Digital Software RIAA EQ for Vinyl

The RIAA time-constants (3180uS pole, 318uS zero, 75uS pole) are specifications that communicate exact magnitude AND phase behavior of the equalizer for any given frequency. Can't have one without the other.

But there's a 40dB-ish magnitude difference between the top and bottom of the audio spectrum with the RIAA curve, and implimenting this purely in the digital domain will greatly magnify the usual noise vs. headroom vs. distortion tradeoffs which are the basis for the challenges in designing high-quality analog phono EQs.

If this is just a "use what you have lying around" kind of application, then it may make some sense . . . but if you're looking for the highest quality transfers or making an investment into new gear, then you will generally get much better results with a high-quality phono preamp with analog RIAA compensation.

Hi Kirkus,

Indeed there is - and care needs to be taken with the gain/amplification methods.... whether analogue or digital.

But even in the final heyday of vinyl in the early 90's - work was done/proposed and presented at a couple of engineers conventions with regards to Digital RIAA either straight digital or Hybrid.

I have seen lots of talk about the constraints of Digital being worse than analogue for RIAA... but no real measurements.

Also I currently have 3 phono stages - none of which have the required adjustability to optimise MM/MI cartridges properly. They are all set as 47k/220pf.... but some cartridges require 68k100pf or 22k/600pf etc...

Also the main difficulty in analogue RIAA (which also introduces distortion, noise and non-linearities) is that there needs to be enough gain to compensate for the 40db loss in the RIAA filter/EQ. - Same problem either way!

Doing it digitally implies using a MicPre at the input to adjust gain properly into the ADC... the digital RIAA processing is far more trouble free in theory than the analogue version, the problems are with keeping the signal in the optimum range without an increase in noise/distortion etc... which can happen if the signal drops too far.

I did happen to have MicPre that suited, and wanted to try this method out before dumping a bunch of dough for yet another Phono stage- phono stages with the necessary adjustability are rare and relatively expensive...

So an experiment in loading for MM/Mi cartridges turned into an experiment in Digital RIAA...

I will post my results from both experiments in various forums....

Since I've actually been down this road before with design calculations and a prototype . . . here are some of the main issues you're likely to face:

The main issue is that of headroom -- this is very critical when feeding an ADC because of its hard-limit clipping . . . mid-band modulation peaks of +15dB (relative to 5 cm/sec velocity) are common in commercially produced records. Now, a well-designed analog RIAA preamp can be set up to have a similar amount of headroom at 20KC as at 1KC, but this is of course impossible if the EQ is done completely after conversion. Further, with a MM/MI cartridge, the load/cable capacitance causes an ultrasonic peak, which compensates for the natural HF rolloff of the cartridge. This peak can easily be 5-10dB at 25-30KHz . . . and overloading an ADC at the very top of the audioband brings out the very worst aspects of its performance, with big-time aliasing an intermodulation components being common.

So if you add together 20dB for the EQ, 10dB headroom for HF peaking, and 15dB for common modulation peaks, this means that if you set 0dB/1KC at -45dBFS into the ADC, you still have very little real-world headroom. So bring it down only another 5dB for good measure, and most of the mid-band modulation is at -50dBFS, which leaves only 50dB or so S/N on a really good 24-bit ADC . . . assuming the mic preamplifier you're using is perfect and noise-free.

Now if the mic preamp is designed for a low-impedance balanced microphone, then its input En/In characteristics are going to be a marginal match (at best) for an inductive MM/MI cartridge, even if you've adjusted its loading. And since you've gone through the trouble of adapting a mic preamp (assuming you've removed loading resistors, phantom-power blocking caps, input-pad resistors and switches, etc. and added an appropriate loading network for the cartridge) . . . wouldn't it simply be easier to adjust the input impedance and capacitance of you existing phono preamps to whatever you want?

In the end, in my pursuits I found that I could get much better performance with a well-designed two-stage RIAA preamp and a typical ADC eval board. To further pursue digital RIAA compensation, I concluded that it would be best to use a preamp with a fixed single-pole (6dB/octave starting at maybe 15Hz) compensation across the entire audioband, and applying only the precision compensation within the digital domain - this preserves both good headroom and noise characteristics, and makes the analog EQ completely non-critical, as then its tolerances only slightly affect level, not frequency response. Also interesting to me was the idea of using an MM/MI cartridge loaded by an I/V converter into an ADC, thus eliminitaing the effects of any amount of cable capacitance. The equivalent of an appropriate load capacitor could be then applied with DSP equalisation after conversion.

Anyway, just a few thoughts . . . good luck.

Hi Kirkus,

your technical capabilites are well beyond mine....

I understand the strengths and weaknesses of the digital environment (computers are where I make my living).

And yep things get worse... as you drop signal levels below the -50db digital zone (which allowing for +18db signals as a max - as per HFN test records torture track means a real signal at -32db....) the THD % starts to rise substantially... and gets worse as the signal gets lower.

So assuming 70db S/N + 18db headroom (88db !) LP's can definitely push the limits of digital's capabilities.

I am however doing the best of digital by using 24bit ADC (real measured S/N of 118db is not too shabby.... around 20bit res.)

Then staying at 32bit and Floating point in the digital domain....

I had to get a new ADC when my previous one went belly up, it was after I ordered the new one that I realised it had a HiZ (High Impedance - 1M ohm) MicPre (for Ribbon Mic's or Instrument input).... which launched my exercise in MicPre based input.

I was originally planning to purchase a new Phono Stage capable of the flexibility I require for tuning MM's... but this has saved me some $$$ and allowed me to begin measuring and trying out the various cartridges.

I've used pink noise tracks to analyse one of my cartridges at 5 capacitances and 5 impedances (all permutations and combinations) - and graphed the results.... it allows me to then focus in on the optimal settings for each cartridge.

Which is not to say that my ultimate result might not be an Analogue stage....

I also find it interesting that you propose something very similar to what was proposed in the early 1990's - a paper was presented at the AES (? I think ?) proposing a hybrid digital/analogue RIAA design - but I believe it was never taken any further.

Today there are digital RIAA stages both at the budget end (basically analogue with USB ADC)- and some very high end units that appear to be truly digital (only? Hybrid?)

Chances are a hybrid approach makes the most sense until such time as Digital Audio can really move into true 32bit+ resolution at the ADC level. (there's plenty of 64bit software around already...)

With regards to peaks/troughs in cartridge / stylus performance... my Shure 1000e with SAS stylus has an electrical generated peak at around 19KHz, and a mechanically generated peak at 13.5KHz - these add up to what is more or less a HF shelf of up to +6db (with 100k impedance) - I can drop the shelf by dropping the impedance load - but then I get a midrange slump.... I still havn't gotten as far as determining the best balance between loading and EQ.

My initial (analogue based) instinct was to get the F/R as flat as possible across the critical midrange, then use digital EQ to adjust the peak (HF) and rolloff (LF)... with the shure I may also need to depress the 100Hz to 500Hz area (there's a reason they are known as a warm cartridge)

But as you've pointed out - the resonances boost the signal and may cause trouble with the ADC headroom (which reduces my available real effective dynamic range) - so I may be better off lowering the resonant peaks with cartridge loading - and then compensating the F/R variations involved with EQ....

Food for thought - Thank you

David

Dear All,

today is a great day for anybody willing to have a fully-programmable state of the art Digital Phono Preamp for perfect Vinil Restoration, or simply (like in my case) to get rid of the old vacuum-tube preamplifier, sounding good, but having only one, or just a few, RIAA presets and taking one hour to warm-up.

I am pleased to inform you that I did an Audiophile Grade implementation of the following Phono RIAA Equalizations/De-emphasis:

Versions optimized for 44.1 and 48 kHz sampling:
- Official Standard RIAA (1964)
- IEC RIAA (IEC 60098) (1974)
- AT IEC RIAA (2012) - Own optimized version to digitally reduce hum & rumble

Versions optimized for 88.2 and 96 kHz sampling:
- Official Standard RIAA (1964)
- IEC RIAA with Neumann Correction (1995)
- AT IEC RIAA with Neumann Correction (2012) - Own optimized version to digitally reduce hum & rumble

The six EQ settings are provided as six XML configuration files to be imported or hand copied (the EQ section only) into your default/target configuration.

For only 19.90 USD I provide the EQ-files and 1 year maintenance support (any optimization and or adjustment to new MOTU FW version release).

The current implementations have been done and tested using a new MOTU 896 mk3 Hybrid and can work with the MOTU devices that fully support the MOTU Vintage Parametric EQ (i.e.: the new 896 mk3, 828 mk3, UltraLite mk3 and all the legacy and older having the same parametric EQ).

In the coming months I am planning to implement all the known 44 different phono equalizations used so far worldwide (among these are 33/45 and 78 rpm versions).
Each one will be optimized/tuned for 44.1-48 and 88.2-96 kHz sampling (due to input anti-aliasing filtering is not possible to have a single implementation perfectly operating under all the sampling frequencies).

Do not hesitate to contact me for further details and purchase.
Kind regards,
Andrea Tarasconi (andrea dot tarasconi at gmail dot com)