Measurement Software - 'Room EQ Wizard'

[alexc]

Looking at the 6N5P (au7 type, mu low-med at 27)  'common plate mix' setup ...

Sig gen remains at 1KHz, 2Vpp to grids

- input to one grid only :  the summed plates'  THD+N    8.19%  at 40Vpp  nearly all H2, virtually no hum
- input to other grid only:                  "                          THD+N    8.73%          "            "                    "

So, not much good from a point of view of low THD  

Very little hum, which is good considering there is 2 inputs and plate resistors of only 27K (low gain setup).

However, if H2 is one's thing, then these are good!  I 'm trying for lower THD for the moment, so goodby 6N5P  :'(

The russian types have a heater difference with the usual a_7 types : paralled heater pins 4,5 with 9 as shield, not center-tap.

I'll try some other russians first, and then other same pinout types I have, even though they are a bit different.

- the 6BZ7 (low-med mu), 6GU7 (low-med mu), 6922 (low-med mu) even though they are all a bit different

-----

First the 6BZ7 - always liked this tube and it kicked off my av7 'thing'  - it has a low-med  mu of 33 (vs 41 for av7)

tube nbr 1 :

THD 4.06%  THD+N 4.06%  at 38.3Vpp output, nearly all H2   

That's pretty reasonable, around 1/2 THD of the 6N5P at a a little less output voltage, despite it's higher mu.

6BZ7 tube nbr 2 :

THD 3.63% at 31.3Vpp output, nearly all H2

So 20% reduction of output level and a little less THD.

Overall this one is a big improvement over 6N5P

-----

- 6N1P (low-med mu of 33)

tube nbr 1 :

THD 2.77% for 37.4Vpp nearly all H2   

tube nbr 2:

THD 2.41% for 37.9Vpp nearly all H2

So a halving again of THD over previous best 6BZ7 and quite consistent (sample size 2 

So far, this is the best.

----

- eh 6922  ( mu 33), a  current issue of a tube well known for hifi

tube nbr 1:

THD 1.48%  THD+N 1.48%  at 38.2Vpp  output almost completely H2


tube nbr 2:

THD 1.75%  THD+N 1.75%  at 38.0Vpp  output almost completely H2  - triode #1
            1.33%                    1.33%        38.1Vpp      "                                            "                - triode #2

tube nbr 3:

THD 1.57%  THD+N 1.57%  at 37.9Vpp  output almost completely H2

Another eh tube with *really* low THD in this setup. Tight spread for this tube sample size of 3.

I'm keeping the two best for use in my latest limiter build and the worst on I keep in this amp.

------

So there it is - for low THD the 6922 is the clear winner  with the 6N1P as a very close second. Selection from a larger sample size would probably equalise that narrow gap.

And this is at 1KHz only - the story is much more involving when looking at the spectral distribution of thd.

I'm pretty sure the 6922 and 6N1P will remain on top when looking at the low end in more detail.
I'll leave that to when I've eliminated the 'outriders'  and settled the configuration of toobs

Anyway, no real surprises here - generally speaking, lower mu tubes of the 'au7' type (mu 17)  have a higher THD than low-med tubes (mu 33) and higher, at least in this kind of non-optimal testing.

I *was *surprised by the magnitude of the increase though - 6x higher  in some cases.

Also, the eh6922 tube I selected has a triode #2 with a quite close match to triode #1 - a good sign.


 

 

[alexc]

Next up, in the third preamp position, currently with an  12AU7. It has one half as eq makeup and other half as aux mix makeup.

I chose the au7 here for low-ish gain, high-ish input capability and hopefully! with moderate thd.

The aux mix makeup  side has 10K on the plate, 1K on the caths and is passing around 6mA.
The EQ make up on the second half is a bit more gain with 15K plate and 5.2mA on the 1K caths.

Supply is a modest 245V, well filtered and dropped from 285V or so. Again, PSU decoupling is important.


EQ makeup GCA
----------------------

With my usual 3Vpp applied to guitar input jack and AY7 grid (REW -12.1dBFS)

signal path is  AY7 -> tone stack -> EQ makeup AU7 stage (gain x5.2)  -> 220K mix resistor

AY7  THD 4.39%  at 79.3Vpp  -- this is my worst AY7  (best was 2.45%)

To find out the contribution of the AU7 stage alone, we do  'sum of squares of THDs'  (not just a straight addition). 

AU7 EQ makeup  THD 6.9%,  at 116V output

Now that's pretty high -  both in output swing and thus THD. Normally this would be adjusted down by the 'Volume' control.

It's a Baldwin (organ) AU7 which is actually a black plate rca from the 60s. Well used but good.

Checking with an eh au7, I get THD 6.8% at 112Vpp output  which is very close.

Now, after the mix resistor, even though the AU7 output is scaled down to 27.3Vpp, THD remains high at 6.5%.

I would expect it to be mostly unchanged, as THD is a relative measure and should be the same despite 'resistive' scaling of magnitudes. 

There can be some additive/subractive effects on the passive scaling components, but by and large it should be the same.

----

Now the other half of the AU7 is the aux  mix make up stage.  Input from the previous stage (6922)  has a THD 2.25%

(ps    - I previously measured this at 1.75% - that was with 2Vpp applied to the inputs, here it is 3Vpp input)
(pps - This tube seems to show reducing thd with warmup, so I expect it to reduce a little here)

At the AU7 plate, THD at 2.72% at 71.3Vpp . Doing the calcs returns THD at 0.89%. Which is very low!

Comparing rca with eh, the eh au7 is 1.5% here, which is worse but still very good.

(ps - need to run more direct au7 tests to confirm thd)

So it appears, the EQ is the main contributor to the high THD of the first au7 triode. Varying the EQ makes major changes to the THD, as one would expect. Thats what an EQ does   Treble control varies THD 7.29% .. 25.2%, Bass 6.86% to 13.7%.

I have pencil'd on the 'flat eq' setting, arrived at by observing the freq response sweeps - which is where I will keep it for further tests.

So it appears the nos rca au7 is in fact very low THD at 0.89%    That one can stay!

 

[alexc]

Now to the 6SN7 driver stage.

I have it configured again as a GCA+CF combo, the so-called 'constant current amplifier' or 'CCA'.

So, low-ish stage gain coupled with a strong drive capability and low output Z to drive the parallel finals, in this case 6L6s (nos rca blackplates).

The 6SN7 CCA has a mu of around 20. I have 27K plate resistor for the first stage and similar as the cathode resistor of the second stage.

 

[alexc]

Some moar toobs on the way ..  quad of 6N1P,  a very tidy performer,  AND a new entrant for me, the 6N23P.

Lot of hype around this one. Not too pricey, so good for 'research' purposes   whilst I save up for a couple of good AY7s.

In future, I'd like to try the eh 'gold pin' 6922 type - as well as their so called  'super tube' 6H30pi, more of a 'super HD' ecc99/6N6P kind of thing, apparently.

Speaking of which, I was kind of disappointed with the 6N6P (sample size of 2   in this, admittedly not really suited setup.

Yes, it's more like an ecc99 or 12BH7+    than a low THD voltage amp as an input stage :/ (lol)

Albeit at a very low gain. Still. *sniff*

Seductive in the specs, the 6N6P  can take quite a lot of signal and with lots of drive capacity).
Like many triodes of this class, there is a  bit more  input capacitance. But not that much!

-----

Enter the 6922 and similar.  In parallel, the cg-a of a pair of 6922 stages is a bit less than the 6N6P, but not too much.
Together, that gives a bit more for drive, into the useful range for 'heavier' loads,  .... plus at ridiculously low rp.

The 6N6P can do all that, each side - and to the next level, in a number ways ... including plate current, rp and so on .... *if* required. Well into power tube vistas!

The jury is still out on that one - I won't give up, even though I thought I picked my operating parameters .. just right.

In the meantime, I'll just get some moar  ... and check the spread.

Like always. Can you beat the median  ;D

More plays required.
-----

Meanwhile,

My pool of nos AV7 has grown, with a quad of 'matched Ip/S measures' and another quad of 'matched date and man' types,  for good measure.

That should take me to a sample size of 15, one of my larger sets.

As well as some 6289 which optimistically are described as 'better' av7-ish; for comparison purposes. Not really expecting too much. I'm relatively new to thd measures, but it looks like the av7 is fairly .. 'divergent' 

But it's profile seems to make a lot of sense to me, so I soldier on.

------

I'm also installing an eq bypass switch, with a matching loss resistor, for the purposes of making the amp more 'hifi', THD wise  :-[
Have to confirm  those au7 thd measure of today. I am cautiously optimistic.

Most everything in this amp is wideband and pretty hifi on the big power stage, SE wise. The big Edcor is a monster. Huge in every way. If there is an adage for Edcor, it's  'you get what you pay for', and I mean that in the best possible way.

Massive iron, flat response  20-20K, even in an prototyping amp like this one. At a reasonable price, and with really good quality control. The sales experience, even on the other, lower, side of the planet, is exceptional.

Completely PRO 'processes'. It's no exaggeration to say a lot of the more .. out there ..  builds I have done, have been made possible  only  by Edcor USA products. 

I'm just a small time customer, but even so - great service and very repeatable. I want to try their 'custom' service and specify exactly what I want That's more like a couple of years from now 

-----

So it makes sense to make this bigSE Amp  into more of a versatile player; instrument/mix  *and* if desired,  hifi - switch off the EQ.

Plug in a 15" audio nirvana full range box and enter your bunker, mythbusters style   Fidelity of a hazaardous kind!

200+ma SE with a wideband traffo like the Edcor  matched to a sensitive 15" with  a *L A R G E* vented box = seismic grade sound!

Of course, I don't do that ...  I only plan a 12" Audio Nirvana box 

---

Anyway - I'm interested in pursuing the low THD route in this amp platform to see how far it goes down. 

After that, the variable nfb. But first, I want to get as hifi as possible, leaving me to then 'dial in' thd as an effect  8)

--------

I have to say, that just several sessions into 'tuning' my amp builds with Room EQ Wizard has transformed me from a wannabe wannabe, to just a plain old wannabe.

Night and day performance wise. Not to mention engaging the interest in a really compelling way 

I've done quite a bit of rework along the way as the results have firmed right up, confidence wise.

 

[alexc]

I reset my HV rails and retuned the 6SN7 - much better now.

The 6SN7 is set up as a GCA+CF, with GCA running at 355V supply, 212V plate over 13K5 and so 11mA plate current at a bias of  +5V. That's very conservatively setup running at 50% of it's plate dissipation max of 3.75W (both plates driven).

The CF is set up with a similar current so as to balance the +ve and -ve phase amplification current draw from psu - the so-called 'constant current amp' CCA.

The +5V bias indicates around 9Vpp before clipping sets in. I set up the front end to provide at 1KHz sine/

OK then - checking the input to the CCA,

6SN7 nbr1 grid    THD  0.43%,  THD+N 0.43%  at        9.0Vpp  (max before clip)
6SN7 nbr 1 plate THD  2.27%  THD+N 2.27%  at  120.0Vpp ( gain x 13.3)
6SN7 nbr2 plate THD  2.30%  THD+N 2.30%  at 116.0Vpp    (gain x  0.97)

Doing the subs shows CCA stage THD 2.25% with the CF contributing  THD  of 0.37% of the total.

Its is currently a reissue TungSol 6SN7 - I have a 4 or so -  here's the comparisons

6SN7 tube #1 - 2.25% THD at 116Vpp output
6SN7 tube #2 - 2.16% THD at 117Vpp output
6SN7 tube #2 - 1.99% THD at 118Vpp output
6SN7 tube #2 - 1.99%THD  at 118Vpp output

So that's pretty good result, particularly for the last two which are the 'matched pair' from my SE hifi amp.
Noise figure is very good too.

Now for some nos 6SN7

RCA          6SN7 tube nbr 1 - THD 3.24%, THD+N 3.24%  at 106Vpp output
RCA          6SN7 tube nbr2  - THD 2.75%, THD+N  2.75%  at 106Vpp

Sylvania  6SN7  tube nbr1  - THD 2.21%, THD+N  2.21%  at 108Vpp
Sylvania  6SN7  tube nbr2  - THD 1.70%, THD+N  1.71%  at 107Vpp

and a re-measure of the original tungsol for sanity check  :

RI TungSol  6SN7  tube nbr1  - THD 2.21%  at 116Vpp 

which is close to my previous measurement, around 0.025% THD uncertainty

I am keeping the 'worst' of the the reissue TungSol 6SN7 here - it's pretty good

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

So that's main and aux paths operating parameters 'tuned' and tubes 'selected' yielding my first useful measurement baseline, from preamp inputs to outputs at maximums, with the EQ in-circuit and flat.


1KHz, 3Vpp input to preamp  -> THD 2.41% at 116Vpp output (around 32dB gain overall).

hum is a bit high at 64mVpp more or less - thats around 65dB down, at the max output before clipping.
I need to do more work there and get that down some more. To around 35mVpp or so  for 70dB down at max output.

But not too bad ..  for now

 

[alexc]

Looking now at the finals : a pair of 6L6s in parallel, fixed bias with independent trim-pots.

UL connection to the Edcor SE output transformer, 200mA rated quiescent current 3200ohms : 8 ohms.

Load box is set for 8ohms nominal, but I run it with a large open back pine box with Jensen 15" 4ohm 50W speaker of >90dB sensitivity.

So, typical usage is 1600:4ohms, which is low enough for a strong pair of finals in triode mode to really push some current thru

6L6s are running at around 60mA each, for 120mA thru the primary. At  B+ 380V to the plates and screens, thats a quiescent dissipation of 21W each 6L6, total 42W.

Around 1/5th of that makes it's way to the load, so around 8Wrms into 8ohms, with 120Vpp or 42Vrms to the 6L6 grids,  as described in  the previous measurements. 

Thats overdriving the 6L6 grids quite a bit - backing off the level to 80Vpp or 28.2Vrms should drive them to full power, at this early stage of tuning

With 80Vpp applied to the 6L6 grids, 1KHz sine, driving load box 8ohms :

bigSE amp, no nfb, input to 6L6              - THD 2.1%, THD+N  2.1% at  80Vpp
bigSE amp, no nfb, speaker terminals -  THD 6.35%,  THD+N 6.35%  at 7.35Vrms output 6.7Wrms

Calculating the contribution of the no-nfb 6L6+SE output traffo stage  -> 5.99%, nearly all H2 

Oh yeah - I forgot to mention, all this is with the finals in 'triode' mode (non UL).  I have a whacking great big switch in the middle which sources the finals'  screen signal from either the plates or the UL connections. Triode mode or pentode-UL mode.

Triode mode has less output power than pentode-ul.

Also may account for the hf droop, there's a few triode class a stages in there all contributing their droopage    I'll be looking at all that next.

From memory, I get around 10W rms into 4ohms for pentode-ul and around 8Wrms for triode.

------

So while I work over the hf roll-off and the gain staging, I think I'll take a listen    Should be pretty sweet.



(SE rules!)

 

[alexc]

So, at the end of this 'first pass' of tuning the preamp section I have a baseline of performance to build on. As well as tidy up a few things and recheck for layout sensitivities wrt to ground noise and what not.

Firstly  ->  reset the load box for 4ohms load, decreasing reflected Z from to intended 1600ohms nominal
Secondly ->  reset the amp to pentode-ul
Thirdly -> add the eq bypass switch

Then I need to do before looking at nfb :

- improve the open loop high end freq response -> it's currently  -6dB down at 20K. Should be -3dB max
- improve the open loop noise floor by 6dB -> hum is currently only -65dB down from max
- improve the open loop max power THD down to 3% or so with eq in-circuit and flat

Should be interesting ...

 

[alexc]

And the open loop SPL+Phase plots

 

[alexc]

Playing a schecter c-1e/a, which is a pretty good studio guitar. It has the LRBaggs 6 element piezos  and mixing preamp on board. Plus its mahogany and chambered - sounds like a super hifi cross between a tele and an sg.

Very extended top end and pure harmonics all the fretboard.

The schecter c-1e/a tends to show up top end that's not smooth. On a good monitor it's very classy.

------

Listening thru the  8ohm  Sansui SP 2500 12" 3 way, single speaker box

In  triode mode 7Wrms 

- it is very nice - quite clean sounding and with some top end rolled off in a good way Relaxed and full.  No complaints - very much cleaner than I had the previous setup. This sounds like a hifi Champ but without the faux breakup.

- hum and noise not audible at silence

In pentode-ul mode 10Wrms - louder, by a lot! starting to tax the speaker a little on the louder passages
Very much more extended high frequency, but no where near brittle. This is starting to sound a lot more like hifi SE at reasonable power should - an even bigger bottom end and a lot of hf detail.

- no hum or noise audible at silence 


All good as it was before this, but surely a lot less preamp THD because the amp sounds significantly cleaner than before despite now having no nfb. Prior, nfb was at 10 to 20dB variable on a pot.

All good stuff and business as usual

 

[alexc]

A few more days straightening out my preamp and driver tubes in this bigSE Amp

Happy to say, that with REW THD measures, I've got them all humming away nicely now.

First pass optimisation for low THD+N and running the amp conservatively with the baseline 'keepers' is showing

- around 1.9% THD+N    for the AY7    tube
-        "        1.5% THD+N        "        6922      "
-        "        2.0% THD+N        "        AV7        "
-        "        2.5% THD+N        "        6SN7        "

And on the balanced output, a 6U8A pentode+triode push-pull cathode follower, an interesting case.
I have it fed by the speaker outputs, (no nfb).

6U8A input      THD is 6.46% THD+N  at 13.3Wrms into 8ohms (120mA primary,  full output power at speaker terminals)
6U8A output  THD is 6.27% THD+N  at +3dBu  (edcor xsm 10K/600 step down to balanced)

So thats sweet fa of THD added by the pp CF being fed by the SE main output - in fact, a cancellation of some THD!
(prololy not significant!)

That really makes me happy - the Alan Kimmel paper - the one on  CFs/MFs -  is a great resource to build different approaches.  If one was a real player, they would try his OTL design. Serious stuff! Indeedy.

----

This amp, with no nfb and 120mA primary quiesscent current,  is making full power at :

- 13.3Wrms into 8ohms at 6.5% THD, almost all H2   

That's around  3% THD+N contribution  from the preamp and around 5.7% THD+N from the poweramp.

------

SO, at a nice setup for instrument mixable with Aux1 and Aux2 line level inputs, it is very managable as a kind of classy 'pa mix amp'. 

My preferred power output is around 10Wrms on the louder passages, but with a reserve of headroom still quite clean.

The bass is all there for sure.

-----------

One thing I found regarding the EQ - it actually adds very little distortion - I wired in the 'eq bypass' switch, with a bypass resistor attenuator. When 'eq flat and in circuit' and 'out of circuit' are compared at same output level, the THD is roughly the same. with the 1KHz input.

SO, no problems really, there.  And I like this EQ - from the Fender PA100 -  which was the inspire for this amp, in some ways, at least.

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

Now to 'dial' out the THD+N - previously I used around 10dB for a 'clean, hifi' sound and maybe '3dB' for a 'high THD' sound.
Also could do with a look at the square wave performance once a 'nfb baseline' is established 

With measurement of the phase margin and so on,  REW can get really interesting. With the impedance measures, if I can get it to work, would be even more interesting!    ...  but for now I mostly want to button up this amp and get it back in use.

---

It's quite a different amp to when I began this optimisation.

On the 'hifi playback', it's quite good. The distinction between a 'guitar amp' and a 'mix playback amp' has disappeared with the overall 'low THD' setup - even for a big SE power amp, as is the case here.

---

ps - I've found a bug or two in REW, at least as it operates on a 64bit Win7 box with asio drivers (and motu box).

It's generally quite stable, but I 'have a knack' with software   .... I can always find the combo that makes it drop like a stone.

---

Here's my mixroom - some great old small 2-way speakers, reworked  ,  with my build of a stereo SE 6550 (no nfb) amp.
I set it up for 6SN7 driver, but can't help run it with some nos Sylvania 6SL7    Really nice. I should set it up for them, add the nfb and have done with it. But, like it is now is so 'engaging'. I wrote it down to change it back to 6SN7s .. but am ..  'tarry'-ing  some.

EQ is a  vintage Tascam graphic EQ for the speaker curve.

'Mastering' tools are my gSSL build and an SPL Vitaliser as well as a digital reverb unit.

I use a Win7 notebook with my software tools, firewire interface and my old but cool Cubase control surface.

 

[alexc]

Also, did a little more ordering of the tubes, following that well worn path, trodden by many others including here at GroupDIY 

I want to do some more tests with 6DJ8 types, the 6072 types , and a whole variety of the AV7 types as well as a variety of AU7 types. Also more testing of the 6BZ7.

Of course, a lot more CFs, MFs and the like, with the dis-similar-pair-in-an-envelope tubes that I really like.
I'm auditioning for the tube set of my 'masterbox'  8) 8) What a labour of love that is!

Lots of time required, though REW is going to help make it my opus magnum!

---

Of course, to complete that whole path, I also need to do some additional builds :

- a really good SE monitor amp with ElectraPrint transformers
- a pair of sensitive full range speakers with AUdio Nirvana drivers and some nice woodwork
- tube headphone amp with jensen transformers

I haven't yet decided on an SE tube or tubes    That's where my mind goes while I wait for the bus and listen to the Byrds.

I'd like to try some of the Svetlana SV types, though. Add that to my b-list. All that should keep me occupied for a while  ;D

 

[alexc]

And I have come to really appreciate the use of a sub with my Sansui SP2500.  They have a really great wide, mid range articulation. 70s style balancing of bass, mid and treble 

Compared to my usual recording set - big power bjt amp + tannoy 8" full range,
....  they seem crawling with organic    life    vs    a glassified image, polished to a mirror finish.    ;D

And I mean *that* in a good way.

For the proper 'kick' level, my vintage speakers have to be compromised - boost lows, cut tops etc  -  to get the right amount of  'push.  HOWever,  by adding a sub, I can use them, with the tube amps,  for what they shine at -  wide, mid range articulation!

The sub does the low note thing,  and the crossover tames the overall  sizzle  - but the midrange, with SE amping, all H2 and very, very little H3> . ..  well it gets scary real  8) 8)

And with not too much effort or heroics. 

 

[alexc]

Looking at the square wave performance now ...

I have a max  'low THD'  power at the speakers of    4.5% THD+N  sine 1KHz  9.15Vrms into 8ohms load box.

At the input to the 6L6s, the (preamp) has about the same THD, indicating the finals+SE traffo is probably very low. I will be doing more detailed testing with a 'reference' preamplifier, to isolate the power amp  THD .... later

When switched to the square wave, (no nfb, no eq) at this level is quite good.

Here's the load box with sine wave under that condition.

 

[alexc]

And the same conditions, switched to square waves  and looking at the 'scope' display on TrueRTA.

Now, a lot of that 'ripple' and 'ringing' is the interface and analyzer artifacts.  ie. compare CRO square wave and PC analyser square wave is quite different, irrespective of the signal sourece    ie  'analog sig gen' or 'analyser sig gen'.

I am doing the detailed square wave measurement on the CRO with the analog sig generator.

 

[alexc]

So the upshot, there isnt too much to concern myself with at this 'first pass optimisation'.

Just doing some more freq response tests before re-connecting the nfb.

- speaker terms output freq response at max power (10Wrms) with 4.5%THD  is  -1dB at 20Hz  and  -3.5dB at 20KHz

Which is pretty good for the low freq extension, but has excessive HF droop by at least a dB, maybe two.

---

Tracing back from the amp output to the input, I am probing the 'cold' side of the coupling caps - the 'probulator' itself is limiting the HF extension, so I am 'direct' probing with no 'dc blocking cap' with great care!

The input to the amp is as expected, flat at 20KHz - no nfb, no eq, 10W rms tests into 8ohm load box.

-  the AY7 stage is -1db,  the AV7 stage is -1dB,  the 6SN7 stage input is -1.5dB, 6SN7 stage  output is -1.5dB
- the SE op traffo secondary is -3.5dB

So I can see the contribution to HF droop is  -1.5dB for the preamp  and  -2.0 dB for the finals+SE opt -> -3.5dB total down 20KHz.

In the preamp, it looks like the AY7 is the main contributor, with the 'Master Vol' at the 6SN7 input the next in line.
The output stage is mostly < 3dB HF droop

I need to do more work here and check the inter-stage coupling networks and so on to improve before moving on.
Could be parasitic shunt capacitances, along the preamp signal path  or could be miller capacitance or something else again.

Apart from the preamp contributed HF droop, there is around 2dB of it in the output stage.

The output transformer has a spec of 20~20K Hz, <1dBu so it is likely in the clear with regards to Hdroop in the output stage.

I'm only running the traffo with a primary current of 105mA at present, differing from the spec for 200mA (max). That may have implications for the HF extension - I don't know (yet).

I think that SE opts have an optimal primary excitation current, and one should try to stay not too far away from that.

Too early to tell for me    I have much measuring to do!

 

[alexc]

Reviewing my coupling networks shows a couple of places where the HF extension is is reduced - it's not the Miller capacitances, in general.

I'm mostly using relatively low gain stages with reasonably good input capacitance figures - AY7, AV7, 6922. The 6SN7 is a little worse, but in most usage, won;t be significant. I am reducing my 68K grid stopper at the AY7 and checking just in case.

HOWever,  I have two mixing resistors feeding a sum node which then drives into a Miller cap.

There's two seperate sum nodes nodes in the circuit, each with two inputs  - I'm using 270K resistors to feed the node.  The parallel value in series with the MIller cap of the following stage forms a high cut filter which is cutting around 1dB at 20KHz.

Also, the 'Master Volume' pot, at the 6SN7 driver input, is contributing a little HF droop when not set to max resistance.

I also think I have an error of around 1dB at 20KHz, when measuring high amplitude signals with my 'HV sig probe'.\

All that is compounding to lop a dB or two off the speaker outputs.

Yes, the amp is in pentode-ul mode, no nfb,  for these tests 

Tomorrow, I'll work to eliminate those issues and recheck everything again.

 

[alexc]

SO reducing the series resistance feeding miller caps at the gain stage inputs has mostly done the trick as far as flattening the HF response. Reducing the mix resistors has done most of it, but also in the first stage grid stopper.

Checking at different settings of the feeding 'Vol' pots shows the HF droop when pot operated at 'less-than-max-resistance' is not really significant but could probably benefit from a bypass cap. I'll try that for the sake of measurement

---

The main output at the SE opt secondary is still a little down, however. And it appears to be level dependent; the higher the output power, the better the HF extension.

So I'll restore the 6550 tubes and rebias for a higher primary excitation current and check the effect of that.

----

With the 6550s back in and biased up for a primary excitation current of 2x75mA -> 150mA of plate current

Measuring now at at Motu soundcard settings of 96KHz 24 bit, with a REW measurement bandwidth of 48KHz.

-> Running sweeps at a  4.5% THD level - output  has increased to 15.5Wrms into 8ohms (vs around 9Wrms for 6L6s)

Measuring now at at Motu soundcard settings of 96KHz 24 bit, with a REW measurement bandwidth of 48KHz.
So it appears that I have a reasonable baseline to move forward - I'll just take a look at the effect of increasing primary excitation  while I'm here.

 

[alexc]

Here's the phase response - you can see it exceeding 90degrees at <20Hz and some resonance jiggery pokery at the HF limits

I'm happy to say that I haven't found any additional issues with REW at the higher sampling rate. It appears as though the response datasets are valid   

Doesn't seem to be any major  differences between high measurement BW and lower, apart from the contribution of the gently increasing noise SPL at the HF end.

For now, I'll keep the amp setup as it is 

 

[alexc]

Looking at the spectra of harmonic distortion, it appears that within the 4.5% THD, the balance has shifted in favour of H3 (3.09%) over H2 (1.75%).      More on that later ....

 

[alexc]

Now to look at hum ...

Not much difference in noise floor with inputs shorted vs open. I have switched jacks which ground the jacks thru 1M when not plugged in.

With the 'Master Volume' at minimum, the hum voltage is 33mVpp - 100Hz ripple with very little riding on it.

-----

With the 'Master Volume'  dialed for 4.5%THD,  15.5Wrms into 8ohms    (27.3Vpp at the speaker terms),  the hum voltage at the speaker terminals is 67mVpp.  Thats a 6dB increase in hum from min to max.

Overall gain under the above conditions is x38.4 or +32dB.  So, +6dB hum penalty for +32dB gain is not too bad really.


And  that's 20log(0.067/27.3) or  down 52dB from fundamental. 

----

Looking at the HV rails, I have a 60uF -> 5H, 200mA choke -> 60uF for B+  that feeds the -R-C- network for B1+ (6SN7 driver), B2+ (AV7 and 6U8A),  B3+ (6922 and Ay7).

B+ ripple is  394mVpp,  B1+ is 37mVpp, B2+ is 27mVpp and B3+ is 11mVpp,  at 100Hz.

And the Vbias has a ripple of around 25mVpp, riding on a slower modulating wave.

Looking at the plates of the finals, there is around 400mVpp of 100Hz hum, riding on a slower modulating wave. That appears at the secondary of the SE op traffo with a divide by 20 stepdown.

All that leads to the 33mVpp minimum hum voltage at the speaker terminals.  At max gain, higher harmonics of the ripple make their appearance.

Finally, checking the balanced line out, there is around 15mVpp  of noise, which shows up as -48dBu on TrueRTA. It's not dependant on 'Master Volume'.

---

So , more work to do there on layout to try to wrestle  a few dBs back from  100Hz hum 

- check grounds
- check psu decoupling between B+ rails
- check wiring sensitivities