REW Update

[alexc]

I just congratulate the new 'young persons' recently getting into music and audio and diy and engineering and the biz.

You'll never know, a world without REW ...

Everything is just so much more detailed nowadays and at the fingertips,  and in every field, all the time.

Often for just the investment in time and study and practice.

I couldn't believe how smart and accomplished, the youngsters at my last work place were. And that was a decade ago.

Back at the start of the eighties, the world of tech was like the flintstones in comparison.

Imagine *actually* recording to strips of plastic tape with rust etc ...  and bank books that the 'tellers' would just write in and *stamp*.

My greatest encounter as a 15 year old was a big teac (tascam)  4-track 1/4"  at the local tv tech shop 

btw how complex are those tube TVs ?

Whoa    I still try to understand some the schematics .. the kind that would unfold into huge diagrams like maps on a ship's light tables  ..

I figure it will take me at least 10 years to just check all the tubes from my past, just to see the change.

< pic of tube tv insides to post >

 

[ruffrecords]

FFT misunderstanding is a subject that comes so often on the LTspice group...

I do wish LTspice would turn off compression by default. IMHO it is the biggest source of errors in its FFT plots. We are no longer so short of memory/storage that we need it and it is a PITA to remember to include winplotsize=0 or is it plotwinsize=0. I always have to check an earlier file to check the syntax.

Cheers

Ian

 

[abbey road d enfer]

I do wish LTspice would turn off compression by default. IMHO it is the biggest source of errors in its FFT plots. We are no longer so short of memory/storage that we need it and it is a PITA to remember to include winplotsize=0 or is it plotwinsize=0. I always have to check an earlier file to check the syntax.

Cheers

Ian

I agree, the syntax in LTspice is not always intuitive. But isn't it the case with most programming languages?

 

[alexc]

I think so.

I used to love the ti tms dsps in the late 80s and later the sharcs in the 90s  ... talk about high powered lean mean machines.

It amazes me that apps with capabilities well beyond that are now common place in java  ... which I'm sure is quite optimised to run on most things nowadays ..  but it's a bit like a 'racing bus'  I think.

 

[alexc]

I wonder where are some of the 'tube guys' from the day, working for tektronix [cro makers] and so  on.

If I had more time I'd like to read a book about the development of those.  Their products were the 'hi end' source of everything tech back then. In terms of tube circuits, they probably still are and always will be - at the 'top of the game'.

...

SO - one of these days, if an old tube cro comes my way, I think I'll take a look inside.

 

[ruffrecords]

I agree, the syntax in LTspice is not always intuitive. But isn't it the case with most programming languages?

Indeed. The other thing that makes me smile is the latest version of LTspice will not run on my old Windows Vista laptop but it does run perfectly on my Linux box under Wine.##Cheers

Ian

 

[alexc]

Not much done lately ....  it's 'grass cutting' season  ....  must  ..  make ..  hay  ..while .. sun ..shines  ...

[no big machines here .. I just have a couple of lawn mowers to do 4acres    downunder in 'the land of grass'  - it's only chest high, after the 'summer rains'.  Meh - I've seen worse sans horsies.. ]

This year is a jubilee year  ..  hence the hand labour ..  fields after horses have been thru  ..  need lot of tlc.

...

Still looking at the wonderful 6922/6DJ8 family  with rew (!) 

.... but the hot weather makes me drowsy and less diy motivated  :  - well suited to grass cutting  ...  where I can think on stuff  ...  and the path ahead  ...  plus my objectives for the year 2019.  All those tomatoes in the orchard  .. with  bug sprayer  .. things.

....

New business, only a couple of purchases  ...  will take a month or so to get here, so I look forward to that.

- 10x  of 6p15p-ev 
  ->  (nice russian examples,  voltages uprated 6bq5/el84 proto - a 12W plate  and 330V plate/scr  pentode)  40usd  'to my door'

- 2x  of matched modern Tung Sol 7591a (also russian, octal beam power pentode 19W plate) 40usd each 'tmd'

I'm hoping the 6p15p are the 'last word' in 6bq5-speak  and I have a hankering to try 4x parallel push-pull  for the hell of it. The nos 6bq5 are sturdy but I want to really bash these ... 

Best to date, in my measurements  (lowest THD, most robust at 330V B+)  has been the 6p14p-ev variant  (the higher 14W plate but with 300V rated plate/scr - some say conservatively).

My average-ish nos sound good no question, but measure a little higher thd  ..  and I am not game at > 300V B+ in 'pentode' mode let alone at 'ultra linear' or 'triode' modes.

I'm interested in the diffs between the various players in this area  ..  there's quite a few variants.

If I had some 'lower B+' power supplies and some 'lower primary impedance' output traffos (any more, that is!)  I'd go for the el86.  But I have to 'pitch' for 'what's in stock'  now ....  more or less.

...

And just wanting  to do some compares of the modern and nos 7591a. As far as my own prefs are concerned, I think this one is prolly 'end of the line'   

Maybe excepting the 6aq5  ..  a very nice little tube 

 

[alexc]

There's been a couple more updates to the REW beta 4    ....  just trying out now. 

Mr Mulcahey and Team are pretty keen  ..  with the releases   

It does remind me of the hurly-burly of the commercial world of 'science instruments' with the fast pace of innovation.

It's great to just be a 'user'   rather than one of the people who do the hard yards, software engineering wise.

The day I see two measurement cursors, well then I'll know  it's the front of the thing.

As 'Krusty' would say  'Hey-Hey' 

 

[alexc]

Another day, another REW thing going ...

I updated to beta4 - everything mostly the same  : good start!

I'm doing some testing on the 6922 family  .... I have one going as a 'grounded cathode amp'  stage + cathode follower stage - a classic functional block and generally pretty good.

While I'm finding my bias points and what not .. REW  is as always, indispensible.

Nothing flash, just workman-like honing in on the zone.

....

SO - I used the 'stepped sine sweep'  button  from the  main RTA window 

Works quite well - it allows the duration of each step in frequency, with a settable interval period  as well as the usual parameters that were available in the 'measurement' functionality within REW previously.

I like it     The CRO shows the proceedings in detail .. all very nice and steady and stable like.  Pro.

It is a big step forward, integrating 'sweep' analysis and reporting,  from within the 'plain old RTA'  window.

The 'frequency response'  results are painted in real-ish time, complete with the fundamental, the h2 and the h3.
Quite intriguing to watch it progress - very engaging to the user.  Top marks for sure.

Once complete, (and with the RTA window inactive), the 'DR' readout  - I think it is 'Distortion Residual' headline figure  ...  is displayed  ...  and it makes sense (even to me). 

This area of 'residuals' is of great interest I think. I can only imagine where this kind of display and reporting is going to go in the future - clearly it's quite cpu intensive - on my 3yr_old AMD box it's pretty flash. 

'like far out man - this is crazy good!' - would be a reasonable approximation of my initial reaction.

After the 'stepped sine sweep measurement process' is complete, you can just continue on your way, RTA wise    ...  the resultant frequency response trace(s) just float at the top .

TOP of the class!

.......

I can see my newest  'annotated REW snaps' with the new capabilities. Hopefully a day or two.

......

The results of the sweep run 'drop in' to the 'dataset handling environment' for historical analysis and post display processing.

Just like before, as far as I can tell. 

This functionality is one  of the most impressive, imho,  and it puts REW on a trajectory akin to the sort of thing one uses in financial and biotech charting,  without question.
...

The other thing is    "how good the 'signal generator' updates are"    ..  the 'level' data entry is superb  ..  smooth and accurate, supporting the measurement thing rather than fighting it.  It's a pleasure to dial around on.

I could probably go for a 'slim mode display' with the big 'wave display' scaled down a bit ..  it's good for 'standard' users but 'advanced users' might like to tailor the display and so on  --  but these are low level suggestions    All good as is!

Whoever worked on that (signal generator updates)  is (are)  (a) real artist (artists).   

I love the dBu display, and the 'little dBFS readout' too    ...  sheer luxury.

Finally but not leastly, the 'input level window' ..  floating at top right corner ...  the 'peak sample' in  dBFS does indeed equal the 'headroom to 0dBFS'

Fan-***-tastic.  The fact it is there at all is very fine. And that is accurate and always-available  ...  well that's just more champagne in the glass.

....

Just another link in the REW chain.  This is what real innovation feels like for the 'common man'.

10 out of 10    tick

 

[alexc]

And on the subject of REW innovations, it does appear to me that there is another kind of mode of display used in the 'RTA window stepped sine sweep' function ..

It's not the same as the 'RTA mode', but is much more like the 'spectra' mode  within REW.

But it's not quite the same as the 'spectra mode' either. 

It is similar in that it has a 'shaded envelope' display for the high freqs. It is different in that the envelope doesn't 'bounce around' but is more 'flat. It reminds me of some of the classic filter responses ..  the chebychevs and the butterworths and so on.

So, it looked to me like a 'spectra mode with flat hf extension'  kind of thing - not sure what I'm seeing yet; could be the device-under-test or could be the new REW functionality.

I'm keen to look further into it - I kind of liked it  ..  this 'spectra mode with flat hf envelope'  reporting 

Thanks kindly Mr REW  ... for all the newest capabilities in addition to the foundations previous.

...

Anyway, onwards and upwards  ....  I have to do more tweeking on the 6922 and so on  ... hope to have a few new snaps for you all shortly. 

I still have another 2acres of grassland to pulverize.

....

 

[ruffrecords]

Make sure you fit grid stoppers to your 6922s. They can squeal in the MHz region.

Cheers

Ian

 

[alexc]

I always do    ...

I wish I could really hone this great example of tube wonder  .. the 6dj8 / 6922  and extended family into modern days  ..  the russkie greats.

Even with all the refinements available, there's still only so many hours to experience it in a 'measurement' sense, let alone 'aurally'.

Still we do as we can.

Thanks Ian - Ruffrecords  ... you've always pulled for the 6922  .... and I can understand why. 

It's my great hope, with REW help, to find some of the other fine performers in the tube world.

...

Here's my initial snap for the 6922  setup as a gca + cf  functional block  ..  driving 4u7 wima into 10K/600 edcor xsm output traffo.

It's pretty good ... the  'low THD test'    gives  0.61%  thd nearly all h2 at the output of the cf.

The input traffo,  which is now a very cheap 10K/10K  contributes    around 0.1% of that  ....

- signal margin over hum is +107dB with 25dB headroom (to 0dBFS) and freq resp is extremely flat 20-20K   

I'm using 47K plate load for the gca stage, with a bias of +3.2V  (input signal is 2Vpp) ..  and the cf stage is cap-coupled thru 0.1uF and has  850R+22K  as the cathode load  .... with a 4.7uF wima to the edcor 10K/600 output.

....

The 'hv probe' attenuation or 'return scaling' factor  is approx -37dB 

ie.  47Vpp (unbalanced measured with cro) at the cf output is  ...    scaled down to approx    0.67Vpp  (div 71)    (psuedo balanced)

The 'psuedo balanced' send to REW  - is  interpreted as 'unbalanced' voltage.

As always, one needs to *check their '6dB factors' with respect to 'single ended' vs 'differential' measurements  * 

...

One other thing REW I noticed today  ...  when 'hovering' over the 'trace legend', the 'trace' that is hovered over  is  *highlighted* in the main window    ...  just another super nice touch from the REW crew!

The 'nudge' amplitude cursor increment remains at 0.2dB  ....  very good, but not yet fully  'REW-good'  (0.1dB) - the underlyings continue to support 0.1dB basic resolutions 'in spades.'

 

[alexc]

So the 6922 is clearly a very good performer in this 'gca+cf' config  ..  especially given the relatively low B+ of 246V.

The cf section is very impressive ...  super low intrinsic distortion  ...  no probs whatever driving the load 

If I push it to say 1% thd, the signal margin over hum rises to > 115 dB  ....  and the 6922 still barely ticking over ...

My impression is that the e83f has more dynamic range but requires a much higher B+ to achieve it  .. and at significantly higher THD.

The 6922 has less THD at much less B+.  The gain of  the 6922 is around 2dB less typically, than the e83f triode mode with the same 47K plate load  ....... 

The downside for the 6922, is that it requires a decently high input signal to really shine  - like 2Vpp  to 3Vpp  or so.

More so than the e83f triode mode which seems to do better with the signal in the 1Vpp to 2Vpp range  .. even with discounting the 2dB  or so difference in gain 
....

However, I'm still refining my comparison      and it all comes down to judgement calls  in a way  ...  both these tubes are very good performers ...  it takes the really small optimizations now,  to get any demonstrable improvs.

Nothing is ever straight-forward ...  so one must always check their apples remain apples, and their oranges remain oranges  ..  at all times.

...

I have to re-check my 'attenuation box'  numbers too ....  it may be that I can 'map' my HV probe voltage range  better  to my interface input range.

So far so good  and I find :

0dBFS = 2.357Vrms -> 6.67Vpp 

and  my 47Vpp signal (measured at the CRO) is mapped down to 0.67Vpp  which is div 71 -> -37dB return scaling.

I still have to check how 'constant' that -37dB scaling factor  (which is position 2 on my rotary atten box) is at either end of the voltage (amplitude)  scale.

It may not be as linear as I would hope.  So  more there to determine. TBD

The other thing is that a return scale factor of -37dB  may be pushing down too far into the interface 'noise floor'.

I'm pretty sure I only have 'about 20-ish dB' down there to 'scale down into' before any 'noise limiting' occurs.

It's best to attenuate only as far as the interface's fundamental noise limit ..  any more is 'noise limiting'; 

The more I compare the results of the

-  'edcor xsm 10K/600 output balancing transformer with no further scaling'  standard
and
- 'hv probe unbalanced with attenuation scaling factor' 

The more I think I am skewing results a bit  ... my attenuation factor should probably be no more than 26dB  :

Basically, by not choosing the right attenuation factor, I'm downgrading the 'margin over hum' perf but not the thd measures so much.

Tomorrow I'll recheck and rejig  - it does take time and effort to really hone in on the zone ..  especially when one's frame of reference get flip-flopped  (for the better).

...

Here's a snap of the REW 'stepped sine sweep' in the 'RTA window'  doing it's thing  ...

 

[alexc]

And when I try to 'LoadSim' the operating conditions, I get a reasonable match of the operating conditions, but nowhere close for THD ..  with a 2Vpp input.

loadsim  says nearly 2% h2  ....  Tube Test Rig  says  0.61%  h2   

 

[ruffrecords]

Simulators are very poor at estimating distortion and they do seem to err on the high side. LTSpice is just as bad. Better to measure.

Cheers

Ian

 

[alexc]

Another day, another calibration run  ....  moved to an attenuator setting of -28.3dB (loopback measurements).

With  plate signals of 10 .. 100Vac my attenuator box 'factor' of -28.3dB is a bit more accurate in REW ...  with respect to the 'signal margin over hum'  or loosely speaking 'signal noise ratio'.

At that attenuation, my 'signal-under-test' noise floor is above the 'noise-floor of the measurement channel', so I'm not 'over-attenuating' and thereby worsening my 'signal under test' dynamic range.

I may need to 'rejig' my atten box to really hone in some more ....  my old -37dB is too much and -28dB is not quite enough  :
I suppose 33 is the magic number  ...  I may try a 100K-50ohm traffo too  ..  it's 33 for sure.

....

I reset my RTA display limits to best show this 'tightened up' view  ....  I can't quite get the display as I want ... I like to have the 'freq-response-sweep' trace fully visible across the top and not obscured by the 'thd display window'. So in order to 'clear' the THD window, I need to 'scrunch up' the RTA amplitude limits  a bit. 

I would like to see the 'THD window'  become  'long' sideways and 'short' downways.  It's a question of 'visual to noise ratio' 
No bout adout it.

*  Never happy are we   *

Also, not at all sure what the 'DR' is telling me .. in the 'peak present' displays.    In the 'silence' I get 'it' more,  and in the 'sweep' too - but the 'plain old RTA mode' THD display window DR field, I don't know what is.

...

I think I get about 2dB more 'accuity' from this ( -28.3dB atten) match of the 'measurement channel' to the 'signal under test'.

ie. signal margin over hum is improved around 2dB

It's a bit like when the optometrist switches the lenses  until he finds the best for your poor eyesight

....

The other thing I have to consider is the reduction in THD one gets in a simple test like this ...  when one sets the gca bias  colder  ...  wrt  to the chosen test input signal amplitude.

ie. 2Vpp sine at the grid  ....  bias at  +3.2V improves the reported THD  than bias at +2.8V.

- gain changes a little when I vary the bias (all other things being equal) ...  when bias 'colder' ->  less gain, THD is also reduced

My thing is to bias as I would in the 'real world' for the expected role ...  and go from there. 

ie. 0 .. 1Vpp range input    or 1 .. 2Vpp, 2 .. 3Vpp, 3V .. 5Vpp, 5 .. 10Vpp and so on

.....

With this in mind, I get a baseline thd of 1.0% from the 6922 gca+cf  at B+ 245V with  +3.2V bias with bypass cap .. 2Vpp input and 47Vpp output  ...  [driving my  '1M' atten box as load]

I could chase that back down to 0.61% by biasing for less gain ...  but it's kind of silly ..  one is generally striving for all the  *basic* available gain.

It's a bit 'artificial' or 'tailored for test' to  get too 'finely biased'    ...  not to mention easier  to go into overloaded grid condition when something apart from the test signal is encountered  ....  ie *music*

....

So now my two tubes types are converging in terms of thd ....  but the very big gulf on the B+ remains. 

The 6922 is designed for a lower B+ and in THD, shows 'no penalty for it'

whereas E83F-triode-mode needs a proper >300V B+ to get similar THD, but with higher signal-hum margin 'for the trouble'

I confess, I am a 'signal to hum' guy  ...  I like to blow any 'hum-gains' made, on 'tube limiting' repeatedly and wherever possible. 

As long as it's <1% that is.

This is what it takes to get there!

....

So it was worthwhile to spend an hour or so rechecking my 'atten factor' to suit the testing I'm doing and to make sure everything 'lines up' properly. It's the kind of thing one does at each test session  to 'keep it real'.

Generally speaking, I get a maximum of approx  0.3dB uncertainty in the measured amplitudes of  I/O with the M-Audio-REW combo 'round trip'.

That's 0.1dB on the Send, 0.1dB on the Return and a further 0.1dB 'for good measure' 

Other items to account for are ....   [apart from 'circuit gain' and 'hv probe box atten+bal factor']

(in order of impact)

-  input debalancing transformer level shifting
-  insertion loss in the input transformer
- uncertainty in CRO measurements 
- loading effects at the audio interface send and return

With that lot in mind, you can really closely measure and compare-to-expectation  the gain structure all the way thru - very important.

....

And finally  ...  the 'takeaway' 

one can get a good match between the 'display', the 'gain of circuit under test' and the 'attenuation factor'  such that they mostly cancel each other out, giving you a 'measurement channel' which is nicely 'normalized' to the 'device under test' and with a 'traceable' degree of confidence.

....

Probably could do with a really great 'attenuator' overhaul  ...  not just the pitiful little lorlin with 4 positions in a 'stomp box' that I currently have  ;D

Might be a good time to buy a proper large multideck switch and do the thing properly.

All good! 

.....

Here's a pic of a suitable eb y  rotary switch for use as an attenuator    ...  12usd  ...  11pos, 4 deck  ..  large open frame


Tomorrow I'll really start to grind and snap  8)  but the 'attenuator box' upgrade will have to wait ...  maybe next winter

 

[alexc]

And on this subject of 'attenuator box with rotary control and hv probe'  thingo ..  for the 'plate-ing' in REW

I wonder could it be done to use a mosfet as a    high but switchable variable input impedance, high voltage probe buffer with some more actives to do the variable attenuation and balanced 600ohm output ?

It would have to have a super low intrinsic distortion (< 0.1% at 1KHz)  and noise (-120dBu)  , however  ...  *and* be linear over a 120dB range of signal  8)

It would be quite nice if it could take a power feed from the 'tube build under test'  ...  further regulated for sure  ...  perhaps with a 'self powering' capability of sorts if really needed.

Fully variable 'attenuation factor' and with a calibrate-able readout of 'fine attenuation factor' please!

Of course, it would require a 'pass thru' to CRO  which can be switched to show 'pre' or 'post' or 'none' (unbalanced) voltages and have a cabled output [to the CRO bnc  input]  with the correct  'range set locating pin' if required.

In addition to that 'cro output' bnc cable, some 'multimeter' connectors might be useful too  ..  for watching on a true rms meter.

I'm very sure HP and Tektronix had this kind of 'supporting test equipment' available.

.....

I have seen things in this area but nothing that grabs me.  I guess it must be possible ...  I could almost do it in tubes     

(say it in 'snow-board'    )

Like the old 'volume indicating amp' from the 50s I have  ...  but with output to REW ]

Mosfets would be a lot sexier for this I think and without all the tube distractions. Could be a project  for one of our younger, more tech-savy (than me)  members.

...  sorry for the bad pic  [of my passive hv test probe with psuedo bal output]

 

[alexc]

and one more Tektronix pic ...  good Lord these people  were good!

 

[ruffrecords]

and one more Tektronix pic ...  good Lord these people  were good!

No, they were superb!!

Cheers

Ian

 

[alexc]

You got that more right    

It's hard to fathom the build 'craftsmanship' in these Tektronix tube oscilloscopes - it's just a joy to behold such dedication to quality.

I suppose these Tektronix 511 series  were from the 1960s.  Must have been great to be a tech in 1965  

...

Here's my 'latest and greatest'  REW snap for the 6922 gca+cf functional block ......  I have my new annotations and display style going  ... 

I'm getting closer to 'normalised' display formats ...  here I still have around +7dB or 'net gain'  end to end  ..  so my -7dBu 'send level' returns as 'near 0dB'  ..... which is easy for the eye 

I chose -7dBu because that results in a 2Vpp input to  the grid of the functional block  - it's one of my 'scenarios' don't you know.

You can see the basic 'input-output' uncertainty of 0.2dB  ...

I like to say another 0.1dB for 'good measure'  ...  like 0.2dB +/- 0.5dB is the basic 'roundtrip'  uncertainty.

.....

Headline figure is 0.94% thd at 116dB signal-margin-over-hum    ..  best yet! 

As I said, by 'biasing cooler'  say +4.0V changed from +3.1V, I can drive the headline THD to 0.61% at the cost of some 2dB in less gain. 

I'll stay with this 'scenario' as my 'standard' for now ...  I like +3.1V bias in a 6922

I think that's about as good as I get for a 'basic setup'  ....  47K plate +3.1V bias on 1K Rk, 0.1uF cap-coupled ,850R+22K cathode

......

The frequency response is mostly influenced by my choice of input traffo  ..  in this case a 5 dollar sino 'altronics 10K/10K' weighing 100g  ...  small but not stupid tiny .. it's actually quite good - it contributes around 0.1% thd to the headline figure and has something like a 1dB insertion loss.

The Edcor xsm,  which is my usual standard,  trounces it in the low register ... it should do at 436g in weight  and  cost .. about twice as much in '2007 dollars'  (now more of course)    ....  but I'm all out of stock  ..  of this 10K/10K ratio.

I figure once I have my 'normalized'  test setup working correctly and 'traceably' THEN I go to the 'magic transformers'

I like to get my 'ducks in line' before parading the greats 

...

NOW ..  back to the E83F-triode-mode ..  which is still as I left it ....  around 1.53%thd for the similar signal-margin-over-hum at B+ supply of 325V  as I recall 

(pre REW beta 4 world shaking release)

  ...  let's see how it 'shapes up'  with my latest and greatest 'normalised' and 'annotated' REW snaps

I predict a  THD 'smack down' in the favour of the 6922  ....  it 'sets the bar' 

  ...  by any account of performance that I have personally measured.