Sorry to start a thread about I/V op-amps. Too many threads about them plus I'm not really trying to revive any specific discussion.
Just FYI for the readers:
Stereophile Feb 2012 reviewed the Bricasti Design M1 D/A Processor.
DAC is AD1955 (with DSP implementing external filter) in mono differential mode. I/V opamps are AD843. Buffers are discrete transistor based.
Results: "Class A+", "State-of-the-art measured performance"
The I/V opamp itself isn't fancy but was carefully chosen and voiced.
AD843 | 34 MHz, CBFET Fast Settling Op Amp | Operational Amplifiers (Op Amps) | All Operational Amplifiers | Analog Devices
Slew rate: 250V/us
Input noise: 19nV/rt Hz
Settling time: 135ns to 0.01%
Please let this be informational only - if you feel strongly about one method of I/V over another (resistor, common-base transistor/FET, your fav opamp), there is no reason to force anyone else to like your taste. As for the technical merits of each method, they have already been debated far and wide.
Thanks.
PS: I have no affliliation whatsoever with any of the above mentioned companies.
Just FYI for the readers:
Stereophile Feb 2012 reviewed the Bricasti Design M1 D/A Processor.
DAC is AD1955 (with DSP implementing external filter) in mono differential mode. I/V opamps are AD843. Buffers are discrete transistor based.
Results: "Class A+", "State-of-the-art measured performance"
The I/V opamp itself isn't fancy but was carefully chosen and voiced.
AD843 | 34 MHz, CBFET Fast Settling Op Amp | Operational Amplifiers (Op Amps) | All Operational Amplifiers | Analog Devices
Slew rate: 250V/us
Input noise: 19nV/rt Hz
Settling time: 135ns to 0.01%
Please let this be informational only - if you feel strongly about one method of I/V over another (resistor, common-base transistor/FET, your fav opamp), there is no reason to force anyone else to like your taste. As for the technical merits of each method, they have already been debated far and wide.
Thanks.
PS: I have no affliliation whatsoever with any of the above mentioned companies.
Last edited:
I am playing now with a few OpAmps (for the I/V stage). My next test subject looks promising:
AD8099. Stable at Gain +2.
Noise: 0.95nV/vHz, 2.6pA/vHz
Slew rate: 475V/µs
Settling time: 18ns to 0.1%, 30ns to 0.01%
Distortion: 2nd Harmonic -92dB @ 10MHz; 3rd Harmonic -105dB @ 10MHz
-3dB bandwidth: 700MHz (G=+2)
I have on my bench a ADA4897-1 that is stable at Gain +1:
noise: 1 nV/√Hz, 2.8 pA/√Hz
distortion: −115 dBc @ 100 kHz, VOUT = 2 V p-p
−3 dB bandwidth (G = +1): 230 MHz
slew rate: 120 V/μs
settling time: 45 ns to 0.1%; 90ns to 0.01%
I did try before to "make" my own discrete OpAmp stage (BJT, JFET, combination) and I realized that I am not capable of even getting closer of values like above ones.
AD8099. Stable at Gain +2.
Noise: 0.95nV/vHz, 2.6pA/vHz
Slew rate: 475V/µs
Settling time: 18ns to 0.1%, 30ns to 0.01%
Distortion: 2nd Harmonic -92dB @ 10MHz; 3rd Harmonic -105dB @ 10MHz
-3dB bandwidth: 700MHz (G=+2)
I have on my bench a ADA4897-1 that is stable at Gain +1:
noise: 1 nV/√Hz, 2.8 pA/√Hz
distortion: −115 dBc @ 100 kHz, VOUT = 2 V p-p
−3 dB bandwidth (G = +1): 230 MHz
slew rate: 120 V/μs
settling time: 45 ns to 0.1%; 90ns to 0.01%
I did try before to "make" my own discrete OpAmp stage (BJT, JFET, combination) and I realized that I am not capable of even getting closer of values like above ones.
Last edited:
I realize the app notes for instrumentation application of high speed DACs do emphasize settling time but it really isn't quite the same importance for digital audio reproduction
in audio reproduction we want/need low pass filtered output to reject the "steps", image frequencies, any noise shaping high frequency modulation
if you take the current technology "sweet spot" as ~96 k sample rate converters with substantial internal upsampling then ~40 kHz low pass is required with order depending on upsampling ratio and any noise shaping
even doubling the sample rate and low pass frequency still gives ~ 2 uS time constants - nS settling time really is irrelevant as a prime I/V performance spec for audio reproduction
like slew rate, settling time can be an indirect indicator of certain op amp qualities that may be helpful - but the actual requirements of 40 kHz or even 80 kHz low pass filtered digital audio reproduction at 2 Vrms consumer line levels doesn't require SOTA max slew rate or settling time per se
what is really needed for audio I/V conversion is Linear treatment of the DAC Iout while performing the required low pass filtering
there a 2 parts of the linear requirement:
low delta V at the DAC out - depending on internal details Iout DAC have limited output compliance V, bigger V causes nonlinear loss of the modulated current in protection diodes, switch on-resistance modulation, collector/drain nonlinear Zout - often not rigorously speced but Order of 10 mV seems a good target without a tighter spec or measurements indicating otherwise
and Linear treatment of the DAC Iout in the I/V circuit - in the face of high current slew rates and tight input V spec - incorporating low pass filtering in the op amp I/V design gives better linearity
even "the best" op amps can't meet both requirements with just R feedback - feedback C greatly reduces dynamics at the I/V input - simple sim does show this clearly
additional topology improvements are available "noise gain" C to gnd at the I/V input is a start- again sims illustrate the principles even if not detailed enough for strictly predictive results
The ADA489x, AD8099 are something new - the "highly linear input" stage properties are poorly advertised but his is a real breakthrough for BJT input linearity with large input Vdiff - just what we should want for DAC I/V
in audio reproduction we want/need low pass filtered output to reject the "steps", image frequencies, any noise shaping high frequency modulation
if you take the current technology "sweet spot" as ~96 k sample rate converters with substantial internal upsampling then ~40 kHz low pass is required with order depending on upsampling ratio and any noise shaping
even doubling the sample rate and low pass frequency still gives ~ 2 uS time constants - nS settling time really is irrelevant as a prime I/V performance spec for audio reproduction
like slew rate, settling time can be an indirect indicator of certain op amp qualities that may be helpful - but the actual requirements of 40 kHz or even 80 kHz low pass filtered digital audio reproduction at 2 Vrms consumer line levels doesn't require SOTA max slew rate or settling time per se
what is really needed for audio I/V conversion is Linear treatment of the DAC Iout while performing the required low pass filtering
there a 2 parts of the linear requirement:
low delta V at the DAC out - depending on internal details Iout DAC have limited output compliance V, bigger V causes nonlinear loss of the modulated current in protection diodes, switch on-resistance modulation, collector/drain nonlinear Zout - often not rigorously speced but Order of 10 mV seems a good target without a tighter spec or measurements indicating otherwise
and Linear treatment of the DAC Iout in the I/V circuit - in the face of high current slew rates and tight input V spec - incorporating low pass filtering in the op amp I/V design gives better linearity
even "the best" op amps can't meet both requirements with just R feedback - feedback C greatly reduces dynamics at the I/V input - simple sim does show this clearly
additional topology improvements are available "noise gain" C to gnd at the I/V input is a start- again sims illustrate the principles even if not detailed enough for strictly predictive results
The ADA489x, AD8099 are something new - the "highly linear input" stage properties are poorly advertised but his is a real breakthrough for BJT input linearity with large input Vdiff - just what we should want for DAC I/V
Last edited:
IMO Linearity is derived principally from a fast feedback loop. A slow loop will induce errors in following the signal and intermodulation products.
Also, limiting the SR at the output of a DAC with capacitors (even if they are in feedback loop) will trow off the compliance voltage. If the DAC current source was ideal, sure you can do that safely. But it's not.
Also, limiting the SR at the output of a DAC with capacitors (even if they are in feedback loop) will trow off the compliance voltage. If the DAC current source was ideal, sure you can do that safely. But it's not.
Last edited:
So let say that you are right and it doesn't matter how fast or slow are the OpAmps.
Why people bother to try tubes, resistor I/V? Or why even the DAC manufactures mention the slew rate and settling time in the aplication notes on datasheets?
Just dump a capacitor on outputs and you are good to go...
PS: I don't belive too much in spice simulations for anything more complex than a couple of transistors.
Why people bother to try tubes, resistor I/V? Or why even the DAC manufactures mention the slew rate and settling time in the aplication notes on datasheets?
Just dump a capacitor on outputs and you are good to go...
PS: I don't belive too much in spice simulations for anything more complex than a couple of transistors.
Last edited:
now you're posturing - I never said slow op amp are good for I/V - you have also blown off suggestions, guidance from Thorsten on DAC I/V - clearly you aren't interested in the learning opportunities here
you may consider any further of my comments in threads you’re participating in as directed to the general readership
you may consider any further of my comments in threads you’re participating in as directed to the general readership
Thorsten is on my ignore list. Because there is nothing to learn from him: resistive I/V and tubes rules. Selling to a niche. Thanks but no thanks, I don't sacrifice noise for speed!
I don't understand what you want, you just attacked the ideea of fast OpAmp for I/V without providing anything in place (besides "the capacitors are great"). You say that 2uS is "enough", that might be fine with you... I don't want nothing that is below listed current settling time for a given DAC.
Maybe I need to learn, but how about engineers that design DAC's - they need to learn too?
Tell all those guys here that are looing into discrete stages that they are wrong too... And even your example above - why do you think is looking for tubes and resistive I/V?
http://www.ti.com/lit/ds/sbas097/sbas097.pdf - page 9
http://www.ti.com/lit/ds/sles117a/sles117a.pdf - page 19
I don't understand what you want, you just attacked the ideea of fast OpAmp for I/V without providing anything in place (besides "the capacitors are great"). You say that 2uS is "enough", that might be fine with you... I don't want nothing that is below listed current settling time for a given DAC.
Maybe I need to learn, but how about engineers that design DAC's - they need to learn too?
Tell all those guys here that are looing into discrete stages that they are wrong too... And even your example above - why do you think is looking for tubes and resistive I/V?
http://www.ti.com/lit/ds/sbas097/sbas097.pdf - page 9
http://www.ti.com/lit/ds/sles117a/sles117a.pdf - page 19
Last edited:
It looks a jolly nice part until you look at its settling transient. Guess that's messy in large part because of the gain peaking (+3dB) at G=+2.
You're going to need a series RC on the DAC output to keep the noise gain high, preferably higher than +2 to mitigate the peaking.
Curious to know how it sounds
I too have played with many OP's but moved on to discretes learning in no circumstance will any OP sound better than discretes even if the specs are way off. but maybe i just haven't tried a good OP, so far the best i've tried is the AD8066 which i prefer over my AD843's.
The specs seem to have little to do with how they sound.
I'd advice you to give discretes another try.
Last edited:
I found out that specs (correct readed) have a lot to do with the sound. For transistors, capacitors, diodes, OpAmps...
For example a 741 has worse specs than a 5532 and it sounds worse.
My jfet based OpAmp sounded and measured better than most of 70's OpAmps, but compared with the top of the line from 2010, it would suck. I did have access to lab-grade measuring equipment back then.
For example a 741 has worse specs than a 5532 and it sounds worse.
My jfet based OpAmp sounded and measured better than most of 70's OpAmps, but compared with the top of the line from 2010, it would suck. I did have access to lab-grade measuring equipment back then.
Last edited:
You made a "blanket" affirmation... If it was true, the specs of 741 and 5532 would have "little to do with how they sound".
Reality is different. Specs always matter. You just need to read them carefully and understand what they mean for a specific application.
Thanks for the ideea. I was trying to find a way to deal with that... But I got lazy, I did order an ADA4897-1, that is stable at Gain +1:
noise: 1 nV/√Hz, 2.8 pA/√Hz
distortion: −115 dBc @ 100 kHz, VOUT = 2 V p-p
−3 dB bandwidth (G = +1): 230 MHz
slew rate: 120 V/μs
settling time: 45 ns to 0.1%; 90ns to 0.01%
Now I have only to worry about the max 10V supply - that's exactly what I have now, kind of scarry close.
Hi, I tried the AD843 in my experimental Opamp Preamp. Well... I much preferred the AD8065/AD8066. There was a noticiable overlay on the audio. Hard to describe, like a noise component. The AD8065's bettered the AD843 in everyway. AD8065's have a very natural sound so I agree with those who like that part. I'll be trying out some others too. I tried the AD8066's in my Hagerman Bugle phono and they are very nice in there too. Plenty low noise. Superb dynamics. Dave
I resurrected this old thread to add some info / fb.
I have this dac (Bricasti M1) and I like it very much, but to my ears the treble (slightly) lacks ultimate resolution and sweetness compared with my other dacs. On bright and older recordings it can tend to be a little "unforgiving".
Taking a look at the pcbs's I too noticed the AD843 op-amps in the IV and buffer stages. These op-amps are not noted for audiophile qualities but they are advertised for high speed applications including 'high speed integration'.
In an effort to add some sweetness I decided to bias all of the op-amps into class A. I did this by adding a CRD with a series resistor from the op-amp output to the -ve rail. (4 per channel) I mounted them on the underside of the pcb so as not to deface the aesthetics. The resistor was used to set the v drop across the CRD to the mid point of the linear region, which for the specific 2mA CRD used was 8.5 vdc.
Because the op-amps are all paired from the diff outputs from the dac, I decided to batch test and match the CRD's to <1%, which was possible because I had a large stock on hand. I also found that the CRD's ranged in value from 1.7 to 2.3mA, but nominally 1.9mA.
I also tested the current sink regulation over a 3V range and it was a creditable 99.8%.
While I had the dac apart I also passed a cursory eye over the linear power supplies which are all discrete types using op-amps to generate references and with tip120/125 darlington output transistors.
I noticed that a 104 film cap had been installed across the bridge rectifier input (transformer secondaries) and another at the bridge output, but there was no series resistor with the input 104, so it obviously isn't an effective "snubber". This seemed odd, so I subsequently unsoldered one side and added a 330r resistor in series ( 2 per channel), though I didn't use the cro to check, from past experience those values are usually effective at damping any ringing.
OK, so how did it sound after this change you might ask?
Actually, I was very surprised to hear such a big change, because frankly, I wasn't able to measure anything substantial on the pc based analyzer. (distortion was already a low 0.001%)
However, the treble is now MUCH sweeter and more pristine than I could have ever imagined possible, and all of those bad old recordings are now very easy on the ear. It's now every bit the equal, and then some, of my Berkeley Alpha 2 in the treble and is easily the best overall and least 'digital' sounding of all my processors. I also now prefer to use the linear filter set rather than the minimum phase set.
It's worth noting though that I made two changes at the same time, so it's difficult to ascribe the contribution from each, but I'm assuming it's the 2mA of bias on all op-amps that's been the major factor.
I can't say that this same modification will be as effective for all op-amps, and / or in other implementations, but it's nonetheless worth experimenting with based on my personal experience.
Last edited:
Should be "CRD" so CDR is a typo.
CRD = Current regulating Diode. It's a jfet with source resistor in a package that resembles a diode (ie. 2 terminals, A and C)
edit:
I decided to create a new dedicated thread..
http://www.diyaudio.com/forums/digital-line-level/288078-bricasti-m1-processor-modification.html
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.