G9 and its frequency response ?? (impedance matching?)

[dmnieto]

They certainly sound great, but I think I could personally trade-off the 6dB in gain for a flatter response in sub 250Hz.

Kingston commented that with the OEP wired as 4:1 he managed to get that from the ECC82, I was just wondering if the same could be done with the LL5402. I guess it will

 

[bruce0]

You could wire it up on the bench and try it out with jumpers to see.

The response sub 250Hz is fine (my test platform is an old p3 laptop and RTAS) it is only down 1db at 200hz.  I saw it down 3db at 50 Hz.

Let me know...I have a G9 wired up with an altec 15095 backwards (it is 4 or 5:1 stepdown) and I have not RTAS tested it yet, but I can.

 

[burglar]

Hi guys,

What I took from all this is that the OEP is a weak option as an input transformer and is a big factor in the low frequency roll off many are reporting.

I have OEP in / out and I am experiencing weak low end. 

I am planning on switching my input trannys to Lundahl.  Are there any success stories out there with this configuration?

Cheers

Ed

 

[Whoops]

I am planning on switching my input trannys to Lundahl.  Are there any success stories out there with this configuration?

That has been covered in this thread many times before.

"Well, it all seems to boil down to the following:
Using the Lundahl output with the standard ECC82 topology results in a big bass roll off due to the highish output impedance of the SRPP stage. The Lundahl doesn't seem to be build for that purpose. (Just look at the datasheet, it pretty much says to use as low source impedance as possible for optimum performance)
Then, on the other side, using the OEPs, the output seems to have enough inductance to get the bass through. (Or at least much better then the Lundahl)
Then again using also the OEP input the roll off happens at the input transformer. This time the OEP input is the problem as you can get the response pretty flat but only with really low source impedances.
So when using the standard topology, I think it would be best to use a Lundahl input and the OEP output.
"

"So, I think it's time (at least for me) to draw a conclusion:

There are two possible secenarios that will cause  low end loss with the G9 design.

1) When using OEP trannies (as I did) the input transformer that seems to lack inductance causes the problem

2) When using only Lundahl transformers, the input is fine and flat but the output transformer again lacks inductance and causes a pretty huge roll off.
    (This transformer wasn't designed with such a circuit connection in mind, at least according to the datasheet)

So worst case scenario would be to use a OEP input and Lundahl output

Best would be to use Lundahl input and OEP output at least with the normal circuit.
"

"I've made many tests on the input trafo's issue.
About that, I can say you for sure that the OEPs input trafos,
if the input impedance is > 50 Ohm, do have a big, audible bass roll-off starting at 200-300 Hz.
That seems to be related to their lack of inductance.

Other input transformers tested [carnhill, edcor, cinemag, marinair, haufe, beyer, etc] have flat frequency response also with 150 and 600 Ohm input impedances.
"

"Let's resume what we were working on.
The basic problem with the G9 freq response was tracked down to the transformers. The OEP input transformer causes roll off and the Lundahl output causes roll off both caused by lack of inductance. (The LL5402 just doesn't seem to be designed for this aplication...)
"

"Even with an OEP output, you'll still loose a couple of db. It seems that the ECC82 is really struggling to drive that transformer. That causes the slightly tilted freq response you can see in an earlier post.
Thats why I think I'm going to use the ECC88 in the ouput stage, as it performs so well in my test.
It will run a little hotter then the ECC82, so maybe I'll have to scale down a resistor or two in the B+ power supply.
Moreover, I will change the 4,7uF cap to a 10uF (as I have those in my junk box) to move the bass resonance down.
I also will have to mod the PCB and power the ECC88 with the 6,3V filament it needs.
"

"If you scroll back the thread (maybe you did already) you will notice the big bass cut is due to the 12AU7 as output tube. It cannot deliver low enough impedance in this SRPP configuration for the output transformer. Adequate tubes with for the task are 5687 or ECC88 (and variants) for example.

"

 

[ln76d]

I would definately try 10uF-15uF cap at the output.
That's for quick
Didn't read all.

 

[gyraf]

IMO, the 5402 is fine for the G9 output stage - I get a low-end frequency response a bit better than what our MTR90 delivers.

I'm pretty sure huge low-end loss in some G9 builds are because of something else

Next time I have a G9 at hand, I'll try to make a reference freq-response plot.

Jakob E.

 

[ln76d]

My currently is in build process, so i'll be able to say something more in two three months.
If i'm not wrong output stage impedance should be around 300-330ohm.
If i'm right it would be a little too much for 5402 but it shouldn't affect low end response that much.

 

[Alexandru marian]

Another resurrect....

Here is how mine does with 1530 and 5402.    -1db at 50Hz on one channel and -1.5 on the other.  I am not bothered at all since I record only acoustic guitars and they are far from going so low. 

I really don't get why the bass responses are different though.  It is built p2p dual-mono, the "better" channel is further away from the PSU, and it has some very slightly optimized wiring but the components are the same.  It is not the tubes as I swapped and it does the same. 

In any case the bass can be bumped to ruler flat with an ECC99 in the output but it accordingly loses some top end and turns to sound a bit sterile. The original is more sweet and polished and matches well the DPA mics I use. 

 

[Whoops]

IMO, the 5402 is fine for the G9 output stage - I get a low-end frequency response a bit better than what our MTR90 delivers.

I'm pretty sure huge low-end loss in some G9 builds are because of something else

Next time I have a G9 at hand, I'll try to make a reference freq-response plot.

Jakob E.

That would be nice Jakob , a Freq Response Plot could then be used as a reference.
A lot of people complained of Low End Loss in this build, it goes back to the first pages of this thread, and normally when there's so many people complaining about the same thing it's because it is happening.

 

[gyraf]

G9 plot - 1dB down at ca. 100Hz and 20KHz - 3dB down around 30-35Hz, slightly depending on gain setting.

Standard setup, Lundahl input and output.

Jakob E.

 

[Whoops]

G9 plot - 1dB down at ca. 100Hz and 20KHz - 3dB down around 30-35Hz, slightly depending on gain setting.

Standard setup, Lundahl input and output.

Jakob E.

Thanks Jakob

 

[Alexandru marian]

I would definately try 10uF-15uF cap at the output.

In the original config, 10 will roll off more bass. Not much but if you do graphs, 4.7 looks better.  10 is good for ecc99. 

 

[gyraf]

The G9 was designed for ECC82, 4u7, and LL5402. Change any of these factors, (even for "better" specs) and performance Will most likely be impaired. I think we discussed this in detail in G9-thread earlier..

Jakob E.

 

[Alexandru marian]

One important thing is that there is noticeably more rolloff with vintage tubes compared to new JJ ECC82.  I've tested old Amperex, Siemens and Philips with the same results.  Despite the slight bass roll, the sound is a bit sweeter with the old ones, I think. 

Here is a graph.  Purple is my left channel with JJ, only 0.7db down at 50Hz.  Blue/green is left with vintage tube and right with JJ, and white is right channel with vintage, last scenario drops 1.5db at 50Hz. 

This left/right difference is driving me crazy

 

[Spencerleehorton]

i might look into getting some transformers designed for my G9 as i've aways had some bass roll off issues, if there are any speifics i need to take into account please fire away and i will pass onto designer.

regards

Spence.

 

[Alexandru marian]

i might look into getting some transformers designed for my G9 as i've aways had some bass roll off issues, if there are any speifics i need to take into account please fire away and i will pass onto designer.

I might be interested in smth too.  There is also this:

From what I've read on the subject so far (I guess everything) it was Kingston who measured better bass using the ECC82 with an Edcor that is 10k:600.  In the end he ditched it because of some whatever sound issue or ringing (can't recall) but the impedance stuck in my head. 

So I found the Sowter 4383 which is 10k:600.  Two problems: I am reluctant in spending the money on one without knowing for sure that it will work "better" than the 5402, and 2:  it is 4:1  so it will have less gain. 

I wish I could figure out what impedance the 5402 has.  What is for sure is that wired 4:1 it has pretty much the same response as before. Maybe 5% better.  As a curiosity I also tried a Sowter 9995 (microphone trafo) and it has the same bass response graph as the 5402 (and slightly less treble)   

 

[gyraf]

Think of it like this:

If you look at the plot, you can see that the low cutoff corner is considerably steeper than 6dB/oct.

This because low end is held up by the resonance between 4u7 and LL5402 primary inductance.

If this resonance "bump" is low-Q and placed exactly one octave below wouldhavebeen-induction/impedance-cutofffrequency, you are rewarded with the full-one-octave downward extension of frequency response. Which also makes the unit sound nice - I'm pretty sure that this trick is why it sounds so unbelievably good on some (DI) bass guitars. But ymmv.

The trade-off being that you get a resulting low-cut of higher order below lowest transition freq.

This why many G9 builds with apparently "upgraded" components comes out sub-par - it's more or less like what mechanically goes on in bass-reflex-ported speakers: You wouldn't blindly "upgrade" port or cabinet dimensions when building speakers, would you?

Jakob E.

 

[Alexandru marian]

Jakob,

What you say there is well above my technical understanding of these things, but I must say that i am really happy with how my stock G9 sounds (with Lundahl (!))  1-1.5db down at 50Hz is nothing much (for me). Compared to perfectly flat solidstate pres, I can hear slightly less meat in the upper bass (very low I don't record), a very slightly lighter tone. But accordingly it also comes out quickly and with nice articulation.  I think that suits an acoustic instrument well. 

I don't know about others but if I tinker with it is for the fun and education, not because I don't love what you designed.  I really appreciate that you shared it and help on these forums.

Maybe I will find some cash to try those Sowters.  If they work nicely the 5402 could be saved for a Pultec    

 

[alexc]

I used the big Carnhills for outputs on mine, mounted off the pcb -  2:1 turns stepdown with 2400:600 impedance.

-3dB point is close to 20Hz - no rolloff really at all.

I used Lundahl 1:7 on the inputs, LL1530 I think. It has a higher step up and a larger core.

Used au7 tubes, eh all round. I subbed some nos AV7 in the first position, and I preferred it, but there wasn't much in it, so I stuck with the au7s

----

What I did find interesting  in re-testing my G9, which I built around 2010 or so,  was the THD numbers.

In my config, it is quite easy to dial in THD from around 0.3% to 4% or so.  ie. quite clean to quite crunchy - the nfb 'gain switch' and the 'level' pot work very nicely and evenly for this.

I formed the impression from the numbers, that it's 'not as clean as I thought it would be'.

Signal-noise was reasonably good, maxing out at around 86dB of margin from the 'hum floor' (using line inputs) to a max of 1% thd, under a variety of gain/level settings.

I have a little 200Hz hash in there worsening that result by around 2dB to 3dB or so - the 50,100,150Hz hum components were quite a bit lower. I think I could improve that with a bit more jiggery and pokery ...

----

Next time I open it up, I'll chase down that little hash at 200Hz,  and while I'm there, beef up the heater regulator a bit.

I can then safely leave these nice old philco av7 in the first position, which use more heater juice.

Ideally, I'd also change the second position to 6dj8 - I would expect that would clean things up some as well as giving a bit more gain, meaning the 'gain switch' could come down a click or two lowering the overall thd a bit ..


Anyway - I have more experience in tube builds since I did the G9 ...  I have found a new respect for the design,  after doing the thd measures  with 'room eq wizard' app 

Getting a decent amount of thd, smoothly variable, and majority 2nd harm, is not always easy to achieve - particularly in a box which caters for mic, line and hiZ inputs  ....

Indeed, I was going to 'recycle' it into some of my current builds, but I ended up keeping it the way it is 

Maybe one of the 'younger generation' of 'The Lab' folk will re-layout this design, with some changes ..

Things I did were :

- seperate the line input and mic input connectors
- added a switched relay bypass from line inputs to  main outputs
- re-inforce the main ground traces
- offboarded with cable the HiZ connector
- off boarding the heater regulator
- offboarding the +48V section with a JLM 3rail psu module
- modified HV section to use a single psu toroid transformer

and just for kicks, I added a led vu meter for each channel, powered off the JLM 3rail.

If I were doing my 'ideal' tube mic pre, I'd probably choose the G9 front end ... 

 

[Alexandru marian]

Some good news for you bassheads:

I tested an output trafo which performs very very well without the need to mess up with other tubes and heater beef-up. But you need to be slightly rich  ;D

It is the Sowter 9063 wired 2:1    http://www.sowter.co.uk/specs/9063.htm

Even in 2:1 the inductance is enough not to cause any bass loss. THD is similar to the 5402 but since there is no saturation the 3rd harmonics are much lower in the bass region.  IMD is lower overall too. 

Test rig is stock G9 with vintage tubes, 10k input Mytek ADC, RME interface.

And not to forget an important thing: new JJ tubes match the 5402 much better.  There will be 1db more at 50Hz.  But you need a bit of luck too. Since the trafos are pushed to the limit, they can differ in response. The plot shows my better one. The other drops more bass and looks worse even if same batch.

Anyway for me the best option is the sowter. Flat with nice old tubes and shielded (the cheap carnhill 2291 might perform as well but no mu case)