Tascam HD P2 Recorder.

For some time I have been making digital recordings using a pair of Pioneer CDRW Audio recorders. These work well, but are limited in two ways. They only record at 44.1ksample sec (16 bit LPCM), and you can only make continuous recordings that last up to 80 minutes. This is for the obvious reason that they conform to the the CDDA standards and have to fit the recording onto a CDRW.

In recent years I’ve increasingly wanted to be able to make longer recordings, and use higher sample rates and sample sizes. Many people use computer ‘soundcards’ for such purposes. But I am wary of that approach. Domestic computers tend to be a bit of a jungle of interference, interrupts, etc, than can easily affect analogue inputs or outputs. And the domestic computer industry tends to ‘churn’ hardware and software. This means that anything you choose may rapidly become ‘obsolete’. You are then sometimes faced with a choice between staying with what you have or upgrading – driven by some requirements – only to find that some earlier hardware or software you rely on no longer works.

I also personally prefer to use operating systems like Risc OS and Linux, not Windows or Mac. Hence there is sometimes a problem finding hardware that works because the makers may simply refuse to provide the relevant drivers, interface, or information.

So I recently decided to start using a new dedicated ‘high definition’ sound recorder – the Tascam HD-P2. Having obtained one I put it though some tests and have since been happily using it.

The HD-P2 can operate at all the standard LPCM sample rates from 44.1ksample/sec up to 192ksample/sec. It can record in mono or stereo. It can also record either 16 or 24 bits per sample. The recordings are written to a FAT32 formatted ‘SD’ removable solid state memory card. Very long duration recordings will automatically be divided into a series of files with no loss. This allows you to make continuous recordings right up to filling the size of the memory available on the SD card you are using. Hence recordings of many hours duration at high sample rates are possible.

It is easy to take the card out of the recorder, use a card-reader, and connect it to your favourite computer to access the recordings. I find this works perfectly with my RISC OS and Linux machines. I’d expect it to also work with the more common operating systems. As a result, combined with suitable software, the HD P2 makes a valuable measurement lab tool in addition to letting me make high quality recordings.

The recorder has RCA analogue line in and out sockets, along with a pair of XLR inputs for microphones or line inputs. It has SPDIF input and output coaxial phono sockets. Hence it can make direct digital recordings at rates up to 192k/24 bit and play them back via an external DAC like the Cambridge Audio DACMagic. This means it can also be used as a small player for high resolution recordings. You can also put recordings onto the SD card to play with the HD P2. So use it as a small compact player for high definition LPCM files. Although the user interface isn’t ideal for that purpose. You also have to ensure any files comform with the file format, etc, which the Tascam records.

The following section shows some of the results I obtained when assessing the recorder. A later section outlines the main details of the recorded files and how they are arranged by the Tascam’s filing system.

Test results
In general the Tascam test recordings I made showed that the recorder worked very well indeed. As you would expect, recordings from SPDIF input were bit-perfect. I didn’t try the microphone XLR inputs as I don’t have any suitable microphones at present. So I just examined the behaviour of the RCA analogue line inputs.

Fig1.gif - 35Kb

Figure 1 shows a spectrum of an 176·4ksample/sec 16 bit recording of a 1kHz test tone. The source was a 1 kHz tone I’d previously generated and written as a track on a test Audio CDR. Played back on a Pioneer CDRW recorder. The test tone was dithered with triangular probability distribution noise to obtain nominal suppression of coherent quantisation for the source. The output level delivered to the Tascam inputs was 2.1Volts. I then adjusted the input gain controls of the Tascam to get a value of just under -6dB on its recording level meters. The spectrum shows two features which attracted my attention. There are quite noticeable components at 2 kHz and 3kHz which looks like harmonic distortion. There is also a clear ‘spike’ in the 40-50kHz region.

The Tascam manual does indicate that the nominal input level should be 0dBm (0·775 Volts) but seems to allow for higher levels. The input gain controls also easily allowed me to adjust the level being recorded to below 0dBFS with a 2·1V input. So I’d initially assumed the high level would be OK. It is common for an input line level to immediately pass though a passive attenuator so overload should not be a problem. To check this I re-did the measurement, this time using a ‘passive preamp’ I’d previously made as an external attenuator. This let me reduce the input level presented to the Tascam down to 0·5Volts. I re-adjusted the Tascam input level controls to get a similar metered level.

Fig2.gif - 34Kb

Figure 2 shows a spectrum of the resulting recording. You can clearly see that the harmonic distortion is much smaller than in Figure 1. The spectrum shows that the recorder (and Pioneer player) were able to work with very low levels of distortion. Hence the key point to note for domestic use is that to get best results you should avoid the level presented to the line inputs rising above about 0·7 Volts and not to rely on the internal gain control to keep a larger input below clipping. I don’t know any of the details of the HD P2 but my guess is that it has an active buffer stage before its gain control, and that this may tend to increase the distortion slightly if you input larger signal levels.

Close examination showed that the spike in the 40-50kHz region was actually at 44·1kHz. This is the sample rate for the CD audio source used for the above measurements. Hence it was possible that this was coming from the source, not the Tascam. So I decided to try using a purely analogue source to discover if the spike would then be absent.

Fig3.gif - 36Kb

Figure 3 shows the result when using a test LP as the source. This was playing a 0dB 1kHz test tone with a Shure V15/IIIHE. The spectrum exhibits harmonics from the LP/stylus, but you can also see that the 44·1kHz spike still appears. As another check I played tones at various amplitudes and frequencies to see if the frequency or amplitude of the spike altered. But it remained essentially unchanged. Since the results shown were all using the HD P2 with 172·4ksample/sec sample rate (i.e four times 44·1k) I concluded that the spike is an artefact of the HD P2 sampling process. Since it is below -80dBFS it should be fairly harmless for most purposes, so didn’t concern me.

File format
I tend to write and use my own software for analysing, processing, and editing audio recordings. It allows me to check that the processes are as I prefer, but it means I have to know the details of the files being processed. So I investigated the file format and structure used by the HD P2.

The recordings themselves are made in the form of a type of Broadcast Wave File (BWF) LPCM file, and the files are given the ‘.wav’ extension.There can be a series of these files in a particular directory structure. If so they have sequential numbers in their file names.

These ‘Wave’ files don’t have the common plain type of WAV header. Instead the header contain a series of extra metadata ‘chunks’ in accord with the BWF extension to WAV. By looking at the files and talking with the TEAC/Tascam people I was able to work out the basic structure of the headers. What I found is listed in the table below. Please note, though, that this may change depending on the recorder you use, or any changes to firmware version, etc, etc. The following is a general guide, not guaranteed to be perfect or set in stone. Nor does it represent an ‘official’ or fixed set of definitions and details from TEAC/Tascam. It is just my understanding of the header details, and allowed me to access the recordings.

Chunk Contents bytes Memo
RiffID[4] 4 'R' 'I' 'F' 'F'
RiffSize 4 file size - 8 (Little-endian)
WaveID[4] 4 'W' 'A' 'V' 'E'
bwf ID[4] 4 'b' 'e' 'x' 't'
Size 4 858
ckData 1 116
Description[256] 256 TAPE=<number>
Originator[32] 32 0
OriginatorReference[32] 32 TASCAM HD-P2
OriginationDate[10] 10 yyyy-mm-dd
OriginationTime[8] 8 hh:mm:ss
TimeReferenceLow 4 0
TimeReferenceHigh 4 0
Version 2 0
UMID[64] 64 0
Reserved[190] 190 0
CodingHistory[256] 256 0
fmt FmtID[4] 4 'f' 'm' 't' ' '
FmtSize 4 18
wFormatTag 2 01
nChannels 2 02
nSamplesPerSec 4 44100/48000/88200/96000/176400/192000
nAvgBytesPerSec 4 nSamplesPerSec × nBlockAlign
nBlockAlign 2 06
wBitsPerSample 2 16/24
cbSize 2 0
tfid ID[4] 4 'T' 'F' 'I' 'D'
Size 4 26
FileName 8 'T' 'a' 'k' 'e' '0' '0' ' ' ' '
Date[10] 10 yyyy-mm-dd
Time[8] 8 hh:mm:ss
pad ID[4] 4 'P' 'A' 'D' ' '
Size 4 70
?[70] 70 0
data DataID[4] 4 'd' 'a' 't' 'a'
nDataBytes 4 size of data chunk

The recorded (nDataBytes) of sound data then follow in the same arrangement and byte ordering as for a standard LPCM Wave file.

Knowing the above I was able to write a simple desktop program that converts the recordings into standard WAV files and handle the results. I have used the default values above.

The HD P2 records ‘projects’ where each project consists of a set of related files. In addition to the metadata in the BWF/WAV headers there are two additional ‘xml’ files that hold information about the recording like the choice of sample rate, meter display dynamics, etc. These xml files are in 16-bit character text form with the least significant byte first. (‘C’ programmers can also handle this as a series of zero-terminated ‘strings’ each containing one meaningful unsigned 1-byte character).

The HD P2 lets you delete unwanted recordings, etc. And as is common for computer GUI desktops it puts ‘deleted’ files into a ‘Trash’ directory or folder in case you want to undo the deletion. But doing this with the HD P2 can be a bit clumsy even though the recorder allows you to attach a PS2 keyboard to enter commands. So I tend to delete the recorded BWF/WAV files from the SD card using my computer. If you do this note that you also need to reset one of the values in the ‘settings/xml’ file held in that project’s directory. Otherwise the file names will contain a sequential identifying number that increments without noticing that older files have been deleted. The relevant setting in the settings.xml file is <FileCount>N</FileCount> where here I have used ‘N’ to represent the number of files the Tascam has already recorded for that ‘project’. Set this to <FileCount>0</FileCount> if you wish to restart the file numbering having deleted the previous BWF/WAV files.

Note that not all SD cards will be fast enough to be able to cope with recording at the highest data rates. Tascam do provide a list of cards that have been tested and can work fast enough. The HD P2 also provides a test function once you have a card and this will tell you what rates the card in the recorder can handle.

Personally, I am very impressed with the Tascam HD P2. It is compact and portable and can be powered via mains psu, or a set of 8 AA batteries, or via FireWire. It has all the analogue and digital inputs and outputs I need. It records all the sampling rates, etc, I require. I can transfer recordings using the SD card and a card reader. But other users may prefer its FireWire interface which should make it appear as an external mass storage device (I’ve not tested this). The cards are FAT32 format so easy to access with Linux and Risc OS machines using a suitable SD card reader or port. I am using a couple of 4GB SD cards, but it will work with much larger data capacity cards if you require. It can also play back high definition recordings provided they are in the format it expects. It has its own internal clock which timestamps recordings to help you keep track. In fact it has a number of timecode/synch options which will be useful for some professionals working with TV or Film, etc. There are also various other features I’ve not mentioned. So although it isn’t cheap, I think it is worth the money if you need to make recordings from analogue and digital sources at high sample rates as well as at the common 44.1k and 48k at 16 bits per sample.

Jim Lesurf
2200 Words
20th Apr 2010

prev.gif - 2352 bytes ambut.gif - 3891 bytes next.gif - 2248 bytes