Reply 400 of 484, by Jepael
wrote:wrote:Yeah, but has anyone tested to see what the actual cut off for the frequency range for the R-2-R covox is? Can it actually output a 22.05 KHz frequency?
The Covox has no filter at all (and the CVX4 also has a no-filter setting). So theoretically there is no cutoff.
I think the easiest way to verify is just to create a test sample, play it back with CvxPlay and record it with a modern sound card that is high-end enough (like a 24-bit 192 kHz card, that should easily pick up 20+ kHz).
But wasn't the Covox a complete system with amplified speaker, so shouldn't it be taken as a whole?
Even if we assume that the resistor network array does not have a filtering capacitor, there could be a low pass filter on the speaker box.
The wire between the plug itself and any input it is connected to has some capacitance. Audio cables can have as high as 500pF per meter, so given the high output impedance of say about 25 kohm as indicated by measurements, even just 4 meters of cable between the plug and sound card input could act as a 3.2kHz lowpass filter.
So the cutoff can happen quite soon. In reality you could have 4 meters of 50pF/m low capacitance cable for 32kHz filter.
The Covox patent also mentions approximately 5nF capacitor and lowpass cutoff of about 3kHz, but that should not be taken as the absolute truth, as even the R2R network is drawn incorrectly. It is missing resistance from data bit 0 summing junction to ground to be a real R2R ladder DAC, and measurements on the real device indicate the custom resistor network has this.
Given the low sampling rates typically used, I would not be surprised it to have a filter approximately at that frequency to filter out the aliasing artifacts, as every DAC should have an output filter to remove these artifacts as frequencies above half the sampling rate should be filtered out. Given that it really is a first-order RC filter, it is not very steep (20dB/decade), so it will attenuate 20dB at 30kHz, while 8-bit data theoretically has 48dB dynamic range.
The high output impedance of the plug makes it difficult to measure the frequency response, as everyone has different input impedance (resistance and capacitance) on their amplifier or sound card line input, and these are typically unknown, so everyone will get a different result.
Well, what could be done, is to build a good op-amp buffer right at the Covox output, so the cable or line input properties do not matter any more.
Or use extremely short wires and a device with known input impedance (such as direct connection without probes to an oscilloscope, the input resistance and capacitance usually reads on the unit).