Topic: Deity PR-2 Pocket Recorder

[breakonthru]

aaronji wrote:
> ^ As far as I know, the 32-bit float "no gain staging needed" thing is actually a product of using multiple gain-ranged ADCs, operating on overlapping bands of input, and stitched back together using regression based algorithms.

That would work only if multiple, gain-ranged mike preamps were linked to those multiple gain-ranged ADCs. It's not impossible, but I would never simply assume that a recorder (especially a low-cost one) has such a feature.

Otherwise the gain of the mike preamp needs to be user-adjustable, and the user needs to take care to set it more or less optimally, just as with conventional record levels.

its more than likely just a single-ranging ADC with a simple IC gain circuit that has the <100 dB available to it well within the 144 dB 24-bit container, and 32 bit file format is just useless window dressing. but wont know until i try one. like all devices these have maximum input levels that can be exceeded. if the max input really is +18 dB (crazy high for an unbalanced signal) then it indeed has some headroom to work with, albeit at the expense of a higher noise floor for those hoping to use it as a mic-in

[voltronic]

That would work only if multiple, gain-ranged mike preamps were linked to those multiple gain-ranged ADCs. It's not impossible, but I would never simply assume that a recorder (especially a low-cost one) has such a feature.

I don't understand why you say there needs to be multiple preamp circuits. It's just one preamp with fixed gain, and the signal is split to the two or three parallel ADCs. Aaronji is correctly describing how Sound Devices, Zoom, and (I'm pretty sure) Tascam, and Stagetec do it.

[Gutbucket]

I see one difference in particular with regards to the architecture of this recorder when compared to the SoundDevices and Zoom recorders, but that difference is not the one we are discussing.

I think by design, all of these recorders use separate analog circuit paths ahead of each separate ADC path.  I don't see how they could otherwise provide the dynamic range that most of them do without some kind of strategy that actively adjusted the preamp gain. Rather, each analog circuit > ADC path is optimized and 'locked down in terms of gain-staging' for its particular voltage range.  That strategy provides a way around both dynamic range bottle necks: That of a single analog input path ahead of the ADC and that of a single ADC path itself.  Otherwise in an SD Mixpre II, a single analog preamp path without any gain adjustment feeding multiple ADC paths for each channel would have to be capable of providing SD's claim of "142 dB of dynamic range" minimum on its own.  That's extreme!  I'm no expert but that level of performance may not be possible from a single fixed-gain preamp stage at any price point.

For that reason I don't see how any of these recorders would work to specification without using multiple preamp paths feeding the multiple ADC paths, with each optimized for a particular range.

Here is the difference I see with the PR-2:
Regardless of which recording mode is chosen, the SD and Zoom recorders provide a sufficient number of parallel signal paths to support their total channel count.  When we were discussing the F6 noise-floor modulation issue that cropped up a while back, it was confirmed that these machines always use their extended-range ADC switching strategy regardless of which output file format the user selects.  I'd originally assumed that when using those recorders in 24bit-fixed mode they would revert to a traditional circuit path arrangement, but apparently that is not the case.  The SD and Zoom recorders always do the auto-ADC-switching thing, even when set to save the output as 16 or 24bit.  Its the only preamp>ADC architecture they use, regardless of how many channels are being used or output format.

The PR-2 appears to use a different strategy.  It presumably only has two signal paths available to it, so it doesn't have sufficient parallel signal paths to support a full channel count regardless of recording mode.  Because of that it needs to use a different strategy ahead of the section that does the "digital switching and stitching" prior to the output file being written.  The two available signal paths can either be used separately in a traditional way for 2-channel throughput (stereo), or used together in a "switching and stitching" parallel mode to provide increased dynamic range for a single channel throughput (mono).  In mono mode, provision for the selection of output file formats other than 32-bit float could be made but isn't, which is simply a design choice.  So unlike the SD and Zoom recorders, the PR-2 uses two different path architectures, switching between "traditional stereo mode" and "auto-ADC-switching mono mode" as a way of maximizing the utility of having just two signal paths available to it.

That strategy reduces cost by eliminating the additional parallel paths that would otherwise be required to provide increased dynamic range functionality for two channels. 

In the SD and Zoom recorders we are never exposed to nor aware of the individual paths.  They are always stitched back together again prior to output.  But with the PR-2 in stereo mode, we are using the individual paths separately.  Diety must set up the mono channel switching architecture in such a way that when mono 32-bit mode is selected it places the two channels in parallel and increases preamp gain through one of the two to appropriately gain-stage it for use as the lower level path.  And after that the stitching back together and file writing takes place.

What is stated in the PR-2 specs is the dynamic range of the two paths used together in mono channel mode.  What remains unstated in the specs is the dynamic range through each of the two individual signal paths.

[aaronji]

As I understand it, after the analog preamp, each ADC has its own analog gain block and filter, at least in SD's implementation. That's based on a brief conversation with an SD tech...

[Gutbucket]

Okay thanks, I accept that.  Can anyone here more electrically astute than myself comment on how difficult (or not) it is to achieve a dynamic range of >142dB through a fixed gain preamp?  Am I interpreting that achievement as being more of a reach than it actually is?

[breakonthru]

Okay thanks, I accept that.  Can anyone here more electrically astute than myself comment on how difficult (or not) it is to achieve a dynamic range of >142dB through a fixed gain preamp?  Am I interpreting that achievement as being more of a reach than it actually is?

sounds impossible as EIN gets amplified when you amplify any signal, you need to jigger it a bit and combine multiple signal paths

[Chilly Brioschi]

The FLOPS are in the maths.
The ADC needs multiple paths.
The preamp is before them.
Math and integrated circuits? I abhor them.


I'm going to the beach!

[mrfender]

... you need to jigger it a bit...

Love the technical jargon, my kind of language!

[justme]

Newsshooter have posted a short review.

https://www.newsshooter.com/2024/07/18/review-of-the-deity-pr-2-pocket-recorder-with-32-bit-float-and-timecode-generator/

[Ozpeter]

Thanks for that link, from which I quote

 -"It can’t fix improper mic placement that will distort the microphone, but the recording level can be adjusted a lot in post, just as long as you don’t blow out the microphone. "

That kind of sums up the whole subject of 32 bit float in one simple sentence. Though perhaps one could add "or preamp".

It seems to me that the best use of 32 bit float is for the implementation of mics that record.  The design of such devices can ensure that the mic can handle high levels and that the preamp is optimised for the mic, possibly including corrective eq if necessary.  That can then be passed on to the dual a/d converters and recorded internally without cable losses in 32 bit float.  Such mics exist but how well the theory matches the reality is another whole matter, probably dictated by the price point. 

[unidentified]

I dunno. For me, the best use of this technology is the ability to record gigs without having to check levels as the music rises and falls in volume in ways that cannot be predicted ahead of time.  The stealth utility is obvious.  In open situations, it means one can set up more than one rig at optimal locations on or near the stage that are not accessible to the taper when the music is playing.  The F3's pre-amps are pretty darn good and the recorder is a good match for Line Audio CM3s. 

[Rairun]

I dunno. For me, the best use of this technology is the ability to record gigs without having to check levels as the music rises and falls in volume in ways that cannot be predicted ahead of time.  The stealth utility is obvious.  In open situations, it means one can set up more than one rig at optimal locations on or near the stage that are not accessible to the taper when the music is playing.  The F3's pre-amps are pretty darn good and the recorder is a good match for Line Audio CM3s.

Yeah, I agree, but I think their point is that the recorder's manufacturer could match the analogue clipping point of the recorder to 0 dBS, and then configure multiple ADCs to write to a certain volume range of a 24 bit file. For this to work, you wouldn't be given the option of gain control either. You'd never clip digitally (unless the analogue stage also clipped), and the dynamic range of 24 bit is said to be enough. But my question is... why? The difference in terms of storage is not huge, and your quietest sounds would still be written at -50, -60, -70dB in the file and demand normalisation. Using 32-bit float is a pretty foolproof way of making sure the user will never run into quantization noise or accidentally clip the audio during post-production.

It's also just crossed my mind that distorting the analogue stage sounds very different from digital clipping. Wouldn't you need more digital headroom there to capture the analogue distortion? And if you do, wouldn't you be writing the quietest sounds closer and closer to the quantization noise? And we must remember: just because a sound isn't completely drowned out by quantization noise, that doesn't mean you can't hear it in the background. At which point I'd ask: isn't it just easier to move on to 32-bit float? There are no downsides other than using a little more storage.

[voltronic]

I still think it's pretty silly to have auto ranging ADCs with 32-bit float on a device with rather poor S/N and dynamic range numbers by modern standards. Just as silly as Zoom doing it with the H Essentials line. True, you get the benefit of no level setting. But if the device is going to be built on this architecture, you should also be getting the benefit of increased dynamic range.

[Rairun]

I still think it's pretty silly to have auto ranging ADCs with 32-bit float on a device with rather poor S/N and dynamic range numbers by modern standards. Just as silly as Zoom doing it with the H Essentials line. True, you get the benefit of no level setting. But if the device is going to be built on this architecture, you should also be getting the benefit of increased dynamic range.

Yeah, I agree with you there re: the H Essentials line, especially since it appears to have worse S/N than the H and HxN lines. That's why I didn't buy one of those. I actually used the Zoom H1 for a long time with a Church Audio pre-amp, and I had no issues whatsoever with noise when the music's dynamic range wasn't too great. A song at -20 dB on average with the occasional -6 dB peak? No issues. But as soon as the band got really quiet at -30 dB or less, the noise was fairly noticeable. Increasing the gain on the Church Audio pre-amp would have fixed the issue, but then the peaks would clip.

But is the Deity's mono 32-bit mode as bad as the Zoom H Essential's? I've never seen a rigorous analysis out there, but I heard their mono 32-bit transmitter (the Theos) was really good - one youtuber said it was 'professional' quality, comparable to Sound Devices. I assumed the PR-2's circuits would be essentially the same, but with the option to record 24-bit stereo as well.

[Ozpeter]

I still think it's pretty silly to have auto ranging ADCs with 32-bit float on a device with rather poor S/N and dynamic range numbers by modern standards. Just as silly as Zoom doing it with the H Essentials line. True, you get the benefit of no level setting. But if the device is going to be built on this architecture, you should also be getting the benefit of increased dynamic range.

I see your point, but I guess the thing is that it seems that 32 bit float chips are cheap now, and we might as well get used to that as the new norm.  I look at it this way - dynamic range and distortion are affectd by the mic, the preamp, and the a/d conversion.  If one of these three at least can be taken off that list, it helps.  But the danger is that people may start to overlook the part played by the first two.  Today I came across a video by a college lecturer in digital audio who said in commenting on a post questioning the preamps of the H1e that any noise problems from the preamps would be dealt with by the 32 bit float format.  Of course I pointed out that those a/d converters would make a very good recording of that noise!