog.

Russ

That part is strange, because I can do it without the need for prolog being loaded at all, because the same AC97 inputs are in epilog.

i.e.
From prolog source code:

macs in0, 0x0, [color=PaleGreen]KX_IN(0x0)[/color], in0vol;
macs in1, 0x0, [color=PaleGreen]KX_IN(0x1)[/color], in1vol;
...

From epilog source code:

...
macs tmpl, RecL, [color=PaleGreen]KX_IN(0x0)[/color], RecIn0;
...
macs tmpr, RecR, [color=PaleGreen]KX_IN(0x1)[/color], RecIn1;
...

From the above, you can see that the same KX_IN's (for AC97) are in both prolog and epilog, so prolog should not need to be connected to epilog at all. Is your epilog source code different? prolog?

As for the rest, I cannot be certain about how it is supposed to work, I only know how it works with my card, and somewhat how it works with your card (based on what you have told me).

<edit>
I would double check your results, as under the default DSP config, prolog is not connected to RecL/RecR of epilog either (xrouting sends it to Stereo_L/Stereo_R).

dj_stick

Russ - IIRC the dsp has a preset microcode that will route ins to outs when the dsp is cleared and prior to the kX mixer loading (this is how you here the windows start-up sound)

perhaps that could explain some of your findings

__________________

Russ

I do not know. I get the same results with different configurations.
Also, going from the microcode of the plugins, the way it works for me, seems like it is the way it is supposed to work, that is why I never thought it was an issue. The only thing that seemed off, was the results of Test 3. So unless it uses completely different source code for you guys, I do not know what to think.

Doug W

Well I tried your experiment with my SB0090 and 3534f

1. In this situation it did record what I fed into but when I connected a FX BUS 0/1 to listen to the recording, it was fluttery and weird but definitely recorded.

2. I saw the signal in the peak plugin.

3. No recording

Russ

Could the strange recording in step 1, be because the signal level was too low (or too high)?
i.e. Without being able to confirm the level with a peak meter, it can be hard to know what level Gain setting to use, etc. The main thing was to see if it recorded.

The 3rd test was just to see if you still saw the signal in the peak meter, after muting the Analog Recording Gain, I did not expect you to be able to record with those settings.

So it seems we having conflicting results (aside from how it works with my own card). I guess we will have to see if any other people test it, to see what's up.

Thanks for testing.

BTW: Do you have the Analog Recording Gain slider in 3534f? (I do not (maybe because it does not recognize my model correctly). But I did get the same results for the other tests in 3534f).

drgrozozo

same results as you here russ

1-records
2-peak signal shows
3-peak signal shows but with less value, not completely frozen

weird

Russ

Here is an example of what Test 3 looks like for me (after muting the Analog Recording Gain).

Here are the relevant microcode dumps (at that time):

prolog

[8000]: 'in0' p=0; translated: 125; current=32c4000
[8001]: 'in1' p=0; translated: 126; current=8c99800
[8002]: 'in0vol' p=20000000; translated: 127; current=20000000
[8003]: 'in1vol' p=20000000; translated: 128; current=20000000

peak

[4000]: 'peak_l' p=0; translated: 125; current=32c4000
[4001]: 'peak_r' p=0; translated: 126; current=8c99800
[8000]: 'result_l' p=0; translated: 141; current=6a588000
[8001]: 'result_r' p=0; translated: 142; current=7064cc00
[8002]: 'resolution' p=1f; translated: 143; current=1f
[8003]: 'tmp' p=0; translated: 144; current=7064cc00

The values and the resulting picture would be different everytime, but it stays fixed, once the Analog Recording Gain is muted.

I am beginning to think that maybe Maddogg6's results (in Test 1 and 2) were a fluke?

@Maddogg6
Is it possible that you missed a step, or that it was just acting up at the time that you were testing? (My results are consistant). That would be good, as it would mean that it works like I always thought it did (and thus my instructions to other users were not wrong), and the only issue is this one (Test 3).

@drgrozozo
What do you think you are seeing in Test 3?
Do you think it is the same signal only lower (but still active), or maybe the noise floor or what? Any idea?

I am still not completely sure what it is supposed to do in Test 3. I suspect that I should see nothing on the peak meter.

drgrozozo

no, just rechecked it, it's the noise

to be honest you've just baffled me with this. what is this? a bug? or a normal behaviour? in which case i've got the whole routing concept wrong.

Russ

It is hard to say, until I know for sure what it is supposed to do in Test 3.

BTW: Which card model do you have?

Thanks for testing.

drgrozozo

i've tested this on live! value (ct4832).

Russ

Thanks for the info.

Tril

I have a CT4832 (SBLive!)

1. I record something.
2. I see the signal on the Peak plugin.
3. I see a frozen value at -75 dB.



I always thought that number 1 is the normal behavior. On my card, the sliders on the recording page bypass the connections on the DSP.

I looked at the code of epilog with AC97 from the recording page muted and at 100 and it changes something. It changes the value of RecIn0 and RecIn1.

The number 2 is normal too. The AC97 slider on the In and Out page changes the in0vol and in0vol registers in the prolog plugin.

The number 3, I don't know. I never muted that slider before.

Russ

Thanks Tril.
I thought that was normal behavior as well (seems logical from the microcode).
For test 3, is it always -75 dB, or does it depend on what the signal level was before muting?

Tril

It moves slightly around -75.

I think that -75 is the limit set by the noise on the AC97 input. If I connect a Peak to "AC97", raise the volume to 100 on the "Ins 'n' Outs" page, close my source and leave "Analog Recording Gain" unmuted, I read -75 dB.

From this, it looks like that muting the "Analog Recording Gain" does disconnect the source from AC97 but it does not deconnect AC97 from the prolog input. So you get the AC97 noise in the DSP.

Russ

Ok, but when you tested it, you left the source playing when you muted it, right?
With mine it seems to act similair, but not exactly the same. It seems to disconnect the AC97 source, but retains it's last value.

I going to have to reinsall the CL drivers on my other drive, and do some testing with them to try to determine what it is supposed to do.

Tril

Yes. I recorded that -75 dB signal and I amplified it to listen to it. It's white noise.

Russ

I did some quick testing with the CL driver (although it is a little more difficult to do with the CL driver, than with kX) and here is what I found.

First off, it does not seem to allow control of the Record Gain at all, with my card. I could not find anywhere to change that. It seems to be fixed at 0 dB.

If I do it manually, it acts similair to kX. It seems to freeze the output (although at a much lower level than kX). So it seems that this may be somewhat normal behavior (although it would not normally happen with the CL driver, becuase it does not let you do it). Increasing the gain does not seem to cause any problems, so I guess the answer is to just not use the mute button. If you want to mute the input entirely, use another method.

So, all in all, I think my results are somewhat normal for all the tests (I just do not understand Maddogg6's results?). Thanks again to everyone that did the testing.

-Russ

Russ

So that is probably just a matter of different codecs doing it a little differently (or maybe my card is flaking out)? Either way, it does not seem to be a kX issue.

drgrozozo

well i for one would like for someone to explain why this is normal.
why would setting recording levels ignore dsp routing and the fact that line in is muted in test 1?
and why would muting recording gain mute line in in test 3?
i don't find this normal, and it doesn't sound like a chip feature to me...

miguel

drgrozozo, this is what I called back then an "inconsistency" with the DSP state (the looks of it at least), but only if you use kX default prolog, epilog, xrouting, etc. Use Lex's (great) ProFX and it all will just behave.

Russ I tried this the other day with a 4830 and 1 and 2 worked (could record, saw normal activity in peak) but since I had no analog gain slider I thought my test would not help.

drgrozozo

yeah, problem is profx ain't controllable outside of dsp afaik, which makes it impractical for some of the usages

Lex Nahumury

Russ,

1. So far these 'test results' are pretty useless
as long as there is no one doing the test with
the *same* card type (+ the same codec) you have.

2. I fail to see the need for those 3 tests and
all the 'voodoo talk' about epilog, prolog microcode,
winmme DSP recording paths etc. etc.

Correct me if I'm wrong, but this issue is about;

Setting bit15 of AC97 Record Gain Register(1Ch)
to mute the analog input record gain
causes abnormal behaviour on your card.

So the only relevant test is:
- clear dsp
- load prolog
- load Peak and connect it to prolog ac97 L&R
- Ins&Out page:
unmute ac97,
set it's slider open
- AC97 page:
Reset ac97!!!
select LineIn (in dropdown box)
- Apply signal to LineIn
- Toggle AC97 Rec Gain mute switch,
and watch Peak for abnormal DC.

Or the easy way:
- clear DSP
- load ProFx3.0x ADC
- optionaly load Peak(for accurate readout) and connect to ADC
- select LineIn source
- apply signal to lineIn
- toggle Rec Mute switch and watch Peak for abnormal DC.


4. My best guess that this is entirly codec related.
Either the codec is 'broken' (faulty batch from factory??),
it might be a typical sb022x hardware issue,
or it simply needs additional register settings/sequence
to mute properly.
Note that analog "Mute" in this context doesn't
nesecceraly means INF dB.

If you want to investigate this issue I strongly suggest
you write a simple test plugin similar to ADC
in which you access/set AC97 registers directly,
bypassing/overriding kxmixer.exe.
(a simple console app would also work ofcourse)

5. I couldn't find a *full* STAC9758 datasheet on the net.
Plz mail it to me if you have it and/or want me to have a
look at the issue.


/Lex.

Russ

@Lex
I did test with the AC97 registers directly (how else could I test with the CL drivers, when they do not have a control for Record Gain, etc?).

The additional tests were to be sure I understood how kX works (i.e. the purpose of the individual sliders in kxmixer), so I could be sure that I wasn't giving people bad instructions when offering help (setting up the MIC for Teamspeak, etc.). i.e. If kX was not working right with my card, then I could not be sure that how it worked for me was correct, etc. (and the guide for setting up the MIC, has a step or two that did not make sense to me, which also made me wonder about it). The additional tests showed that the way it works for me, is most likely correct, and thus my understanding of the purpose of those sliders is correct. The thing about the microcode was only to affirm that the microcode is used in a way that makes sense with my understanding on what those sliders are for.

That only left the issue with Test3. I tested and determined that it works basically the same with the CL drivers with a different OS all together (WinXP and WinME). This shows me that the issue is not kX related, which is what I wanted to know. Other people's test results also showed me that the results may be different for different codecs, which is useful information to me.

As for the codec datasheet, it used to be available for download, but was removed (I am not sure why), but I do have it, and can send it to you if you want to take a look at it.

Russ

It is not ignoring DSP routing, nor the levels.
It is just that both prolog and epilog contain the same hardware inputs (the extra routing (and controls) is hidden inside of epilog, so it is not obvious), so you can use either one, depending on which sliders are active in kxmixer.

<edit>
The sliders are located after the inputs in the DSP (or plugin). They do not mute the hardware inputs themselves, rather they mute the plugin's output (i.e. they modify one of the pathways for the signal (or modify the signal itself at that point only), within the DSP (or plugin)). If you include the same hardware inputs within another plugin, than you have a seperate pathway for that signal (i.e. ProFX:ADC is not effected by either the 'AC97 In' slider, or the 'AC97 Rec Level' slider. It creates it's own pathway for the signal within the DSP).
</edit>