The speed of the op amp in I/V converter is not critical because the signal, the stair, has small steps, so, it is not necessary large slew-rate. You can use almost any good op-amp. You can use opa1612, opa1656, opa2156, opa1642, AD826, LM6172, LM4562, LME49860, LME49720, LT1028, etc, etc, etc...
Its not that simple for a sigma-delta dac. What do you suppose accounts for the distortion shown below?
High speed OP with a large slew rate is absolutely necessary in I/V because of a high frequency components in a signal.
Or to use some LPF between the DAC and I/V which is not usually used.
It is not simple for any DAC without LPF inside, even for so lovely R2R.
In sigma-delta, the frequencies are simply even higher.
What can be done with that impulse signal from the op amp perspective ? It looks to has aprox. 10khz-20khz frequency for fundamental signal with a lot of harmonics. Any decent op amp can reproduce that but the problem is not at op amp. Most likely that parasitic signal need to be filtered. Can you share us entire paper ?
High frequency components from the signal has a small amplitude, I don't know, maybe 1mV. There is not a real problem for a decent op-amp with slew-rate of 10v/us.
High frequency components from the signal has a small amplitude, I don't know, maybe 1mV. There is not a real problem for a decent op-amp with slew-rate of 10v/us.
1st of all - it is not a voltage, it is current (we are talking about I/V?).
2nd - even if it will be a voltage - 1mV from 2Vrms signal -is -66 dB, which is equivalent to ~11bit conversion.
3rd -these are frequencies, together with these harmonics - tens of MHz!
This frequencies creates additional noise and distortions in OP's input stage, especially if it is bipolar.
For DAC with the voltage output (AK440x, for example) is possible to put passive Low Pass filter (RC) before OP, bit for the DACs with the current output it is more complicated.
1st it's not about current, it is about voltage. At the output of the I/V converter is voltage and the opamp must achieve a virtual ground (very low impedance) at its minus input. The op amp itself has nothing to do with the current coming from the dac, it works only with the voltages both at the input and at the output. The resistor Rf should work with that current, not the op-amp. The op amp just keeps the voltages where they should be. A capacitor, Cf, is placed over the resistance Rf, which has the role of having a very small capacitive reactance at the frequency where there is noise comming from the DAC. So, in that situation, the voltage of the noise at the output of the op amp will be low enough so that the reaction speed of the decent op amp will not be exceeded. The noise with Mhz range it should be removed with the next stage, LPF. In my opinion, the op amp is not the only important thing, but rather we need to see the wiring and the parts. That is, the parasitic elements involved there become extremely important. These parasitic elements must be minimized so that the filter scheme works correctly. The grounding route is essential as well as the energy supply.
What you said is a basic model of a dac and an opamp and how they interact together. It turns out the reality is more complicated, especially in terms of what kind of noise and transients are present on the analog outputs of a modern dac chip such as ES9038PRO.
The distortion residual image in post #42 was taken from an internal study at ESS. They were trying to understand what was causing the so-called 'hump' distortion. The remainder of the study is not available.
Anyway, if you take an RF spectrum analyzer and an E-field probe and measure the emissions at the top surface of a dac chip, what you find is garbage extending up into the hundreds of MHz, if not GHz. Some of it is clock edge harmonics, and some of the harmonics appear at the analog outputs of the chip. In addition, dac modulators are very complex high order feedback circuits that are subject to instabilities, and that produce noise artifacts dependent on the nominal instantaneous DC level of the analog output, while at the same time also being dependent on the present values of the modulator state variables (i.e. its history). IIUC some the noise effects can be measured and some can only be modeled, with the models then being studied.
We know that more recently designed opamps are specified for RFI/EMI tolerance at their inputs. LM4562 and its variants such as LME49720 are relatively poor at RFI/EMI immunity. They are especially bad for I/V use. There are a couple of threads over at ASR by member PMA that looks into measurements of conducted and radiated EMI/RFI sensitivity, if you want to take a look.
Empirically, we also know the both ESS and AKM have settled on OPA1612 as their I/V opamp of choice. As I tried to explain earlier in this thread, if OPA1612 doesn't sound good for I/V then there is some other problem in the dac board design that is wrong.
Last edited:
Combine this with glitch-killer capacitor you get a nice 2nd-order lowpass with ESS' "current" output which appears to be switched currents cells to GND pulled-up with an on-chip silicon resistor, so actually medium impedance. The series resistor appears as series input impedance at audio frequencies and can be counteracted with a little bit of positive feedback from the output (ESS seems to do this feedback for the ES9039pro to even compensate bonding/metalization resistances so that the silicon resistor effectly see zero voltage modulation, for lowest distortion).
Yes, because ESS9038Pro output is not real "current output" but voltage output with 200 Ohm series resistor.
Without the actual study the picture in post #42 is useless. We have no information of the DUT or the test setup. AFAIK nobody has been able to publish similar results either with ES9018 or more modern ESS dacs. Besides looking at the graph the noise peaks are at about -20dB. That should be easible visible even with a scope (and audible).
Since the pic in #42 shows a distortion residual, presumably there may have been a makeup-gain amplifier involved. The -25dBFS refers to the level of the sine wave test signal. It was probably chosen because the distortion artifacts were worst there. Also IIRC, in the early days ESS was using AD797 for I/V. Moreover, Scott Wurcer declared it was a picture of hump distortion.