Measurement Software - 'Room EQ Wizard'

[alexc]

Hi All

This is worth checking out :

- Room EQ Wizard

http://www.roomeqwizard.com

It appears to work well and have some super features.

I just bought a Dayton Audio measurement mics for doing some speaker work.

http://www.daytonaudio.com/index.php/omnimic-v2-precision-measurement-system.html

-----

Room EQ Wizard  has the first RTA with THD+N measurements from an asio driver on Windows that I've come across in the free world.  I can finally do some proper distortion tuning, easily and in real time. 

Rightmark does a 'static' display of THD I know, but it is unbearably clunky imho.

 

[alexc]

The sig generator in Room EQ Wizard is of lower distortion than my test bench staple, TrueRTA.

1KHz sine wave at -10dBu output from Motu 828MkII bal analog into bal analog
(loopback test)

    THD      0.0004% Room EQ W vs  0.0104%    for TrueRTA
  THD+N  0.0165%                            vs  0.0391% 

I'll check my other interfaces  ... some  Motu 2408MKIII

That's a bonus 

-----

Now to look at calibrations with Room EQ Wizard ....

With True RTA is's pretty straight forward.

You can offset the output amplitude of the sig gen output from True RTA such that is lines up with an external (trueRMS) voltage level when driving a known 600ohm load.

Then, hookup for loopback test and offset the input gain at TrueRTA such that the measured input amplitude is displayed as that same dBu for the frequency of the sig gen.

I wind up with an error according to my external measurement accuracy - in my case I have the measurement cursors at the CRO and a trueRMS digital volt meter.

I think I can get to around +/- 2mV on the CRO and a  it less  on my DVM - thats with no additional test equipment at around the 0dBu level.

So, down at the noisefloor of around -76dBu (for the noisier Motu 828MKII) and -84dBu (for the quieter Moto 2408MkIII), which is uV of AC measurement, only the DAW RTA is able to measure.

I think I have an error of around 0.5dB or so, which is the flickering of the RTA least significant digit.

------

So - next to look at (basic) calibration of the  Room EQ Wizard RTA ...

 

[Andy Peters]

Hi All

This is worth checking out :

- Room EQ Wizard

http://www.roomeqwizard.com

It appears to work well and have some super features.

It doesn't do transfer function measurements in a reasonable way, by which I mean using a reference input in addition to the measurement input.

 

[alexc]

RoomEQWiz does do phase plots (and impedance plots) and I think it does use one input channel as 'measurement input' and another as 'reference input'. I'm early with this, so not certain.


It's good to have a second detailed real time (!) RTA that can run concurrently with my regular  - provides a viable reality-check and cross-calibration capability.

There are some other intriguing features I have yet to check out - IR modelling, Thiele and Small design, room and eq modelling.

Fairly detailed stuff.

 

[alexc]

Some more testing with RoomEQ Wizard app ..

using the real time analyzer module, the THD% measurement capability is just fantastic.

The task - check the distortion of my EL34 builds into 8ohm dummy load at each of :

- output of the input debalancing transformer (Jensen 10K:10K) and attenuator (10K pot across secondary termination)
- output of the stage 1 voltage amplifier  (6N6P one triode)
- output of the phase inverter (6SN7 long tail pair)
- output terminals  (secondary of Edcor 6K:8ohms)

THEN redo the finals' bias, re-balance the phase inverter and check for improvement with the THD meter.

I rigged up a THD probe with a 450V film cap coupled to a 10K:10K transformer (cheaper model! but 20-20KHz up to 15dBu or so with 0.1% distortion at 1KHz)

AND - awesome! I thought I had done a good job of balancing a couple of years back BUT I dropped the power amp output distortion from 3% to 0.6% with the aid of  real time THD+N display.

That sets my new benchmark and I can now go on to do further noise and distortion analysis.

This will be so much better than manual estimation and/or qualitative comparisons.

Fantastic result for my amps - a routine 3yr service has really hit the spot  with this new tool 

 

[alexc]

Next tasks are to :

- revisit each of the amp stages to characterise the THD+Noise
- revise the NFB factor; I currently have 10dB only
- revisit the square wave performance


The aim here is to get to 1% to 2% THD at 30W  (full rated power) or thereabouts with the EL34 finals.

I will then do some tube rolling (with re-adjustment of bias etc as necessary) followed by some mulling over thereof 

 

[alexc]

Good start

I set up this test so that 15.5Vrms ( = 43.9Vpp) appears on the output of the amp which is driving a 8ohm dummy load.
I used a +4dBu 1KHz from the REW app 

(this amp has a low sensitivity and is designed to be driven balanced with nominal +4 signal)

( 2x16ohm 25W alu types parall'd, it's reasonable close to 8ohms at operating temperatures )

Measured the THD+N at stage 1 voltage amp (single triode grounded-cathode gain stage)
0.48% - input attenuator wide open  and  0.27% with the input atten at typical midway on the dial

No problem there!  The low mu 6N6P  tube in that stage is biased pretty cold at around +6.5V and it's happy with levels up to 12Vpp before things begin to get too squashed!

Next, the PI. I balanced it for min THD at the speaker terminals then looked at the THD at the plates of the 6SN7.
Here I had to back off the level so as not to OD my analyser.

THD+N at each plate was around 0.47% for one side and 0.2% on the other - a little bit more optimisation maybe here, but that is not too bad for now 

Next, looking at the THD+N at the speaker outputs. 

Also  freq response of the EL34s with the Edcors (60W 6K/8ohms) is very, very flat from 20-20KHz. Square wave shows some ringing but not bad at all.

So, I did the procedure again - rebalanced the 6SN7 (has one bal pot) by voltmeter, and then rebiased each EL34 (has one pot per final) by voltmeter.

Noted down the values, then by visual on the THD+N meter adjusted first one bias pot, then the other for the minimum values.
Then did the same for the same for the 6SN7.

8) 8)  The reward was quite cool - I honed in a long way from my previous method of measurement. Nothing like a cool new meter that improves your lot in life!

Now this all with the nfb removed.

The 6N6P  is a bit little high THD in this low-gain role, but the 6SN7 is quite good in it's role.

Thats about 4x less distortion than when I started out today!

NOW for the EL34s 

I have them in ultralinear 43%, 462V plate and bout the same at the screens. As one would expect.

It's a high-ish voltage way to run them (without going crazy! that is) and not all the 25W and beyond power tubes can deal with it. I have Sventlana Winged C in them - they are supposedly rated for 800V plate max and 500V screen. So they are a fairly rugged tube and I haven't had any issues after 3years as my lounge room hifi  8)

So biased for 20W plate - that's 43.3mA each side quiescent @ 462V.  And that's what I have - fixed bias with individual pot adjust and a 10R in the cathode circuit to ground, over which I can measure voltage on the (0-100mA) meter.

------    tbc

 

[alexc]

Just rejigged my 'THD probe' with an attenuation network that maps around 254Vpp to the analyzer range of around 10Vpp before distortion sets in. Its a divide by 25 thing.

Pretty poor, and I have pencil'd in a project for a decent active probe circuit -  HV input, High Z with variable attenuation and low Z output 

But for now, she's good.

Again setting up for max speaker terminals output at 1KHz -> 15.5Vrms into 8ohm dummy load

Voltage amp  1.03%    at 12.2Vpp output  (gain x14.2)
Voltage amp + phase inverter  1.35% at 77.3Vpp output (gain  x14.2 + x6.4)


That's about what I expected - pretty reasonable really.  There's some THD attributable to the Jensen (quite low
as well as the 'measurement iso transformer' (higher than the Jensen but surprisingly OK).

Then theres some THD attributable to the crap probulator  ...

So all in all, I think the THD attributable to the gain stages would be in reality lower - that's the challenge, to set up for better measurement. But this real time THD meter is a huge step forward!

(For some good comparisons, the 'Angelfire' tube designer web site is a good resource. )
(He is very detailed and useful with THD figures of various common stages)

----

Now. the main output is something else altogether 

I have the cro measuring, along with the digital voltmeter and both REW analyser and TrueRTA analayser all going together.

So I can reality check and all that.

With the 15.5Vrms into 8ohms, I get a crazy THD measurement - too sad to mention here!

The cro shows a pretty good sine wave - I think the threshold for visual signs of distortion on a sine wave is around 10% THD.
If so, then I believe my THD is < 10% but the analyser shows MUCH higher.

TBC - I have to do some reality checks with my solid state amp and perhaps rejig my setup if required.

Good fun but some trepidation      about what the THD figure will turn out to be! (actually)

It's not impossible that I have a gross issue in the setup of the finals and nfb  that I have not up till now detected.

I have only done rough calcs of THD noting that, of course, THD rises with level!

But we will see, because now I'm checking more closely. 

----

The data sheets has some good info for an el34 pair, in p-p with 43% UL taps for screen with 1K dropper.

Around the 450V mark for plates and screens, at full rated 30Wrms into 6K p-p traffo load, Dtot(%) should be around 0.8% when driven with 43Vrms.

I'm in that ballpark, but with higher plate/scr voltages - as well as less drive voltage at 27.5Vrms.

Perhaps my operating point is not what it could be - B+could be too high for the 6K plate-plate load leading to some unsafe excursions into >20W or even more! dissipation on the plate.

Or it could be the screen is dissipating a bit too much - perhaps raising the 1K dropper could help.

Or or, the UL connection isn't set up right  (? it's pretty simple right

And my nfb could be a bit crap  too. A more careful network may be required.

My 8ohm dummy load could do with some more work - it's more like 8.3 ohms!

THEN maybe I'm right and the measurements are wrong (2 in 10 chance  maybe?)

So!

time to get measuring some more ....  I hope it's the measurement technique, not my amps!

 

[alexc]

Couldn't stand the suspense !

I hooked up my pioneer av amp to the loadbox and set it for 30Wrms into 8ohm dummy load.

Phew! It shows similar figures to the tube amps. ie > '10%' THD+N. Very, very close, actually.
Except it can drive 40Vrms into 8ohms no trouble, which is 200Wrms  The figures the brochure quotes are definately super low. Like < 1% at highish output levels.

I back it off to find the 3% THD point, then the 1% and it's very similar to the tube amp.

So, I conclude, my original 3yr old tube amp measurements and listening sanity are not 'crazy'. 

Stage 1 complete 

I need to do more checking of my measurement tools.

My tube amps are still good to go!

 

[alexc]

It appears that what I am seeing is simply the Motu input overloading poorly.

The THD seems to have 3 or so 'bands', with very distinct onset rms voltage inputs.

'Quantised' if you will

Doesn't seem to depend on the amp I connect.

If I remove the 'load box' with it's attenuator, it merely shifts the threshhold  a little.
But even without it in circuit, same behaviour  ..  ending with a crazy (>50%) thd figure and tears all round!

So, the useful, low distortion input range of the Motu is around 0 ..3.2Vrms or 9.0Vpp.

Which is pretty reasonable

-----

Next up :

- scale the load box such that 30Wrms into 8ohm dummy load is mapped to 0 .. 3.2Vrms at Motu and THD meter 
- verify that the solid state amp at 30Wrms has a 'reasonable' measured THD figure  ie. < 1%

That should demonstrate that within the available Motu input voltage range, THD is correctly measured and displayed

- return to tube amps and check again!
- verify toob amps = gud        ( ie. around 1% to 2% worst case)

Awesome!

 

[alexc]

I rejigged my load box such that it  can 'map' an ac signal of around 50Vrms (142Vpp) onto the Motu with a low distortion max ip of  3.5Vrms (10Vpp).

Retesting my Pioneer AV amp (my 'gold standard') with output of  1KHz, 15.5Vrms (43.9Vpp) into the 8ohm load box gives a THD of 0.031% and THD+N 0.045% respectively.

That's pretty good  

 

[alexc]

The ss amp with  35.86Vrms (101.5Vpp) into 8ohms for 160Wrms

THD+N is 3%

This also the THD+N at which it starts being visually observable on the CRO.  3% - the magic number!

 

[alexc]

Now for the 'money shot'

EL34 pp tube amp with 10dB nfb at  15.5Vrms -> 30Wrms into 8ohm load b ox

THD        8.80%
THD+N  8.84%

Not too great! But I think it's realistic. 

So far, I have confidence in the THD measurement capability in REW.

 

[alexc]

and el34pp 10dB nfb 3% THD + N is at 13.5Vrms -> 22.7Wrms into 8ohms load box

Again, I think this is a believable figure.

That's what I'd like to have as a minimum baseline of performance at 30Wrms  into 8ohms load box

finally,  el34pp 10dB nfb 1% THD + N is at 10.9Vrms -> 14.9Wrms  into 8ohms load box

So, more 'tuning' to do  ;D

 

[alexc]

Looking at the other el34pp amp (a pair of monoblocs) - this one I have not touched for years.

Right off the bat, I get a 1%THD+N at 14.8Wrms, 3%THD+N at 24.1Wrms and 30W RMS THD+N  at 8.35%
Which is pretty close to the other.

-----

Now to all-round rebias (again!) and rebalance (again!  ;D) with new measurements for voltage amp (VA) and phase inverter (PI) stages, as well as overall THD+N for the amp.

Once I have my baseline, I will 'do this and the other things'  ....

- optimise nfb network
- check THD for different load impedances
- check THD with speaker connected
- check THD at typical usage levels of around 10Wrms into 8ohm, 88db sensitivity 3 way vintage 12" types

then look at some more of REW functions, like phase measurement and impedance measurements.

Once the amps are 'tuned' and characterised, that will complete the '3year service' of the amps.

At that point, I am doing some crossover/speaker work with the measurement microphone and solid-state amp.

-----

This round of measurement and tuning all started to check my Sansui SP-2500 c1975 speakers.

They are perfect condition, little used and I've grown accustomed to them, over the years,  in my lounge room.

BUT they need a lot of EQing - I use a Finalizer2 for that purpose. My 'house curve' invariably winds up being big +6dB at 70Hz broad Q and a big -6dB cut on the high at around 3KHz with a mid-narrow Q. Plus some boost air at 10HKhz.

So I started to look and listen more closely to the speakers and amps on the bench.  I know the amps' response is ruler flat, so I wanted to look at the damping factor of the tube amps and the cross-overs in the speakers.

I'm just recapping the crossover to replace the ancient electros (bipolar rubycon). While I wait, I thought I would do the service on the amps. I didn't expect to find much really.  But here I am now finally doing the fine tuning on both!

So, a big win to be able to reliably and believably (not to mention easily have the real time THD measurement capablility in REW's RTA module. 

9.5/10 is my score    so far

The only thing that bugs me is I can't seem to turn off the 'persistent' averaging trace. It's very useful, but not ALL the time please! So far, I can't seem to find a setting for that.

Also, I've been doing my measurements with a 20-20KHz bandwidth, 24bit.

I need to check how it works with the wider 20-48KHz measurement bandwidth.

Apart from that, it has given me some new motivation to get detailed and more competent about the next level of audio testing. I

 

[alexc]

So I have 'discovered' I have a couple of probs giving rise to the 8.5% THD I'm measuring at 30Wrms into 8ohms.

- questionable design of VA  - this is doing the majority of the trouble at 7.45%
    -> need to redesign for lower THD

- poor headroom at the PI - at full power, the PI is approaching clip contributing around 1%
    -> need to optimise for more headroom

as well as the cheap iso traffo in my probulator is contributing quite a bit of hum and thd itself   

Now I'm more confident with THD probing at plate levels, imma dispense with it and rejig the probe to be a simple resistive atten after the chunky HV wima cap that couples the tip.

Off with me to sub the VA tube and rejig the PI 

-----

Did the rejig of my probulator to remove around 1% of reported THD at plate level voltages.
-> accurate THD monitoring of plate level voltages. Working well!

With my new test setup, I'm repeatably getting :

-  VA plate with 7.3% THD+N  at  an output of 22.9Vpp signal (1KHz , amp at 30Wrms into 8ohm load box)

Did subbing of same type tube 6N6P in VA
-> No luck - turns out I originally had my best one in there at the onset.

Thus I need to redesign the VA to achieve a target of around THD+N  0.65%.

Given I want to increase the nfb to 20dB, that may see me choose a medium u tube in there, something like an AV7 type,  instead of the 6N6P which is more of an 12bh7/ecc99 type. I like it there for it's low gain and low sensitivity - suitable for a hot signal with minimal attenuation required at the grid.

I do just happen to have some nice nos rca 12AV7s here, so I think that's probably the choice in a medium gain, medium sensitivity setup.

-----

BUT first, I need to figure out why the 6N6P seems unhappy right here, right now.

Could be the nfb network to it's cathode is poorly done, or it could be the plate resistor load is not quite right. Even the cathode bypass must be looked at. Perhaps also the input capacitance and of course, the way the grid is coupled to the 10K attenuator at the Jensen 10K/10K input traffo secondary. Could just be not a particularly linear triode

The waveform at 7.5% THD looks visually to be 'triangulated' and is composed almost entirely 3rd harmonic. 

 

[alexc]

Bingo!

The problem is the nfb network, as coupled to the cathode circuit of the 6N6P VA stage. Disengaging the nfb network drops the VA stage plate THD to normal levels.

VA plate THD+N  0.84% at 22.5Vpp signal      (1KHz, amp at 30Wrms into 8ohm load box)

That's what I expected to find more or less (but with the nfb connected!)

In fact, the 6N6P VA stage shows 1% THD+N at 77Vpp, which at x12.5 gain, is an input signal of 6.2Vpp. 
That's what I set it up as - with a very low Rplate to boot.

I can do more on the VA but for now, I am a VA happy man.

-------------

Looking at the amp now, with nfb removed, I get 13.1% THD+N

So, although the VA stage has 'cleaned up' when not bothered by the (probably very poor) nfb network, the overall amp has reverted to the stone age, THD+N wise - as expected but not at all welcome!

Onwards now, sans nfb, to the PI   

 

[alexc]

The PI needs to provide the finals with around 75Vpp max for full output. It is clipping badly at that level.
As it stands, the 6SN7 here has a  1% THD level of  a mere 36Vpp.

So this is a redesign of the PI jobby.  Tomorrow  <yawn>

 

[alexc]

Corrected my 6NS7 PI such that it has more headroom - it's still not the best possible, but is improved a lot.

I balanced it up nicely with CRO, voltmeter then looked for the thd minimum around the balance point.

at rated power 30Wrms into 8ohm load box and no nfb :

VA          thd+n  0.85% 
VA+PI  thd+n    2.85%  and h2  0.254%,  h3  2.88        h5+  0.155%
amp        thd+n    1.68%            h2 0.303%,    h3  1.38%,  h5+  0.714%

and full power hum and BW :

hum            4mVrms  (71.8dB below fundamental)
20Hz          -1.5 dB
20KHz      -3.0 dB

So now I have a good starting point for basic performance.

Further optimisations could be :

- check  effect on THD+N  of changes to psu filtering /decoupling  for  B1+(VA) and B2+ (PI)
- investigate the freq response of each section
- investigate square wave ringing of each section

and of course

-  tube selection

NOW, I can look properly at the nfb network and factor  (

 

[alexc]

Did some tube selection on the 6SN7 PI  (full amp power into 8ohms) :

- from a sample size of 6x current issue TungSol (russian), the spread of THD+N was 2.65% .. 3.30%
- from a sample size of 2x nos RCA (usa), the spread of THD+N was 2.75% .. 3.55%

These figures are for the VA+PI together - I haven't checked the PI in isolation - that will be a refinement in my next builds
Test-THD-as-you-go approach 

I kept the best 2 current issue Tung-Sols in these mono-blocs for a small improvement in THD+N.
Probably more due to burn -in than anything else. The selected ones have been used for the past 3years!

I think I can again refine the operating point of the PI to get a bit more - imma ordering some new resistors now.

Apart from coupling cap rolling (I have a few I can try here!), that's about it for the PI for now. 


With the tube selection, I get a full power THD+N% at the speaker terminals of 1.50% with no nfb.

Not a bad improvement from the nfb figure of 13.5% I began the optimisations with  :