• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Zoom F6 Portable Field Recorder Review

jerryfreak

Active Member
Joined
Jan 12, 2020
Messages
125
Likes
131
Location
Reno, NV
There will of course be a microphone preamplifier stage in the unit.
the block diagram has the input going right into the ADC. Per zoom, there is a fixed gain stage in there, however there is no user-adjustable gain prior to the ADC. This gain stage also appears to be not bypassable in line-in mode, which, looking at noise floor, looks like the same signal path with the 20 dB pad slapped on top, reducing the effective dynamic range of the unit.
 
Last edited:

Tks

Major Contributor
Joined
Apr 1, 2019
Messages
3,221
Likes
5,494
ADCS have been exceeding the resolution of the signals they are used to record, for decades now. they are out there.

Even for the low sensitivity dynamic mics? I suppose proper interfaces with 50dB+ worth of gain are going to work well with something like a Shure SM7B, but something like a Focusrite Solo that maxes out in the high 40's if I recall, may leave something to be desired. Though still good enough nonetheless.
 

PeteL

Major Contributor
Joined
Jun 1, 2020
Messages
3,303
Likes
3,838
the block diagram has the input going right into the ADC. Per zoom, there is a fixed gain stage in there, however there is no user-adjustable gain prior to the ADC
That makes a little more sense, but still an oddity. If a sm58 and a condenser mic gets the same gain before hitting the ADC, the bits your are not hitting with the 58, you'll never get them back, no matter what digital manipulations you do. I believe there could some digitally controlled gain settings that are not user adjustable, but fully fixed gain?
 

jerryfreak

Active Member
Joined
Jan 12, 2020
Messages
125
Likes
131
Location
Reno, NV
me:
so is the analog preamp stage fixed gain? and then digitally attenuated by the trim?

zoom:
Yes, the analog stage is fixed. Digital attenuation is adjusted by the trim.
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,524
Likes
37,057
it will clip like any other device if the input level is exceeded (+4 mic +24 line), but yes in 32bit float, level control really does nothing and 'Full Scale' is pretty arbitrary. you can record something consistently over 0dBFS and the data is all there if you reduce the volume in software



this would be ideal to see these tested as stand-alone recorders in the same manner, as they are really same market segment

particularly if they were tested with mic level signals in the -40 dBV range typical of condenser mics

the mixpre has discrete class A preamps and I would expect it to shine in this application.

Both the mixpre and the zoom f6 have similar EINs but the mixpre can take a much hotter signal (+10dB on mic and +16dB on line) so perhaps it would make better utilization of the 32-bit float.

Sound devices also uses a completely different architecture for its multi-ADC implementation, utilizing 3 converters:

http://patentimages.storage.googleapis.com/a7/6b/f5/77e31e68cca8b7/US9654134.pdf

Yes, I know there is no getting around clipping the analog front end. I wonder if the device would perform better in 32 bit float with look ahead limiting engaged.
Even after reading the manual for the F6 it wasn't at all clear to me how it would work best. One of the problems doing a review of such a device unless you have time to delve into it quite a bit. That in itself is a knock on Zoom for not being more clear.

Even then if proper setup let me record field items with no (or nearly no) chance of clipping even these poor noise levels might be acceptible. OTOH, why should any device perform this poorly these days? Answer is it shouldn't and especially not a piece of gear like this. Very disappointing.

Tests like this are a real service. I almost purchased one, but held up because of what they didn't list as specs and turns out that was a smart decision the way it looks now.
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,524
Likes
37,057
that mic will still need gain, but at the end of the day the >110dB dynamic range of modern ADCs still exceed the dynamic range of real-world recording situations, limited by self-noise of the mic, and background noise (as well as max SPL)

That is what people who haven't done some recording don't understand. If you get 105-110 db of dynamic range you'll never get a recording that can use all of that. So not much impetus to improve ADCs as quite a few inexpensive ones already reach this level. One useful improvement in use of it in the field was a 32 bit recording capability to allow one to be relaxed about setting max levels and worrying about exceeding them. Of course I thought they would be fixing a 100 db DR into this wide window on a sliding basis. Not 60 or 70 db.
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,524
Likes
37,057
the block diagram has the input going right into the ADC. Per zoom, there is a fixed gain stage in there, however there is no user-adjustable gain prior to the ADC. This gain stage also appears to be not bypassable in line-in mode, which, looking at noise floor, looks like the same signal path with the 20 dB pad slapped on top, reducing the effective dynamic range of the unit.
So these results would look like maybe a 50db fixed gain? Which as was pointed out eats up the available EIN severely limiting performance.
 

restorer-john

Grand Contributor
Joined
Mar 1, 2018
Messages
12,580
Likes
38,282
Location
Gold Coast, Queensland, Australia
the block diagram has the input going right into the ADC. Per zoom, there is a fixed gain stage in there, however there is no user-adjustable gain prior to the ADC. This gain stage also appears to be not bypassable in line-in mode, which, looking at noise floor, looks like the same signal path with the 20 dB pad slapped on top, reducing the effective dynamic range of the unit.

Block diagrams are not schematics. An electret requires an active stage before anything useful to an A/D is produced. It makes sense to passively attenuate and protect a low noise mic stage rather than risk A/D front end damage from a too hot line signal.
 
OP
amirm

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,370
Likes
234,436
Location
Seattle Area
Even after reading the manual for the F6 it wasn't at all clear to me how it would work best. One of the problems doing a review of such a device unless you have time to delve into it quite a bit. That in itself is a knock on Zoom for not being more clear.
To be clear, I spent quite a bit of time with level adjustments, both at source and on the device itself. Throughout all this, SINAD remained more or less the same. I assumed the low SINAD must be a config issue so I dug in unlike normal reviews. I just could not get better results.

Now, the answer may still be out there but I did try. :)
 

jerryfreak

Active Member
Joined
Jan 12, 2020
Messages
125
Likes
131
Location
Reno, NV
So these results would look like maybe a 50db fixed gain? Which as was pointed out eats up the available EIN severely limiting performance.
that maybe on the high side. i wouldnt go by the measured dynamic range and compare it to the RME and assume the difference is all gain. There might be other bottlenecks in there that only allow it to hit 100 dB in absence of the fixed gain stage.

the company is not very forthcoming about the architecture of the units, would probably take a teardown to reveal the ADC chip(s) used and look at the datasheet and back out the amount of fixed gain

that said, if +4dBU is the real, undefeatable max input sensitivity, and the 20 dB 'line in' pad just tacks on the front of it, yes there is no way this unit is close to SOTA for 24-bit recorders with balanced input

if you want to take a deep dive into how the F6 handles very low level signals you can check out this suite of tests i did mic'ing a monitor played at a very low level in a quiet room

https://taperssection.com/index.php?topic=190161.msg2331113#msg2331113
 
Last edited:

PeteL

Major Contributor
Joined
Jun 1, 2020
Messages
3,303
Likes
3,838
That is what people who haven't done some recording don't understand. If you get 105-110 db of dynamic range you'll never get a recording that can use all of that. So not much impetus to improve ADCs as quite a few inexpensive ones already reach this level. One useful improvement in use of it in the field was a 32 bit recording capability to allow one to be relaxed about setting max levels and worrying about exceeding them. Of course I thought they would be fixing a 100 db DR into this wide window on a sliding basis. Not 60 or 70 db.
If we where to use the full 110 dB of dynamic range, the quietest parts of the recordings would be at the same level as the noise floor.
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,524
Likes
37,057
To be clear, I spent quite a bit of time with level adjustments, both at source and on the device itself. Throughout all this, SINAD remained more or less the same. I assumed the low SINAD must be a config issue so I dug in unlike normal reviews. I just could not get better results.

Now, the answer may still be out there but I did try. :)
I don't doubt that you did. I know myself with such picky and adjustable devices I've messed around with them, and then later saw something that is a headslapper. Oops no wonder it wasn't working better.

I'm pretty sure that isn't the case here because all the poor results are the specs they omitted in the manual.
 

Rja4000

Major Contributor
Forum Donor
Joined
May 31, 2019
Messages
2,682
Likes
4,220
Location
Liège, Belgium
a 32-bit float can resolve a maximum of 23 bits of data resolution, the rest of the bits are the sign and mantissa. read this thread starting at this post:

https://www.audiosciencereview.com/.../new-member-field-recordist.10858/post-303744

in any case, for the practical limits of audio interfaces and hearing, at any recording levels close to correct, a 24-bit fixed, 32-bit fixed, and 32-bit float all have ample dynamic range to capture signals. 32-bit is used on all workstations to prevent rounding errors, primarily
Well, first, if I'm not mistaken, you have 23 bits + sign in PCM 24 and you have the same in 32 bits IEEE 754 single-precision binary floating-point format.

But that's not my point:
Even if you had one bit more with an integer coding, that means that you could resolve the lowest level component from a bi-tone signal, with a close-to 0dBFS tone combined with a -138dB tone.
How usefull is that?
Anything real-life, where you keep the level comfortably below -6dBFS, to avoid surprises, this "benefit" is gone, and the lower the recording level, the more the 32 bits float pulls ahead.
 

jerryfreak

Active Member
Joined
Jan 12, 2020
Messages
125
Likes
131
Location
Reno, NV
Even if you had one bit more with an integer coding, that means that you could resolve the lowest level component from a bi-tone signal, with a close-to 0dBFS tone combined with a -138dB tone.
How usefull is that?
in the real world, not useful at all. the -138 dB tone you recorded would already be buried in noise. youd have a faithful 24-bit representation of said noise
 

shal

Active Member
Joined
Nov 30, 2019
Messages
119
Likes
122
Location
Rennes, France
I'm confused by this, I can't think of an adc chip that would operate properly at mic level voltages, you have to gain it up somehow no?

A good page on 32 bit float
https://www.sounddevices.com/32-bit-float-files-explained/


Sound devices control automatically the gain of preamp for avoid clip. This gain is encoded in the 32 bit float.

A fun video and good demonstration that show the possibility at
at 1:16
 

Francis Vaughan

Addicted to Fun and Learning
Forum Donor
Joined
Dec 6, 2018
Messages
933
Likes
4,697
Location
Adelaide Australia
a 32-bit float can resolve a maximum of 23 bits of data resolution, the rest of the bits are the sign and mantissa. read this thread starting at this post:

https://www.audiosciencereview.com/.../new-member-field-recordist.10858/post-303744

in any case, for the practical limits of audio interfaces and hearing, at any recording levels close to correct, a 24-bit fixed, 32-bit fixed, and 32-bit float all have ample dynamic range to capture signals. 32-bit is used on all workstations to prevent rounding errors, primarily

Even that thread is bit naive. The claim that 32 bit float can only resolve to 23 bits is misleading. To avoid confusion, the sign bit is clearly part of the resolution. Audio simply biases the sampled values so that they are always positive and has no sign bit. Just bias the samples and you get the same answer.
Next, floating point can actually represent an additional bit exactly. This is because of the hidden bit. Using the lowest bit of the exponent means that the hidden/implicit bit becomes part of the available integer representation. Add the sign bit and the equivalent is 25 bits. Effectively the lowest order bit of the exponent becomes part of the integer resolution. To be clear, there are 2^25 exact integer values representable with ieee 32 bit float.
But it is a more nuanced than that. For use in audio the dynamic range available in floating point is absolutely useful. You don't care about resolving 150dB below your instantaneous signal. Sure, the value will be slightly imprecise, but your error is always 150dB below your signal. Which is not the case for integer representation. For 25 bit integer resolution you can only get that resolution at maximum amplitude, and it is always worse otherwise.
Now there remains devil in the details, and DSP algorithms must be crafted not to throw resolution away, something that can happen all to easily. But, if anything, 32 bit float is generally more robust. Otherwise, if you stay in the integer world you quickly find you need things like 53 bit arithmetic to avoid compounding rounding errors.
 
Last edited:

bennetng

Major Contributor
Joined
Nov 15, 2017
Messages
1,634
Likes
1,692
If we look at the history of some typical studio levels...

In the open reel era the professional grade machines have +26dBu max level, with 0VU calibrated to +4dBu there is 22dB of headroom to deal with the slow average nature of VU meters.

In early digital era, for example, you can see that this DAT machine has marker on -20dBFS when looking at the meter.

Nowadays +24dBu = 0dBFS is common used, albeit some lower-end or bus-powered equipment often fall into somewhere between +10 to +20dBu.

So more or less in the ballpark of 20dB headroom in a half decade based on past experience and that's a good starting point to deal with recording levels.

Floating point is very useful if you are dealing with a lot of track and plugins in DAWs. For example, when you use a virtual orchestra with separately sampled release tails, multi-velocity crossfade and so on. You can easily have >500+ internal voices mixing simultaneously even when the DAW project only has tens of tracks, and that's before additional bussing, effects and so on. With floating point the audio engine only needs to take care of levels when dealing with plugins sensitive to a particular level (e.g. limiters) and before quantizing to integers. With generic CPUs we have nowadays and hardware level floating point support it is the most efficient way to deal with the data.
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,524
Likes
37,057
In the linearity test it says 1 volt rms. The device is said to accept input values up to 24 dbu which is a bit over 12 volts. Shouldn't the input for linearity be at this level? I'd think that actually would add roughly 20 db onto the curve before it become off on linearity. Which then isn't such a poor result.
 
Top Bottom