• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). Come here to have fun, be ready to be teased and not take online life too seriously. We now measure and review equipment for free! Click here for details.

Zoom F6 Portable Field Recorder Review

bennetng

Major Contributor
Joined
Nov 15, 2017
Messages
1,324
Likes
1,269
Sony PCM-D100 I mentioned in another thread. @jerryfreak also replied me as well:
https://www.audiosciencereview.com/...est-spec-adc-chip-currently.13469/post-405504

RMAA results:
https://hydrogenaud.io/index.php?topic=107528.0

Read the txt file in the attached 7z file, and see the dynamic range vs THD+N plots, you can clearly see that they have different noise floor. The RME involved in the test was from the studio I worked from, and the use of headphone output was because all other outputs are balanced TRS.

[EDIT] For convenience I attached the plots here.
summary.PNG

dynamics.png

thd.png



As you can see the results are not very different from one of jerryfreak's posts:
https://www.audiosciencereview.com/...table-field-recorder-review.15668/post-499930

Floating point from the "math" perspective:
https://www.audiosciencereview.com/...depth-should-i-set-my-dac-to.8956/post-255104

Actually, ESS uses 48-bit accumulator on 24-bit DACs and 64-bit accumulator on 32-bit DACs, but the one who prepared that PDF didn't bother to talk about these things.

Audio "resolution" in practice if you really listen to it.

Kind of funny that 16-bit half precision float for example is actually a newer format for something like GPU shaders, so you have 16 bits in one channel and 64 bits in a RGBA vector for example.
 
Last edited:

jerryfreak

Member
Joined
Jan 12, 2020
Messages
99
Likes
106
Location
Reno, NV
To add to what I said earlier he is a screen grab directly from the F6 manual.

yes, as an F6 owner, I have read that

doesnt mean i am buying what they are selling. While that may be true theoretically, it doesnt appear to be in context to the use of their unit. Yes maybe in the example above where you are going to a very small bit depth like 8 bit. but 32>24 bit will show no real advantage for the vast majority of material, especially with a unit with a high noise floor like the zoom. Any of the fine details you are talking about will be completely buried in noise. Perhaps the sound devices unit is better equipped to make use of said range.


Again i am posting some real world low-level tests, the key to the blind tests is in the attached file
We already know the zoom is somewhat kneecapped on high level signals relative to pro gear due to its low max input clipping level, relative to its EIN
so really low-level tests are where we expect the unit to shine

Some of these are extreme tests and would be indicative of someone who had literally zero understanding of proper use of their equipment. Some of them are pushing the zoom to its max recording level. In all cases we run into the noise floor of the mic. in a few cases we run into the noise floor of the recorder. Now these arent the worlds lowest noise large diaphragm mics, but they are typically used small condenser mics in the 14-17 dBA self noise range, designed to record material in the 90-120 dB range

ch1 (L): DPA 4011A
ch3 (R): Schoeps CMC64

Source material: Time Difference by Hiromi's Sonicbloom, SACD
playback: Benchmark DAC2 > balanced out > single Dynaudio BM5A monitor at low volume

source material volume was the same for all tests. Volume level (SPL) estimated at 64-70 dB peak, 44-50dB RMS*

room noise estimated at less than 41-47 dB peak, 28-34 dB RMS*. it was difficult to tell if i was running into the noise floor of the recorder though

I also measured room noise at 31 dBA with iphone app. Room was quieter than my tinnitus. Room is in quietest corner of house, nothing electrical in the room except for the dynaudio monitor, the system feeding it via xlr was 2 rooms away.

on one sample (car_noise.flac) you can hear a vehicle accelerating through its gears all by its lonesome at 2AM. Thats through a closed window on a highway about a 1/3 mile away. Long story short, you would be hard pressed to be recording any source this quiet outside of nature settings

recorder was mounted right on top of stand, mics sticking right out the side, about 16" from the monitor
 

Attachments

  • zoom tests key.txt
    1.8 KB · Views: 63

jerryfreak

Member
Joined
Jan 12, 2020
Messages
99
Likes
106
Location
Reno, NV
and id be happy to do any other tests you guys would like once i get my zoom back. unfortuantely we are limited in the dynamic range we can attempt to provide to the unit, as even the best DACs will have limited dynamic range when you turn them down to achieve quiet passages.

If you guys can think of any other way to feed a full range low-level signal to the zoom im all ears (perhaps dac at higher volume>passive attenuator?)

i think amir's AP could have successfully fed a low-level signal with full dynamic range but the zoom is already on its way back to me.
 

jerryfreak

Member
Joined
Jan 12, 2020
Messages
99
Likes
106
Location
Reno, NV
probably 1.3 or 1.4 for both my tests and amir's

1.5 wasnt out when i sent amir the unit

none of the firmware upgrades address any issue in my tests, however
 

DLS79

Member
Joined
Dec 31, 2019
Messages
50
Likes
30
none of the firmware upgrades address any issue in my tests, however

As a developer, I would warn against trusting the release notes. Who knows what someone might have slipped in under the radar to cover up a poor job they did.

I've busted a lot of junior developers over the years for doing exactly that.

you also need to consider they might have fixed one thing and broke something else, due to poor or non-existent regression testing.
 

jerryfreak

Member
Joined
Jan 12, 2020
Messages
99
Likes
106
Location
Reno, NV
it seems to be operating normally per my recording and listening tests, theres nothing 'wrong' with the performance, its just operating inside of the very real constraints of its design

i can try to duplicate amirs distortion measurements with various firmwares, should be easy to replicate something that coarse
 

Blumlein 88

Major Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
12,328
Likes
18,488
How about feeding the microphone inputs the thermal noise of a resistor at 100 ohms, 10 kohms, and 1 megaohm? The megaohm should give you a noise signal of about -95 dbV.
 

bennetng

Major Contributor
Joined
Nov 15, 2017
Messages
1,324
Likes
1,269
yes, as an F6 owner, I have read that

doesnt mean i am buying what they are selling. While that may be true theoretically, it doesnt appear to be in context to the use of their unit. Yes maybe in the example above where you are going to a very small bit depth like 8 bit. but 32>24 bit will show no real advantage for the vast majority of material, especially with a unit with a high noise floor like the zoom. Any of the fine details you are talking about will be completely buried in noise. Perhaps the sound devices unit is better equipped to make use of said range.


Again i am posting some real world low-level tests, the key to the blind tests is in the attached file
We already know the zoom is somewhat kneecapped on high level signals relative to pro gear due to its low max input clipping level, relative to its EIN
so really low-level tests are where we expect the unit to shine

Some of these are extreme tests and would be indicative of someone who had literally zero understanding of proper use of their equipment. Some of them are pushing the zoom to its max recording level. In all cases we run into the noise floor of the mic. in a few cases we run into the noise floor of the recorder. Now these arent the worlds lowest noise large diaphragm mics, but they are typically used small condenser mics in the 14-17 dBA self noise range, designed to record material in the 90-120 dB range

ch1 (L): DPA 4011A
ch3 (R): Schoeps CMC64


Source material: Time Difference by Hiromi's Sonicbloom, SACD
playback: Benchmark DAC2 > balanced out > single Dynaudio BM5A monitor at low volume


source material volume was the same for all tests. Volume level (SPL) estimated at 64-70 dB peak, 44-50dB RMS*

room noise estimated at less than 41-47 dB peak, 28-34 dB RMS*. it was difficult to tell if i was running into the noise floor of the recorder though

I also measured room noise at 31 dBA with iphone app. Room was quieter than my tinnitus. Room is in quietest corner of house, nothing electrical in the room except for the dynaudio monitor, the system feeding it via xlr was 2 rooms away.

on one sample (car_noise.flac) you can hear a vehicle accelerating through its gears all by its lonesome at 2AM. Thats through a closed window on a highway about a 1/3 mile away. Long story short, you would be hard pressed to be recording any source this quiet outside of nature settings

recorder was mounted right on top of stand, mics sticking right out the side, about 16" from the monitor
h sounds best to me. The noise floor is less annoying. I mean less high frequency white noise and less heavy low frequency thump sound in general, something simple RMS can't tell. Perhaps the weighting curves in BS.1770 may do a better job but some also think the original ReplayGain loudness algorithm from David Robinson can do a better job in this respect. Well, perhaps it is also related to the music being played so this specific combination bring out the music signal better. Anyway, my opinion only and I may pick a different one another day. Also the occasional traffic noise and such could be distracting on some other files.

I converted the file to 8-bit in foobar with its built-in noise shaped dither, I can't tell 8-bit one from the original apart even though the noise shaping curve is a rather simple one. The screenshot shows the null spectrum against the original file.
8bit.png
 

scott wurcer

Major Contributor
Audio Luminary
Technical Expert
Joined
Apr 24, 2019
Messages
1,503
Likes
2,762
To add to what I said earlier he is a screen grab directly from the F6 manual.
View attachment 81466

This graphic is grossly misleading and borders on nonsense. There are no floating point A/D's the original data is 24 bit (IMO 32 bit A/D's are 24 bit with 8 bits of noise) with at best 21-22 bits of SINAD. There is no way to create headroom by recording well below full scale as this might imply.
 

Blumlein 88

Major Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
12,328
Likes
18,488
This graphic is grossly misleading and borders on nonsense. There are no floating point A/D's the original data is 24 bit (IMO 32 bit A/D's are 24 bit with 8 bits of noise) with at best 21-22 bits of SINAD. There is no way to create headroom by recording well below full scale as this might imply.
Yes which why I wish someone would do tests to see what it is they are doing and what if anything it accomplishes. Even Zoom isn't claiming it creates a file with 32 bit dynamic range. My assumption would be the low gain of the pair of the ADCs is only used for the higher input signals that clip the high gain ADC. This done well might slightly increase the real dynamic range, but mainly makes sure you almost never clip the result. How they combine the two is an interesting question.
 

jerryfreak

Member
Joined
Jan 12, 2020
Messages
99
Likes
106
Location
Reno, NV
its difficult to understand an dual-adc implementation when single ADCs from two decades ago are right there in performance
 

DLS79

Member
Joined
Dec 31, 2019
Messages
50
Likes
30
This graphic is grossly misleading and borders on nonsense. There are no floating point A/D's the original data is 24 bit (IMO 32 bit A/D's are 24 bit with 8 bits of noise) with at best 21-22 bits of SINAD. There is no way to create headroom by recording well below full scale as this might imply.


It's not about headroom, as I've said several times now it's about resolution!
 

scott wurcer

Major Contributor
Audio Luminary
Technical Expert
Joined
Apr 24, 2019
Messages
1,503
Likes
2,762
It's not about headroom, as I've said several times now it's about resolution!

The resolution is determined by the A/D which is binary there is no way to improve it by simply changing the numerical representation. Someone mentioned multiple converters summed in some way I don't see where that came from. This was proven as virtually impossible decades ago in medical imaging applications, the gain tracking just didn't work.
 

DLS79

Member
Joined
Dec 31, 2019
Messages
50
Likes
30
The resolution is determined by the A/D which is binary there is no way to improve it by simply changing the numerical representation. Someone mentioned multiple converters summed in some way I don't see where that came from. This was proven as virtually impossible decades ago in medical imaging applications, the gain tracking just didn't work.

Tell that to Zoom and Sound Devices!
 

Blumlein 88

Major Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
12,328
Likes
18,488
The resolution is determined by the A/D which is binary there is no way to improve it by simply changing the numerical representation. Someone mentioned multiple converters summed in some way I don't see where that came from. This was proven as virtually impossible decades ago in medical imaging applications, the gain tracking just didn't work.
Zoom says they use two ADC's set for different levels running in parallel. Sound Devices uses 3 ADCs. Someone posted a link to the Sound Devices patent on it earlier in the thread. Neither are exactly forthcoming about precisely how they are combining them in the actual product. Even if they have a way of doing this you'll be limited to maybe 130 db or so effective range by the physics of the situation.

From Zoom's F6 page:

The Zoom F6 is the first professional field recorder to feature both 32-bit float recording and dual AD converters, providing an unprecedented amount of dynamic range. With 6 inputs, Zoom’s solid time code, multiple power options and wireless control, the F6 is poised to be your new secret weapon.
 

AnalogSteph

Major Contributor
Joined
Nov 6, 2018
Messages
1,681
Likes
1,548
Location
.de
Someone's got a slight problem with either a noisy voltage regulator or a jittery clock generator, I reckon. Not a practical dealbreaker at all as the device seems to perform well otherwise (I doubt you are going to care much if your noise floor can only be about 85 dB below current signal level), but a sub-85 dB SINAD seems a bit disappointing in something that you can still buy today for a sizeable amount of money, even if it was released 6 years ago. I wonder what happens if you hit it with 10 kHz instead, but again, nobody ever records high-amplitude high frequency tones in real life...
 

bennetng

Major Contributor
Joined
Nov 15, 2017
Messages
1,324
Likes
1,269
Someone's got a slight problem with either a noisy voltage regulator or a jittery clock generator, I reckon. Not a practical dealbreaker at all as the device seems to perform well otherwise (I doubt you are going to care much if your noise floor can only be about 85 dB below current signal level), but a sub-85 dB SINAD seems a bit disappointing in something that you can still buy today for a sizeable amount of money, even if it was released 6 years ago. I wonder what happens if you hit it with 10 kHz instead, but again, nobody ever records high-amplitude high frequency tones in real life...
RMAA has a dual tone IMD sweep tell which call usually reveal jitter issues if it is bad enough, but not in this case.

This recorder also has a dual gain structure (two ADCs with 12dB offset) and I was showing the effect in action. So not "noise floor modulation" kind of distortion. However not very practical since a 24-bit data path is more than enough for a product like this so if one sets the level carefully there is no need for such a gain structure.

But then I agree it is a rather disappointing result, especially at this price. For example, the very first Asus Xonar D2 or EMU 1212m soundcard can do better than this, but they are not field recorders.
 

Blumlein 88

Major Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
12,328
Likes
18,488
RMAA has a dual tone IMD sweep tell which call usually reveal jitter issues if it is bad enough, but not in this case.

This recorder also has a dual gain structure (two ADCs with 12dB offset) and I was showing the effect in action. So not "noise floor modulation" kind of distortion. However not very practical since a 24-bit data path is more than enough for a product like this so if one sets the level carefully there is no need for such a gain structure.

But then I agree it is a rather disappointing result, especially at this price. For example, the very first Asus Xonar D2 or EMU 1212m soundcard can do better than this, but they are not field recorders.
Where did you find out there is a 12 db offset? Just wondering. And thinking that is all, 12 db? Using one good ADC and being careful in choice and design could have provided a lot more than 12 db vs what the tests are showing.
 

bennetng

Major Contributor
Joined
Nov 15, 2017
Messages
1,324
Likes
1,269
Where did you find out there is a 12 db offset? Just wondering. And thinking that is all, 12 db? Using one good ADC and being careful in choice and design could have provided a lot more than 12 db vs what the tests are showing.
No, I did not find it out myself, it is advertised as such, for example in this review:
https://www.solidstatesound.co.uk/sony_pcm-d100.htm

As I said previously, I posted the full review archive with notes to describe the test procedure in this post:
https://hydrogenaud.io/index.php?topic=107528.0
 
Top Bottom