1073 type input stage caps...why??

[HenryL]

On this basis it doesn't look like there is any problem to get rid of with zobel. Playing around with these zobel values on the mic input looks on these plots more like creating some sort of 'air' eq... The cap makes more difference on the -40 setting, and adding the 2n2 across the line input tx as per the original Neve schematic and the original Golden Age implementation looks distinctly worse. All of which is thankfully consistent with what I saw on the scope where the line input cap rang like crazy and the mic input was more overshooty/ringy at the -40/-45 positions and more subdued at -20.

So the fatal flaw is...?

 

[RoadrunnerOZ]

I was aiming to maintain a typical mic source impedance into the pre73 after attenuating. Intriguing that you suggest going higher - could you clarify the rationale?

Looking at the loading of the 18i20 when I saw the 430Ω in the op manual and assuming the 1200Ω setting on the 1073, making an attenuated signal that presented a lower impedance (470Ω (- up to 970Ω with the trimmer if needed)) than the expected source device impedance at the input of the 1073 (in this case <1200Ω) - if using the 1073’s lower input device impedance setting of 300Ω I would alter the attenuator values accordingly ie 12K, 12K, 120Ω

 

[RoadrunnerOZ]

On this basis it doesn't look like there is any problem to get rid of with zobel. Playing around with these zobel values on the mic input looks on these plots more like creating some sort of 'air' eq... The cap makes more difference on the -40 setting, and adding the 2n2 across the line input tx as per the original Neve schematic and the original Golden Age implementation looks distinctly worse. All of which is thankfully consistent with what I saw on the scope where the line input cap rang like crazy and the mic input was more overshooty/ringy at the -40/-45 positions and more subdued at -20.

So the fatal flaw is...?

Looks like you don’t really need any additional filtering and that the 2n2 on the line input is a dud. I don’t think there’s a problem with the 1073, just noise from the signal source being amplified and initially some sort of mismatch between the interface and the 1073. Also not sure what impedance the GA73 is expecting to see on its output - the line input impedance of the 18i20 being 60KΩ

 

[HenryL]

Looks like you don’t really need any additional filtering and that the 2n2 on the line input is a dud. I don’t think there’s a problem with the 1073, just noise from the signal source being amplified and initially some sort of mismatch between the interface and the 1073. Also not sure what impedance the GA73 is expecting to see on its output - the line input impedance of the 18i20 being 60KΩ

Great! Thanks for confirming the interpretation - and thanks for all your timely help with this along the way - it's been much appreciated.

( I did quite a few permutations of tests previously - with the scope - trying different loads on the output of the GA73 (600R / 2k2 / 47k ) and the 600R and 2k2 clearly rolled off the top end to greater and more moderate extents (and with it any hf ringing in the top end) so with REW I decided to just go straight into the high input impedance of the Focusrite as it should be better to show up any ringing (and with no zobel I didn't see anything). I haven't tried it yet as I have permutation fatigue but if I'm feeling keen tomorrow I might run a couple of last measurements with no zobel and different loads but I'm confident adding more load on the output will only have the same effect of rolling off the top. )

 

[RoadrunnerOZ]

Great! Thanks for confirming the interpretation - and thanks for all your timely help with this along the way - it's been much appreciated.

( I did quite a few permutations of tests previously - with the scope - trying different loads on the output of the GA73 (600R / 2k2 / 47k ) and the 600R and 2k2 clearly rolled off the top end to greater and more moderate extents (and with it any hf ringing in the top end) so with REW I decided to just go straight into the high input impedance of the Focusrite as it should be better to show up any ringing (and with no zobel I didn't see anything). I haven't tried it yet as I have permutation fatigue but if I'm feeling keen tomorrow I might run a couple of last measurements with no zobel and different loads but I'm confident adding more load on the output will only have the same effect of rolling off the top. )

So you could try using the mic input on the 18i20 with the pad switched in, which has the lower impedance of 3KΩ rather than the line in. Maybe the GA73 will work better with a bit more loading. After all you need to work out a setup that gives optimum performance without any parasitic ringing that will colour the audio and if the mic input works better on your interface it may mean no mods at all to the 1073.

 

[HenryL]

So you could try using the mic input on the 18i20 with the pad switched in, which has the lower impedance of 3KΩ rather than the line in. Maybe the GA73 will work better with a bit more loading. After all you need to work out a setup that gives optimum performance without any parasitic ringing that will colour the audio and if the mic input works better on your interface it may mean no mods at all to the 1073.

Probably some slight ambiguity in my wording. From the REW technique I don't see any sign of ringing or anything bad in the frequency response when there are no zobel type components ( I think you are agreeing with that interpretation), and when using a scope I found no clear reason to do anything either - so I'm planning to put the unit back together without any zobel type stuff on either input tx (if that's what you mean by mods). (Edit: sorry now I reread my words from last night they did imply carrying on looking for ringing - must be the effect of sleep but now I think I'll just put the lid back on )

I'm not sure what the Zin on the mic inputs is on the 2nd gen version of the 18i20 it's not in the manual and I haven't measured it myself, but I prefer not to patch into the front panel mic inputs (which have the pad option), they are usually being used for inst DI etc - I could use the rear mic inputs and trim the output level down as required as one option but in any case I have a few solutions for adding load via my patchbay setup that I've been using. For example I have a 600R load to 12dBpad patch lead I've used in the past when playing around with the GA73 which also allows driving the output hard. Another loading method I've used is simply to plug various shunt load resistances into the second output socket on the GA73, which I've brought out to the patchbay as an additional GA73 output - the TRS and XLR outputs on the GA73 are just paralleled.

I haven't tried 3k as such though generally I did tend to like the sound of the GA when loaded so I'm going to experiment further.

As it usually get's EQ'd all over the place anyway in the mix then the importance of the general tonal changes created by the preamp output loading is a moot point. But if the interaction with the physical loading is doing something more useful in terms of getting rid of ringing type stuff than digital EQing will achieve later (even if we can't see any problems with REW up to 96kHz...) then it might be another reason to keep loading it down.

I believe GA introduced a built-in 600ohm load in a later version, that you could enable/disable with an internal jumper.

 

[RoadrunnerOZ]

I believe GA introduced a built-in 600ohm load in a later version, that you could enable/disable with an internal jumper.

I haven't tried 3k as such though generally I did tend to like the sound of the GA when loaded so I'm going to experiment further.

Yeah that was the output I was envisaging - 600Ω .
By my previous post I meant that it seems like you need nothing at all on the GA to have it running fine - both line and mic inputs.

Just maybe looking to see if heavier loading would perhaps yield better sonic results. Also turning down GA output and 18i20 mic input may further reduce noise - who knows.

 

[HenryL]

 

[ruffrecords]

On this basis it doesn't look like there is any problem to get rid of with zobel. Playing around with these zobel values on the mic input looks on these plots more like creating some sort of 'air' eq... The cap makes more difference on the -40 setting, and adding the 2n2 across the line input tx as per the original Neve schematic and the original Golden Age implementation looks distinctly worse. All of which is thankfully consistent with what I saw on the scope where the line input cap rang like crazy and the mic input was more overshooty/ringy at the -40/-45 positions and more subdued at -20.

So the fatal flaw is...?

Just one thing to be aware of, the interface is almost certainly not flat up to 100KHz. You should first do a loop back plot of the interface on its own. REW allows you to store this in a calibration file and subtract it from subsequent measurmenets to obtain a more accurate measurement.

Cheers

Ian

 

[HenryL]

Just one thing to be aware of, the interface is almost certainly not flat up to 100KHz. You should first do a loop back plot of the interface on its own. REW allows you to store this in a calibration file and subtract it from subsequent measurmenets to obtain a more accurate measurement.

Cheers

Ian

Thanks. No indeed and yes I did that - probably should have mentioned - just trying to keep the wordcount down a little

(I used the blue plot to generate the cal file)

Edit: I mean the graphs shown above have already had the blue line below subtracted from them.

 

[RoadrunnerOZ]

I prefer not to patch into the front panel mic inputs (which have the pad option), they are usually being used for inst DI etc - I could use the rear mic inputs and trim the output level down as required as one option

Do the inputs 3 - 8 not have pad switches? Looking at the manual it seems so but not sure if this manual applies to your particular model. I know also it’s a bit dicey using mic inputs when you have phantom power required on two inputs of a group of 4 (I believe your interface switches phantom to 1 - 4 and 5 - 8) and want to use the other two for something else. Decoupling caps solve this problem of course. A lot of consoles used to have phantom permanently connected to mic inputs - not a problem though for any transformer isolated gear including dynamic mics or your GA73.
However for the purposes of a quick test to see if the lower impedance inputs provide a better match to the GA you could try the front inputs.

 

[HenryL]

Do the inputs 3 - 8 not have pad switches?

Not on the 2nd gen I have, but from the manual for the 3rd gen they evidently included a pad for each channel in that version.

I believe your interface switches phantom to 1 - 4 and 5 - 8

That's right - though as you say it wouldn't hurt the PRe-73 and if using the inputs in TRS line input mode for other inputs they are OK as the phantom pwr is not applied to the TRS. I haven't found it much of a problem in practice, though it's not ideal.

However for the purposes of a quick test to see if the lower impedance inputs provide a better match to the GA you could try the front inputs.

Yes I think it's worth a try as it would save putting a pad together and there could be some difference in sound compared to shunt-loading with a 3k resistor into a line input (also easy to do).

 

[RoadrunnerOZ]

I was wondering also about how the cal compensation in REW works - whether or not this leads to an increased reading of the noise floor by ramping “apparent” input gain to compensate for loss of frequency response? Then there is also the possibility of filter leakage at very low level from the D/A converter……

 

[HenryL]

I was wondering also about how the cal compensation in REW works - whether or not this leads to an increased reading of the noise floor by ramping “apparent” input gain to compensate for loss of frequency response? Then there is also the possibility of filter leakage at very low level from the D/A converter……

I'm not an REW expert, so just my interpretation so far of how it is working from using it, with rather limited reading...

I don't think it is doing anything as fancy -or potentially confusing- as that. It just subtracts one curve from the other, and the curves are just straight measurements.

The cal has to just reflect what REW finds digitally between what it sends out and what it gets back and take that as an adjustment curve to factor out any quirks of the interface itself. It's not a perfect system as it can't easily account for all factors such as the different electronic characteristics of the gear you might subsequently connect to the interface afterwards which might change it a little (as you can see in the different bass rolloff in my two cal curves for example). Other calibrations can be added to the measurements though, especially -when it comes to acoustics meaurements- the measurement mic's freq response. Probably just stating the obvious here. I think when it comes to acoustic measurements then effects due to small details of the interface, are, ahem, in my experience lost in the noise

I don't see how REW can make any assumptions about the frequency response - that would defeat the purpose of actually trying to measure it wouldn't it?!

Re the noise issue I ran into, I was slightly horrified at first how noisy the interface seemed to be at those frequencies, but I'm not used to working at those frequencies and sampling rates etc and I wonder how the interface would measure when set back to the humble 44.1 I normally work at. If I get time I'll try that.

The two 'cal' curves above do not look anywhere near as bad as the first-efforts picture I posted. I experimented with averaging over multiple sweeps but in the end - having discovered the 'sweep length control - just used a longer sweep which has the effect of removing the noise effect on the sweep signal (by 3dB per doubling of the sweep length IIRC) . This produced curves which looked smooth enough to work with except at the very end of the line, even the unpadded one. I then used the same longer sweep setting as well as using the 40dB pad when taking the readings with the pre-73 in the loop.

'Filter leakage' (I'm not really sure what that is exactly TBH) is the kind of thing I was suspecting as being an artefact from the DAC which might be the source of the noise, and hence I was wondering whether the behaviour might change at lower sample rates. ?

 

[RoadrunnerOZ]

When your interface sample rate is 192KHz then it’s filtering out anything above 96KHz in the ADC path so an effective read up to 100KHz is probably where there’re going to be issues at the very top end - both in the DAC and the ADC. Maybe the best would be a sweep to 80 or 90KHz as we don’t know what the true upper frequency performance limit of the 18i20 is, especially when it hits less than 2 samples per cycle of outgoing and incoming signal waveform.
I’d be more inclined to look at analysing only up to 50KHz at the higher 192KHz sample rate. After all, when looking at the 18i20 stand-alone there are issues at the higher end of the spectrum and it’s perhaps better to keep the test range well below the starting frequency of those issues. If working at lower sample rates you’re going to also bring down your effective top frequency for performance tests.

 

[HenryL]

REW will not sweep to more than half the sample rate. So at 192kHz it is automatically limiting the highest frequency of the sweep to 96kHz. If I could run at a higher sample rate than 192kHz then could have measured the response to higher frequencies. (Edit: ie. the top frequency in the above plots is 96kHz - the fact that it looks like the graph finishes tidily on a 100kHz boundary is just a coincidence)

To analyse the response up to 50kHz running at 192kHz sample rate, for example, then you just need to ignore the portion of the above graphs to the right of 50kHz. You could run the sweep again at 192kHz and set the sweep range to start at, say, 12kHz and end at 50kHz if you wanted, and you would get a reduced set of measurement data, but it would just be a section of the same graph.

 

[RoadrunnerOZ]

REW will not sweep to more than half the sample rate. So at 192kHz it is automatically limiting the highest frequency of the sweep to 96kHz. If I could run at a higher sample rate than 192kHz then could have measured the response to higher frequencies. (Edit: ie. the top frequency in the above plots is 96kHz - the fact that it looks like the graph finishes tidily on a 100kHz boundary is just a coincidence)

To analyse the response up to 50kHz running at 192kHz sample rate, for example, then you just need to ignore the portion of the above graphs to the right of 50kHz. You could run the sweep again at 192kHz and set the sweep range to start at, say, 12kHz and end at 50kHz if you wanted, and you would get a reduced set of measurement data, but it would just be a section of the same graph.

I see. Also the mess at the top end is likely to be irrelevant as the interface is only rated to 20KHz in the specs and seems to fade off and get noisy at the high end of the 100KHz scale anyway without the GA in circuit. I was more thinking of a test that only includes a frequency range that the 18i20 is capable of.

 

[HenryL]

The spec for the 2ndGen 18i20 for both the line inputs and the line outputs I'm using is "20 Hz to 20 kHz, +/-0.5 dB" (it's a bit tighter for the 3rdGen) and I'd say the blue curve supports that very clearly, in fact if anything it exceeds that spec when you consider that the 'cal' test includes the rolloff of a line output and a line input added together.

It doesn't mean it can't handle frequencies beyond 20kHz though, it demonstrably does and is only 3dB down right out at 50kHz, and right down to 4Hz. They're just spec-ing the flatness in the audio range.

The REW sweep ran up through all the audible frequency range and we have that portion of the response curve.

It does raise an interesting question re the noise and distortion specs. They quote "THD+N" figures for the line outputs and inputs down in the tenths of a percent magnitude. Another bit of homework I have to do I guess about what that terminology means exactly... but anyway I think it must imply either that it is quantifying an amount of noise/distortion that is detectable within the audio band only, or the 'noise' effect visible at high frequencies in the above REW efforts is actually only generated when you try to get the interface to work at those frequencies. Or both.
The latter is rather close to what ruffrecords Ian was saying though I didn't quite understand what he meant as the reason for it. (It's there in the high level full resolution signal too, not just the low level one).

 

[RoadrunnerOZ]

I think this is where a combination of stepped signal generation of a sine wave, sawtooth and square wave in increments up to say 50KHz and the good old trusty scope looking for deformation and ringing superimposed on the wave forms of higher frequencies. This being done first with the interface only in a loopback then the same with the GA73 - see if there’s a difference. Testing up to 96KHz is I think working too far outside the range of any condenser microphone or any other line level gear.

 

[HenryL]

I think this is where a combination of stepped signal generation of a sine wave, sawtooth and square wave in increments up to say 50KHz and the good old trusty scope looking for deformation and ringing superimposed on the wave forms of higher frequencies. This being done first with the interface only in a loopback then the same with the GA73 - see if there’s a difference. Testing up to 96KHz is I think working too far outside the range of any condenser microphone or any other line level gear.

I've been coming to a similar conclusion.

On the one hand the 192kHz sample rate 'cal' measurements of the interface look pretty well behaved out to ~50kHz, and when I added 250pF across the inp tx on the Pre73 it made a bump in the REW graph at around 40kHz, and this seemed fairly consistent with what I saw on the scope as a wiggle along the square wave (albeit I didn't try to check the actual ring freq on the scope). But too many other aspects of the REW approach are causing more questions and doubts rather than inspiring confidence, and I'm not really sure what 'bad' behaviour looks like in the REW frequency response any more than with the scope. Less clear if anything.

I've played around with running some other REW sweeps on the 18i20, including using a different DAC as the output device, with tests at different sample rates down to 44.1kHz and it has not made things clearer! I get wiggly response curves at the top of the frequency range, of various magnitudes, with both output devices, at different sample rates, much better looking filter cutoff behaviour in some cases, but not others... I'll probably kick off a separate thread on these aspects.