I am looking for a more suitable solution for the stereo recording setup which I currently use to record classical music and jazz. I have a pair of Sennheiser MKH800 microphones, and I want to record to Compact Flash card. I currently use the mic preamps in my Yamaha i88x, but I have to use an unbalanced line signal from the insert point using a cheap jack-to-RCA adaptor, to plug it into my Tascam HDP2 stereo recorder. 
The Audio & Design DMA2 (reviewed in SOS January 2008) is a firm favourite among the currently available two-channel mic preamps. What's more, it's got an AES output, so it can be connected to Ian Simpson's Tascam HDP2 stereo recorder.
If I could utilise the S/PDIF from the i88x then I would have no problem here, but the S/PDIF input and output on the i88x is mLAN exclusive, so you can't get the direct mic preamp signal from there. You can only route it to the unbalanced insert point or the monitor output.
I am after the best possible quality in terms of A-D conversion and mic preamps for around £1000. I would also like to convert from analogue to digital as early in the signal chain as possible, so ideally I'm looking for a high-end, two-channel mic preamp with an S/PDIF output. Alternatively if there are any other rackmount Compact Flash products in the pipeline with AES input, then a mic pre with an AES output would be even better.
I am also keen, despite scepticism from others, to record at 88.2kHz rather than the now almost standard 96kHz as I believe that, as the final destination is audio CD, then the algorithmic processing involved in bouncing down to 44.1kHz is much simpler and therefore likely to be more accurate. I have proposed this to various industry contacts and had very mixed responses. What's your take?
Ian Simpson
Technical Editor Hugh Robjohns replies: The first and simplest option, which I presume you have already considered and discounted, is to use the mic preamps in the Tascam machine itself. These aren't bad for a machine of this type, and they can provide phantom power for your Sennheiser mics. That would obviously be the lowest-cost solution, although not the absolute best quality route.
After that, it really depends on how much of your budget you want to spend. There are numerous two-channel mic preamps around with built-in A-D conversion. At the high end (but stretching your £1000 budget) you could look at the Neve 1073DPD, for example. But being a little more realistic, I would suggest the Audio & Design DMA2 (reviewed in SOS January 2008, and on-line at www.soundonsound.com/sos/jan08/articles/audioanddesigndma2.htm). This is an extremely neutral and accurate preamp, with resettable digital controls, excellent A-D converters, and a really well thought-out feature set. Best of all, it is astonishingly well priced and will meet your needs perfectly.
The output meets, technically, the AES3 standard, but you won't have any problems coupling it to the Tascam's S/PDIF input provided you keep the cable reasonably short and wire the RCA connector between pins 2+3 (use twin-core screened cable, but leave the screen connected to pin one at the XLR end, and isolated at the RCA end). If you want to be technically pedantic, you should use a 'balun' to convert between the AES3 output and the S/PDIF input — and appropriate in-line adaptors are available from suppliers like Canford Audio and Studiospares — but I have found that they are rarely required in practice.
Regarding your desire to record at 88.2kHz rather than 96kHz, the DMA2 can operate at either sample rate, so you'll have no problem there, although in fact the 88.2kHz mode wasn't included in the original design and was added later following user feedback (which explains the slightly odd sample-rate indicator arrangement!)
The idea that integer sample-rate conversion is more accurate than non-integer sample-rate conversion — which is what you are assuming by suggesting that 88.2kHz is easier to convert to 44.1kHz than 96kHz — is no longer true. The best of the modern sample-rate converter (SRC) systems perform astonishingly well regardless of the input and output sample rates, because they calculate the precise sample amplitude required at the precise time it is needed. Older (and we are talking 15 years ago) oversampling-based systems did favour integer-related sample-conversion rates, but the technology has moved on a very long way since then. Indeed, the current crop of single-chip SRCs easily out-perform the best D-A converters in terms of noise and distortion.