Measurement Software - 'Room EQ Wizard'

[alexc]

Some more ruminating ..  and a lot more testing of my circuits ...

I finally measured my Motu 828MkII bench interface balanced analog line in-out XLR connectors with the CRO and DVMM to determine the available range in dBFS, dBu, Vpp, Vrms.

That allows me to 'calibrate' Room EQ Wizard (REW) measurements appropriately to dBu and 'keep it real'. 

REW has a max of '-3dBFS' in the sig gen output. The inputs accept up to 0dBFS in the software.

Happy to say it all went well - I get a value of  21dB  in converting dBFS to dBu. 

Everything is consistent and no anomalies at all. I can say with confidence that with the Motu 828MkII, REW dBFS measures are dBu less 21dB.    Or, dBu is REW dBFS plus 21 dB.

Also, REW 'avg dBFS' live display in the 'RTA' window is accurate for the measurement bandwidth in use. You can also see the expected characteristics of a good FFT implementation when changing bandwidths, windows types, FFT lengths and so on.

All great news and I finally leave behind for good my previous RTA. The only extra thing I would like is a capability to display in dBu, along with an settable calibration factor,  like I have just described. Same reasoning for those dBV heads out there!

I also figured (duh!) how to switch the peak display off

--

Now I have a proper handle of the dBFS measures that REW likes to work with,  against dBu which I like to work with....

I have a truer measure of my bigSEamp at the various points in the signal chain ...

-  throughout the gain stages from input to output
-  with several different gain stages for comparison and dev 
-  connected to different speakers (8ohm and 4ohm) and load boxes (16ohm down to 2ohm 50W)

Now in this amp, the intended 4ohm 15" speaker box is the final arbiter of what's what noise-wise.

It turns out that my 'threshold of audibility with ear-to-speaker' test results in a required 'hum' of -115dBFS at a max when measured at the speaker terminals with 4ohm speaker connected.

That means -94dBu is the max for any of the hum components 50, 100, 150, 200 Hz.

Which is a pretty low figure - the speaker is very sensitive and the amp's input-output gain is approx 27Vpp/1Vpp or x27 or +29dB.

At present, with 'All My Circuits', I get a personal best in this bigSEamp of some 82dB signal-to-noise ratio.

The 6SN7 driver has 87dB signal-hum ratio. There's not much more there to gained, maybe a little with an even higher plate resistor in there. The output amp costs about 5dB in hum. Hence 82dB s-h-r 

So, that means 'silence' piped in the input (everything wide open) needs to have max hum components of -95dBu at whatever gain makes it  achievable, with a max of 3%THD at any point,  from input to speaker terms  ...  with signal applied then  82dB higher than that.   

Good news is that, I do have a personal best baseline - like I described above!

Bad news  is that  I'm  right at the bottom limit of that spec.      And I can only get it without anything upstream of the 6SN7 driver and finals.

With the AV7 in front of that, I lose a further 6dB to noise, taking me down to 76dB of 'dynamic range'. Which is too low.

Albeit, with the AV7 running some +29dB gain, and at 2.35%THD or so. That includes  the imperfect v.e.m plate follower which is not too bad, considering. Maybe I can improve that a little more. I'll know after I implement it in 6922 

Which is actually very good performance, for a AV7 type (as far as sample size of around 20 has allowed).

However ....

If I use an Edcor xsm 600:15K in place of the AV7, I achieve that 82dB 'magic' minimum dynamic range.  Yessss!

AND it sounds better   It's distortion is less than the 2.35%THD of the AV7 in this circuit.

Freq response is still all there - it sounds a little denser on the bottom end and a bit clearer too. In short - better.

OF course, I'm using an input signal from the Motu analog outs at a fairly high level - something like 4.25Vpp, stepping up to around 22Vpp for the grid of the 6SN7.

(there is nfb at the cathode of the 6sn7 so the input signal gets knocked down 6 .. 14dB depending on the 'nfb control' setting.)

----

AND I noticed, for the first time EVER some cancellation of hum, at the speaker terms, when flipping the phase of the Edcor xsm600/15K  secondary. It really dropped the h100 like a stone.  Primary flip don't do anything much. The secondary phase flip reliably clubbed the h100  down, not here-there-gone-one-second-later type of thing. 

Also, adding the thick alu bottom plate has a slight improvement in hum floor - nothing much, perhaps 0.5dB only.

That 100Hz was my worst offending hum component previously;  and nearly always is! I am mighty impressed with this  8)

So there is indeed some value in thinking about tailoring phase inversions thru the chain, such that it cancels out a little at the end.

Next step would be to use rheostats and so on to incrementally 'mix' in psu noise - whilst checking for reliable improvements. 

I'm not going there in this build, however, I am seriously considering prototyping a John Broskie 'Aikido' stage in this build, if necessary, to ensure my beloved 82dB of signal-hum ratio is delivered safely at the end


----

Here's the 'silence floor' spectra of what I'm talking about ..    averaged  'hum and hiss'  noise  is  -97.8dBFS in this example. Meaning -76.8dBu is the actual 'noisefloor' headline figure, 10Hz-48KHz measurement bandwidth. 

A lot of that figure is 'hiss'  ie noise > 600Hz or so - there is more energy in the components at higher frequencies.

For the hum alone, I have a worst case  from the 50 to 200Hz components of -115.4dBFS, meaning an headline figure of  -94.4dBu 'hum' at the speaker terminals. That's  0.01476 mVrms.

Very low and is virtually inaudible, with the ear-to-speaker and amp wide open. (+27dB end-end gain)

---

Directly, I can measure to 1/2 significant digit giving 0.5mV at best, with my true rms DVM or I can measure with 2mV per div on the CRO, meaning  1/2 significant digit of 0.2mV at best.

The HV and high signal probe I use, has an as yet, undetermined  accuracy   and a range arbitrarily chosen to attenuate down from 350Vpp to 12Vpp or so, to be suitable for connection to the Motu analog inputs.

Finally, the load box I use interfaces the Motu analog inputs to speaker loads  ie.. 10K load to be presented to the Motu inputs, whilst attenuating the signal-under-test such that it results in 0 .. 12Vpp or so.  With 2,4,8,16 ohms 50W dummy load for amp outputs feeding that attenuation network.

 

[alexc]

So I have the AV7 wired up - I leave it for now wired up but not in use,  and move on for now.

I have the REW sign gen -> Edcor xsm 600:15K -> 6SN7 -> 6550 finals  chain

I'm moving to implement the 6922 tube. I know it's a good, low noise, low THD tube. I want to compare it with the AV7 results, in a similar configuration. I know it will be 'better' measuring - the question is, can it beat the Edcor ?

Meaning, can I achieve a signal-noise ratio of  > 82dB which is my personal best inside of this amp, at the speaker terms ?

I doubt it, but will see. The thing is that the SE ouptut stage has quite a lot of B+ ripple in there. So one is already behind the 8 ball at the get go, short of using a second full rated choke.

I have yet to try any kind of 'snubbing' or special filtering in the PSU HV rail - it's just a 5AR4 rectifier and standard C-L-C filtering to the SE opt. My B+ is not particularly high, at some +370V loaded, so series resistance is low by necessity. Also at 150mA for the SE opt, any significant resistance adds up to lots watts.

---

So, the driver has to be a lot less noise than the SE output stage so as not to worsen the already  poor SNR.

Hence my addition of a second low-ish rated choke for the driver+preamp of this amp.  It's  a 10H, 100mA unit  - to knock down ripple  EXTRA,  extra low, without losing too many V -  the extra filtering is a necessity in the preamp.

And the voltage amp before the driver, especially, especially  needs to have a good SNR so as not to contribute anything significant to all that downstream yutz. That means a really good tube, high plate resistor and a good size R-C local B+ filtering.

I'm currently dropping about 50V here for the AV7 - I'm going to drop by more like 100V or so , with the 6922's plate supply  - to around 230V.

Things are a little easier upstream of this critical position - it is this 'choke point' for SNR that is challenging  ;D

Bascially it's the 6922 vs Edcor  to determine who is to replace the AV7 in this amp's driver-driver stage!

I expect the 6922 will be quieter wrt hum  than the AV7 but still not as good as with the traffo.

If I have to, I could drive the primary of the Edcor push pull, or more likely, get a proper parafeed interstage traffo.

While I'm at it, I could also try putting using a  'plate choke' instead of the 30K resistor in the AV7 now      This is one place it might be interesting to try,  induced hum not withstanding. 

The traffo step up of x5 or +14dB for 'free',    noise-wise is the main attraction here.  The rest of the gain will have to come from upstream in the instrument preamp section.

So far, not seeing any induced hum into this xsm 600:15K but I'm watching and listening all the time

I have the discretion to raise or lower the input to this stage, providing I don't exceed 3%THD at any point downstream.

-----

Anyway, I'm light years beyond where I was at the start of using REW and the start of this thread! 

A little bit of positioning of the edcor xsm  and voila - here's the latest spectra of the speaker terminals

....  90dB s signal-to-hum  margin  for 3%THD or so, 3Vpp from the sig gen for 6,72Vrms or 11.2Wrms  into 4ohm speaker load.

So that is my new personal best of the best   Now, that's with the amp just turned on - after a while that will settle in and the noise will rise a dB or so higher.

Still  ..  this is what I have to beat now!  Thats a pretty high bar as far as I am concerned.

 

[alexc]

When one looks at the 'silence' spectra, the hum components  h50, h100, h150, h200 increase - it's a feature of the 'bins' used in the fft computations - the energy at silence is redistributed into the display bins a little differently than the display with H1 and harmonics signal present.

It can vary quite a bit, so I always take 3 measurements when calculating my signal-to-hum margin :

- the 'silence' hum components
- the 'signal present' hum components
- the 'signal present' H1 component

I take the worst of the hum components and subtract it from the H1 value. There's prolly other ways to treat that. I find this way is pretty honest.

Of course, for the true signal-to-noise ratio, the hiss stuff needs to be added, so I use the 20-48KHz bandwidth 'averaged' value, such as displayed at the top RHS of REW's RTA display  (also at the 'Levels'  signal  'In'  bar meter and readout).

So that means I have a real noise floor of  -77.5dBu and a real hum floor of -98dBu for the amp running at 11.2Wrms into 4ohms.

Completely inaudible hum with the ear-to-speaker . Which is the desired outcome.  8)

That's with 3Vpp input signal and 88Vpp to the grids of the finals.

This is starting to compete with my pushpull el34s - since I rejiggered them post-REQ   

Also, the ratio H2/H3+ of the speaker output has shifted towards a more pronounced odd order harmonics.

With the AV7 in circuit, is a lot more H2 than H3+. With the Edcor XSM in circuit, the odd orders look more like 'pentode' mode.

... speaking of which, lets take a look there, now     

 

[alexc]

Have been testing the 6922 position - wired up as a grounded cathode amplifier, very standard.

Have settled on 235V plate supply, 70V plate, +2.1V bias on 47K plate resistor and 680R cathode resistor, bypassed.

Stage gain is x24 or +27.7dB at 1KHz driving a high-ish Z, some 500K or so.

My circuit, which has a 1M audio taper pot+ 0.1uF cap hanging off the grid circuit, which consists of 1M grid leak resistor and 1K grid stopper.

Signal is connected to 'aux' guitar jacks, identical for both triodes. 

With the Motu feeding 1KHz for 3.0%THD at the plates,  'Vol' control fully open ....

- I can get a best signal-to-hum ratio of  77.0 dB

With the Motu feeding the Edcor xsm600/15K to the jacks,  'Vol' control fully open ....

- I can get a best signal-to-hum ratio of 83.3 dB

I checked again with the Edcor 600/15K secondary phase flip, and same as before, I get 3 .. 6dB decrease in the hum floor, when adjusted for same THD at the plate.

That old 'cancellation of 100Hz hum component' thing again 

Now those figures can be improved by about 0.7dB by decreasing the grid-leak circuit resistor from 1M down to 300K or so.
At the expense of input impedance, which is fairly important here (instrument inputs) - so I'll stay at 1M for now. But duly noted.

Also, judicious balancing of the 'Vol' control and input level, for same THD at plate increases the signal-hum-ratio a bit.

'Backed off 20% rotation' and increasing input level gives an improvement to 83.0 (traffo) and 79.0 (direct) dB

----

So it appears that the 6922, as I have it set up - main parameter is best signal-hum margin at 3%THD  - is about the same  as the AV7/6829  similarly setup, in terms of  signal-hum margin :  around  80dB  for each  ..  more or less ..  (+/- 1dB)

I'm pretty sure that's close to my limits here -  can't seem to 'goose' the signal-hum margin further without some major 'heroics'.

SO,  It's looking like I can keep the AV7/6829  and the 6922 where they are .....  for now    A lot of this measuring is to characterise these med-mu tubes somewhat - it's my first usage 'in earnest' of either.

----------

NOW .....  back to AV7/6829 tube  - I'm going to rework it   

-  from  a      'v.e.m plate follower' + grounded cathode gain stage'
-  to      the  John Broskie  'two triode aikido grounded cathode plate follower'

Imma try for some 'psu noise cancellation' with and without traffo involvement   

----

Super interested to try my second JB circuit - the first, his variant 'CCDA' stage has always been a favourite and has worked well  for me.

Here's hoping my first 'aikido' type circuit is a winner  I have some more stuff upcoming!


PS - so far, the nos 6DJ8,  6N1P,  6N23P that I have tried, have all measured repeatably worse than the humble 6922 steel pin from EH.   

 

[trobbins]

Well done Alex. 
It's a long thread now, when one starts from the beginning!

I agree that REW is well worth the initial effort in getting to work.  For most testing I just use a converted $1 USB soundcard, and use a Creative E-MU 0404 for reference or lower noise floor or wider bandwidth checks (not keen on damaging soundcards in valve amps!).

Are you able to track an increase in mains harmonic levels as a signal output power level is increased?

Ciao, Tim
PS. Good to see a Vifa kit still in use - I was trying to chase down tech details the other day, but not much on line.

 

[alexc]

Hi - thanks for the interest.  ;D

Mains harmonic levels, also known as 'hum',  increases with gain and also the degree to which a stage amplifies the low frequencies.

ie. not all stages amplify 'coherently', the hum components.
eg. +10dB stage gain does not necessarily amplify hum by the same amount.

There can be some degree of 'positive reinforcement' of hum or 'negative reinforcement' of hum in a gain stage.

Basically some cancellation or some additive effects result from a particular configuration - the 'equivalent noise resistance' and the 'power supply rejection ratio' are but two reasons why.

Signal output power increases generally do mean an increase in hum - basically the same as increasing overall gain.

However, the cancellation effects are often more evident in the output stage - there is step up/step down gains, transformers, varying inductive loads, negative feedback etc - all of which have an effect on how coherently the hum is amplified.

So, preserving signal-to-noise-ratio is key to maintaining an end-to-end acceptable level of hum. That means careful choice of gain staging throughout the signal path.

Not so easy, as I have found out!, particularly in a unit with co-located psu and signal components. And even worse in a high HV current, single ended configuration without regulation and only passive filtering of supply rails!

And tube rectifiers is another subject altogether.

REW is indispensible for really tuning a signal path - hence the length of this thread! The ease of continuous, realtime cursors and measures means you can really track the 'devolution' of signal from end to beginning of the signal path.

----

What I have found does surprise me - even allowing for the 'error margin' incurred with my primitive highZ, 'HV probulator' - an error which I have not yet quantified, but is likely to be substanstial, in terms of %THD measures.

For example, Vacuum Tube Valley includes similar tests of stage THD for lots of common tubes. They consistently achieve THD figures far lower than my measures. They use pro distortion analysers

When I finish with this amp, I will do more work in quantifying the measurement errors and making an active HV probe - one which loads the circuit less and does not impose additional distortion - if possible.

----

What I've found is that most tubes are pretty poor in noise and distortion. I've found that getting less than 1.75% or so THD from a tube, with decent gain,  is not easy, at least for for a simple grounded cathode gain stage.  [big surprise hey ?]

And achieving a signal-hum ratio of >80dB is not so easy either!

Only the EH 12AY7 and the 6SN7 has been 'easily' able to achieve these two objectives in my platform (psu+passive filter, co-located psu/signal etc).

The 6922 is good on distortion but not so great for hum.  The AV7/6829 is generally good but needs careful selection operating parameters. AU7 and AT7 are pretty ordinary. I don't 'do' AX7.  8). 6SL7 is pretty good too.

I have yet to do detailed measures on small signal pentodes - that's upcoming ...

The 'ease' by which one can achieve these desirable goals of >80dB noise margin and sub 2%THD could be improved by having a test rig allowing variations of supply, plate load and bias parameters. One can then 'dial in' the required performance and 'find' any interesting 'nulls' for hum or THD.

----

In my bigSEamp platform, there is a couple of mixing stages, as well as a couple of gain stages in the preamp. So it's pretty challenging to get a decent signal -hum ratio in the preamp.

The output stage is pretty good, with it's big step-down transformer, the 6SN7 and 14dB of nfb around. It's capable of >90dB with a clean input.

The AV7/6829 virtual-earth-mixer plate follower + grounded cathode gain stage combo is pretty good now too. Capable of around 84dB of signal-hum margin with +30dB gain.

The 6922 is my current problem - it is very coherently amplifying the hum ie. +30dB gain giving similar amount of hum gain. I currently get something around +76dB signal-hum margin at best. It's not the PSRR - thats pretty good. It just likes to coherently amplify hum!!

That means, for example, that my 'starting hum floor' of -105 dBu is degraded to -75dBu by the  +30dB 6922 gain stage. Already behind the eight ball  - I am striving for minimum of 80dB, preferably more 

------

As always, the quality of wiring and layout and the proximity of PSU components to signal components,  is the main determinant of hum.

My 'schtik' is not quality craftsmanship, so I am always let down by my 'experimenter's expedience'.

The question for me is how to get quality performance given my generally poor abilities in constuction!

The answer is to carefully check each stage for hum and THD, 'as-you-go',  working backwards from output to input.

Hence REW and it's wonderful capabilities are essential.

And I haven't but scratched the surface of it's capabilities, using so far only the sinewave spectra and THD measures. There's a heap more REW out there I haven't used.

Doing quanititative THD 'offline' or 'manually' is a pain, as are  'visual estimation of waveform distortion' or 'lissajous figures' etc. Good for the early 20th century though!

----

Couldn't live without REW  now - and my previous RTA, once apple of my eye,  is now consigned to the dustbin of history 

 

[alexc]

Oh yes - no joy with any 'aikido' stuff' - admittedly hastily tacked on and all.

The hum is just not coherent enough to allow for any real 'inverse cancellation' within a stage - SO FAR.

I am going to do more in this area in future, but so far .. zip.  Any visions of magically eradicating hum is, unfortunately for me, just a fantasy. 

What's made a positive difference so far is increasing PSRR by increasing plate loads and decreasing currents. I can catch a little goodness from the increased gain and decreased psu hum injected - those together add up to a few dBs improvement.

However, too much drives down current too far, reducing stage mu, increasing rp and eventually increasing THD.

 

[trobbins]

Hum contributors are certainly an interesting topic - I enjoyed putting some dialogue down in the link below. 

http://dalmura.com.au/projects/Hum%20article.pdf

Scott Reynolds (TavishDad on diyaudio) has been doing some surveys on tube noise.  Hum is eliminated from his test environment, but he is contemplating including AC powered heater benchmarking at some stage.  I guess like microphonics, distortion, and hum, low noise level is well worth tube-rolling for if you are trying to minimise certain artefacts.

http://www.diyaudio.com/forums/tubes-valves/274670-list-tube-noise-measurements-please-nominate-lowest-noise-tubes.html

 

[alexc]

Thanks for that!  Very interesting ..  just going thru it now ...

I'm still stuck on one tube position ...  just can't seem to get beyond 76dB signal-to-hum margin for 3% THD or less.
It doesn't seem to be the HV rail - I've jiggered and poked and lifted and seperated ad nauseum :-\

Not the signal ground either. Rerouted with no luck.

And it's not the tube  - I rewired the position from EH6922 to EH 12AY7 both behave the same. I'm thinking I have a 'noisy' socket    If so, it would be my first!

Tomorrow I'll do up one of the remaining positions and go again 

 

[ruffrecords]

REW is a very interesting program, not least because it runs under Linux. I have been trying it out with my Focusrite Scarlet 2i2 and my DIY test interface box. For me, the most frustrating part is calibrating REW to the 2i2 or rather understanding how the levels in the 2i2 get set because REW is pretty consistent  in dBFS terms. What would be really nice would a be a no nonsense USB audio interface with fixed and known input and output gains so we could make a once and for all calibration with REW. Does such a thing exist?

Cheers

Ian

 

[alexc]

The Motu line inputs and outputs have a fixed gain and no adjustment - mine are Firewire but I think the usb models would be similar.

I get very consistent factor of 21dB to convert from dBFS to dBu - that's with basic measurement gear : CRO on 2mV/div and a True RMS meter.

I wish that arbitrary 'calibration factors' could be set for REW's RTA display, similar to TrueRTA, allowing, for example, dBu usage rather than dBFS.

I'll be doing some more testing in future of  REW's THD measures with some better equipment : a measurement  amplifier and an 'active HV signal probe' - I'm pretty sure my crude HV signal probe, used to measure at the tube plate, pre-coupling cap, is affecting the results.

My results are all very much higher than others' measurements  ???    VTV results routinely show numbers like 0.5% THD at 1KHz, something I can only dream of. My best, so far is something like 1.7% for an EHAY7 and EH 6922.

From the work of 'TavishDad', looks like I should get me some JJ 5751 tubes and maybe some of those russian 6J52P.

Anyhow - back to the bigSEamp and that troublesome gain stage!

 

[alexc]

Found my prob - chassis 'ground' for the un-isolated 1/4" guitar input jacks was a bit high resistance compared to the 'single chassis ground' lug near the iec power connector - a length of maybe 30cm of alu 2.2mm thick.

I connected the input jack ground (which is connected to chassis) to the signal ground and that sorts it out  - I have some isolating jacks coming - perhaps it will make a difference in this app.

For now .... good enough. My preamp tubes are far happier now - getting around 75dB signal-hum margin for the first preamp tube position, for a nominal 1Vpp input with the AY7 and 0.86% THD+N.

Pretty typical for a low-ish level input - raising the input level improves signal-hum margin significantly, peaking out  at 88dB for a hot input signal of 3.25Vpp at 2.75% THD with the AY7.

It remains top of the heap, along with the equally common EH6922, under similar conditions.

Moving to an AX7 ..  peaks out at 75dB with same THD ... obviously both the max output and hum components are quite a bit higher than the AY7, but the margin is not better.

And then back to AY7 with no changes -> 77.5dB signal-hum margin  ... with 1.31Vpp input and 36Vpp output for 1.1%THD.

SO - I'll keep the AY7 there - with 100K on the plate and biased for +1.65V on 1K1 cathode resistor, bypassed. About right for a typical-to-slightly hot guitar input.

The AY7 has the edge over the AX7 so far.

--

For the aux inputs mixing spot, position 2, I can setup for a 6922 and a much hotter input, like a fuzz pedal or compressor pedal  source.  OR a pedal steel   

There, where the input can raise up, even to 3Vpp, the AY7 and 6922 start approaching a really good signal-hum margin ... 88dB or so, as I described above.

..

SO - the reason I have been hitting a 'wall' of signal-hum margin in the preamp, despite choice of tube,  has more to do with the required 'sensitivity' than psu or tube noise.

Where I must cater for a lowish 1Vpp nominal signal, I hit the '1Vpp wall' of around 76dB.  Where I can specify a hotter signal, I approach my 'max possible margin' wall of 88dB.

The first position is obviously pretty important, so I have to 'grind away' and beat  my '1Vpp input signal - hum margin'  of 76dB  .....by another 6dB or so  ...    and get to 82dB  ...  which is just  barely acceptable  ;D

If I can snake a really good 5751 in there, I think I can achieve it ..  then pad down the output a little to 'clear the bar' of low hum with really low THD comfortably.

---------

SO - now to fool around a bit more ...  lock all that in ....  and  wait for a couple of JJ 5751 low noise selections to materialise here, then resume tools. Maybe a couple more selected AY7s for good measure.

I'll save really pushing the grind  until my next boxes get going - with modular psu : regulated HV, dc heaters and the like 

 

[alexc]

A few other things I noticed in traversing my preamp tube position nbr 1

1.

The sovtek ax7 wc variant  have really different 1Vpp input spectra at the plate.  H3+ signal harmonics are a  lot higher than H2. Right off the bat.  Quite consistently across 2 tubes (4 triodes). Really H3-ish   

Something to note. ANd AVOID unless for an effect.


The AY7 signal spectra I posted above, are just dandy - most all H2, very low H3 and H3+ almost nonexistent - a 'fairly-damn-low' THD

(even with my measurement rig, which is likely to have THD measurement error of maybe +/- 0.25% in the 'sub 3%' space )

Thats what I likes    Almost totally simple harmonic structure. 

Just another 6dB quieter and I'll be in clover  8)

2.  Most of the other AX7 types have a spectra with a lot denser H3+ components than the AY7, which is kind of a 'bare cupboard'.  So where some H3+ is really desired, the AX7 is probably a better bet.

3. nos rca 7025 (seventies) is the cleanest spectra of my ax7 bunch - it is a little less strong than new, and it measures out almost identical to the eh ay7. spectra, gain, noise margin - all that was almost the same. It's a great tube. Sits up there alongside the best I have.

----

Listening tests bear this out too  - you can really hear the harmonic spectra 'transition' points really well on a SE output stage.

Twerking gain stages, changing from 'predominantly H2' types to 'some H3+' thru to 'significant H3+' signals, is clear as bell thru this bigSEamp and 15" speaker    ... literally, it's like tuning a big bell with a 25" led display ... and a 20W output 

The predominantly H2 scenarios are pure like a bell, as H3 sets in, a more complex tone starts to cut thru, brightening the sound then 'densifying' it with ever higher edgy overtones - thats the majority H3+ scenarios.

ps - I have 32dB Class 5 noise reduction ear protection  for the speaker-connected,  tone testing  !

 

[ruffrecords]

The Motu line inputs and outputs have a fixed gain and no adjustment - mine are Firewire but I think the usb models would be similar.

I get very consistent factor of 21dB to convert from dBFS to dBu - that's with basic measurement gear : CRO on 2mV/div and a True RMS meter.

The 2i2 is much the same. With the input gain turned down the calibration factor is about 21dB. The trouble with that is that it is 20dB of resolution thrown away.  The outputs that are also because they are 'monitor' output s so they go via a big volume control knob on the front panel.

I wish that arbitrary 'calibration factors' could be set for REW's RTA display, similar to TrueRTA, allowing, for example, dBu usage rather than dBFS.

Yes, that would make it just about perfect.

Cheers

Ian

 

[Andy Peters]

REW is a very interesting program, not least because it runs under Linux. I have been trying it out with my Focusrite Scarlet 2i2 and my DIY test interface box. For me, the most frustrating part is calibrating REW to the 2i2 or rather understanding how the levels in the 2i2 get set because REW is pretty consistent  in dBFS terms. What would be really nice would a be a no nonsense USB audio interface with fixed and known input and output gains so we could make a once and for all calibration with REW. Does such a thing exist?

(Shill) ]Smaart I-O.

Clarification: input gain is known and repeatable. Output level is fixed.

 

[MicDaddy]

I'm using a TEF25 and GoldLine's SoundLab, but thinking of offloading it for something like this.

FWIW there's a guy on gearslutz that was getting very good results with either REW or FuzzMeasure using time delay spectrometry or similar and getting higher resolution results than I can with the TEF.  At least it would take very long sweeps (linear) for me to get the same data points with the TEF.  I'll dig in my emails and try to recall what he was using and any details.

 

[alexc]

As is my wont, I am waiting on parts to go the next level on my bigSEamp. 

The challenge - improve the preamp signal path signal-hum margin 

- with it's gain stages, eq, mixers etc. it stubbornly sits at a best of around 73dB margin for 3% thd.

The output SE amp is setup for 90dB signal-hum-margin with the use of an Edcor as an interstage traffo - my best yet margin in this big ass amp platform.

So - wanting 7+ dB improvement from the preamp, any which way I can   8)   Time will tell where the noise floor will end up.

My changes, which with luck will help are to be :

1.  'selected', 'low noise' JJ 5751 in pos 1 and 2  (pos 3 remains EH AY7)

2. refinement of bias settings and plate supply voltages

3.  isolated input/output jacks with no unwanted chassis-to-audio_ground connects

4. refinement of psu-decoupling


And in the 'heroics' category :

1.  a couple of  better quality potentiometers  - see if that helps with the considerable THD my 'Volume' controls seems to be introducing

2. addition of Schaffner IEC inlet filter

3. improvements to my passive  'hv signal probe'  - using a cro probe, a rotary 'attenuation' setup and a  nice jiffy box!

---

The good news is that the amp is really starting to recombobulate *just right*  after all these  REW    inspired  measure-teardown - rebuild    cycles   

It is sounding so very touch sensitive, with nicely dialled preamp thd  or  poweramp thd. 

Really easy on the ears at around 10Wrms - no real treble harshness issues at all with a really superb mid-range detail.

Just rich and detailed, all at an easy, loafing along power levels. 

It's definately shaping up as my best amp, with even  more refinement than my other builds.

Just needs the final stretch now to make it really quiet.

Also I need to do up a speaker-cab closed rear-baffle, for the times I want even more 'chest' impact. 

The 'open back' is good for guitar, not too scary -  the 'closed back' for more bass impact. That has the power to move you 

 

[alexc]

SOOOoooo, back to the L'Amps - the el34 monoblocks, whose optimisation with REW I have also been banging on about in this lengthy thread ..

Speaking of which, any serious measurement work requires good 'snapshot annotation' - I use 'SnagIt' which is quick, easy to annotate and has whole lot of great features.

Cataloging spectra, being methodical with descriptions and filenames is a must - for the person with several amps on the go, it can all get hard to follow after a while.

With good filenames, meta descriptions and methodical snapshots of baselines, it is all easy to recall and continue after time and a few more brain cells have elapsed ..

----

Where I left off, I had only the first tube position to complete  - the voltage amp and nfb network.

Back in there today, and what a pleasure of a starting point 

At 'first light', with the EH 6922 wired in, I get a 'hum floor' of -103.9dBu, signal-hum margin of 113.7dB, nearly all H2.

Here it is - EH6922 plate spectra 

 

[alexc]

Just twerking the 6922 for a little more gain with 20K on the plate, then bias 'hotter' to around +2.5V and get the plate supply to +250V or so.

Plenty of input signal room to move, with low THD, some 4Vpp , once the 10dB nfb is re-connected.

That should be about right.  THD will be very low, < 1.0%  when driving the rest of the amp to full power.

Right then - that's tomorrow.

 

[ruffrecords]

I love the way you annotate the screen captures from REW. What software do you use for that?

Cheers

Ian