k brown
Well-known member
I was being 'arch'.
Upcoming 32-bit float Audio Interfaces in 2024
- Thread starter Tranasaurus Reks
- Start date
Help Support GroupDIY Audio Forum:
zamproject
Well-known member
- Joined
- May 11, 2010
- Messages
- 1,580
This is where a masselec transient limiter rules. It’s like tape as Brian mentioned, the hysteresis of the tape eats the transient when you elevate the record level. But also management of your signal through mic choice and position is key. (I’m thinking live music recording.) I remember switching to a ribbon mic on a female voice during a recording session to minimize sibilant voice problems. This was a dialogue session on a tape series about Psychology. The condenser was not good on her voice.
I was thinking of the 1812 Overture, which is why I said cannon and not gun.
Complaining about the dynamic range of digital recording is pretty strange to me. There is way more than can be used on a finished recording.
Complaining about the dynamic range of digital recording is pretty strange to me. There is way more than can be used on a finished recording.
Last edited:
Wavelength
Well-known member
Gold is correct this is just a way of sending or receiving the data. In USB land you can setup the enumeration to support 32 Float, interger PCM etc... and then there is underground stuff for DSD like DoP. But most DSD is just using 32bit interger in a PCM slot to do DSD as DoP is a bit of a pain in the a** and requires 2x the sample rate.
All of these require some DSP, Processor or FPGA to convert the PCM or DSD signal to Float 32. This has been around in macOS since 10.6.8 when Apple supported High Speed USB UAC version 2 protocol.
Both AKM and ESS have 32 bit ADC now, but since the signal to noise ratio is still below 144dB (24bits) that only can contribute to nuance which people do seem to hear.
On playback both AKM and ESS are getting real close to 144dB with their 32bit DAC chips.
But really in the end why are we striving for that when we all know and have some special piece of gear that to us sounds better and probably what like ~16bit signal to noise???
All of these require some DSP, Processor or FPGA to convert the PCM or DSD signal to Float 32. This has been around in macOS since 10.6.8 when Apple supported High Speed USB UAC version 2 protocol.
Both AKM and ESS have 32 bit ADC now, but since the signal to noise ratio is still below 144dB (24bits) that only can contribute to nuance which people do seem to hear.
On playback both AKM and ESS are getting real close to 144dB with their 32bit DAC chips.
But really in the end why are we striving for that when we all know and have some special piece of gear that to us sounds better and probably what like ~16bit signal to noise???
I think that you missed Tascam Portacapture series… very versatile item, and shines in the area where 32bit has real benefit: field recording. And of course you can use them in the studio occasionally… in case if nobody knows how to record anything properly.
Deadly Mix
Audio-engineer
What requires you to need 32bit float? I find sound quality to be far more important in a converter. Even if you're talking about hobbyist-grade all-in-one USB interfaces, where the converters are usually mediocre at best, finding the least harsh/thin/lifeless sounding preamp would be more useful in most cases. I'm not saying you're wrong for wanting what you want. I'm just curious to understand.
For studio use, 32 bit float isn't bringing anything to the table, imho.
For nature recording, or live news events, sports etc. it will prevent some overload in occasions where it's impossible to predict sudden loud noises. 'till now, a second channel at -20 dB is often used, but that's not so flexible...
For nature recording, or live news events, sports etc. it will prevent some overload in occasions where it's impossible to predict sudden loud noises. 'till now, a second channel at -20 dB is often used, but that's not so flexible...
rogs
Well-known member
I use my Zoom F3 as a field recorder for recording bat ultrasonic echolocation signals.... It's 32 bit float format is really useful when you're dealing with wild variations of signal source location , and really wide dynamic ranges that can be encountered, on occasion.
Not having to try and make gain adjustments outdoors - and in the dark! - is a really useful feature.
OK, bit of a specialised function I grant you, but 32 bit float format is really useful where you are making this kind of unpredictable nature recording... or indeed trying to record any type of highly unpredictable sound source....
Not having to try and make gain adjustments outdoors - and in the dark! - is a really useful feature.
OK, bit of a specialised function I grant you, but 32 bit float format is really useful where you are making this kind of unpredictable nature recording... or indeed trying to record any type of highly unpredictable sound source....
Wavelength
Well-known member
Gang,
Remember what 32float is: 1 bit sign, 8 bits exponetial, 23 bits fraction. In most cases this is enough for even a good converter reaching close to 144dB signal to noise ratio (full 24 bit accuracy). PCM is 1 bit sign and up to 31 bits fractional from -1 to +1.
The only reason a DAW uses 32float is because it is summing and doing math with multiple stems which can exceed the PCM -1 to +1.
Remember what 32float is: 1 bit sign, 8 bits exponetial, 23 bits fraction. In most cases this is enough for even a good converter reaching close to 144dB signal to noise ratio (full 24 bit accuracy). PCM is 1 bit sign and up to 31 bits fractional from -1 to +1.
The only reason a DAW uses 32float is because it is summing and doing math with multiple stems which can exceed the PCM -1 to +1.
Just curious about practical use -maybe I'm overthinking this and I might be using the ZOOM F3 for an upcoming project. Does the interface avoid clipping during recording by scaling the gain and conversion—resulting in "auto-gain" rides? Or does the file get normalized after the record pass so that the maximum level is some predetermined level (below clipping)? Or do you import the file at 32-bit float to a DAW and then scale the gain to lower peaks below 0 dBFS? Or do you just set the input level to something like -50 dBFS and let the box do its thing and just gain up the file later? I guess I should download the ZOOM and SoundDevices user guides to get this info...
rogs
Well-known member
That's my approach. One advantage of 32 bit float is that - unlike conventional A/D conversion - low level signals are recorded without any loss of resolution. I have found that even using a mic connected to the line input, and then increasing the gain in DAW can still give quite amazing results.
With absolutely no danger of overdriving the input preamp.... The analogue line input on the F3 has an clip level of +24dB ----which should be enough for most microphones!
So, importing a 32 bit file into a DAW and then either reducing - or increasing - levels as required both seem to work well.
That's the only advantage I see. The total available dynamic range will be set by the analog input stage. There is no way to scale that. I can see how having full 24- ish bit resolution at very low levels is a big help in field recording.
Back in the 1990s, I recorded sound effects for a video game and I got to travel to a ranch where the owner had several vintage tanks, cannons, howitzers, and all kinds of rifles and machine guns. He didn't want to waste a lot of cannon shells so I wound up setting up something like six microphones at different distances from the gun (outdoors, of course). I recorded into three different field mixers and DAT machines to make sure I had a wide range of gain settings and pads so as not to overload the recordings. The noise floor was limited by the ambient sound of the field we were in, and the headroom by the mic capsules. We managed to get some excellent recordings—along with lots of overloaded recordings that couldn't be used. A 32-bit recorder would only improve that situation a small amount, since the environment, and not the electronics, dictated the dynamic range. For production sound on a quiet movie set, 32-bit could be set to record shouting and still sound good when capturing whispers or breaths.
For studio productions, I don't see much advantage to using a 32-bit interface. In the studio, I still see the biggest challenge as getting the engineers not to needlessly overload the analog front end (mic pre/comp) before the interface.
I'm in the market right now for a new field recorder for interviews and podcasts and I feel it's still early to invest in the 32-bit tech, but I'm sure the 32-bit SoundDevice MIXPres are solid and will be a good investment for at least a few years into the future. I've owned the original MIXpres and they had excellent preamps.
Barry Hufker
Well-known member
I can't say anything about the newer Sound Devices products as I have not had any first-hand experience with them, but I've had a 788T for 15 years and it has been outstanding and faithful all this time. I recently sent it in for full refurbishment because I intend to use it another 15. But if I were purchasing a new recorder it would be Sound Devices. For my needs it would probably be a MixPre II.
k brown
Well-known member
Just curious - what about it needed refurbishing?
Barry Hufker
Well-known member
Some XLR connectors had seen a lot of use, some rubber and foam (internal “shock mounts” had worn down. I broke a “header” of a ribbon cable for the internal hard drive connection…. It was mostly minor stuff and things that just wear with age.
It has been a wonderful machine. And the service was top-notch and very friendly. I have a machine as good as new!
It has been a wonderful machine. And the service was top-notch and very friendly. I have a machine as good as new!
Wavelength
Well-known member
The format (i.e. float vs PCM) will have no effect on clipping or adjusting the input signal. Some field units have what's called an autoleveler which is somewhat like a compressor in that it measures the incoming stream and creates a DCV output and that goes into a VCA to adjust the signal to best fit the ADC input capabilities. Other systems rely on the user to adjust the level going into the unit.
This has been a big problem with digital effects for guitar and bass. Depending on where the unit is in the chain or the in the FX loop the input signal can be less than 1Vac rms or up to 12Vac rms. That's a problem unless your unit has a compander or autoleveling circuit of some kind.
Most manuals and advertising will say this since it's a big selling point.
Deadly Mix
Audio-engineer
From what I remember of the topic back in college, the added headroom is the result of the "float" portion of "32-bit float". It's short for "floating integer". The additional headroom is simply a result of mathematics.
As for why the feature is readily available on cheap low-grade/hobbyist all-in-one interfaces and rarely, if ever, available on high-end professional converters; the answer is, once again, mathematics.
The nature of floating-integer guarantees it can never be as accurate.
The problem with USB mics like RØDE, Blue, and all the others (except the one Neumann makes) is the entire USB-mic concept itself. A digital mic is an all-in-one device, like a Scarlett 2i2. That means the Mic, the Preamp, the AD Converter, and its Interface are all included. Not only do they all need to be cheap enough to appeal to the budget hobbyist market they appeal to, they also need to fit inside a little microphone chassis. Subsequently, compromises and sacrifices have to be made to accommodate both of those matters. All the limitations those factors impose, on top of the individual devices usually being made by companies who don't excel at some or all of the types of devices in them when space, shape, and cost are less of a hurdle, impact the quality.
The exception I mentioned for Neumann's short-lived D-01 is only a partial exception, due to the fact that they set-out to make a premium product at a premium price, which required less compromises.
As for why the feature is readily available on cheap low-grade/hobbyist all-in-one interfaces and rarely, if ever, available on high-end professional converters; the answer is, once again, mathematics.
The nature of floating-integer guarantees it can never be as accurate.
The problem with USB mics like RØDE, Blue, and all the others (except the one Neumann makes) is the entire USB-mic concept itself. A digital mic is an all-in-one device, like a Scarlett 2i2. That means the Mic, the Preamp, the AD Converter, and its Interface are all included. Not only do they all need to be cheap enough to appeal to the budget hobbyist market they appeal to, they also need to fit inside a little microphone chassis. Subsequently, compromises and sacrifices have to be made to accommodate both of those matters. All the limitations those factors impose, on top of the individual devices usually being made by companies who don't excel at some or all of the types of devices in them when space, shape, and cost are less of a hurdle, impact the quality.
The exception I mentioned for Neumann's short-lived D-01 is only a partial exception, due to the fact that they set-out to make a premium product at a premium price, which required less compromises.
