Topic: 24 bit recording in 32 bit float devices with digital gain controls - pointless?

[Ozpeter]

Someone asked me in a YouTube comment whether the Tascam FR-AV2 has gain controls in the digital or analog domain.  After some googling and manual reading, it seems pretty clear particularly from the FR-AV2 discussion in these forums that the device only has digital gain / level controls.  The new Zoom H5 Studio is the same, according to the circuit diagram in the manual.  Probably it is the same with some other devices.  (I'd be interested to hear of any).

In these devices then, if you are using 24 bit float you therefore have to set the gain so that the level does not exceed the 24 bit file maximum level.  But the gain setting has no bearing on getting the best signal to noise ratio out of the preamp because it isn't adjustable.  So by using 24 bits you are giving yourself a problem, and you are getting no sound quality advantage in terms of choosing how you drive the preamp (because you don't have the choice).  So the only point in using the 24 bit setting is if you have an issue with the capacity of the memory card and want to keep file sizes down, or you can't handle 32 bit float in your DAW, or you are having to immediately hand over the recording to someone who has specified 24 bit.

But aside from those rare requirements, the whole argument about the merits of 32 bit float recording if using one of these devices become moot.  If you really want to record in 24 bits because you think you will get a better result, only use a device which either doesn't do 32 bit float at all, or use one which definitely has the level controls in the analog domain.

Or am I wrong?

[rastasean]

Am i reading this correctly that it also applies to mixpre ii series as well?
https://www.audiosciencereview.com/forum/index.php?threads/re-review-of-sound-devices-mixpre-3-ii-multichannel-recorder-review-advanced-mode.50437/

Now a word of explanation:

Usually, a Mic preamp interface has an analog preamp stage followed by the ADC.
And the analog preamp stage has a variable gain.

> So the only point in using the 24 bit setting is if you have an issue with the capacity of the memory card and want to keep file sizes down, or you can't handle 32 bit float in your DAW, or you are having to immediately hand over the recording to someone who has specified 24 bit.

Those reasons are what i can think of as well and make sense to me.


here's a several year old review of the mix pre 3ii (which has MSRP of the recorder at $650. Now it's $300 more!): https://nofilmschool.com/sound-devices-mixpre-ii-review
It has this to say about dual AD converters, which you may find interesting regarding the recent tascam recorders:
Either way, 32-bit float recording is nothing without dual AD converters. It’s the combination of these two that creates a wide dynamic range. Without both working together, 32-bit float files would clip just like any 16-bit or 24-bit recording. With that said, let’s dive into the MixPre II series.

[aaronji]

^ No, I think the MixPre recorders actually have analog gain stages in 24-bit mode. See the last post in that ASR thread, for example. I was also told this directly by Paul Isaacs. The final files, I believe, have a combination of analog and digital gain (SD termed this "secret sauce"; I think it applies to how/where digital gain is applied).

[Joop]

Or am I wrong?

You are wrong. For a very long time you can set levels digital via an IC(-set) which is special made for this purpose. It is done with or decoders (DR-701-D) or with an processor (probably FR2-AV). In this  way you don't have to make use of the classic potmeter which always go bad after some time and you can make more steps than with a switch, so the recordlevel can be set smoother. It is not the same digital thing as the digital end of an AD converter, that's a whole other party. 

[Ozpeter]

Or am I wrong?

You are wrong. For a very long time you can set levels digital via an IC(-set) which is special made for this purpose. It is done with or decoders (DR-701-D) or with an processor (probably FR2-AV). In this  way you don't have to make use of the classic potmeter which always go bad after some time and you can make more steps than with a switch, so the recordlevel can be set smoother. It is not the same digital thing as the digital end of an AD converter, that's a whole other party.

Hmm.  Well, in the case of the H5 Studio the diagram shows AD converters immediately after the inputs, followed by "DC Cut", followed by gain, followed by "Low cut" and so on.  But we don't have a diagram of the Tascam devices, only what Tascam said when asked, which was "REC LEVEL is done at the digital stage" - although I concede that is open to interpretation.  However, when the meters show clipping, in the manual they don't advise reducing the record level, they suggest moving the mics further from the source or reducing the level fed to the line input. That would fit in with the digital gain theory as digital gain controls won't help with clipping in the analog stage, of course.

The Zoom F3 follows the same pattern - "waveform amplification" well after the ADC - but as it doesn't offer 24 bit recording, the scenario is different.

[Sebastian]

Don't confuse digital gain *controls* with digital gain!

Most (if not all) of these devices make use of very small ADCs/audio codecs. These ICs are controlled via a serial data bus like I²C or SPI. The audio codec usually contains an (analog) op-amp that is controlled in coarse steps via this data bus. The fine gain steps can sometimes be digital.

Here's an example of a Texas Instruments chip that follows this pattern. It has 0-42 dB of analog gain that's digitally programmable in 1 dB steps. This gain has to be set before the channel is powered up. It also has an additional digital volume control with 0.5 dB steps. That volume control can be used to digitally reduce/increase the volume after the analog gain stage. This reduction is unproblematic because it just truncates data that would be too quiet anyways. Here's the detailed explanation from the datasheet: https://www.ti.com/document-viewer/TLV320ADC5140/datasheet#programmable-channel-gain-and-digital-volume-control-t5097758-21/t5097758-21

[Ozpeter]

Thanks for that link - which kind of goes right over my head, but I've read the relevant bit several times.  My totally amateur interpretation is that there is a preset analogue gain stage - which it says should not be varied after powering up - and then a digital gain stage (altering the digits not the analogue sound) after that which can be varied.  It seems to me that the first analogue stage is calibrated in the factory to match the expected input levels and would not be variable by the end user.  The later digital gain stage would be accessible by the user to manipulate the digital values emitted by the ADC.  So that to me fits in with what I've been saying.  However, I concede that I may be totally misunderstanding how it works. 

In the Zoom H5 Studio diagram, the circuit is described something like this -

Input > Dual ADC > DC Cut > Gain > Low Cut > Advanced limiter > etc etc

If I understand you correctly, they should describe it as -

                Gain
                 ^
Input > Dual ADC > DC Cut > Low Cut > Advanced limiter > etc etc

thus showing the gain as part of the Dual ADC rather than a later stage.

Hopefully when I return to my "studio" next week I could experiment a bit with the various 32 bit float devices I have there to see whether I can identify how their gain stages are working at a practical level.

[Sebastian]

Thanks for that link - which kind of goes right over my head, but I've read the relevant bit several times.  My totally amateur interpretation is that there is a preset analogue gain stage - which it says should not be varied after powering up - and then a digital gain stage (altering the digits not the analogue sound) after that which can be varied.  It seems to me that the first analogue stage is calibrated in the factory to match the expected input levels and would not be variable by the end user.  The later digital gain stage would be accessible by the user to manipulate the digital values emitted by the ADC.  So that to me fits in with what I've been saying.  However, I concede that I may be totally misunderstanding how it works. 

In the Zoom H5 Studio diagram, the circuit is described something like this -

Input > Dual ADC > DC Cut > Gain > Low Cut > Advanced limiter > etc etc

If I understand you correctly, they should describe it as -

                Gain
                 ^
Input > Dual ADC > DC Cut > Low Cut > Advanced limiter > etc etc

thus showing the gain as part of the Dual ADC rather than a later stage.

Hopefully when I return to my "studio" next week I could experiment a bit with the various 32 bit float devices I have there to see whether I can identify how their gain stages are working at a practical level.

Yeah, you seem to have understood it correctly.
That TI chip was just an example of how things *might* work and what's possible these days. I have no idea what the Tascam uses internally. It most likely has a different chip with a different setup. It could be completely analog, completely digital or a mix like in the TI chip. They might even be making their own chips (like Sony does with the A10).

I would like to point out that the combination of analog and digital gain is superior to having digital gain exclusively. Let's assume the input signal is boosted by a fixed 20 dB in the first analog gain stage. Let's also assume that the user wants to set the total gain of the recorder (what he sees) to a total of +10 dB gain. Since we have already boosted the signal by 20 dB, we now actually have to decrease its amplitude by 10 dB to achieve the target value of +10 dB. This decrease is done digitally. What actually happens is the least significant information (the lowest 10 dB of acoustic information) is removed and everything else is lowered by 10 dB. The user ends up with exactly the same information as if they'd used a single, analog gain stage with a +10 dB gain setting. I hope this makes some sense.

[Ozpeter]

Thanks.  It made sense after I read it twice!  I do wish there was more transparency from the manufacturers of these devices about how they actually work internally (exactly how gain is controlled and how they choose the initial analog gain if there's only digital gain setting, not to mention my other dual / single ADC obsession!).  But then if they told us too much they would have to shoot us - hopefully with a video camera...

Meanwhile Curtis Judd has just reviewed the Zoom H5 Studio and took pains to point out that the gain controls only operate in the digital domain, and he included the block diagram circuit description in the video.  But to my slight surprise he didn't seem to have a problem with that at a practical level, despite being a bit traditional at times in his judgements over 32 bit float vs 24 bit - overall he seemed to rather like the device.  I guess in one way one could say, never mind the theory, does it work in practice at an audible level?