Hi KYW,Kuei Yang Wang said:
If we are talking about PCM1704, the limit is +/-1.2mA/200ns, with the usual 3k resistor it means 2.4mA*3k/0.2mks=36V/mks max
Don't Sign magnitude DACs have the glitch free output?
It can't ... we still have the limit 200ns ... for the full scale from -1.2mA till +1.2 ma ...
Right, but to follow correctly 36V/mks it will be enough to have 55V/mks, isn't it?
Yes, it is correct for the DACs with glithes, IMO ... those require faster OP amps ... but for PCM1704 I've found OPA627 as better sounding then faster op amps ... at least for me (tried AD843K, LM6172, next step AD811, AD8610, AD8065).
IMO the reason is in lower levels of produced higher harmonics ...
Please note, that AD8620 has HIGHER distortions (a bit, but still higher) then OPA627 at 20kHz, then at 30kHz it has less distortions ...
Very interesting paper, thanks. The problem is that Mr. Hawksford took the op amp with very low (only 60dB) open loop gain ... (fig. 3-1) Anyway, it is a perfect article ...ojg said:Also a very interesting paper here (as posted in a different thread), where the author demonstrates that slew-rate induced distortion is the same as jitter distortionand then compares two op-amps (pages 11-18) one with 500v/us and one with 50V/us for different sampling rates. Some real eye-openers here (for me at least...) Notice how the non-os case is much more sensitive to slew-rate than the higher sampling rates.
Unfortunately it happens too oftenKuei Yang Wang said:
How quickly? For PCM1704 it is 200ns, but we have to know that is its 36V/mks ... it means that OPA627 with 550ns settling time (we have to notice that it means completely different voltage range, it is 55V/mks and still exactly as you said, it settles faster then PCM1704 DAC, correct?)
"Sounds better" is a subjective argument, unfortunately. As we know, some people prefer more even harmonics to have more pleasurable, "warmer" sounding, some people prefer no distortions at all, and some people prefer a bit "bright" though more transparent sounding ... the simple example - Legato Pro ... (a lot of people use it, even knowing that there is nothing good in it)
Exactly
Good morning everybody,
I am a bit overwhelmed by all those high-powered digital filtering posts. Didn't have time yesterday to do further research, too many household chores (you know, supermarket, get haircut, make love to the missus) but I looked up Julian Dunn's excellent book he wrote for Audio precision "Measuring digital audio devices".
In the digital filtering chapter he shows how a recorded square wave is reproduced because of digital filtering. We all know a sq wave is (ideally) made up of an infinite number, in level decreasing, odd harmonics to the carrier. By adding them all up you get more and more a perfect shape. The converse is also true: by taking them away one by one you get slowly back to the carrier sine wave. That's what he shows for digital filtering: The square wave starts to show ringing, which is a natural result is you start taking away higher harmonics. It's obvious when you think about it. And the general wave shape moves more and more from a rectangle to a sine. And, what do you know, the risetime gets slower and slower also.
It seems to me however you cook it, low pass filtering *by definition* decreases the rise time.
Jan Didden
I am a bit overwhelmed by all those high-powered digital filtering posts. Didn't have time yesterday to do further research, too many household chores (you know, supermarket, get haircut, make love to the missus) but I looked up Julian Dunn's excellent book he wrote for Audio precision "Measuring digital audio devices".
In the digital filtering chapter he shows how a recorded square wave is reproduced because of digital filtering. We all know a sq wave is (ideally) made up of an infinite number, in level decreasing, odd harmonics to the carrier. By adding them all up you get more and more a perfect shape. The converse is also true: by taking them away one by one you get slowly back to the carrier sine wave. That's what he shows for digital filtering: The square wave starts to show ringing, which is a natural result is you start taking away higher harmonics. It's obvious when you think about it. And the general wave shape moves more and more from a rectangle to a sine. And, what do you know, the risetime gets slower and slower also.
It seems to me however you cook it, low pass filtering *by definition* decreases the rise time.
Jan Didden
Delta-sigma = dreck???
Mark your calendars:
We agree on something.
OK all you theoretical types:
It is possible for the above mentioned '1704 DAC to go from -1.2 mA to +1.2 mA in 200 nS.
Just what kind of signal.........and remember that it has to have been band-limited in the A-D stage.......could you possibly have on a CD to create such a monster?
Curious minds want to know.
Jocko
Mark your calendars:
We agree on something.
OK all you theoretical types:
It is possible for the above mentioned '1704 DAC to go from -1.2 mA to +1.2 mA in 200 nS.
Just what kind of signal.........and remember that it has to have been band-limited in the A-D stage.......could you possibly have on a CD to create such a monster?
Curious minds want to know.
Jocko
Please be more specific
Jan,
Maybe we are referring to different things and talking past
each other. Digital filtering does not affect the rise/fall time
of the samples output from the DAC. It does of course reduce
the rise/fall time of the signal.
Edit:
Assume we had a theoretical perfect DAC with zero rise/fall
time on the output. For a non-filtered signal, it will switch
between sample values in zero time. This sequence of
samples will then later be smoothed by analog filtering.
Now consider a signal that has gone through a digital
LP filter. The DAC will still switch between samples in zero
time, but the difference between the sample values will
be reduced compared to the first case. Also in this case
we will smooth the signal by an analog filter. If we don't
talk about the time to change between two adjacent
sample values, but instead define a kind of slew rate on
samples by taking the difference between two adjacent
sample values and divide by 1/fs, then you are correct
that this "slew rate" will be reduced by a digital filter.
Maybe we are referring to different things and talking past
each other. Digital filtering does not affect the rise/fall time
of the samples output from the DAC. It does of course reduce
the rise/fall time of the signal.
Edit:
Assume we had a theoretical perfect DAC with zero rise/fall
time on the output. For a non-filtered signal, it will switch
between sample values in zero time. This sequence of
samples will then later be smoothed by analog filtering.
Now consider a signal that has gone through a digital
LP filter. The DAC will still switch between samples in zero
time, but the difference between the sample values will
be reduced compared to the first case. Also in this case
we will smooth the signal by an analog filter. If we don't
talk about the time to change between two adjacent
sample values, but instead define a kind of slew rate on
samples by taking the difference between two adjacent
sample values and divide by 1/fs, then you are correct
that this "slew rate" will be reduced by a digital filter.
Christer,
I have just read the Hawksford paper referred to earlier, I think I see the light now. The step sizes between samples may be reduced, but the 'sharpness' of the steps remain. The same 'sharpness' at lower level is a small reduction in rise time, I guess, but I agree that there are still basic steps even after the digital lp filtering. I also underdstood from the Hawksford piece that indeed the cap across the I/V feedback resitor limits the required opamp slew rate, but only to a limited degree.
Interesting reading, that paper. This Hawksford guy knows his stuff. The I/V converter in fig 4.2 is a prime example of simplicity, elegance and high performance. Smart guy!
Jan Didden
I have just read the Hawksford paper referred to earlier, I think I see the light now. The step sizes between samples may be reduced, but the 'sharpness' of the steps remain. The same 'sharpness' at lower level is a small reduction in rise time, I guess, but I agree that there are still basic steps even after the digital lp filtering. I also underdstood from the Hawksford piece that indeed the cap across the I/V feedback resitor limits the required opamp slew rate, but only to a limited degree.
Interesting reading, that paper. This Hawksford guy knows his stuff. The I/V converter in fig 4.2 is a prime example of simplicity, elegance and high performance. Smart guy!
Jan Didden
Re: Delta-sigma = dreck???
As it is stated in the datasheetJocko Homo said:
Why only CD? How about DVD-A with 24-192 resolution? Though you are absolutely right, in the real world we don't need to deal with the max abilities of DACs ... Therefore, if we don't need such fast op amps in theory for the max possible signals in DACs, limited by their settling abilities, why do we need them in the real life with significantly slower signals? IMO for PCM1704 op amp with >50V/mks is OK ... we have to think about other things as open loop gain, level of distortions, their structure (higher harmonics, odd and even ones) but we don't need to have significantly faster op amp just because it is faster ...
New chips, new chances ...
I got marketing news a few weeks ago from TI about their new transimpedance opam the opa380.
The datasheet reads its application is for amongst others 'precision I/V conversion'.
I briefly compared it with the specs of the opa627BP, I currently use for I/V, and must conclude the 380 :
- is nosier (from 10Hz to 10Khz, for >10Khz the 627 datasheet has no noise info)
- has a higher slewrate (80 compared to 55V/us) and
- the settling time is less 2us 0.015% against 0,55us, 0,01%.
When thes are the I/V conversion critical parameters I don't expect the 380 to be better than the 627. What do I miss here?
Has anybody tried the opa380 yet??
How about the opa675, The TI-datasheet tells it only is better for the settling time 450ns them the opa627.
Henk
I got marketing news a few weeks ago from TI about their new transimpedance opam the opa380.
The datasheet reads its application is for amongst others 'precision I/V conversion'.
I briefly compared it with the specs of the opa627BP, I currently use for I/V, and must conclude the 380 :
- is nosier (from 10Hz to 10Khz, for >10Khz the 627 datasheet has no noise info)
- has a higher slewrate (80 compared to 55V/us) and
- the settling time is less 2us 0.015% against 0,55us, 0,01%.
When thes are the I/V conversion critical parameters I don't expect the 380 to be better than the 627. What do I miss here?
Has anybody tried the opa380 yet??
How about the opa675, The TI-datasheet tells it only is better for the settling time 450ns them the opa627.
Henk
Agree with you, Extreme_Boky.
I use them with a CDP XB 920 ... but ... without the "current converter" not good ! Or at least, with a small battery supply for the reference. So i use it for the differential output of the converter (V mode) with LP filter at the same time.
Did you try the AD 8620 ?
I use them with a CDP XB 920 ... but ... without the "current converter" not good ! Or at least, with a small battery supply for the reference. So i use it for the differential output of the converter (V mode) with LP filter at the same time.
Did you try the AD 8620 ?
Hi Richard,
No, I haven't tried 8620. It looks very promising especially as an I to V converter. There is always DIL to SOIC 'adaptor board' hassle. I do not like these adaptor boards, they introduce capacitive problems to these OP's, they are suspect to RF, and in particular afect noise performance (which is excellent fot this OP) of the input. There is also a contact resistance, if you use sockets. Never use sockets - NEVER!
Cheers
No, I haven't tried 8620. It looks very promising especially as an I to V converter. There is always DIL to SOIC 'adaptor board' hassle. I do not like these adaptor boards, they introduce capacitive problems to these OP's, they are suspect to RF, and in particular afect noise performance (which is excellent fot this OP) of the input. There is also a contact resistance, if you use sockets. Never use sockets - NEVER!
Cheers
Actually we can use sockets, but good ones, like here, for example: http://www.soundlabsgroup.com.au/components/ics.htm#icsockets They are not expensive, and even for our DIY requirements are not bad.
Concerning the RF, 2 pf means nothing ... I also used this one http://66.216.121.197/index.asp?PageAction=VIEWPROD&ProdID=12 - They had been designed correctly from RF point of view, therefore IMO can be used for AD8620 or other SOIC chips.
Concerning the RF, 2 pf means nothing ... I also used this one http://66.216.121.197/index.asp?PageAction=VIEWPROD&ProdID=12 - They had been designed correctly from RF point of view, therefore IMO can be used for AD8620 or other SOIC chips.
I would strongly suggest one try the THS4082 dual chip for the i/v conversion stage. It's a very fast opamp which has 175Mhz bandwidth and is ideal for audio applications with proper implementation. That includes component layout, especially with digital part - digital and opamp sections should be kept apart as much as possible. Also, the ground plane is needed, a compensation cap can be placed on the negative input and output of the chip. It can be socketed but soldering it directly to PCB saves time and money and reduces RF and parasitic capacitance even more.
I replaced my PPA with single-channel THS4081's and can't be happier. The improvement in sound is not subtle and can be noticed after a second of listening with decent headphones.
Check my thread on head-fi.org regarding this matter: link to post
The chip is 20 times less expensive than OPA627/637 and sounds way superior and cleaner than its audio competitors retaining the well-known warm Burr-Brown signature.
I replaced my PPA with single-channel THS4081's and can't be happier. The improvement in sound is not subtle and can be noticed after a second of listening with decent headphones.
Check my thread on head-fi.org regarding this matter: link to post
The chip is 20 times less expensive than OPA627/637 and sounds way superior and cleaner than its audio competitors retaining the well-known warm Burr-Brown signature.
Why so fast?
Why such a fast wide bandwith opamp???
I found the somewhat slower opamps like OPA2604 and OP275 perform much better sonically as IV-converter.
Edit : the reason this opamp does not work with the diamond buffer is, I think, the opamp is too fast for the buffer. Something like a dog chasing its own tail. This is explained in application note 47 on Linear Technology site:
http://www.linear.com/pub/document.html?pub_type=app&document=50
Hi,IpsilonSound said:
Why such a fast wide bandwith opamp???
I found the somewhat slower opamps like OPA2604 and OP275 perform much better sonically as IV-converter.
Edit : the reason this opamp does not work with the diamond buffer is, I think, the opamp is too fast for the buffer. Something like a dog chasing its own tail. This is explained in application note 47 on Linear Technology site:
http://www.linear.com/pub/document.html?pub_type=app&document=50
Re: Why so fast?
Hi Elso!
It wasn't my idea initially, I only proved that theory once again.
Many audio designers already paid attention to using fast opamps for audio to obtain the possibly highest quality and virtually distortion-free audio signal.
For example:
http://www.sound.westhost.com/highspeed.htm
Also, Aos chose LM6172 fast opamp for his Ally DAC on purpose as well:
www.aoselectronics.com
You're right, DB from Larocco wasn't initially compatible with the THS4081 chip, it was oscillating, especially at lower gains(~5). But putting a bandwidth cap to rolloff the frequency a bit after 1.6Mhz plus increasing the gain fixed the problem for 100%. Limiting the bandwidth didn't affect the sound in any negative aspect, i tested both cap/no cap versions and, of course, didn't hear any difference.
Like I mentioned before, it's worth a try to use THS4082 as an I/V opamp stage or as a headphone amplifier as well.
Elso Kwak said:
Hi,
Why such a fast wide bandwith opamp???
I found the somewhat slower opamps like OPA2604 and OP275 perform much better sonically as IV-converter.
Hi Elso!
It wasn't my idea initially, I only proved that theory once again.
Many audio designers already paid attention to using fast opamps for audio to obtain the possibly highest quality and virtually distortion-free audio signal.
For example:
http://www.sound.westhost.com/highspeed.htm
Also, Aos chose LM6172 fast opamp for his Ally DAC on purpose as well:
www.aoselectronics.com
You're right, DB from Larocco wasn't initially compatible with the THS4081 chip, it was oscillating, especially at lower gains(~5). But putting a bandwidth cap to rolloff the frequency a bit after 1.6Mhz plus increasing the gain fixed the problem for 100%. Limiting the bandwidth didn't affect the sound in any negative aspect, i tested both cap/no cap versions and, of course, didn't hear any difference.
Like I mentioned before, it's worth a try to use THS4082 as an I/V opamp stage or as a headphone amplifier as well.