Topic: Tascam FR-AV2 new compact 32bit float recorder.

[Sebastian]

Curious if anyone can lend some insight on the following:


PCM A-10 AUDIO in specifications:
Input impedance: approx. 4.7 kΩ
Rated input level: approx. 2 V
Minimum input level: approx. 18 mV

The problem with LLMs is that they are good at summarizing text. But they can't think. And they can't make sense of what they output in the same way any human child can. It simply repeats what it has been trained with. In this case: marketing materials as it seems.

You write that you're running DPA 4061s. These have a sensitivity of 6 mV per Pa (that's Pascals, the unit for measuring atmospheric pressure). The max. input level of 2 V on the A10 gives you 2000 mV / (6 mV / Pa) = 333.3 Pa. That equals approx. 144 dB SPL (here's a calculator). This means the input stage of your A10 will clip when your DPAs are recording sounds louder than 144 dB. These are SPLs you will find at rocket launches, but not when recording music. So you're safe with the A10.

You can repeat the same calculations with the other values. 1 dbV = 1.12 V and 24 dbU = 12.27 V (calculators).

1.12 V / (6 mV / Pa) = 186 Pa (= 139 db SPL)
12.27 V / (6 mV / Pa) = 2045 Pa (= 160 dB SPL)

[Kyle K]

Curious if anyone can lend some insight on the following:


PCM A-10 AUDIO in specifications:
Input impedance: approx. 4.7 kΩ
Rated input level: approx. 2 V
Minimum input level: approx. 18 mV

The problem with LLMs is that they are good at summarizing text. But they can't think. And they can't make sense of what they output in the same way any human child can. It simply repeats what it has been trained with. In this case: marketing materials as it seems.

You write that you're running DPA 4061s. These have a sensitivity of 6 mV per Pa (that's Pascals, the unit for measuring atmospheric pressure). The max. input level of 2 V on the A10 gives you 2000 mV / (6 mV / Pa) = 333.3 Pa. That equals approx. 144 dB SPL (here's a calculator). This means the input stage of your A10 will clip when your DPAs are recording sounds louder than 144 dB. These are SPLs you will find at rocket launches, but not when recording music. So you're safe with the A10.

You can repeat the same calculations with the other values. 1 dbV = 1.12 V and 24 dbU = 12.27 V (calculators).

1.12 V / (6 mV / Pa) = 186 Pa (= 139 db SPL)
12.27 V / (6 mV / Pa) = 2045 Pa (= 160 dB SPL)

Appreciate those calculators and details, thanks.

Did you mean to say "you are safer" with the A10? I know I'd be fine, it has been my primary device for awhile, what I'm trying to figure out is the actual practical benefit I'd receive from 32bit float / ADC of the FR-AV2 given the lower max input level of the 3.5mm jack.

In the 24 bit/A10 "world", let's call it, I am using a preamp to apply gain and get as loud as possible without clipping so that I don't have a hissy recording (if I don't send the signal to the recorder with enough gain, and raise in post, I'll likely get hiss)

In the FR-AV2 world, I don't have to be concerned with digital clipping on the device -- I can essentially run as hot as I want and lower after the fact or so I thought. This "max input level" is throwing me for a bit of a loop. In a practical sense, does it mean that if I apply too much gain via the preamp, I am screwed? No amount of post processing could repair the exceeded input levels, right?

But the thing is, I shouldn't really have to worry about applying gain much at all.. I can apply little to no gain, set conservative levels and thanks to the ADC almost certainly not have to worry about hiss if I have to bring it up a bunch later. This is where I am landing in theory -- will see the reality once I am able to actually test. I am suspecting that in effect it means I do still need to worry about setting levels for louder/quieter shows, depending upon how noisy the pres are. Unless I use the XLR/TRS. There, I actually have enough headroom that I can more or less set it up once and never worry about it again.

[Rairun]

In the FR-AV2 world, I don't have to be concerned with digital clipping on the device -- I can essentially run as hot as I want and lower after the fact or so I thought. This "max input level" is throwing me for a bit of a loop. In a practical sense, does it mean that if I apply too much gain via the preamp, I am screwed? No amount of post processing could repair the exceeded input levels, right?

You're right you don't have to be concerned with digital clipping, but that doesn't mean you can run as hot as you want. If you run it too hot, you will overload the analogue stage of the recorder - you still won't get digital clipping, but you very much will get analogue distortion.

The idea behind the better 32-bit float recorders (the MixPre II, Zoom F3/F6, Tascam FR-AV2, etc, as opposed to for example the Zoom H1essential) is that you won't have to use an external preamp. They all have very clean internal ones.

[Kyle K]

Cool, thanks. I knew this was a concern, I just wasn't expecting the max input to actually be lower than the A10. Was surprised once I did the conversions and realized. Fingers crossed they're as clean as we hope.

[grawk]

Have you actually tried running lower and raising the levels to see if you get hiss?  With DAT era devices, that was definitely needed, but I regularly record at -30db and then normalize, and haven't had a problem of noisy recordings.  The main difference with 32 bit devices is the recorder basically forces you to run low.

[unidentified]

In the FR-AV2 world, I don't have to be concerned with digital clipping on the device -- I can essentially run as hot as I want and lower after the fact or so I thought. This "max input level" is throwing me for a bit of a loop. In a practical sense, does it mean that if I apply too much gain via the preamp, I am screwed? No amount of post processing could repair the exceeded input levels, right?

You're right you don't have to be concerned with digital clipping, but that doesn't mean you can run as hot as you want. If you run it too hot, you will overload the analogue stage of the recorder - you still won't get digital clipping, but you very much will get analogue distortion.

The idea behind the better 32-bit float recorders (the MixPre II, Zoom F3/F6, Tascam FR-AV2, etc, as opposed to for example the Zoom H1essential) is that you won't have to use an external preamp. They all have very clean internal ones.

Do we have good info on the quality of the Tascam FR-AV2's pre-amps?

[TheJez]

I'd be particularly interested in recording level functionality when using 32 bit float. With the F3 there is a fixed gain in the analog input, and then the 32 bit float samples are stored with an amplification depending on the 'magnification level' as was set at start of recording.
I can't find any details in the FR-AV2 manual about something like this. It seems as if the record level still applies to the analog input, regardless if 24bit or 32bit float is used. This would make it still possible to set the record level 'too high' when using 32bit float, resulting in overloading the analog input stage, kind of defeating the merrits of a 32 bits float recorder (set-and-forget). I might be wrong, of course...
Hope to hear soon about real-life experiences with this interesting device! I'm about to replace my Edirol R-09HR, and it seems I will choose between the F3 and the FR-AV2...

[goodcooker]

Do we have good info on the quality of the Tascam FR-AV2's pre-amps?

The info available so far says that it has Tascam's HDDA mic preamps that are presumably the same discrete design as the 701d and some other devices in their family. Here's the marketing speak from the 701d page -

"HDDA mic preamps - the differential mic preamp circuits use discrete architecture, built with carefully selected components. For example, Texas Instruments OPA1652 op amps were selected for their high quality and low noise – with even better audio performance than the previous DR-70D model. As a result, the unit boasts an equivalent input noise of -124 dBu or better"

I ran a 701d for a few years and it sounds pretty good. The preamps are just fine for what many users are going to do - live music, film dialog, etc.
There's a bunch of recordings in my LMA signature link that use the 701d preamps if you want to have a listen in different scenarios (inside, onstage, outside, matrix) and with different mics (I had several preamps at the time and often used them in front of the recorder but there's some with the stock device Pres in there). That device is marketed towards videographers like this new one is and has many of the same features.

[Kyle K]

Have you actually tried running lower and raising the levels to see if you get hiss?  With DAT era devices, that was definitely needed, but I regularly record at -30db and then normalize, and haven't had a problem of noisy recordings.  The main difference with 32 bit devices is the recorder basically forces you to run low.

Yes e.g. https://ia802308.us.archive.org/4/items/tmg2021-09-23/05_Hair%20Match.mp3

I think this may have been running my DR680MKii low as opposed to A10, but same principle. Certainly didn't kill the tape or anything, just something I try to avoid since.

[vanark]

I would very much hesitate to run at -30dB even at 24 bit. The noise floor will come up with the music when you amplify it in post. I tend to shoot for -14 to -12, even -16 if I think the band is super loud and may amplify by 6 dB at some point, especially at the end of the show.

[grawk]

 SBD  > dr680mkii + CA-14 omni (healy on stage) > CA-Ugly 2 > pcm m10
So the noise could have come from the soundboard, the ca-ugly, the m10, something done in the mix, etc.

The 680mk2 has 100db SNR, so amplifying the noise floor by 30db should still have the noise floor of the recorder at -70db.  Once you add an external preamp into the mix, you need to determine the right levels for the preamp and recorder to minimize noise, as that's a lot different for every device.

Given that most 32bit devices are applying 0 analog gain, if you're recording music that doesn't incorporate a jet that's taking off, you're running at significantly lower than -30db peaks.

Have you actually tried running lower and raising the levels to see if you get hiss?  With DAT era devices, that was definitely needed, but I regularly record at -30db and then normalize, and haven't had a problem of noisy recordings.  The main difference with 32 bit devices is the recorder basically forces you to run low.

Yes e.g. https://ia802308.us.archive.org/4/items/tmg2021-09-23/05_Hair%20Match.mp3

I think this may have been running my DR680MKii low as opposed to A10, but same principle. Certainly didn't kill the tape or anything, just something I try to avoid since.

[Rairun]

Given that most 32bit devices are applying 0 analog gain, if you're recording music that doesn't incorporate a jet that's taking off, you're running at significantly lower than -30db peaks.

I'm not sure this is true? Just because you don't set the gain, that doesn't mean 0 analogue gain is applied. Take this from the Zoom F3 specs:

Equivalent input noise: −127 dBu or less (IHF-A) when waveform magnification is ×1024 with 150 Ω input.

As far as I can tell, this sort of EIN can usually only be attained when your gain is of at least +30 dB. If you take a look at this table, you will notice the EIN is always measured at the maximum available gain settings, which in the F3 is equivalent to the waveform magnification of ×1024.

A recorder with no gain will never have an EIN of -127 dBu. The converter's self-noise normally sits around -95 dBu. What improves the EIN drastically is the use of a preamp, which does add a little noise, but proportionally adds a LOT more gain than noise. So if the preamp adds 30 dB of gain and only 5 dB of noise, then your EIN will be -95 +5 -30 = -120 dBU

AFAIK, the trick with these 32-bit float devices is that they essentially add a lot of clear analogue gain when the signal they receive is weak, and little or no gain when the signal is strong (and then, digitally, they account for the different amounts of gain when merging these two paths).

What I'm trying to say is that using one of these devices is not at all like running -30 dB peaks with a recorder with a single ADC. If you're running -30 dB peaks with a regular recorder, the converter stage self-noise will instantly drop to -65dB or so when you normalise. The peaks themselves likely will be fine, but the quieter parts will suffer for sure.

[Kyle K]

I'd be particularly interested in recording level functionality when using 32 bit float. With the F3 there is a fixed gain in the analog input, and then the 32 bit float samples are stored with an amplification depending on the 'magnification level' as was set at start of recording.
I can't find any details in the FR-AV2 manual about something like this. It seems as if the record level still applies to the analog input, regardless if 24bit or 32bit float is used. This would make it still possible to set the record level 'too high' when using 32bit float, resulting in overloading the analog input stage, kind of defeating the merrits of a 32 bits float recorder (set-and-forget). I might be wrong, of course...
Hope to hear soon about real-life experiences with this interesting device! I'm about to replace my Edirol R-09HR, and it seems I will choose between the F3 and the FR-AV2...

Just received mine. Just to answer this, with 32 bit float enabled, you can still set gain independently to each channel (1 or 2). From 0 to +60 DB. It is set to +18 by default. Can adjust while recording though the menu is a bit clunky. Expect it's easier with app.

Going to be testing some items for my purposes (battery time recording with my mics, effective Bluetooth range, app controls, running hot with loud stereo approximate concert volume and reducing in post, etc) but if anyone wants any info I may be able to provide feel free.

As an aside, I'm a dumbass and just clicked the suggested cards when purchasing it from BHP, didn't realize it suggested the wrong size. I bought two bigger ones. Need to return em.

Edit: note you need to assign both channels to EXT input manually if you plan to use the 3.5mm. Flip the STEREO LINK toggle under INPUT to do so. No auto detection or stereo assumption - threw me for a loop for a second.

Edit 2: another note, two drawbacks to the lack of explicit HOLD.
1- you hold down 1 button for several seconds to stop the recording. Could conceive of this occurring in a pocket, though unlikely.b
2 - there is one unlocked button that could fuck up the recording, but you'd need to hit that button, then a separate button, then a third different button several times, in a particular order. Again, unlikely, but technically possible. All the buttons are pretty hardy and difficult to press accidentally.

[unidentified]

"with 32 bit float enabled, you can still set gain independently to each channel (1 or 2)".  Sorry to be slow, but how is this useful?  I own a F3 and I can always adjust the relative volume levels in post as needed. 

[Sebastian]

As far as I can tell, this sort of EIN can usually only be attained when your gain is of at least +30 dB. If you take a look at this table, you will notice the EIN is always measured at the maximum available gain settings, which in the F3 is equivalent to the waveform magnification of ×1024.

A recorder with no gain will never have an EIN of -127 dBu. The converter's self-noise normally sits around -95 dBu. What improves the EIN drastically is the use of a preamp, which does add a little noise, but proportionally adds a LOT more again than noise. So if the preamp adds 30 dB of gain and only 5 dB of noise, then your EIN will be -95 +5 -30 = -120 dBU

This is correct. That's why EIN figures are useless if the conditions under which these were achieved are not known. Tascam lists the EIN as "<= -127 dBu", but doesn't mention at which gain and with which source resistance, or whether or not it is a-weighted. But since they probably want to be able to list the best possible EIN, we can safely assume this was measured with the recorder at max. gain (50 dB). This is a pretty big amplification factor and I guess most people will never really need to apply this much gain. So this measurement does not tell us anything about the pre-amp's performance at, e.g. unity gain. The other parameters remain unknown, so we can't really use their EIN number to compare the pre-amps to other devices.