Hi George,
I may be getting confused here. I know some DACs don't handle FS at all well and need to be maxed at -.1dBFS but that is completely seperate from 'overs' as discussed by Benchmark?
I've just scanned the benchmark document which explains they attenuate by 3.5dB in the digital domain then bump up the analogue gain. Makes sense. I do have some (recent) recordings where audacity shows actual clipping. May do some tests on those to see if the attenuation helps.
billshurv,
How can you possibly correct for a recording with peak clipping in the recording, it seems that would be the worst case situation with little to no real way with attenuation you could ameliorate that situation, you can't exactly remove that type of recorded distortion so I just don't think we should look at that as anything but a poorly mastered final mix. Now overdriving the dac with to much level I see what you and George are talking about, that seems like a simple solution with the correct gain structure.
How can you possibly correct for a recording with peak clipping in the recording, it seems that would be the worst case situation with little to no real way with attenuation you could ameliorate that situation, you can't exactly remove that type of recorded distortion so I just don't think we should look at that as anything but a poorly mastered final mix. Now overdriving the dac with to much level I see what you and George are talking about, that seems like a simple solution with the correct gain structure.
The Groner survey from 2009 can be found here: http://www.nanovolt.ch/resources/ic_opamps/pdf/opamp_distortion.pdf
Bill it is me that I have confused you.
“Inter-sample "overs" is a sub category of DAC input overloading.
What I experience with my set-up is most probably DAC input overloading and not “Inter-sample "overs" per se.
I generate 11025Hz sinus files from 0dB to –6dB rel FS, in 0.5dB steps to use them as a test signal for to see what is going on.
I’ll report back when I have results.
George
An oscillator that presents common mode voltage to the op-amp is compromised we have moved on from there. As for verifying performance at these levels we found moving leads around or changing benches i.e. everything mattered.
Sorry was rushing out a reply before so hadn't had a chance to explain.
Audio That Goes to 11 - Benchmark Media Systems, Inc. introduces the concept of 'overs', where the interpolation between 2 points would go over 0dBFS. So I just wondered aloud if the clip was between two samples then would a 3.5dB attenuation allow the reconstruction filter to avoid outputting a clip.
I have possibly missed something important in my musing though. But worth looking at some music to see what there is there.
billshurv,
So there is not actually a clipped bit in the recording as much as an interpolation error caused by to high a gain sent to the dac. Seems you would be correct that the dac wouldn't interpolate that the same at a lower output and though the peak interpolation could still reach 0dbv you wouldn't exceed that if the overall level was lowered as you say. Now I follow your logic.
So there is not actually a clipped bit in the recording as much as an interpolation error caused by to high a gain sent to the dac. Seems you would be correct that the dac wouldn't interpolate that the same at a lower output and though the peak interpolation could still reach 0dbv you wouldn't exceed that if the overall level was lowered as you say. Now I follow your logic.
The maximum interpolation "over" is computable and IME a fairly small head room allowance.
Thanks John. [my internet was down last several hours] As you all know, the upgrade on the 339A was not a redesign but what could we do to make it better. later, david (davada) started his own design from scratch and it is extremely low distortion and Variable freq as well (not spot freqs).... and fast response time at low freqs. The work can be followed in the test equipment forum. SG has a great performing osc/gen also there. A rather complex design -- pcb will be avail.
I am Not interested in how to make the 797 work in any particular app. I have read the huge effort by SG in his opamp tests a few years ago. Not recently. Just want information or explanation which may be useful later or to others here. I like that people are starting to get involved and ask better questions of their own. Jitter too.
The LM device which worked so well is in the 339 upgrade section.... I dont recall it now off the top of my head. I might look for it later and let you know. it wasnt the last word in noise though.
Thx-RNMarsh
Last edited:
I have had the same endless experience when looking below the -100 dB mark. What they did in the LIGO must be heroic to keep all the gremlins out.
Getting below -100 dB with real usable audio equipment in a system is very hard.
Keith Johnson pointed out an issue (one Philips is guilty of on early chips) that the digital filter can increase levels above FS in some cases. Its a side effect of the sharp rolloff. Perhaps the Benchmark has one of those filter implementations.
Pretty much every DAC is better at -3 to -20 dB. Given the dynamic range exceeds anything a normal human can tolerate I see no downside to using a DAC from -6 down. A decent digital volume implementation, probably as part of a digital filter could make this pretty seamless and invisible to both the users and the designers who can't see deep into the chips (pretty much all end product designers).
Pop music can have transients that hit FS and some of the RAP I have looked at has a lot of bass that hits FS. I'm sure there is a limiter effect that smooths the clipping and stops it just (1 bit?) before FS. Some of the pop I have looked at has an awful lot of -2 dB peaks.
Of the big shows that I work at, those that are running digital desks often run an external/outboard clock.
Other FOH engineers insist on analog desks and state that although digital desks enable myriad of functionality they don't sound 'right'.
In conversations with digital desk FOH sound engineers, they insist that their sound/mix is better with their (usually own/personal) external clock.
General consensus is that desk internal clocks are not be be relied upon subjectively.
Some good info here -
http://pinknoisemag.com/pink-papers/pink-paper-002
http://www.cranesong.com/A_Matter_%20Of_Time_The%20_Audibility_Of_Clock_Jitter.pdf
jitter_1
jitter_2results
Dan.
Other FOH engineers insist on analog desks and state that although digital desks enable myriad of functionality they don't sound 'right'.
In conversations with digital desk FOH sound engineers, they insist that their sound/mix is better with their (usually own/personal) external clock.
General consensus is that desk internal clocks are not be be relied upon subjectively.
Some good info here -
http://pinknoisemag.com/pink-papers/pink-paper-002
http://www.cranesong.com/A_Matter_%20Of_Time_The%20_Audibility_Of_Clock_Jitter.pdf
jitter_1
jitter_2results
Dan.
Last edited:
The DAC et al needs to be much lower in distortion so that we can stay away from clipping and still have good dynamic range and good S/N. Classical music may have one acoustic instrument playing softly up to the full orch playing full blast. Recording levels (without compressor) would have to be carefully set to capture the full dynamic range without noticeable noise and distortion added.
Another reason is, as I have said many times, is to keep the accumlated distortion of the whole process - from mic to speaker in the home. Every opamp or amp in the process adds (not necessary a direct sum) - how many? - contributes additional distortion to the signal. So we need to start with the best we can do so that the sound at the home end is closest to the original acoustic signal.
Next subject: here is the first gen LIGO PIN and amps - complete with schematics. I can see that the 797 and maybe others would do well in an update.
http://dspace.mit.edu/bitstream/handle/1721.1/85355/45011890-MIT.pdf?sequence=2
MAX --- nice info re jitter.
THx.
THx-RNMarsh
Another reason is, as I have said many times, is to keep the accumlated distortion of the whole process - from mic to speaker in the home. Every opamp or amp in the process adds (not necessary a direct sum) - how many? - contributes additional distortion to the signal. So we need to start with the best we can do so that the sound at the home end is closest to the original acoustic signal.
Next subject: here is the first gen LIGO PIN and amps - complete with schematics. I can see that the 797 and maybe others would do well in an update.
http://dspace.mit.edu/bitstream/handle/1721.1/85355/45011890-MIT.pdf?sequence=2
MAX --- nice info re jitter.
THx.
THx-RNMarsh
Last edited:
Have you checked out THAT corp.? I don't know if they are making the old DBX VCA's which should be workable in this application. I know Rochester is good for getting real parts not fakes but they have large min quantities on some. Unfortunately it is not one of the parts I stocked up on, I only grabbed a pile of MAT02's and MAT03's in TO cans before they were scrapped for non-RHOS.
- Status
- Not open for further replies.