Measurement Software - 'Room EQ Wizard'

[alexc]

One thing I am going to look at - is mod the phono stage to defeat whatever curve they put there - I can't even remember the designation riaa something or other ...

Clip that out - I'm not a vinyl fan - and I remember tape fondly, as a cheapo thing, but only ever really got up to 1/4" tape, hifi wise. Haven't personally heard the wonder of 1" master tape and higher ... Oh well.

Keeping the high sensitivity, I plan to use a passive sum box onto the first AX7. The less sensitive stuff starts on the second AX7.
The tone controls and what not, on a third. Thence to a further ax7 (I think) as a PI.

So - mod the first AX7 flat and simple for a small mix of 4 channels of outboard.

-----------

The other interesting thing, is I'll be making my own first el84 amp, kind of similar to the core of this -  and with old school Japan pp transformers ...  not sure the actual age but slate sixties, very early seventies I think. Nice construction and all, without being crazy expensive  (positively cheap, actually) - be interested in how that goes ...

I will use a seperate psu - the same one as for my 'Esquilax Masterbox' - in my other thread.  ;D

 

[alexc]

Here's the basic perf data for the Encel japan 6BQ5 pp amp :

1% THD at      5.53Vrms ->  3.8Wrms
2%  "              11.74Vrms -> 17.3Wrms
3%  "              12.13Vrms -> 18.4Wrms
4%  "              12.23Vrms -> 18.7Wrms

Max output at 4.5%THD is 12.25Vrms into 8ohms -> 18.7Wrms

Typical listening levels : 3.89Vrms into 8 ohms -> 1.9Wrms at at 1.1Wrms

----

4%THD (full power) freq response is -2.6dB at 20Hz, -2.1dB at 20KHz.    HF -3dB point = 36.4KHz

And the 4%THD  magnitude and phase responses  ...

 

[alexc]

And the 4%THD harmonic distortion spectra  ...

 

[alexc]

So this amp is not too bad at all, even in it's current, non re-capped state.

The 'tone shaping' controls are interesting :

- 'treble and bass' shelving controls (prolly an active bax)
    ->  +8dB to -15dB max (20KHz)  breaking 2KHz, and +14  to -15dB max (220Hz) breaking 500Hz

- 'scratch filter' - 6dB cut breaking around 1KHz
- 'rumble filter'  -6dB cut  breaking around 200Hz

- 'MFB bass boost' - gives from flat to +12dB boost centered on 35Hz 

----

I typically use it at around 2Wrms, for 1.2% THD and with a 6dB bass boost, treble flat.

So not too bad at all, really - this will be the baseline for my upcoming 7915a based p-p hifi amp with late 60s Pioneer 20Wrms trasnformers.

I also have the psu traffo, which is used as a  'full wave voltage doubler'  config. Also from same period. Just the same as this Encel japan amp - full wave voltage doubler to take 140Vac at supply to 380Vdc at the choke input.

For this build, I am maing speaker cabs - my first hifi effort. I have the original mint condition 12" alnico woofers that went with the transformers too  ;D ;D 8)

Super stiff paper and fibre cones, cast alloy frame, large vented voice coil ...  60Wrms high sensitivity.  What's not to like?

I plan to have a seperate psu chassis (all my new tube stuff going this way), with ability to power seperate preamp.

The preamp I am making to have active filters of the discrete opamp kind, so that I can do some bi-amping and mess with the  crossover points.

That should be interesting : my first attempt at active crossovers.

Awesome!  This will allow me to use my big EL34 pp in conjunction with the 7915 pp for a biamp system.

 

[alexc]

Back to the Encel  ...

At the first stage, there is an AX7, one half for each channel L, R setup with plate resistors of 235K.

So - high gain with 1.0V on the grid and 2K2 cathode resistors - originals are carbon comp.
I replaced with modern carbon - they had failed when I bought the amp.

And 0.05uF coupling caps. These look to be electro, but could be something else.

Input to the amp  is attenuated from 2Vpp at the signal-source-selector switch,  down to 229mV at the grids. I can't yet tell the nature of the atten - resistive pad or part of a feedback network?

THD at the AX7 plates is 1.67% and 2.07%THD - the difference is due to error in the grid attenuation network. Something to look at when I recap. Gain is x47 or 33.5dB

So - my first AX7 THD measures - not too bad. I'll sub a few others and check the diff.
I'll look at one plate only for simplicity

orig tube is mid 60s AX7 'matsushita' brand grey plates, large 45degree halo getter.
  -> these are 'Made in Japan on Mullard equipment' - not cheap either!

original matsushita                    AX7 -> 2.07%  THD    x47 gain

sub  1 - 1990s 'groove tubes'  "            1.74%              --  good
sub  2 - 2010 'mullard'                "              2.20%              --  very good                         
sub  3 - unknown                            "              3.01%              -- high   
sub  4 - 1990s mesa                      "            1.92%                -- average   
sub  5 - sino old                              "              1.83%                --  lower  with low output
sub  6 - sino modern                    "              2.20%              --  higher and low output
sub  7 - sovtek wc                          "            1.92%                -- average 
sub  8 - sovtek wc                          "              2.16%              -- higher
sub  9  - 1970 rca blackplate    "              1.75%              -- good
sub 10 - 2010 mullard                  "              1.97%              -- higher but with higher output
sub 11 - 60s Matsush*ta #2      "              2.06%              -- higher
sub 11 - 60s Matsush*ta #3      "              1.94%              -- higher

Not much in it for the low THD, so I'll keep the originals as they are. For now 

Goes to show that AX7s are pretty low THD : 2% for x47 gain is pretty good.

 

[alexc]

Spent a while digesting the results of the THD+N measurements on the bigSE Amp that I've been writing about here ...

I unwound the amp's preamp section completely to redo now that I have the config and gain structure a  bit clearer 

The output section is the same, I reworked the driver 6SN7 for higher gain and sorted out my nfb around it to be 6 ..14dB on a front panel pot.

At the same time, I settled the chassis, grounds and HV supply some - I ended up adding a second psu choke of 15H at 50mA for the feed to B1+, which is everything downstream of the finals.

Checking the results of the back end redo, I'm really pleased.

The PSRR of the 6SN7 is now much higher than before, the B ripple to it is a lot lower than before.

Here's the spectra of the 6SN7 output  (reissue TungSol) with the amp is it's 'triode' mode at 7.3Wrms.

 

[alexc]

The 6SN7 is doing 70Vpp  at THD 1.26%. Hum is way! down, 100Hz is  80dB below H1.  The thd is nearly all H2.

The finals need about 65Vpp to drive to full power at low-ish thd of 3% at the speaker with min nfb dialled.

I have around 7dB of headroom at the 6SN7 plate beyond this before the 6SN7 moves to 3% thd itself.
It is biased for a max low thd case input of 9Vpp at the grid and a gain of x15.9

The cathode follower stage adds more or less nothing in the way of THD and noise. 

The ripple at the grid of the finals, due to the 6SN7 driver is around 20mVpp, including a little 'motorboat' style 10Hz modulation. The larger plate resistor (36K) and  second choke made a big difference here to the noise. 

I will do the second choke  in all future SE builds (that don't use regulated HV, that is!)

The  'min nfb' case (6dB nfb)  ripple at the speaker terminals is around 6.7mVpp for full power into 8ohm load box, lowering to 4.26mVpp for the 'max nfb' case  (14dB nfb).

That's 100 to 110 dB below fundamental at 100Hz, from  the speaker terminals with 8ohm load box connected.

Which is pretty good  8)  ..  for now

 

[alexc]

So that's the final button up on the output section - quiet, gain and nfb both are smoothly dialling in, with stable operation.

Output noise is dominated by the B+ ripple at the SE transformer primary - not much scope left there for improvement. It turns out to be 186mVpp at the plates of the finals.

I'm trying to reduce the slight 10Hz rolling modulation on the B1+ ripple a bit. , but it's not much left as it is.

----

Moving downstream to the preamp sections .....

I'm giving a try, this time around, to some 'psu noise cancelling' techniques ...  as described in VTV constributor John Artwood's 'The Randall Amplifier' .

It's that trick of arranging an inverting gain stage to feed the output section driver such that the psu noise is cancelled out.

The supply voltage for the inverting stage is sourced from a 'raw-er' HV takeoff point, scaled with a resistive divider.
You try to null out some hum by varying the divider ratio on a rheostat.

I'm not optimistic of a useful result, but it's worth exploring. Particularly here,  since the front end for this bigSE Amp is a little elaborate - with a pair of mixing circuits, 3 inputs and an eq

It all adds up to noise if not carefully done (NOT admittedly, my strong suit!)

----

Anyway, I can feel this amp is on the last frontier - the concept is prewtty schwweet for a busy recording room.

Again, thanks to Room EQ Wizard for giving me the  tools, and hence the motivation, to run this amp all the way to it's potential.

 

[alexc]

Looking at the noise floor for the edcor xsm 10K:600 balanced output traffo mounted inside, underside of SE OP traffo ...

It is 'passively' connected for now : the hiZ primary winding is connected single ended to the speaker terminal hot.
I'm just wanting to see the basic noise floor of the thing in this config.

----

It has a pretty good noise floor - something like 2.8mVpp across the secondary, with the winding grounded thru the probe.

The audio ground has around 0.5mVpp of hash in it, so it's more or less pretty good. Hum is essentially the speaker hum (input of 12mVpp)  divided by 4 due to the stepdown.

It can stay    Which is good cause I don't have any other free space!

-----

The top deck switch will be an 'source' selection, between the 'speaker output' and the 'preamp output' options.

Last time, I used a CF as the traffo balanced output driver. I found I wanted some more gain and a 'Output Level' pot, independant of the main amp volume.

SO - this time round, I'm dong a mu-follower, using the 6U8A pentode - med triode tube. I'll add a 'top deck' control for the level. This will be my first mufol with the pentode-on-top thing


---

So, a great platform to do the final wireup of the preamp, whilst check-as-I-go  ..  to keep the great noise floor and low disortion to completion.

 

[alexc]

This AV7 stage is proving a little challenging - spec is simple enough : 70Vpp output at < 3.0% THD+N

I'm sure I've had it right in the past, but I'm having difficulty getting right where I want, in the bigSE Amp preamp re-work.

Firstly, the AV7 is reasonably high THD to begin with, secondly there is a good spread of performance across tube samples.

I've checked around 12 samples and settled on the nos 'Philco' av7, the best THD with output in the upper range of samples tested. The 4x Philco's were remarkably similar to each other - more so than nos ge, rca.

The other thing is I have some 6829 to sub and compare - these should be good :  5star ge (double mica).

---

Running lower plate loads gives a poor PSRR and distortion remains stubbornly high.

It's hard to get AV7 decent gain (x25) with plate resistors in the range of 24K to 48K - the mu drops quite a bit at lower currents as well as the rp increases quite a bit to from 8K to 12K.

The gain does increase going from 24K to 48K plate,  but not enough to outstrip the rising THD and rp.

So, I need to rebias for a 'hotter'  cathode of +2.75V, and move to 48K plate. I already have the cathode bypassed.

The noise performance at the av7 plate is pretty good though; 80dB below fundamental and trending better as I hone in on my preferred operating conditions.

-----

What I really need is a decent 'setup' jig : it's on my list of 'to-do' : tube test fixture

- stepped, regulated +HV
- rheostat for varying plate load
- variable regulated -LV for bias

as well as

- HV capable signal probe interface with attenuator and feed to balanced XLR of soundcard
- patchable dummy load

-----

Here is the spectra of the Philco AV7, where I'm at so far with it.

 

[alexc]

The 6829 measures up a bit better than the Philco AV7 but not by too much.

I've worked the bias to +2.5V with a 475ohms for 5.3mA at 160V plate for a gain of x32.
I get 3% THD for 38Vpp output for the 6829.

Now measuring at the speaker terms, I get my 1KHz  3% THD  at 10.6Wrms into 8ohms with max NFB (14dB).

The output amp THD is effectively cancelled out by the NFB applied, and the resultant THD is thus dominated by the AV7/6829.

So - in the right direction.  I don't think improvements in this position will yield too much more good.
I will await a time when I have my test jig ready for my next revision of this AV7 stage.

I'm just doing a few quick checks for hum sensitivity before moving on.

 

[alexc]

I also noted that I was getting about 0.75% error in the THD at the speaker terminals due to my 'hv probe' being connected and floating    Mea culpa.

So that translated into some extra  bit of output power, whilst still being <3.0% THD.

I really do need to characterise my 'hv probe' a bit better, with respect to THD measures.

ie. do some comparisons of THD losses at different, safe, signal voltages with/without the 'hv probe'.

---

Anyway - I'm disappointed somewhat by the AV7/6829  'on paper' so far, and with my current 'level of uncertainty' on my 'hv probe' AND with my still-crude operating point choice.

I think it's a tube that really does need some nfb to reduce the THD or replacement with a lower THD med-mu triode.

I'm already using a 6DJ8 type in this amp - don't really want to have 2 of the same tube in series, that ol' 'too much of a good thing'  ...  thing.  8)

 

[alexc]

FINALLY, I have my baseline config for the bigSE Amp with the AV7 stage all buttoned up

I have refined my measurement practices to trace the evolution of  :

    ground noise -> input-shorted 'noise floor'  -> input-active 'silence floor' -> and signal 3%THD spectra

for
  speaker terminals and AV7/6829 plate signals 

with
    8ohm load box and 4ohm rated speakers

8) 8) 8)

One needs to be rigorous in measurement, as far as practice, skill and tools allow  ;D

-----

My baseline config with 'NFB' control set at max (14dB NFB) gives me :

- speaker terminal THD of 3% dominated by the power amp SE stage THEN the preceeding AV7/6829 stage at 2.37% - which is as it should be.

- For 3% THD into rated 4ohm speaker, I get 20.7Wrms with hum and noise 76.8dB below H1 (fundamental).

Hum components are -102dBFS and lower, avg silence floor' is -96dBFs, which is just audible 'ear to speaker box' hum test.

Threshhold of  'ear to speaker box' hum audibility test  is -96dBFS avg, so there is still 6dB room to improve on the hum front to be absolutely pristine quiet.

Not sure I will get any better than at present but will do my best not to degrade it further 

---

That translates to VERY LOUD! and VERY QUIET with the >90dB sensitivity of the 15" 50W Jensen C15 4ohm open back speaker box.

Sound is still quite clean at high SPL (speaker testing to come later!), as would expect with 3%THD.

With the 'Level' control backed off to a reasonable listening level, hum is completely inaudible, speaker THD drops off < 2% and the sound is 'clean with a good balance of thickness and chime'.

Quite different to a push-pull amp - much smoother crunch with a lot of touch sensitivity. I can surely hear and feel the difference over my standard amp, a Mercury Magnetics/Jensen C10 equipped Super Reverb clone.

Here is the final spectra for the speaker box connected

 

[alexc]

Oh yes, the nos GE 5star 6829 is a keeper, with lower THD and higher output than all my other AV7s (ge, rca, philco).
Sample of 4 shows very close gain and THD matches.  (I bought these as a tested close matched set)

I have found a decent op-point (though not necessarily the best possible) for the 6829 :

->  +2.5V bias with 475ohm bypassed cathode for 5.3mA, 30K plate load, 160V plate,
        1.2Vpp input, x32.5 gain, 39Vpp output for THD+N 2.35% almost all H2


Here is the final AV7/6829 plate spectra corresponding to the above attached speaker terminal spectra.

 

[alexc]

Now I can move on to the stage feeding the AV7/6829 grounded cathode amp described above.

-> 2 input, virtual-earth plate-follower mixing stage      using the other half of the 6829.

Essentially same setup as before but with a lot of local feedback.

 

[alexc]

I have my basic 'virtual earth mixer' plate follower stage going ... just tweaking a bit more for lowest hum.

I have 2x inputs each with 100K Rin resistors for 50K parallel'd with Rfeedback of 68K. That ratio sets the negative feedback factor.

For the 6829, I get x1.13 gain for the plate follower, knocked down from x28 or so open loop. THD is something like 0.2%.

With the second 6829 stage gain of 33, that gives an overall gain of x34 overall. That provides 50Vpp output at 3%THD, which is virtually all from itself.

SO - at present I have the gain numbers more or less correct for 3%THD.

Problem is the hum noise margin is around 65dB down from fundamental - it's still a bit low. I'd like to improve that by about 6dB.

The limiting factor is the Rfeedback - needs to be at least 4x the plate resistor, which is 15K at present - about the lowest I dare to use for the 6829/AV7.

Rfeedback is presently at 68K, which is x4.5 - any higher, and the noise of the plate follower starts to shoot up, quickly!

I'm now trying out lower values of Rin  : I'll drop the parallel resistance of the sources and check the effect.

Here's the spectra of the combined 6829 stages : vem plate follower + grounded cathode gain stage.

 

[alexc]

I'm not yet convinced that there is value in using the vem plate follower to mix 2 signals, over say, a common-plate mixer config or even a simple summing node into a single grounded cathode gain stage.

In terms of noise, it seems quite a bit worse.

Still - I'll soldier on with it before making the comparison.

----

Success! 

Increasing the series resistance into the virtual earth summing node is beneficial.

Increasing from 50K to 200K for the parallel resistors' combination increases the noise margin for the same THD  at the same output level.

I get >75 dB of 'noise margin' at the output of the v.e.m plate follower+ grounded cathode gain stage combo for 3%THD

---

Now, the freq response should be pretty good - one of the problems with the previous approach of a summing node feeding into the grid of a single triode gain stage, was the high freq roll off due to the series resistance working into the miller cap of the AV7 (which is reasonably low).

Testing the freq response of the v.e.m plate follower+grounded cathode gain stage shows -3dB 20Hz and -0.8dB 20KHz.

Which is great - no sign of hf droopiness as before, with the sum node into the miller cap. 

It is a keeper!

 

[alexc]

Looking at the speaker terms with 4ohms speaker connected, I can now achieve max power of 18.8Wrms with 2.36% THD (max NFB of 14dB).  The v.e.m+gain stage is at 19.2% THD. Freq response is -3dB 20Hz .. -0.8dB 20KHz

I could go further here, to extract more power at similar THD but I'm happy to stop now 

Here's the speaker output spectra ... of the keeper setup  ...

 

[alexc]

Righto! Now I'm excited   

First time I've been with a full power noise floor I'm happy with! Even with the NFB control dialled to min (6dB nfb), the 'ear-to-the-speaker' test is inaudible. That's quite a bit of power at the speaker with no noise.

Here's the spectra of the speaker terms, passively balanced to an XLR, using an edcor xsm10K:600 traffo. The amp is running with the speakers terms connected to 8ohm load at 9.7Wrms.

Pretty good really ; hum at 81dB below H1, 3% THD, which is close to the speaker term spectra, but with quite a bit more H3 and above. Sounds really good - pristine quiet and ready to work with the next steps.


-----

Moving on now to the next bit - the mu-stage for the edcor xsm output traffo. 

 

[alexc]

Also if interest is the pentode-ultralinear mode - output power jumps up quite a bit from around 9.5Wrms to 15.5Wrms into 8ohm load box. H3 and higher order harmonics increase a bit and the ratio slides  to    H2  2.6% and H3  1.17%.

Full power  noise margin is a tad worse at 79.6dB below H1, compared to triode mode  -80.7dB.

You can hear the change in timbre of the 1KHz tone 'singing' thru the iron - a more complex, brighter tone on the pentode mode than the deeper, more 'focussed' singing with the triode mode.

All in all, a useful change of mode