Is that without the GA In circuit?
1073 type input stage caps...why??
- Thread starter HenryL
- Start date
Help Support GroupDIY Audio Forum:
Yes, straight from line output 3 to line input 3, and straight from a Lavry DA10 to line input 3.
Differing but similar results. Seems like the interface is the ringing culprit.
Tried a different line input?
Tried a different line input?
Not tried a different line input so far.
I don't think it's ringing. It's a periodic undulation in the frequency response across a wide range of frequencies.
Ok here's a quick example with a selection of some of the cases at 88.1kHz, I blew up the HF area and spaced the curves, and it's perhaps worth highlighting that these variations in the frequency response are only about a tenth of a dB at audio frequencies. I did fast sweeps as I was interested in seeing where the noise was, not hiding it. The Lavry has an internal analogue attenuator so the output is always at full resolution.
And here's a couple of 44.1 cases Lavry versus Focusrite. A bit more zoomed out and spaced out. The difference in filter behaviour is notable. Nasty looking spike appearing in both cases at the limit of the measurement.
The undulations in the red 18i20 curve at 44.1 looked rather like a frequency beating with the rising test signal frequency to me, with the beat getting slower towards the top end - so I speculated it is some artefact linked to the sampling frequency. But in the 88.1 red curve it seems to be going the other way, and the Lavry undulates gently over a wider range with a less obvious increase/decrease in the spacing of the peaks/troughs. Jitter / clock frequency recovery variations shouldn't be a factor in the 18i20 to 18i20 case as it's all on the same internal clock...
I don't think it's ringing. It's a periodic undulation in the frequency response across a wide range of frequencies.
Ok here's a quick example with a selection of some of the cases at 88.1kHz, I blew up the HF area and spaced the curves, and it's perhaps worth highlighting that these variations in the frequency response are only about a tenth of a dB at audio frequencies. I did fast sweeps as I was interested in seeing where the noise was, not hiding it. The Lavry has an internal analogue attenuator so the output is always at full resolution.
And here's a couple of 44.1 cases Lavry versus Focusrite. A bit more zoomed out and spaced out. The difference in filter behaviour is notable. Nasty looking spike appearing in both cases at the limit of the measurement.
The undulations in the red 18i20 curve at 44.1 looked rather like a frequency beating with the rising test signal frequency to me, with the beat getting slower towards the top end - so I speculated it is some artefact linked to the sampling frequency. But in the 88.1 red curve it seems to be going the other way, and the Lavry undulates gently over a wider range with a less obvious increase/decrease in the spacing of the peaks/troughs. Jitter / clock frequency recovery variations shouldn't be a factor in the 18i20 to 18i20 case as it's all on the same internal clock...
There’s a noticeable big difference between 18i20 - 18i20 and Lavry - 18i20. I think you’re getting internal feedback/leakage of filter noise at the threshold of the filters when it’s the 18i20 running solo which drastically reduces when it’s Lavry to 18i20 to what appears to be just filter noise attributable to both units.
I wonder if the spike is just the result of instantaneous cessation of delivered signal at the top of the frequency range - I don’t imagine a fade-down?
ruffrecords
Well-known member
An alternative method of measuring frequency response that does not rely on sweeps is to use white noise from the generator and look at the response on the RTA. I often use this for testing EQs because you can see the response change as you vary the controls. There used to be a feature in REW that allowed the creation of correlated white noise which gives much smoother response curves in the RTA but I am not sure if it is still there. Maybe worth a try though.
Cheers
Ian
Cheers
Ian
Yup - the Lavry output looks right to me with an overall flat response then consistent steep anti-aliasing cut, whereas the 18i20 looks pretty ropey, often in multiple ways! Especially at the lower sampling rates 48kHz and eg the 44.1kHz example I posted above, where it looks like the anti-aliasing filter is essentially bust.
I've been trying a number of further 18i20 test permutations - different outputs, different inputs, different sweep levels, different sampling rates - and to cut a long story short have come to the conclusion that I don't trust the 18i20 outputs.
I don't think the messy behaviour has significantly affected the use of the 18i20 at 192kHz to look for resonances in the Pre73, just added a confusion factor to the process which I could really have done without - but I definitely don't trust it's output behaviour for general audio at 44.1kHz (or 48kHz) sample rate.
Optimistically I suppose it might be something occurring in the loop scenario which wouldn't normally occur otherwise (some kind of feedback-related effect as you've speculated), and the apparent undulations in frequency response are pretty small magnitude.
However, happily all this is not a worry in practice since I decided long ago that I didn't like something about the sound of the 18i20 outputs. My monitors are always driven from the Lavry and I only use the 18i20 outputs for cue mix and occasional re-amping, and I don't mix using outboard gear...
I'm hoping so! REW is a pretty mature implementation though, so I'd imagine it would at the very least stop on a zero-crossing. I have seen a display of a sweep waveform in REW that looked like it had a fade down at the end but I'm not sure that that necessarily represents what it does in the kind of measurement sweeps I"ve been using (I saw it in a different section of the tool). Otherwise the spike looks like ... well... as I suspect the 18i20 anti-aliasing is flawed then maybe it is some aspect of that.
I've done that too for examination of EQs... just in the audio range up to 20kHz though and I just used my DAW facilities not REW.
There is a 'noise' option on the measurements page in REW next to 'sweep' but it was inactive in all my tests and I didn't delve more into the ins and outs of REW to find out if/how/when it became useable as an option. There's a "full-range" noise option in the 'Generator' page which output something when I activated it but then the input didn't seem to appear in the RTA and I gave up poking around at that point
I agree - which makes the 18i20 seem like it has a ringing at cessation of signal output perhaps due to its filters.
So... I'm wrapping up the Pre-73 Zobel Saga - again I've actually put the screws back in the lid this time!
I went back to ye trusty scope to double-check the conclusion reached with REW that no zobel components were needed on the line or mic input tx.
My signal generator is pretty basic but it goes up to 200kHz sine at stable amplitude, and at 100mV pk has a Zout of 100R. I used 2 scope channels and with the input signal on one and the output at various test points on the other. Matched the amplitude and polarity of the test trace to overlay the input trace exactly then manually swept up through the frequency range to 200kHz multiple times, watching for any level changes or divergence of the traces. I did this at (1) the output from the mic tx, (2) the 'trim pot' before the final drive stage and (3) at the normal line output from the output tx into a 47k load (for the line input TX I just monitored at the normal line output).
The result was the same at all 3 test points and for both tx inputs: no resonance peaks detected anywhere, a slight phase shift beginning to be noticeable at around 2.5kHz and increasing as the frequency increased (reaching roughly 6.5uS at 20kHz - haven't bothered working that out in degrees) the output remaining level up until about 50kHz at which point it started rolling off.
I've actually put the screws back in the lid this time!I went back to ye trusty scope to double-check the conclusion reached with REW that no zobel components were needed on the line or mic input tx.
My signal generator is pretty basic but it goes up to 200kHz sine at stable amplitude, and at 100mV pk has a Zout of 100R. I used 2 scope channels and with the input signal on one and the output at various test points on the other. Matched the amplitude and polarity of the test trace to overlay the input trace exactly then manually swept up through the frequency range to 200kHz multiple times, watching for any level changes or divergence of the traces. I did this at (1) the output from the mic tx, (2) the 'trim pot' before the final drive stage and (3) at the normal line output from the output tx into a 47k load (for the line input TX I just monitored at the normal line output).
The result was the same at all 3 test points and for both tx inputs: no resonance peaks detected anywhere, a slight phase shift beginning to be noticeable at around 2.5kHz and increasing as the frequency increased (reaching roughly 6.5uS at 20kHz - haven't bothered working that out in degrees) the output remaining level up until about 50kHz at which point it started rolling off.
Time to make some music……..
Exactly !Time to make some music……..
