Support and General Use > Recording

Recording Enhancements Pack

<< < (93/189) > >>

BayTaper:
Should I put this in the AGC thread? Anyway...

First of all, thanks guys for all of your hard work. I'm a taper who uses a couple of different recorders, but I'm in the process right now of transitioning one of my recorders from a JB3 to an H120. I'll be just going analog line in for most of my use (not digital, use something else for that). Anyway, I've got Rockbox plus REP installed and working fine it seems no problem, and while I am not a fan of AGC in general for the kinds of things I tape (music), I think the "Safety" AGC mode is a stealther's dream, and it's so easy to turn it off if you don't even want that. So, a big thanks goes out to all of the developers that have turned the H120 into a much greater thing than it was out of the box.

Now to my main question... I'm trying to understand how the gain works on the H120 with RB. From what I've read, I think it just uses the analog gain to take a big step up and then the decimator to shave some off db digitally so it increments in steps of .5 db and in a way that uses as much analog gain as it can while only using digital gain to lower the level back for .5 steps. So, from what I've seen when the gain goes negative it is simply a digital scaling down of the signal. Do I have that right?

If so, then when you're clipping, and you reduce the gain into "negative"  territory in the UI, you'll still have distorted audio with flat waveforms at a reduced db level. I totally understand that there's nothing the H120 can do about this situation because it doesn't have true analog "pads" or anything like that. But, from the AGC's "Safety" mode, I'm thinking a slightly better algorithm might be to "safety it down to 0 gain and then stop". Now maybe I've got this wrong, but my thinking from a recordist's perspective is once I'm clipping, I'm pretty screwed, and all lowering the gain down past zero into the negative gain territory is just bringing my noise floor up at that point. My thinking goes that, hey you're clipping, that data is lost forever, so it doesn't really matter to lower the gain digitally, it's not doing me any good past zero is it? Now I can see where as long as you aren't clipping, you could use negative gain to simply lower the signal in the recorded file, but why would you want to do that if you weren't already clipping cause aren't all you're doing is raising the noise floor? One idea I had was to make this an option, where you can put a floor on the gain as an option. I'd set mine to floor = zero (zero gain that is, not db).

Well, let me say that I'm not asking anyone to develop anything here, and I am grateful for what's already been developed, and I love RB and the REP as is for sure. I was just thinking about all of this negative gain thing while foolin' around with my H120, and I wanted to try to better understand how the gain staging was working because as a taper, I just want to know those kinds of things.  :P

Anyway, thanks again for your code guys!



petur:

--- Quote from: BayTaper on August 08, 2006, 10:59:43 PM ---Now to my main question... I'm trying to understand how the gain works on the H120 with RB. From what I've read, I think it just uses the analog gain to take a big step up and then the decimator to shave some off db digitally so it increments in steps of .5 db and in a way that uses as much analog gain as it can while only using digital gain to lower the level back for .5 steps. So, from what I've seen when the gain goes negative it is simply a digital scaling down of the signal. Do I have that right?
--- End quote ---

correct up to the point of the digital scaling down. We don't know for sure what the decimator does and works on.


--- Quote from: BayTaper on August 08, 2006, 10:59:43 PM ---If so, then when you're clipping, and you reduce the gain into "negative"  territory in the UI, you'll still have distorted audio with flat waveforms at a reduced db level. I totally understand that there's nothing the H120 can do about this situation because it doesn't have true analog "pads" or anything like that. But, from the AGC's "Safety" mode, I'm thinking a slightly better algorithm might be to "safety it down to 0 gain and then stop". Now maybe I've got this wrong, but my thinking from a recordist's perspective is once I'm clipping, I'm pretty screwed, and all lowering the gain down past zero into the negative gain territory is just bringing my noise floor up at that point. My thinking goes that, hey you're clipping, that data is lost forever, so it doesn't really matter to lower the gain digitally, it's not doing me any good past zero is it? Now I can see where as long as you aren't clipping, you could use negative gain to simply lower the signal in the recorded file, but why would you want to do that if you weren't already clipping cause aren't all you're doing is raising the noise floor? One idea I had was to make this an option, where you can put a floor on the gain as an option. I'd set mine to floor = zero (zero gain that is, not db).

--- End quote ---

Good question.... I still need to do some tests on this. I once had a recording where the AGC turned the gain down negative (not much) and I didn't see clipping I think. So what the decimator does exactly is still a bit of a mistery. It does come right after the ADC and operates on the full ADC output (which has more accuracy than the data the chip sends to the cpu)

Mmmm:

--- Quote from: petur on August 08, 2006, 07:22:44 PM ---Mmmm: new AGC patch on the tracker.

Let me know if I missed something.
I think it's getting close to ready for commit.

Took over your screen handling changes,
I'm still looking for a better solution.

--- End quote ---

Great...
What is it you're not happy with?...I'm away from my dev environment for a while so I won't be able to do any testing/improving for a week or so... but I'll be near a computer until tomorrow at least, so I could offer some suggestions (or excuses for how i did it :P).

BayTaper:

--- Quote from: petur on August 09, 2006, 03:50:40 AM ---
--- Quote from: BayTaper on August 08, 2006, 10:59:43 PM ---Now to my main question... I'm trying to understand how the gain works on the H120 with RB. From what I've read, I think it just uses the analog gain to take a big step up and then the decimator to shave some off db digitally so it increments in steps of .5 db and in a way that uses as much analog gain as it can while only using digital gain to lower the level back for .5 steps. So, from what I've seen when the gain goes negative it is simply a digital scaling down of the signal. Do I have that right?
--- End quote ---

correct up to the point of the digital scaling down. We don't know for sure what the decimator does and works on.


--- Quote from: BayTaper on August 08, 2006, 10:59:43 PM ---If so, then when you're clipping, and you reduce the gain into "negative"  territory in the UI, you'll still have distorted audio with flat waveforms at a reduced db level. I totally understand that there's nothing the H120 can do about this situation because it doesn't have true analog "pads" or anything like that. But, from the AGC's "Safety" mode, I'm thinking a slightly better algorithm might be to "safety it down to 0 gain and then stop". Now maybe I've got this wrong, but my thinking from a recordist's perspective is once I'm clipping, I'm pretty screwed, and all lowering the gain down past zero into the negative gain territory is just bringing my noise floor up at that point. My thinking goes that, hey you're clipping, that data is lost forever, so it doesn't really matter to lower the gain digitally, it's not doing me any good past zero is it? Now I can see where as long as you aren't clipping, you could use negative gain to simply lower the signal in the recorded file, but why would you want to do that if you weren't already clipping cause aren't all you're doing is raising the noise floor? One idea I had was to make this an option, where you can put a floor on the gain as an option. I'd set mine to floor = zero (zero gain that is, not db).

--- End quote ---

Good question.... I still need to do some tests on this. I once had a recording where the AGC turned the gain down negative (not much) and I didn't see clipping I think. So what the decimator does exactly is still a bit of a mistery. It does come right after the ADC and operates on the full ADC output (which has more accuracy than the data the chip sends to the cpu)

--- End quote ---

Thanks for the response. I guess I forgot to mention that I DID TEST this already. As far as I can tell negative gain just digitally lowers the signal strength. I ran a hot signal into the H120, it was clipping at 0 db with 0 gain. The AGC safety mode then moved the gain into negative territory (or I'm sure I could have just done that manually myself), and I then inspected the waveform in SF. The top of the wave was below 0 db, but it was chopped-off flat just like if it had clipped at 0 db. To me that spells only one thing, that negative gain is a digital scaling down of the signal and nothing more. Thus, I don't think there is really any advantage to using it (negative gain that is, I understand why its kind of nice to use the decimator to achieve smaller .5 increments, that makes sense to me, but whenever possible it's probably better to stick with pure analog gain, which I think I read was in 3 db increments starting at zero, right?). The wavform is just as screwed as it would be up at 0 db, and I think by lowering it from that point the signal to noise ratio suffers, not to mention that any digital scaling at all at 16-bit creates quantization noise from the rounding that must occur to do it. Anyway, just wanted to add more flavor to my post above considering I did actually test this yesterday. THANKS!

Wolf:
Very interesting BayTaper. I also bougth my H140 mainly for recording live gigs and this information helps to get hi quality recordings.

Do you by any chance has any screenshots documenting your analysis?
Thx, Wolf.

Navigation

[0] Message Index

[#] Next page

[*] Previous page

Go to full version