REW Update

[alexc]

And what's really ..  serendipitous, is that some old traffos have come up, that have a v.big step down to a super low impedance rating coil    .. like *scary* low.  Like a guitar DI traffo but much bigger and more crazy.

Most people wouldn't look twice.  But me ?  Yep.

A couple of years ago I was blabbing on about some grounded grid circuits  and all that ...  and how it would be hard to find an input traffo to suit.

These guys really would fit the bill to a tee.  And very much like  a small opportunity to be taken, I think. Historical, like.

I'll have to ask the guy if he has them still  ;D  I'm feeling a  little historic tonight.

 

[ruffrecords]

The little 6AU6 is a good tube. I used them triode strapped in my very first mic pre. Only NOS available now and you have to select for low noise - maybe 20% are good enough for the first stage.

Cheers

Ian

 

[alexc]

I'm looking forward to going thru this 'Volume Indicator' box ...

Interesting the way the VU meter has some dc offset applied courtesy of some 1st stage cathode resistance ..  sort of a nfb loop of sorts I think and a 'zero adjust'  .. I suppose the meter has some rectifier in there ... it's not on the schematic that I can see.

Other than the vu meter load, and whatever efffect it has, it's pretty basic really  ..  quite a bit like the NYD One Bottle preamps above; 2 stages cascaded with nfb.

First stage for high gain and with nfb into cath, second stage in triode mode biased for some higher current.

I have used '6AU6 like' pentodes such as the russian 6K4P in triode mode for my 660 style limiter, without any real probs.

I'll probably move to them if needed - I have a bunch of them already tested up.

And the 6X4 is a groovy rectifier  .. I have one or two lying about.

I'll need to replace the electrolytic can caps of course ..  prolly I'll try that trick where one 'drills out' the content of the old cap and glues in a modern, smaller cap, saving the old aesthetics and mounting arrangement.

If it seems worthwhile and do-able, a small inductor wouldn't hurt for some B+ filtering.

Should the psu traffo traffo allow, I'd like to add another 9pin mini for a small line amp; it looks to be reasonable sized, but I don't have any details of current specs. Otherwise, some mods to use 9pin dual triodes all round.

It may not need, that second stage can do a reasonable amount of drive, possibly enough for driving a 15K or 20K winding sufficiently to  modest line level.

Freq response should be fine, perhaps a little bandlimited in the bottom end from the 0.047u coupling cap  ..  and not sure what the nfb network results in.

So all in all, an interesting little project for affordable $

 

[alexc]

Some more testing today ....  finding any 'sensitive' wiring  .. in this Quad NYD preamp ... 

I've already done my best 'ground scheme', which is the 'sensible or common sense' scheme plus a tweak or two. 

Now that basic  'hum' is firmly placed in it's box AND a good 'ground scheme' is implemented and tweaked .....

The main 'twitchy-ness' is in the wiring      ie.  tapping/jagging wiring with a chopstick and watching for jumps in noise floor

the next most important factor wrt to noise, is  related to the front panel 'Gain' pot and 'DI jack' positioning.

So I watch REW while I find a stable orientation that gives best improvement, then I do a 'mock up' in alu strip to hold in the position I think is lowest hum. And check again!

Turns out there is 10dB right there ....

My new baseline for the humble AV7 mic pre is :
    > 93dB of signal margin over 'hum' at THD+N 0.35% nearly all h2

 

[alexc]

I re-checked my 'calibrations'  for the M-Audio Profire 610 with balanced inputs and outputs, quite an affordable unit ..

I get the factor  'add dBFS + 12.7dB -> dBu'   

So that is included in my annotations to my REW snap.

Also, I added a factor for 'return signal scaling'  which indicates if any  'attentuation' is applied to the return signal  ..  like from the 'hv ac probe box'  I use to 'scale' down high ac signals, like at tube plates.

In this snap, I'm not using any scaling of the return signal, I'm driving an edcor xsm10K/600 with a 2K2 secondary termination - this is the 'normal load' for this build - that goes to XLR connector and on to the M-Audio and REW.

So I add the note 'no return signal scaling'  in the green box in the annotated snap.

Normally for 'scaled return signal' measurements of a tube plate, I use a -37dB 'scaling factor'  which maps something like up to 500Vac signal down to 0 .. 8Vac or so.

So for those measurements, I add the annotation 'signal return scaling factor -37dB' in the green box.

That 'keeps me honest' with respect to tracking the gain structure of my device-under-test and my measurement setup and how it relates to both  'absolute voltage' measurements (with CRO) and REW readings.

....

While on that subject, and during my 'calibration' measurements, the effect of loading needs to be considered ...

The hv probe doesn't do much, a couple of percent,  *but* the process of switching from a standard line level return to the 'hv probe' return makes a large difference to proceedings.

ie. my M-Audio interface has a large-ish input impedance, like 10K per phase differential ...so it doesn't load down low Z signals under test, much.

But when you disconnect that M-Audio balanced input from the  'output of the circuit under test',    the output of the 'circuit under test'  jumps up    because you now have an open-circuit load.

SO one should 'terminate the circuit under test' such that the output amplitude remains reasonably steady when connecting/disconnecting the M-Audio interface bal input, while watching the signal on a CRO.

In my case, I have a 10K/600 output traffo ... terminating the secondary in 2K2 means that when I connect/disconnect an xlr to the M-Audio (20K input Z more or less)  the output signal stays similar  ...  as measured by a CRO.

NOW when I switch to my 'HV ac probe', I don't get huge jumps in the 'circuit under test' output  due to  connecting/disconnecting of my measurement probes (and their associated loads)

....

When all that is measured up and taken into account, my 'gain structure' calcs all line up    I can compare a signal using either a 10K/600 output traffo with no other scaling, or my 'HV ac probe' which has a -37dB scaling.

...

Of course, all that is flip-flopped when considering balancing/debalancing calculations  ...  those ol' 6dB  factors that crop up.

One needs to figure out how the software application, like REW, interprets 'balanced' or 'unbalanced' lines.

For this, one can use a dual trace CRO, with each trace giving a 'phase-to-ground' voltage measurement. You then 'INV' one trace and 'ADD' both traces for a correct 'unbalanced' voltage measurement.

Basically, I find that 'REW' coupled to M-Audio Profire 610 reads 'unbalanced' voltages, rather than balanced.

That's fine, you just need to account for it if you want to calculate gain structures and then link that up to REW readings 

That's why I always do 'signal margin over hum' calculations ...  that doesn't change by much in all of these 'level shifting'  types of measurements ...  for a given THD.
....

It's a little confusing, but the maxim is test, measure, verify, calculate    and repeat!  Till you think you know what's what!

Then do it again.

 

[alexc]

And finally, all that is why it's good to have a 5-position rotary attenuator on the 'hv ac probe box' . ..

You can select the 'scaling' factor or attenuation and tailor it for the amplitude of hv signal under test.

ie. you want to maximise your return signal on the REW rta window ...  the closer it is to the maximum allowable signal in REW (-3dB from full scale), the more accurate your measurements are.

maxim is you don't want to attenuate your signal more than the 'hum floor' peak is above background noise.

ie. you can reduce a signal, and it's amplitude *and* the hum floor will both reduce.  Until the hum floor peak drops below 'background noise'. 

Any further attenuation is just throwing away signal-hum margin, a big no no. You can't get 'quieter' than background!

Sooner or later, you need to rise above the floor  and then you have to pay the piper. There's no tricks.

..

I did that trick, when matching a 'tone stack' loss to a gain make up circuit  ...  you don't want the tone stack's  typical  'loss' to exceed the  driving stage's  'hum peak  margin above background noise'    otherwise you are throwing away margin.

If you get it right, you are rewarded with exemplary performance over hum ... vitally important when chaining in some EQs and Limiters 

 

[alexc]

Now that I have my 'boundaries', I can do some rollin' of tubes and transformers ....  and find the best suited to this circuit.

I don't think I'll do much better with this preamp build ... 93dB margin is not bad .. in typical low THD use.

edit .. some days later ...

I did compare a bunch of av7 and at7 ...both nos and modern ...  rca, ge, eh, philco    .. I checked about 10 from a field of 20, mostly bought as 'sleeves' of 5 ...

Very tight spread .... in this NYD One Bottle application (with minimal nfb setting)

-  output variation between av7s  was less than 0.2 dB
- thd varied from 0.35 .. 0.50 %, nearly all h2 with the h3 varying from around 0.015 to 0.005 %.
- AT7 had something like 0.3dB more gain than the AV7 both at the peak and the hum floor
- hum floor varied less than 0.3 dB all round

So, I stayed with the philco set of 4 that I began with 

The general best were the nos ge, almost identical among the 5 and consistently at the low end of the thd scale.

Obviously, the nfb present, even though it is at the minimum in this test, irons out the variations across tubes.

I could take out the nfb and repeat, but here I'm just looking for a good set of av7.

...

UNfortunately,  when I previously  said I was 'closing in on opamps' - wrong!    Sadly, I confused myself when I conflated  the 'scaled down' hum floor with the regular 'unscaled'

**mea culpa  - got that measure 20dB  wrong *** I'm still remaining well away from the fabled -120dBFS barrier.  :-X

BUT, yes, closer than ever before.  Still 22dB away from that wonderful  op-amp noise floor. 

...

I may exceed the 100dB margin with a really low distortion tube,  but that 'hum floor' looks like its a hard one. Like titanium.

Now I'm looking much more sharply! I have my 'Hubble' vision going.

....

HOWever,  the Quad NYD with Edcor  does sound very nice. Very, very nice.

Not even a hint of hum from my  regular  hifi speakers ...  from a hifi stereo amp  into 8ohms at > 15Wrms levels.

I may find a slightly more optimal way to drive it, by twerking the  'send level' and 'nfb Gain' control to maximise my margin over hum ...  at the lowest possible resultant THD.

...

But it will have to wait until my next tube build to make a serious assault on the '100 dB signal-hum margin'  barrier.

I'm pretty sure the best single  'tube-gain'  stage -biz I've ever seen,  is something like 103dB margin over hum  at 1.5% THD+N  ....    the plate of a 5751 tube into 1M load or such.

...

So that's the goal - mostly better tubes, better transformers and better component/wiring placement, I think, to get me there.

PSU and grounding  is now about at my basic  limit,  as far as I currently know.

As always, it's that last couple of dB that is the hardest.

That, and honest accounting    ;D      Only REW stands between one's wavy opinion and reality!

 

[alexc]

And all of this fixation with hum floor is *historical*  and  *geographical* , in my case.

....

When I first came to this place, down here,  way down south, I found the electricity supply has a low level 'tick' at 1Hz or so.

Rarely noticeable, but with distortion pedals and compressors doing lots, which I did do a lot of the time!  , it's bad.

Maddening it was,  when I first transitioned from the city,  and their city-clean elec supply!

I did measure that tick with a storage cro .. it was there. Not just my imaginary tinnitus or my actual tt.

...

But what do you do, except learn to live with the outrageous impingement on one's personal e-noise floor!

That and measure, quantify and educate one's self!

...

Nowadays I don't run into it much ...  I've lowered my own noise by 10dB  and I get out in the orchard a lot more!

 

[ruffrecords]

The one addition to REW I would really like is the ability to alter the dB scale to something meaningful.

Cheers

Ian

 

[alexc]

I'd like a nice 'active' and 'smart'  probe box  ...  a robust thing  that you can attenuate at will  and  load the way you want!

You really have to be vigilant on performance measures ... even if you are only a part time player!

 

[trobbins]

The little 6AU6 is a good tube. I used them triode strapped in my very first mic pre. Only NOS available now and you have to select for low noise - maybe 20% are good enough for the first stage.

Ian, was that a change in hum harmonics, or noise floor level (and was that mainly higher or lower f range)?
And was that a typical triode strap, or the special 6AU6 hum minimisation triode strap?

 

[ruffrecords]

Ian, was that a change in hum harmonics, or noise floor level (and was that mainly higher or lower f range)?
And was that a typical triode strap, or the special 6AU6 hum minimisation triode strap?

It was a change in noise floor alone. No hum because for those tests the 6AU6 heaters were powered from batteries. In those days I was not measuring noise spectrum only rms level so I don't know in which frequencies it changed, However, it is fairly well known that the most likely noise variable amongst tubes of the same type is in flicker noise so I suspect that is what it was. It was a regular triode strap because I was using dc heaters. No need for the clever 6AU6 strap using the screen grid as anode and grounding the anode.

Edit: I may still have the original results somewhere if you are interested.

Cheers

Ian

 

[alexc]

I have noticed one new thing in REW I don't like - if you exceed the max input level of -3.2dBFS or so, it ceases to function (everything) while displaying a message, requiring acknowledgement by clicking a box, then requiring switching the sig gen back on.

And it does it after 3secs or so everytime until the input signal is reduced.

Pretty painful when you normally have your attention of your circuit under test, to have to stop and 'go fix the analyzer'.

It's not like an input signal overload is uncommon, after all. Why all the hoopla ?  (in this new revision).

Oh well, small matter in the grand scheme of things.

A few other unser interface quirks still there .. like when using two monitors ..  REW can't figure that out so one has to 're-arrange' the second screen each time. Never seems to remember sig-gen output assignments either .. always have to select the same thing on startup.

We're never happy, are we ?  ;D

 

[ruffrecords]

I have noticed one new thing in REW I don't like - if you exceed the max input level of -3.2dBFS or so, it ceases to function (everything) while displaying a message, requiring acknowledgement by clicking a box, then requiring switching the sig gen back on.

Is this only with the latest version? I am a couple of revisions behind and I have never had this problem. Have you reported it to the developer? He is very good with bug fixes.

Cheers

Ian

 

[trobbins]

I am pretty sure my in and out driver settings are kept between sessions, but can revert to a default - i think that was when the program wasn't closed properly, or the usb device was powered down before the app.

 

[alexc]

No - not the basic 'i-o' assignments in the 'setup' window - I'm talking about the 'sig gen' window where you have to have the outputs checkboxes assigned. It remembers the 'test signal' out but not the 'reference' - I always have to select it on startup.

And the 'overload nanny' seems to be a new thing ..  exceed -3.2dBFS return for about 1s or so and it pops up the 'nanny overload'  and will do so until you reduce, stopping the whole show until.

If you are 'zoning in on the max' things can sometimes be twitchy on a vol knob ... easy to overload ..  and with a basic RTA refresh latency of praps 1/3 of a second, it can take a few adjustments to get back on track .., meaning can take a couple of 'nanny taps' to resume what I was doing ..,  plus a couple of seconds to regain my 'unit under test' focus of mind and eyes!

WHen you are squizzing about with a CRO and HV and such, with probes probing  and chopstick tapping  ...  focus is everything, right ?

There should be an 'nanny opt out' setting  ;D ;D  It's a deal breaker for my testbench, so back to the previous version I go.

I can still use the new version on some other machines, where I don't do such 'hands inside' testing ...

....

Another thing I dislike, but in this case, I may be doing it wrong. After doing a 'measurement' sweep, I can't return to 'plain old' rta realtime mode. I have to shut down REW and restart  and  resetup the sign gen and then move all the windows over to my 2nd monitor ..  which is next to the cro ... 

I need the monitor close otherwise I can't read each of them ...  old eyesight is not so great anymore.

It may be me being clumsy, and not noticing that a 'measurement' run is active in the LHS tree view of historical runs or something like that.

I check some more.

One other thing I get, is *glitches * in the realtime rta ...  like a 'buffer overrun' kind of thing. The spectra jumps to crazy high and then returns perfectly . about 0.5s in duration.  Not that often, perhaps once every minute  or so ... but consistent, in a random-ish way.

It happens on all versions, as far as I can recall ..  on most all my machines.

Other asio apps don't do it ..  like cubase, media layer etc .... these are all music machines, with asio etc and very little issues. I have measured that latency biz with that applet and there are 'dlp interupt' thingos going on, but it could do it a little smoother.

.........

Relatively benign things , but in a world class app, like REW is approaching ...  well there's always work to be done, isn't there. Mostly though, it's unbelievably wonderful to use.

 

[alexc]

And here's a snap of my M-Audio Profire 610 balanced inputs and outputs, driving my 'Opamp Test Rig'

Its an opamp rig +/-17V with 317/337 regulated psu, a JLM  5way ...

It has  LM49860 debalancing -> Alps Rotary 100K log dual gang pot -> LM49860 non-inverting gain stage -> DRV134 balancing

It's my  'real life' unity gain scenario or 'basic starting point' for opamp work ...

....


Now when the 'unity gain' scenario has say +15db added (to normalise it with my baseline NYD One Bottle), I would get an expected hum floor or around -114.7 dBFS. 

In practice, it's a little better than that... maybe -5dB of 'extra traction' to be had .. 'seredipitous gain', I call it.

.....

So that means my NYD One Bottle, with it's 'real-life'  -98.7 dBFS hum floor (from earlier snap) is +16dB (to 20dB) worse hum.

SO  16dB 'to go' ..  downwards  ... on my quest to see low I can go  8)    ...  in tubes.   

.....

In the 'regulated HV' category, so far the  winners are :

LOWEST HUM

1.    NYD One Bottle Edcor xsm 10K/600 + ModPSU with 2m cable .. and  ebay reg module  -  is the clear winner 

...  by about  2 dB more or less    ... over the similarly historical . ...

2. Orange86  ...  cast shield output traffo ..  with onboard semi-reg and co-located psu spacious-rack- with-shielded-toroid

.. yep, it overcomes nearly all the obstacles in a rack unit  ..  but magic traffos do help a bit!

LOWEST THD

1. Orange 86 - by a long way, lower THD and it sounds like it ..  very pro

2. NYD - very charming h2 ...  hard to resist ....  but higher measuring

....

Going forward, I think I can do -3dB better from wiring,  at best , before I get over 'it' regarding basic stuff. 

The foundations of it all are psu and measurement. Then craftsmanship!  After that, magic components and topolgies.

And I can live with that. Just as long as I know 'what is'. 

...

Thanks Mr  REW.  I'll be doing a small contribution for Xmas.

 

[alexc]

The great news is that this M-Audio Profire 610 is more than 10dB quieter than my previous Motu 328Mk2  ..  which had developed some late life 'hum' problems of it's own!

Possibly as a result of being my bench interface for a couple years.

A juicy >10dB spike at 100Hz creep-ed in in it's last days.  And the firewire was never right either .. always with the clacking in and out and what not.  Eventually it checked in ..  but not out.  Must have done about 8yrs of service or so.

...

The M-Audio is super cheap, so it's a good choice for this kind of  'hazaardous' work  ..  a touch more THD than the Motu had, but a hugely superior noise floor and generally very stable even on my 'older' machines 

I think the M-Audio was already around 6 or seven years old when I bought it.  It's one year bench time and going strong  8)

(time to get another one !)

...

Looking back over the years of my measurement 'baselines' is really interesting ..  like the DIY audio equivalent of 'family' snaps.

I can see some great improvements and also a convergence of results in the last few builds ... 

Some good refining trend which I can revisit and  reference myself to,    without too much effort.

That's what these REW snaps and annotations  are all about.

 

[alexc]

Here's a snap that I'm playing about with now ...  it's a 15 dollar eby module ...  an stereo  preamp with bax style 3 band eq ...  uses ne5534 chips for the front end and a discrete  JLH style back end  ..  with onboard +/- psu circuit

I haven't optimised anything ..  it's a first test; I think I can improve the psu hum a bit ....  and use it with some other modules added ..

Huge signal margin over hum ...  generally low THD  ...  that h3 spike is from the eq section, I think. JLH generally have h2 dominating.

Pretty interesting for cheap  - the eq is really quiet and works really easy and the JLH can drive 600-600 traffos with ease.

 

[alexc]

Next test is some JLH 'mini modules'  ...  pcbs about 5cm square .. 12usd each eby ...

These are the 'old school' JLH .. single supply rail .. 12 to 24V usually .. at anything from 100ma and up depending on bias and what not.

4 transistors, with the 'finals' edge mounted for easy heatsinking. I'm using tip41c for current drivers and a moderate decent heatsink. Intended as line amps driving some 600-600 traffos.

In this test, I'm using a  'class a' discrete regulator pcb of some 12usd or so.  Enough to to 200mA if the voltage drop isn't too crazy. I'm driving that psu module with another psu regulator mosdule so I can experiment some  .. and not burn everything up!

:  So hopefully I will get that thing ...  essentially a 'tube's tube' ...  better thd, lower hum and mighty output power, early JLH style.

We shall see.