Using the AD844 as an I/V

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.

Thanks for the idea Ken. I tried it with a 10kHz sine wave and yeah the distortion does go up compared to lower frequencies (but that's fairly normal I think), but not dramatically. Using a 49713 instead of a 49710 made the THD slightly worse. At 50kHz, the 49713 was slightly better.
 
Last edited:
Thanks for the idea Ken. I tried it with a 10kHz sine wave and yeah the distortion does go up compared to lower frequencies (but that's fairly normal I think), but not dramatically. Using a 49713 instead of a 49710 made the THD slightly worse. At 50kHz, the 49713 was slightly better.

Hi, Dirk, I suggest simulating a high frequency intermodulation test. That would come much closer to revealing the potential issues presented by a DAC chip output stair-step waveform than would utilizing a single sinewave source. In your sim tool, you should be able to connect two sinewave sources is series, each with a slightly different frequency. Reduce the signal amplitude of each source by half (-6dB) relative to the amplitude you would otherwise have used with a single source. In addition, reduce by half the internal impedance of each source so that their sum equals the desired source impedance.

Typically, sinewave generator frequency pairs are separated by 1kHz as that will produce an in-band IMD spectral line at a convenient 1kHz. For example, frequency pairs could be 19kHz + 20kHz, or 39kHz + 40kHz. While this doesn't nearly replicate the full broadband spectrum content of a DAC's stairstepped output signal, it can offer you some notion of how poorly (more so than how well, I think) a given I/V circuit under simulation may handle the broadband signals with which it will be presented.

I agree with Richard about the problem of correlating simulated audio circuit results with the actual sound of the circuit. The whole point of circuit simulation software is that it's supposed to save us time and money. However, as has been famously said, the proof of the pudding is in the tasting. Looks like Thomas Edison's dictum about invention being 1% inspiration and 99% perspiration still holds for audiophile circuit development.
 
Last edited:
Hi, Dirk, I suggest simulating a high frequency intermodulation test. That would come much closer to revealing the potential issues presented by a DAC chip output stair-step waveform than would utilizing a single sinewave source. In your sim tool, you should be able to connect two sinewave sources is series, each with a slightly different frequency. Reduce the signal amplitude of each source by half (-6dB) relative to the amplitude you would otherwise have used with a single source. In addition, reduce by half the internal impedance of each source so that their sum equals the desired source impedance.

Hi Ken, thanks for your response. My software does not allow to signal sources. I can put in a custom wave file but don't know how to create one.

So, I gave it a 10usec pulse and this is the response. The response was the same with either the 49710 or the 49713. I don't see any advantage to using the 49713.
 

Attachments

  • 49710 IV pulse response.JPG
    49710 IV pulse response.JPG
    37.4 KB · Views: 1,137
Hi Ken, thanks for your response. My software does not allow to signal sources. I can put in a custom wave file but don't know how to create one.

So, I gave it a 10usec pulse and this is the response. The response was the same with either the 49710 or the 49713. I don't see any advantage to using the 49713.

Okay. The problem with observing waveforms is that you cannot ascertain much about potential frequency domain distortion characteristics unless they are gross enough to obviously change the shape of the signal. On the other hand, time domian distortions appear well controlled.
 
Here's my attempt at correlating simulation with SQ. This LTSpice file pulls in an external waveform which is a multitone (all tones >13kHz) file 'noisemodmono.wav'. The sim schematic turns this file into a staircase waveform and applies it to the circuit you'd like to test - in this instance just a complementary pair of emitter followers. Then there's a steep filter with around 8kHz cut off to remove the original tones. What comes out of the filter is the noise modulation - folded down noise.

I found that playing with the bias of the complementary pair does vary the output of the filter - as you'd expect lower bias gives greater noise mod. I tried this process with some opamps and found the models aren't really representative enough, if you'd like to test it with circuits of your own I suggest trying to build them out of discrete transistors rather than waste time with opamp macromodels.
 

Attachments

  • noise-mod-jig100dB.zip
    1.9 KB · Views: 142
  • noisemodmono.zip
    219.9 KB · Views: 97
Thanks for the files but I'm not sure I can use them in Tina. I'd have to learn LTSpice and then create this circuit in there in order to use them.

I'm not saying that the LME49713 is not superior to the LME49710 as an IV converter. I am saying that I cannot see any evidence in my software that shows it is better. I can only work with what I have unfortunately. I can't afford to order new circuit boards that accommodate the 49713 so that I can test it in real life, even if I had the gear to do that.

I have pretty much given up on discrete circuits because of what John Curl said in the thread about the open discrete op amp design. He said that unless you have a lot of money to throw at a project, then it is better for amateurs to stick with integrated circuits because they give much more predictable performance. I don't have a sophisticated lab to test a lot of different circuits in real life. I can only do basic checks right now in real life. I tested this IV circuit I have in real life using a good quality high impedance input transformer to provide the balanced input signal since I don't have a balanced current source available. I can only measure up to 22kHz, but what I could see showed that the circuit did not add any discernible distortion not already present from the transformer.

So, if I had a big lab and an EE degree I might be able to do something more sophisticated and precise but that's not my situation right now. I really appreciate all the support and help you all have provided, as well as your patience with this simple minded mechanical engineer.
 
Its very unlikely indeed either your sim or mine will show up the SQ superiority of a CFB amp over one with a traditional LTP when the models are manufacturer-supplied macromodels. If you built the two opamps out of discrete transistors inside the sim then its possible you might begin to see differences which correlate with SQ, but then not by THD testing.
 
Its very unlikely indeed either your sim or mine will show up the SQ superiority of a CFB amp over one with a traditional LTP when the models are manufacturer-supplied macromodels. If you built the two opamps out of discrete transistors inside the sim then its possible you might begin to see differences which correlate with SQ, but then not by THD testing.

OK, but where does one get an accurate schematic for either op amp? I know they often show an abbreviated version in the datasheet, but that's not going to necessarily give the same performance as the real life chip. I understand that we can make up our own op amps in sim and study the effect, but how would that provide evidence to use a specific CFB opamp over a specific LTP opamp? It seems like a Catch 22 to me. I can see it could give general guidelines and clues to their respective behavior, but I really just wanted to make something that has a reasonable chance of success. My circuit is based on the standard IV converter as shown in the PCM1794a datasheet. It's nothing special.
 
I guess we two come at simulation from different angles - I use it to get a feel for how a circuit is going to work, what aspects of it are important. So its more an aid to understanding than anything else - as one experienced EE put it, simulation is mainly to gain insights into circuits, rather than specific results.

If all you want to make is something that works, why not copy the TI application circuit for that DAC? They use NE5534 which will suck sound-wise, but then that DAC's nothing special anyway so spending time optimizing for an S-D DAC isn't going to pay that many dividends in the end. Or by 'success' did you really mean sounding great? If so then the sooner you get away from your computer and behind a bench with an iron in hand, the better.

As for a specific CFB opamp, architecturally they're very similar across a whole range of manufacturers so your choice would come down to things there's no need to sim for, like noise or power consumption and required supply voltages.
 
If all you want to make is something that works, why not copy the TI application circuit for that DAC? They use NE5534 which will suck sound-wise, but then that DAC's nothing special anyway so spending time optimizing for an S-D DAC isn't going to pay that many dividends in the end. Or by 'success' did you really mean sounding great? If so then the sooner you get away from your computer and behind a bench with an iron in hand, the better.

I doubt I can hear well enough to differentiate between superb and just good audio, so "sounding good to me" is enough of a goal. I can't hear higher than 10kHz and I have tinnitus so high end audio is pretty much lost on me.
 
actually op amps designed, marketed for DSL service already are intended to operate with complex, wideband, "noise like" multitones in the QAM Mb modem signals

likewise medical imaging, ultrasound demands bandwidth, low distoriton to avoid wiping out contrast - and many superior performance analog op amps are made now for both these markets

and are speced to perform well with signals 1-2 decades above audio


I suspect some of them can even handle audio DAC output - if you incorporate any filtering at all
 
Last edited:
I tried the LME49710 as the I/V from the PCM1704 it wasn't good, the OPA627 was a bit better, but the LME49710 not used as an I/V is quite good, better than the OPA627 but neither as an I/V can touch the magic of the no global feedback of the Pedja Rogic's AD844, not even close.

Cheers George

Did you perform any measurements? The OPA627AP and BP are very expensive.
 
No Dirk, just how it/they sounded, and I did look on the scope for occilations and how the square waves looked and noise, and measure offsets and such to make sure nothing was going on that shouldn't have been, but no distortion measurements were done.
They (all feedback opamps) sound bad in comparison to the "feedback less AD844". Just just they all have a different type of bad and it comes down to which you don't dislike the most, I have found after listening to the AD844.


OK here goes some poetry to describe the sound, has to be done even for the measurement only freaks. If you don't like it don't read it I tried to refrane, but no longer.
It (the 844) gives you time to hear eveything, there seems to be a greater sens of space between the notes so nothing smears together, yet there is an abudance of detail and dynamics you didn't think possible, you get better fullness to each note even at lower volume settings, even cymbals seem to have more body. (forget about thinness and music being squeezed out at you) this doesn't happen with the AD844 at any volume setting. And I haven't even metioned the bass which is powerfull, tight and detailed and seems to go down another octave, yep there is some magic going on in the AD844 Pedja's way compared to all other opamps i/v's.

Cheers George
 
Member
Joined 2009
Paid Member
Here is what works great for my PCM1704, if your doing a sub for another I/V opamp, don't forget to disconnect the feedback loop around the old opamp.
And give thanks to Pedja Rogic for letting this out of the bag, as I think it would be hard pressed for even a discrete I/V to better the sound this can give. And kudos Barrie Gilbert for designing the AD844 and abraxilito for the dac dc offset nulling circuit
Cheers George

If I understood correctly instead the 15v psu + 2sk170 (2sk369) + VR1 + C1 can't be used, I must use the 2nd pic with the +-5V, right? with 20K preset I have to ajust 0VDC at pin 2 of AD844, right?


Thanks for letting me know George, could you confirm my question?

Cheers
Felipe