• 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
Any way we can get you to test in 32 bit float the input to the SD card. Then export the file and see if it really can't be clipped. Not that I expect it will improve the noise situation or get close to 25 bit dynamic range. Just wondering if it will scale the various input levels so it works not to clip any signal as claimed. That is the reason anyone who purchases this recorder would choose it over another Zoom model or other brand recorder.
in 32 bit float mode it does in fact 'declip' files over 0dBFS when volume is adjusted in an audio editor
however any signal over +4 dBU will still be ruined
i havent ever used the limiters

at the end of the day per my tests the 32 bit offers no audible benefit over 24 bit, even if levels were set low enough to allow for headroom. if you know your gear its not too difficult to set your levels where it will clip the input before it clips @ 0dBFS
 

DLS79

Addicted to Fun and Learning
Forum Donor
Joined
Dec 31, 2019
Messages
528
Likes
441
at the end of the day per my tests the 32 bit offers no audible benefit over 24 bit

Maybe I'm just naive, but I don't think it was designed to offer more range/benefit. I was under the impression the only reason they used 32bit was to ensure enough resolution existed in the file to handle adjustments in post because of not setting gain.
 

jerryfreak

Active Member
Joined
Jan 12, 2020
Messages
125
Likes
131
Location
Reno, NV
you can do the same in any audio editor. they all do calcs in 32bit FP

noise floor of the source file would be the same
 

DLS79

Addicted to Fun and Learning
Forum Donor
Joined
Dec 31, 2019
Messages
528
Likes
441

AnalogSteph

Major Contributor
Joined
Nov 6, 2018
Messages
3,334
Likes
3,278
Location
.de
Here is my Sony MZ-R3 (1995) MD walkman's line-in directed to toslink output [the opposite: toslink input directed to line-out] without ATRAC compression. Obviously everything is 16-bit but still better than what I see in this review.
View attachment 81340
That's actually somewhat better than the kind of soundcard we would use in the early 2000s (Terratec DMX XFire 1024 w/ CS4294-JQ, a pretty decent little AC97 codec in and out of itself). Not bad for a '95 MD player.

BTW, could you try the same exercise with a 24 bit test file dithered to 16 bit using copious noise shaping (e.g. Audacity should be happy to do this for you)? That should bring out the analog noise floor much better. (BTW, that's one application where bit-perfect output does actually matter.) EDIT: After rethinking, we're probably sufficiently far away from theoretical limits to see much of an improvement at this point. Maybe half a dB.

And yeah, there's something seriously buggered up in this F6, maybe some sort of firmware bug that keeps the ADC combo from working right or whatever. I might get it figured out if I had one to play with myself, and although I could basically use a mobile recorder, that sort of cash outlay seems a bit high for seriously subpar performance if I can't.

As for the excessive low-frequency noise, that would be one of my lesser worries but I'd guess someone liked their coupling capacitors on the small side.
 
Last edited:

jerryfreak

Active Member
Joined
Jan 12, 2020
Messages
125
Likes
131
Location
Reno, NV
You missed my point. It doesn't matter what precision you use in your algorithm, you can't do better than the precision of the source file.

Take a look at this, it kind of puts it in audio terms, but still not good enough from a math/software standpoint imo.
https://www.sounddevices.com/32-bit-float-files-explained/


I understand your point perfectly. a 32 bit representation of high noise floor offers no advantage over a 24-bit representation of same high noise floor. when you bring a 24-bit file into a DAW, it uses 32 bit float in all calculations to prevent rounding errors and then renders a final product

in no case is a 1500+ dB dynamic range useful in a practical sense, as no audio program material exceeds 120 dB, at best
 

bennetng

Major Contributor
Joined
Nov 15, 2017
Messages
1,634
Likes
1,692
That's actually somewhat better than the kind of soundcard we would use in the early 2000s (Terratec DMX XFire 1024 w/ CS4294-JQ, a pretty decent little AC97 codec in and out of itself). Not bad for a '95 MD player.

BTW, could you try the same exercise with a 24 bit test file dithered to 16 bit using copious noise shaping (e.g. Audacity should be happy to do this for you)? That should bring out the analog noise floor much better. (BTW, that's one application where bit-perfect output does actually matter.)
The test was done several years ago, at that time the LCD was partially dead, now it is completely dead so I can't do any test anymore. That's also why I made a mistake and claimed it was line-in directed to toslink output. Luckily I took a photo and some notes in my test archive, just realized that I was talking about something exactly opposite.
MZ-R3 setup.jpg


I suppose you suspected it can do better by using noise shaping... too optimistic. Here is 16-bit J-Test, the same one that Stereophile uses.
1208JittfIg02.jpg


My result with 24-bit recording:
jtest.PNG
 

DLS79

Addicted to Fun and Learning
Forum Donor
Joined
Dec 31, 2019
Messages
528
Likes
441
I understand your point perfectly.

when you bring a 24-bit file into a DAW, it uses 32 bit float in all calculations to prevent rounding errors and then renders a final product

I'm sorry but you don't.

if you take a high resolution data and then scale/round it to fit into a lower resolution file format, you have no way of getting the resolution back out of the file (it's gone).

People here seem to be stuck on the noise floor issue....
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,523
Likes
37,056
In regards to the F6 until you get something from the SD card you could have a very decent ADC with a barely good enough DAC mucking up the results. You can go listen to recording of the F6 which doesn't sound like the measurements indicate.
 

DLS79

Addicted to Fun and Learning
Forum Donor
Joined
Dec 31, 2019
Messages
528
Likes
441
In regards to the F6 until you get something from the SD card you could have a very decent ADC with a barely good enough DAC mucking up the results. You can go listen to recording of the F6 which doesn't sound like the measurements indicate.

I agree, and I would say the review is incomplete until it is done!
 

scott wurcer

Major Contributor
Audio Luminary
Technical Expert
Joined
Apr 24, 2019
Messages
1,501
Likes
2,821
in no case is a 1500+ dB dynamic range useful in a practical sense, as no audio program material exceeds 120 dB, at best

There is no 770dB or 1500dB dynamic range, the noise floor of FP numbers is determined by the length of the mantissa not the total range of representable numbers. The usefulness of FP here is the ability to minimize errors when results are normalized for conversion to a binary output format.
 

jerryfreak

Active Member
Joined
Jan 12, 2020
Messages
125
Likes
131
Location
Reno, NV
correct. and i am submitting that recording with the zoom in either 24 bit or 32-float will yield identical results when brought into a DAW and results are normalized in 32bit FP in either case

I'm sorry but you don't.

if you take a high resolution data and then scale/round it to fit into a lower resolution file format, you have no way of getting the resolution back out of the file (it's gone).

yes thats correct, the extra data containing all that low-level noise is in fact gone. will you miss it?
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,523
Likes
37,056
correct. and i am submitting that recording with the zoom in either 24 bit or 32-float will yield identical results when brought into a DAW and results are normalized in 32bit FP in either case



yes thats correct, the extra data containing all that low-level noise is in fact gone. will you miss it?
We don't know. It may record with more range to the SD card. Until that is checked we don't know how well it records or at what level or whether switching to 32 bit expands its range of operation more than 24 bit recording gives. Not that in any case it would achieve 24 bit levels, but we don't know if the operation of the parallel DACs is in use with both 24 and 32 bit recording to the SD card. All of these are the key reasons to get the F6 and were ignored in the review. We know that in a loopback situation the device does an exceptionally poor job.
 

DLS79

Addicted to Fun and Learning
Forum Donor
Joined
Dec 31, 2019
Messages
528
Likes
441
yes thats correct, the extra data containing all that low-level noise is in fact gone. will you miss it?

No no, you are so fixated on low level noise you don't see it.

Here is a similar scenario from the video world. Take a cheap camera that only records in 8 bit, and then drastically under expose a shot, so you are only using a small subset of the file formats available range. Then shoot the exact same shot but recorded at 10 and 12 bit.

In a 32 bit NLE (davinci resolve for example) when you pull the 8 bit footage up to the appropriate exposure level you will get horribly ugly banding 99% of the time. When you pull up the 10 bit footage you will only get banding in the worst case scenarios, and virtually never get it with 12 bit footage. This is because the higher the bit depth, the more discrete values there are to represent a given segment of dynamic range.


Zoom and SD are using 32 bits to store the audio, not because they need a huge dynamic range, but because they want the highest number of discrete values possible within the dynamic range the recorder is capable of working with. That way when the sound guy sets the gain level super low (so he doesn't have to worry about clipping), he won't get the audio equivalent of banding in post when he sets the gain appropriately.


nobody who has listened to those has determined that 32 bit offers any audible advantage with this hardware
Again its not about getting some form of audible advantage.
 
OP
amirm

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,388
Location
Seattle Area
Thanks for responding. Say I don't send them all at once, I just send the Zoom F4 for now. How long do you think the turnaround would be? I'm very curious how it performs as Zoom doesn't list any specifications, just dynamic range, ADC dynamic range is 120dB or greater.
3 to 4 weeks.
 
OP
amirm

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,388
Location
Seattle Area
You really should test field recorders by recording a file to the SD card and analyzing said file. Otherwise you'll be testing the performance of the headphone amplifier instead, which usually on these Zoom products isn't implemented as well as the ADC.
Not in this case. The low frequency rising noise was there with digital capture from ADC (using USB) and pass through its line out/headphone out. It is a design flaw and you can't get around it by recording to SD card. In the original review we thought maybe USB noise was causing it but now we know that is not the case as it shows up in stand-alone mode as well.

Anyway, I shipped the unit back to its owner.
 

DLS79

Addicted to Fun and Learning
Forum Donor
Joined
Dec 31, 2019
Messages
528
Likes
441
Not in this case. The low frequency rising noise was there with digital capture from ADC (using USB) and pass through its line out/headphone out.

Did you validate the firmware was up to date?

https://www.zoom.co.jp/news/F6-update
New Firmware V1.5 for F6 MultiTrack Field Recorder has been released to correct the following issue:
  • Noise is sometimes generated when F6 outputs low-level signal to LINE OUT.
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,523
Likes
37,056
guessing you didnt heck out any of the low level recordings i did here https://www.audiosciencereview.com/...table-field-recorder-review.15668/post-500023

nobody who has listened to those has determined that 32 bit offers any audible advantage with this hardware
I did, but it wasn't clear to me what some items in the chart were showing. I intended to ask about it, and it slid thru the cracks. So can you make clear what the chart is showing us if it is not too much trouble. Or do you cover that in later replies in that other thread on the other forum?
 

DLS79

Addicted to Fun and Learning
Forum Donor
Joined
Dec 31, 2019
Messages
528
Likes
441
To add to what I said earlier he is a screen grab directly from the F6 manual.
manual.png
 
Top Bottom