Using the AD844 as an I/V

I posted this by accident on another thread anyway here it should be.

Funny you should bring up the AD844 Jan.
I found this circuit of Pedja Rodic's (attached) and used it as is (NO FEEDBACK) and it worked a treat, best opamp based I/V so far by a long way, it blew the OPA627 completely away I though it would never be beaten except for a discrete I/V of some sort. (And I’ve tried just about every opamp out there)
I used it without feedback as per the diagram and was very very happy. Used with TZ as he has it 1.5K with 1000pf to ground, I think this gives a 1st order rolloff –3db @100khz.
Does the 1.5k resistor have a bearing on what input resistance the dac (PCM1704 in my case) sees?
Because I changed the resistor for 2.2k to get some more gain but this time with a 260pf cap –3db @ 270khz (still all nice and stable with a bit more gain) still very very good.
But I had the feeling it was puchier and more dynamic with the 1.5k resistor even though I had less gain. I feel that the input loading for the PCM1704 is better with the 1.5k than the 2.2k resistor or am I imagining things?.

Cheers and thanks for your input George


Added on the 15-09-2023: This should be gone to after reading all this about the AD844, as it gets very interesting when I start stackng them here on the 31-03-2013
https://www.diyaudio.com/community/threads/using-the-ad844-as-an-i-v.227677/post-3434718

Cheers George
 

Attachments

  • ad844_c-b_i-v_sch.gif
    ad844_c-b_i-v_sch.gif
    3.1 KB · Views: 11,103
  • AD844.JPG
    AD844.JPG
    73.5 KB · Views: 9,909
Last edited:
Does the 1.5k resistor have a bearing on what input resistance the dac (PCM1704 in my case) sees?
Because I changed the resistor for 2.2k to get some more gain but this time with a 260pf cap –3db @ 270khz (still all nice and stable with a bit more gain) still very very good.
But I had the feeling it was puchier and more dynamic with the 1.5k resistor even though I had less gain. I feel that the input loading for the PCM1704 is better with the 1.5k than the 2.2k resistor or am I imagining things?.

Probably you're not imagining things but no that resistor won't change the loading on the DAC which is determined by the input (common-base) pair. I guess you heard lower noise modulation from the AD844 with the lower valued resistor. if I was playing with this circuit I'd experiment with different grounding of those decoupling caps, or preferably using shunt regs for both supplies.

Also worth trying a ferrite bead prior to the AD844 -ve input (with shunt resistor for damping - say 100R) to help reject some of the glitches. I use MPZ2012S102. Do this when removing C1 and see what you hear.
 
I posted this by accident on another thread anyway here it should be.

Funny you should bring up the AD844 Jan.
I found this circuit of Pedja Rodic's (attached) and used it as is (NO FEEDBACK) and it worked a treat, best opamp based I/V so far by a long way, it blew the OPA627 completely away I though it would never be beaten except for a discrete I/V of some sort. (And I’ve tried just about every opamp out there)
I used it without feedback as per the diagram and was very very happy. Used with TZ as he has it 1.5K with 1000pf to ground, I think this gives a 1st order rolloff –3db @100khz.
Does the 1.5k resistor have a bearing on what input resistance the dac (PCM1704 in my case) sees?
Because I changed the resistor for 2.2k to get some more gain but this time with a 260pf cap –3db @ 270khz (still all nice and stable with a bit more gain) still very very good.
But I had the feeling it was puchier and more dynamic with the 1.5k resistor even though I had less gain. I feel that the input loading for the PCM1704 is better with the 1.5k than the 2.2k resistor or am I imagining things?.

Cheers and thanks for your input George

I am wondering about the AC voltage at pin 2 of the opamp. The SK170 CCS is basically the equivalent of a really large impedance resistor between the DAC output pin and the +15 volt supply (which is an AC ground). It seems to me that this CCS would act as an IV resistor, but don't know for sure.
 
Hi Dirk, I don't use the SK120 ccs circuit for dac dc offset nulling. I use the one Abraxilito gave me below. I don't know what proplems it creates but it works fine, as my PCM1704K's gave out 17mV and 50mV dc offset and this nulled it to zero and holds perfectly. And then the AD844 (i/v) as I have used it above gives out only a couple of mV, direct coupled all the way through an active (LP) filter OPA627 also direct coupled to an LME49710 as a (single to balanced convertor) then a two fet output buffers on each phase to the balanced or single ended outputs.
Everything is direct coupled from dac to outputs.

Cheers George
 

Attachments

  • PCM1702%20DC%20Offset%20null%20circuit.JPG
    PCM1702%20DC%20Offset%20null%20circuit.JPG
    249.5 KB · Views: 8,835
I I have to spread the joy on this Pedja Rogic's way of implementing the AD844 I/V with no overhaul feedback, it is a game changer.
I just for interest I put on the cd's that I love the music of yet never played, you know the ones, that are massed on another shelf marked never to be played.
Well I got out an old Crosby, Still's & Nash the one where they're sitting on an old lounge on the porch. Every time this was played with any of the other i/v opamps with feedback, it was thin, shrill and smeared. Now with this Pedja Rogic AD844 way of using it, it's like the first time you heard it and loved it on the best analog setup you had back then, gone are the nasties and it's almost like it was done yesterday but now with body and detail and dynamics that you never though possible.
I'm not going to say it's the equal of the discrete I/V stages that are being developed, but it's hard to believe it can get better than this.
And it's been said that the output buffer in it is not that good in the 844 and the output should be taken from the TZ output. Can someone say what the output impedance is at that point, and if it able to drive my next stage which is the opamp based LP filter at the moment a OPA134.

Cheers George

Cheers George
 
Hi Serg, I mean what is the output impedance at TZ without the resistor, the data sheet seems to me to say it could be 1.3megohm to 1.5megohm!!!!! That would mean the input stage of the OPA134 LP filter would have to be at least 10megohm or more as not to load the i/v input stage down.
I have no circuit of the LP filter I assume it's typical of what CDP manuafcturers use, as it is made by Cary. CD-303/200

Cheers George
 
You're right that the DC output impedance will be the output impedance of the current mirrors in the chip, however its audio so the AC output impedance is the one to consider and that will be determined by the shunt capacitor. You're not doing yourself a favour by following up a NFB-free stage with a common-or-garden opamp!
 
You're right that the DC output impedance will be the output impedance of the current mirrors in the chip, however its audio so the AC output impedance is the one to consider and that will be determined by the shunt capacitor. You're not doing yourself a favour by following up a NFB-free stage with a common-or-garden opamp!

Yes I am at the moment but at least it's not trying to do I/V which I think has the more impact on if it's an opamp or not.
But later I will try something discrete I have here ready made (attached) straight off the TZ pin if I know it's not going to load down the TZ output.
Let me know what you think.
 

Attachments

  • fetbuffer 2. circuit.jpg
    fetbuffer 2. circuit.jpg
    41.7 KB · Views: 3,509
  • fetbuffer2.jpg
    fetbuffer2.jpg
    124.4 KB · Views: 2,965
You're right that the I/V stage is the most critical to get linear, so using a normal opamp there is indeed the worst possible place for one. I have no experience of using FETs so I'll leave appraisal of your circuit to others with more relevant expertise. A FET buffer does seem to me to be the right way to go but I'd want to apply at least some passive filtering with a series inductor first.

PS forgot to mention - if you don't load down the TZ node then there's too much gain, you can't just delete the 1k5 resistor, it'll just clip the rails.
 
You're right that the I/V stage is the most critical to get linear, so using a normal opamp there is indeed the worst possible place for one. I have no experience of using FETs so I'll leave appraisal of your circuit to others with more relevant expertise. A FET buffer does seem to me to be the right way to go but I'd want to apply at least some passive filtering with a series inductor first.

PS forgot to mention - if you don't load down the TZ node then there's too much gain, you can't just delete the 1k5 resistor, it'll just clip the rails.

Thanks for that abraxalito, I wondered if I had to keep the resistor to ground on TZ I use a 2.7k with a 470pf across it that should cut the highs -3db @125khz.
So then I'll come off TZ to the discrete buffer, should I take off anything on the input of the buffer to make it easier for the I/V to drive?
Also what do you suggest as filtering on the buffer input? I'm in your hands.
BTW I think it was John Curl who said that the buffer circuit I posted was a good sound design.
Cheers George
 
Well the way I'd go would be to work at a lower signal level rather than a higher one - so reduce that 2k7 down to something at least 10X lower. This is because its hard to get high enough series impedance with inductors coming from such high impedance as 2k7. Inductors always have shunt capacitance which takes over at HF. If your source impedance is low enough then you can start using ferrite beads as inductors which are small, cheap and don't need winding :) These days I'm using long strings of such beads (more than 10 in a row) to get small, cheap inductors - I've recommended a TDK one further up this thread. Each bead needs a shunt resistor - say 100-200R. Experiment with how it sounds as you add more and more beads, you might find this fun, I did :)
 
I mean parallel each bead with its own resistor of that value. So at low frequencies its the inductance which dominates the combined impedance, but as the frequency goes up the resistor takes over. I've had unexplained weird effects without including the resistor - even though the beads are lossy at much higher frequencies there seems to be something odd going on in intermediate freqs (say 1-10MHz).
 
Of course once you're resigned to working with signal levels around the 100mV range then what's the point of doing any I/V with active components? :D Just put the 50-100R resistor directly on the output of the PCM1704 and do the first stage filtering passively. Then once the signal is securely band-limited add a gain stage.