I believe that any compliance voltage at 1704 Iout pin is measured by DC servo, correction voltage calculated and then added to compliance voltage at BPO internal on-chip connection to keep Iout pin at exactly 0 V DC; however I need to know what the compliance voltage tolerable range is to be able to design a (correct, I hope...) I/V with current feedback IC.
So, any opinions or experience with this is welcome!
Boky
So, any opinions or experience with this is welcome!
Boky
I'm not sure I fully understand the context for what you are asking, but perhaps the following answer may help. If you will be using a current feedback I/V amplifier it should inherently present a minimal impedance (certainly so within the audio band) virtual ground to the PCM1704 output. Certainly low enough in impedance to not be concerned with issues of the DAC's output voltage compliance, I should think.
Last edited:
doesn't appear to be any usable spec in the 1704 datasheet for Vos
I think the ADA4898 is the current best option on tech specs for audio DAC I/V - AD doesn't properly advertise the significance of the "highly linear input" innovation
I beleive the ADA4899, AD8099 have the same linearized diff pair input but higher GBW - a 300+ MHz 'scope might be needed to assure you properly compensated these faster versions
I think the ADA4898 is the current best option on tech specs for audio DAC I/V - AD doesn't properly advertise the significance of the "highly linear input" innovation
I beleive the ADA4899, AD8099 have the same linearized diff pair input but higher GBW - a 300+ MHz 'scope might be needed to assure you properly compensated these faster versions
Several of us tested this a few years back on the other forum. Most found <30 mV to have the least distortion, beyond that I couldn't measure. I believe most agreed 10 ohms load min.
At the time I had a PCM63k with a 47 ohm passive I/V and swapping to the PCM1704 obviously gave less output but also higher distortion proving to me that the PCM1704 just isn't meant for passive I/V
So this chip has to be loaded down, it even struggles with the Jfet D1. It shines with a "opamp" I/V with a bit of feedback but no so much that the input impedance skyrockets at medium rf, so its a rare case where a discrete opamp is easier than store bought.
You may want to look at the output slewrate and design from there, keeping the input impedance below 10 ohms past the slew rate's equivalent frequency.
So looking at the input impedance of the ADA4898 I/V vs freq to 3mhz and comparing it to the 1704's step frequency would be interesting.
At the time I had a PCM63k with a 47 ohm passive I/V and swapping to the PCM1704 obviously gave less output but also higher distortion proving to me that the PCM1704 just isn't meant for passive I/V
So this chip has to be loaded down, it even struggles with the Jfet D1. It shines with a "opamp" I/V with a bit of feedback but no so much that the input impedance skyrockets at medium rf, so its a rare case where a discrete opamp is easier than store bought.
You may want to look at the output slewrate and design from there, keeping the input impedance below 10 ohms past the slew rate's equivalent frequency.
So looking at the input impedance of the ADA4898 I/V vs freq to 3mhz and comparing it to the 1704's step frequency would be interesting.
Patrick these are back of the envelope numbers.
But you should multiply by 2 (its +/-1.2mA), the PCM1704 outputs a 2.4mA signal.
So 2.4ma x10ohms = 24 mV
10 ohms is what many target probably because it falls in line with the only "public" spec we have for old DAC's which is if I remember right is 25mV on the large version Tda1541 datasheet.
And to be honest if you could get lower than 10 that would probably be better.
On Oleg's webpage there is a very good study of the effect of load on the TDA1541, we know that the PCM1704 is an even tougher bird. As I have found trying various passive I/V loads.
But this more than 10 ohm load target is easy to achieve with an opamp, the problem is that the output of the PCM1704 (even at 8xfs) has steps.
As I am sure you know even as the chip is spitting out a simple 1khz signal it has steps that build the "sine" wave which are close to vertical, but not quite. So the I/V stage has to deal with these high frequency steps, meaning that the input impedance should be targeted to stay below the 10 ohms number well into the mid rf range. I don't remember what the slope of the steps are from the PCM1704 but I think general rule of thumb is keep the input impedance below 10 ohms to at least 2MHZ? Which the last I checked chip opamps have trouble with.
You guys know more about modern opamps than I do, would be interested to know if this is still a problem.
Last edited:
The 47pF capacitor from DAC output to OpAmp output should take care of the RF.
OPA627 shown in datasheet (55 V/us) is a different animal from the usual suspect 5532 (8 V/us)...
ADA4898 has a similar 55 V/us and low noise. Carfull with the heat paddle that is on bottom of package - it is connected to V- internally.
OPA627 shown in datasheet (55 V/us) is a different animal from the usual suspect 5532 (8 V/us)...
ADA4898 has a similar 55 V/us and low noise. Carfull with the heat paddle that is on bottom of package - it is connected to V- internally.
Last edited:
Thanks, Have you tried that shunt cap method? I've thought of trying to build a low impedance filter prior to the opamp but the PCM1704 has such a low output impedance for a current source. I vaguely remember this being the issue with the shunt cap and the fast opamps. Need to look at that again. Its being more common to see 4x parallel PCM1704's per channel which I believe helps the issue?
Firstly I’d like to thank everyone for replies.
Here are my findings; I hope they might be useful to others:
I have tried several CFB OP’s: AD811, AD844, AD812 and LME49713. Initially I was using AD8066 which worked REALLY well as I/V… but I wanted more.
The 1704 are connected as dual differential DAC’s, 2 per each channel, 4 DAC chips total. I/V conversion is executed by 2 dual OP’s, so my choices were somewhat limited and I did not want to use a separate board. AD811’s were very good, but AD811 in SOIC case was running way to hot, the chip was NOT oscillating, but it was causing the change in specifications to the extent that I could easily hear the overheated AD811 only after few minutes of it being powered-up. I wanted to keep + and -12V supply, so the next step was {2 AD811 DIL’s to single DIL adaptor board} X 2. This produced very good sound and the heat dissipation that was under control – but I swear I could hear loss of high frequency resolution – this was very easily picked up not only by me, so this option (2 DIL’s to a single DIL adaptor board) was not chosen neither. AD 844 in same 2 DIL’s to single DIL adaptors sounded very good – but too coarse. Each of these chips were getting Rf resistor optimised for min peaking and max resolution, as per datasheets. The capacitor in parallel to Rf (that was used with VFB OP’s and was originally 100 pF) was removed for CFB OP’s. The power supply noise was very low – sitting at the same level as surrounding ground noise potential, with special care taken to decouple Vcc / Vee pins to achieve these results.
Then I tried LME49713. This was very good, with just right amount of heat dissipation – around 50 deg C fully warmed-up. The adaptor boards were latest Browndog’s 2 SOIC’s to DIL. Rf was 1.2k and I could not see any improper behaviour with my 100MHz CRO. I’ll try to get faster CRO, but everything seemed OK. Compared to my reference DAC, I could not hear any difference with this chip. Maybe the input impedance is low enough for 1704 AND constant across important “for typical 24/96 digital audio range” application… don’t know… the specs say that this chip was optimised for audio application, I am just not sure was it 20 – 20k range, or a typical digital audio range after DAC chip…. Anyhow, it sounded excellent.
AD812 (as 2 DIL’s to single DIL) was VERY close….. I liked LME’s better because of slightly better sparkle, separation, and harmonic truthfulness to the original. LME’s sounded natural with a full extension of the whole audio range.
The second filter stage, after I/V conversion, is differential voltage to single ended converter with the filter. For this application, LM4562 sounded best by far. I tried AD8066 here (was very good – but “too fast…”), tried also the LM6172 (too fast and dry), AD826 (not bad, bit sterile, but very good, my second best after LM4562), OP2134 (very good, but slightly “loose” at the bottom end; however -> it produced very natural presentation with excellent top resolution… also my second best).
None was “really” bad – LM4562 was just that tiny bit better sounding compared to the rest. No doubt, different system may “prefer” AD826 (valves) or OP2134 (solid state AB). I use single end class A mosfets with a lot of silver throughout my system for interconnects, speaker cables and power cables…. and LM4562 sounded really good.
The CD player is Denon DCD 1650SR which I’ve been trying to get to the same level of sound quality as my transport/DAC combo purely because it is a single unit approach to CD playback. This CD player was heavily modified… the highlight was very low noise levels across the whole player, even around upsampling chip (that was NOISY!).
The OP chips' roll / tuning / test were done with the full range active speakers worth 30K – very revealing.
Boky
Here are my findings; I hope they might be useful to others:
I have tried several CFB OP’s: AD811, AD844, AD812 and LME49713. Initially I was using AD8066 which worked REALLY well as I/V… but I wanted more.
The 1704 are connected as dual differential DAC’s, 2 per each channel, 4 DAC chips total. I/V conversion is executed by 2 dual OP’s, so my choices were somewhat limited and I did not want to use a separate board. AD811’s were very good, but AD811 in SOIC case was running way to hot, the chip was NOT oscillating, but it was causing the change in specifications to the extent that I could easily hear the overheated AD811 only after few minutes of it being powered-up. I wanted to keep + and -12V supply, so the next step was {2 AD811 DIL’s to single DIL adaptor board} X 2. This produced very good sound and the heat dissipation that was under control – but I swear I could hear loss of high frequency resolution – this was very easily picked up not only by me, so this option (2 DIL’s to a single DIL adaptor board) was not chosen neither. AD 844 in same 2 DIL’s to single DIL adaptors sounded very good – but too coarse. Each of these chips were getting Rf resistor optimised for min peaking and max resolution, as per datasheets. The capacitor in parallel to Rf (that was used with VFB OP’s and was originally 100 pF) was removed for CFB OP’s. The power supply noise was very low – sitting at the same level as surrounding ground noise potential, with special care taken to decouple Vcc / Vee pins to achieve these results.
Then I tried LME49713. This was very good, with just right amount of heat dissipation – around 50 deg C fully warmed-up. The adaptor boards were latest Browndog’s 2 SOIC’s to DIL. Rf was 1.2k and I could not see any improper behaviour with my 100MHz CRO. I’ll try to get faster CRO, but everything seemed OK. Compared to my reference DAC, I could not hear any difference with this chip. Maybe the input impedance is low enough for 1704 AND constant across important “for typical 24/96 digital audio range” application… don’t know… the specs say that this chip was optimised for audio application, I am just not sure was it 20 – 20k range, or a typical digital audio range after DAC chip…. Anyhow, it sounded excellent.
AD812 (as 2 DIL’s to single DIL) was VERY close….. I liked LME’s better because of slightly better sparkle, separation, and harmonic truthfulness to the original. LME’s sounded natural with a full extension of the whole audio range.
The second filter stage, after I/V conversion, is differential voltage to single ended converter with the filter. For this application, LM4562 sounded best by far. I tried AD8066 here (was very good – but “too fast…”), tried also the LM6172 (too fast and dry), AD826 (not bad, bit sterile, but very good, my second best after LM4562), OP2134 (very good, but slightly “loose” at the bottom end; however -> it produced very natural presentation with excellent top resolution… also my second best).
None was “really” bad – LM4562 was just that tiny bit better sounding compared to the rest. No doubt, different system may “prefer” AD826 (valves) or OP2134 (solid state AB). I use single end class A mosfets with a lot of silver throughout my system for interconnects, speaker cables and power cables…. and LM4562 sounded really good.
The CD player is Denon DCD 1650SR which I’ve been trying to get to the same level of sound quality as my transport/DAC combo purely because it is a single unit approach to CD playback. This CD player was heavily modified… the highlight was very low noise levels across the whole player, even around upsampling chip (that was NOISY!).
The OP chips' roll / tuning / test were done with the full range active speakers worth 30K – very revealing.
Boky
Last edited:
I may either lower the power supply rails with SOIC AD811, or build external I/V board with 4 DIL AD811 in the near future when the time permits... this OP did sound very good as well and I am sure that my current set-up did not allow this chip to really shine - it has very low input impedance that can approach around 24 ohms, combined with very low Rf -> if the Rf is around 350 ohms.
But, that LME sounded soooo right and soooooo brilliantly extended at both ends even with around 55 ohms impedance ….. don’t know…. I’ll try AD811’a again....
Boky
But, that LME sounded soooo right and soooooo brilliantly extended at both ends even with around 55 ohms impedance ….. don’t know…. I’ll try AD811’a again....
Boky
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.