You're right, we recently bought a 4e SYS-2 unit...
And not to forget that we have 5 units ATS-2 (and one APX515) for production....If you want to know the quality of this product, have a look at the FFT graph of a 5Khz signal on page 12 of the datasheet.
Yes, both THD and noise performance is state of the art. It doesn't get much better than this.
And all measurements are done from analog to analog, so it will be even better if you use the digital input!
You're right, we recently bought a 4e SYS-2 unit...
Has Bruno one at home?
And the DLCP spec's are almost touching the THD+noise floor of the AP system:
AP High Performance Audio Analyzer & Audio Test Instruments : 2700
AP High Performance Audio Analyzer & Audio Test Instruments : 2700
Maybe, but it is mainly in the design for the integrated measurement system!
care to elaborate ?
I think one big difference between the minDSP and the Hypex is that the miniDSP kits (expet the openDRC) operate on fixed point arithmetic for its biquad calculations.
Rounding errors consequences will be different between the two...
I am awaiting for the miniDSP minisharc (based on openDRC hardware) with all the filtering done in one FIR convolution (direct convolution!).
Rounding errors consequences will be different between the two...
I am awaiting for the miniDSP minisharc (based on openDRC hardware) with all the filtering done in one FIR convolution (direct convolution!).
i listen mainly to 80´s hard rock , this happens at any volume level , compared to my cd player connected directly to amp it lacks clarity (even with eq all off) , like there is a curtin muffling the sound.
Yes.......
Why should you use Hires 24bit 96khz with a computer connected to your system by USB???? All the advantages will be gone because of all the noise your computer transmits in your system...........
In Holland we call it putting the horse behind the carriageMaybe because a well-implemented USB connection can be more reliable and have less jitter than S/PDIF?
Yes.......
Why should you use Hires 24bit 96khz with a computer connected to your system by USB???? All the advantages will be gone because of all the noise your computer transmits in your system...........
Why should USB be inherently more noisy than any other protocol, isn't it just dependant on how the connection is realized on the receiver side?
It's not like SPDIF cannot transmit noise into the system if you direct couple everything.
A USB connection is not galvanic isolated, a proper spdif implemantation is and AES EBU for sure!
Both the TAS in the DLCP and the SigmaDSPs in miniDSPs are fixed point. The difference is the SigmaDSPs are 28 bit coefficients, feedback, and data path with 28 or 56 bit MAC whereas the TAS3108 is 28 bit coefficients, 48 bit data path (and presumably feedback), and 76 bit MAC. SHARCs are 32 bit floating point with 40 bit math or 32 bit fixed point with 80 bit MAC---I believe with either it's 32 bit coefficients and 32 bit feedback. miniDSP's miniSHARC blurb refers to floating point but doesn't actually say if they'll use floating or fixed point. From a numerical precision standpoint fixed is preferable due to the deeper MAC but I'm not familiar enough with the SHARCs to speak to the implementation tradeoffs.
FIRs are not usually particularly numerically sensitive so you may not care about this. IIRs are computationally quite a bit more efficient, albeit at the expense of being numerically fussy when realized as biquads centered at low normalized frequency---as a first order approximation, numerical error is inversely proportional to the center frequency. When outputting 24 bit data 32 bit coefficients and feedback with 64 bit MAC with typical fixed point scaling tend to result in more than half a bit of noise coming into the 24 bit samples at normalized frequencies around 0.02---call it 900Hz for redbook, 1.9kHz at the DLCP's resampling frequency. As higher end audio DACs are in the 19 to 23 bit range this is nominally OK but somewhere around a normalized frequency of 0.002 maintaining precision requires more coefficient and feedback bits.
So, for subwoofers, room correction, midbass crossover and EQ, and similar things all of these platforms are numerically a little sketchy for biquads. Loosely speaking, the order from worst to best is SHARC floating point, SigmaDSP, SHARC fixed point, TAS3108. I'm not aware of any blind test results as to how audible the numerical noise floors are---plus it'll depend on what else is in the signal path (DAC, amps, speakers, room, listener's ears)---but my experience with other numerical issues in audio processing leads me to consider 32 bit coefficients and feedback with 64 bit MAC the min bar for general purpose IIR. 48 bit coefficients and feedback with 64+ bit MAC is probably good enough for just about anything currently in audio, though I haven't looked at it too closely as it's computationally more efficient to implement 64 bit coefficients and feedback with 96 bit MAC on the platforms I'm targeting. So something along the lines of the TAS3108 is needed for the biquad numerics in a DSP implementation to be good enough that performance limits on the DAC (if one's going to shave bits the coefficients are the place to do it).
In comparison FIR is, well, inelegant. Bruno's pointed out---and I happen agree---that if one's after good sound quality most FIR synthesis isn't really in the right direction. That'll improve over time but, at the moment, I would say the approach used on the DLCP is fairly difficult to beat and the miniSHARC isn't really competitive on quality. Price, maybe, but that hasn't been announced yet.
Last edited:
Yes. But if you use it you will need a complete seperate power supply and grounding scheme in front of it.
And you have to intergrate it on the pcb, means more space> more costs.
I think Hypex made a valid decision not to implement a state of the art USB audio streaming option. A lot of people are moaning about the price already!
BTW, with a Minidsp product you have to buy a seperate USB streaming board!