Hi,
or the OPA861 which is the 860 wo Buffer.
The sim-model for the 861 I have says it´d be not reliable regarding THD.
Still though with 4mAp-p Input current, Riq=100Ohm and an RL of 2kOhm it sims at 0.07% THD@1kHz.
It sims a low and very constant input impedance and allows for Iq current setting.
Seem to be of similar or equal internal structure as the AD844, but looks more promising on paper than the AD844.
jauu
Calvin
or the OPA861 which is the 860 wo Buffer.
The sim-model for the 861 I have says it´d be not reliable regarding THD.
Still though with 4mAp-p Input current, Riq=100Ohm and an RL of 2kOhm it sims at 0.07% THD@1kHz.
It sims a low and very constant input impedance and allows for Iq current setting.
Seem to be of similar or equal internal structure as the AD844, but looks more promising on paper than the AD844.
jauu
Calvin
I'm not capable of designing my own spice model, sorry.
I remember I had previously tried that opamp and got worse THD. I just tried it again and got good results, so who knows. I'd have to redesign my circuit board for this opamp though since it is not available as a DIP.
Oh THD is one of the least reliable things to get out of a Spice sim of any flavour - take distortion figures with a huge pinch of salt. Particularly ignore THD in an I/V sim if you're not testing with a staircase waveform (as comes from a DAC) rather a sine (as many do on DIYA).
There's a great paper on how to design Spice opamp macromodels by Bowers and Alexander - both of whom were at ADI when they wrote it. Oddly though ADI doesn't appear to adopt their approach in their own models, go figure
There's a great paper on how to design Spice opamp macromodels by Bowers and Alexander - both of whom were at ADI when they wrote it. Oddly though ADI doesn't appear to adopt their approach in their own models, go figure
I just threw THD out there as part of an overall picture. Obviously I don't expect the simulation to exactly reflect reality. It's just a guide in my opinion. I don't have a source or a way to create a staircase waveform.
Current feedback opamps typically can only have some high value of feedback resistor, like 1.2kohms in the case of the LME49713. So, a buffer in the feedback loop is required since a DAC like the PCM1794a puts out a lot of current and thus requires a lower feedback resistor value in the typical IV opamp circuit.
Just thought I post this PDF by Pedja Rogic if nobody has seen it, about the AD844 being feed from a TDA5141 and compared to a discrete I/V he did, very interesting.
http://www.google.com/url?sa=t&rct=...RlHNoW8V3Xgq5Er4eTSsg&bvm=bv.1357700187,d.aGc
Cheers George
http://www.google.com/url?sa=t&rct=...RlHNoW8V3Xgq5Er4eTSsg&bvm=bv.1357700187,d.aGc
Cheers George
Richard, that's an excellent point. For those of us who may not know how to set up the type of simulated signal source Ricahard is talking about, you can rather quickly gain a first order sense of the potential problem simply by applying high-frequency sinewave test signals to the simulated circuit. Even a 10kHz sine test signal will often reveal dramatic increases in higher order distortion products versus those produced by a 1kHz sine test signal, in particular with feedback based I/V circuits. Ultimately one would need to build, measure, and listen to a promising circuit.
There's a way to do it in LTSpice - but as you use Tina I guess that's only of academic interest.
You might find this appnote from TI enlightening (I most certainly did) in respect of opening new vistas for CFB amps : http://www.ti.com/lit/an/slyt099/slyt099.pdf
Indeed Ken, listening is the best way to develop an I/V circuit, in my experience. I've simulated for hours, days even and my conclusions have typically been that the models suck because, for example, its hard to impossible to replicate in simulation the SQ differences between opamps. I have made a very small degree of headway but without more representative models I'm not returning to sims any time soon for I/V.
Simulation with OP amp models are not very accurate, but I have very good results with discrete circuits. And generally all my designs tend to be discrete. But of course the accuracy of the simulation depends on the quality of the models. spice is a indispensable tool and is been around for long time now , and the semiconductors company keep on providing such a bad models. Is just sad.
Edit: the simulation that Calvin did with opa861 worth nothing as in the model it is a warning saying that distortion simulation is not accurate with the model .
Edit: the simulation that Calvin did with opa861 worth nothing as in the model it is a warning saying that distortion simulation is not accurate with the model .
I'm curious to know what correlations you may have over time noticed between I/V technical circuit elements and subjective SQ. For example, feedback vs. non-feedback, voltage feedback vs. current feedback, discrete device vs. monolithic device, passive resistor I/V node vs. active virtual ground node, those sorts of things. Thanks!
My most recent explorations (in the past couple of weeks) were in relation to attempts to provoke noise modulation. I did get a slight glimmer of it but on further runs with different opamps decided it wasn't reliable enough and abandoned the approach.
What I did was create a multitone waveform (using Audacity) as stimulus and fed this into LTSpice. The waveform had over 100 tones at -40dB level individually, closely spaced into two blocks, placed around 15kHz. It does have to be a staircase waveform so not the normal way LTSpice sims from an external .wav file. Then I'm looking at how much of this gets folded down to lower frequencies by IMD so I put in a steep LPF and look in the time domain at how much comes out of this filter. With this set up, I did notice a small change when the dynamic performance of the opamp changed (slew rate in particular).
As regards listening vs circuit elements, then I've only tried with passive I/V of late. When I was running my AD1955 I found LM6172 sounded best and then with only a very low value (<10pF) feedback capacitor. Nowadays I interpret better sound as lower noise modulation so I figure the lower cap value improved this factor (hence the work with the simulation I mentioned above).
What I did was create a multitone waveform (using Audacity) as stimulus and fed this into LTSpice. The waveform had over 100 tones at -40dB level individually, closely spaced into two blocks, placed around 15kHz. It does have to be a staircase waveform so not the normal way LTSpice sims from an external .wav file. Then I'm looking at how much of this gets folded down to lower frequencies by IMD so I put in a steep LPF and look in the time domain at how much comes out of this filter. With this set up, I did notice a small change when the dynamic performance of the opamp changed (slew rate in particular).
As regards listening vs circuit elements, then I've only tried with passive I/V of late. When I was running my AD1955 I found LM6172 sounded best and then with only a very low value (<10pF) feedback capacitor. Nowadays I interpret better sound as lower noise modulation so I figure the lower cap value improved this factor (hence the work with the simulation I mentioned above).
I always thought it was its bad or at least touchy unity gain stability that stopped Walt from using AD797
I suggest you read his reports a bit closer - he only got the issues when the reg was powering something digital from my reading of it. If it had been stability he could have fixed that up, he knows enough about Bode plots I reckon - in a reg its the stability of the whole loop, not just the opamp.
Walt and I also had issues with 797 stability in our 1994 regulator series. I tried to convince Walt to use another opamp which he eventually agreed to.
The 'touchiness' of the 797 was related to the internal neutralisation, which is a form of positive feedback which can, depending on the feedback loop characteristics, cause instability. That doesn't mean its not a good chip - its a very good device, but you must know what you are doing.
BTW I just read earlier posts about people biasing the Tz node of the AD844 up to 10mA to force it into class A. Please be aware that at those currents, the Tz node is quite nonlinear, causes gross distortion and may also exhibit slew rate limiting. Don't get hung up on class A too much - the 844 as-is is much better than almost all discrete circuitry.
jan
