ECC99 in G9

I'm thinking about trying ECC99 for the SRPP stage of my G9.  I have the space (3U) and I have the power (3A from the trans). In order to reduce strain from the regulator I can build an additional 12V supply with its own regulator and heatsink. 

First of all the cathode resistor will have to be lowered for a bit.  That will be nice an easy, add the appropriate value in parallel to R29 plus a on/off switch (so ECC82 can still be easily swaped)

But my very limited tube knowledge stops here, and I don't know about R30.  It would be quite difficult to mod this one as it is wired under the tube socket. I'd have to cut and resolder a lot of things to get in there (yeah my convoluted first try at p2p building)

Thanks! 

It is unusual for the two cathode resistors in an SRPP to be different. For maximum output swing you would normally aim for the output to sit close to half the HT supply so it is normal for both cathode resistors to be the same in order to achieve this. Here is a link to my headphones amp design which uses an ECC99. As you will see both cathode resistors are 300 ohms. Quiescent current is around 15mA. I don't know how much HT current the ECC82 takes but it may well be less than this.



Cheers

Ian

A larger cathode resistor for the other triode seems typical for Jakob.  In his ECC88 srpp he has 1k and again 470R for the lower triode. 

By simple logic  a larger resistor eats more juice leaving less for the tube internals. Perhaps Jakobs aim was to reduce distorsion? All the resistors around the lower tube are higher than in your srpp.

I might guess that I could simply swap in a 99 and nothing will blow up but the tube will not sound as nice as it can - having too little current in it.  I've read ECC99 really needs the juice. 

Other than this I've looked at a bunch of ECC82 SRPPs and it seems every designer has it's own idea.  Some have 1.5k for both cathodes. Others have different values....

Whatever the values of the two cathode resistors,  the current in both triodes of an SRPP must be the same simply because they are in series. The different value resistors will therefore create different values of grid bias for each tube so they run at different operating points. The one with the higher grid voltage will have a higher static plate resistance and drop more of the HT voltage across it. The two tubes will therefore share the HT voltage based on their operating points. Since the tubes are at different operating points, the distortion produced by each one will be different. Perhaps this is why some people use different cathode resistors - to produce a specific tone?

As to the values used, assuming both are the same, higher values lead to lower quiescent currents. I run my ECC88s at about 5mA which, with 150V across each triode means the plate dissipation is 750mW per triode which is comfortably within their dissipation spec but is high enough to drive over +20dBm (100 milliwatts) into a 600 ohm load via a 2400:600 transformer. For that sort of current in an ECC88 you need a cathode resistor of around 680 ohms. The 300 ohms I used for the ECC99 results in a quiescent current of around 15mA. This is sufficient to output 500mW into a headphone load via a suitable transformer.

Basically you need more current for more power.

Cheers

Ian

FWIW my current ECC82 voltages look like this:

HT 240.5
left and right of output cathode resistor:  83 - 89V
at input cathode resistor: 1.9

Could there be any problems for the tube (premature wear/death) by running it with too large cathode resistors?    If not, I have no problems doing some ear tests myself.  Swap the tube then progressively lower resistances until it sounds nice?
Another thing that might need a change is the output cap currently at 4.7uF

Thanks to an old Jakob post, here is a really long and complicated read on SRPP.  I've read it trice and understood little but I'll keep trying  
http://www.tubecad.com/may2000/page1.html

Well I guess experiment is the mother of most development so let's see what a bit of messing around with Rak (R30) does for ECC82 before I actually receive an ECC99. 

Voltages with original Jakob's value of  1.5k :  89 and 83 V before and after

Reduced to 1.3k:  93 and 88 V

Reduced to 540 ohms:  121 and 117. 

Sound?  I would not dare try to judge by recording things quickly, I believe such changes can be judged reliably only with extensive recording and listening, so let's look at RMAA frequency response graphs instead. 

Difference between 1.5 and 1.3k?  Almost zero. 

540 ohms:  now the bass starts to decrease.  About 0.3db worse at 50Hz, and also an increase in noise by 0.5db
Also THD is up  a bit




 

Here are the results: 

With the ECC99 simply put in, sound lost clarity and focus. It is a thirsty tube so cathode resistors need to be lowered.  R29 down to 300, some improvement but not enough.  R30 down to about 500, all clarity is back. 

Overall gain increased around 3db, maybe a bit higher.  It is like getting one extra step on the gain switch. 

RMAA distorsion tests show a reduction of 2 to 3 times. 

The bass response has improved, maybe too much.  From about -2.5db at 30Hz, it jumped to +2.5.

I could partially fix this by adding a 100n cap in the low filter section.  Not ideal to add one more cap though.  Perhaps a better way is to reduce the input cap by I didn't try it. 

Looking back at the Millett SRPP, output cap is 10uF.  Added a 4.7 in parallel to C13 and it worked, bass hump decreased by 1db and also seems to have moved below 30Hz.  Perhaps a total of 12uF would  do the trick for ruler flat. 

To counterbalance the bass improvement, treble response has diminished a bit from the original.  This might be and might be not a problem.  It is easy to tweak it by lowering (or removing) or increasing C6.  With C6 cut off it is ruler flat even past 20k, but a bit cold sounding.

Overall the sound has kept the same character as the original, on the other hand it is slightly more sterile, more solid-state.  I do not really appreciate this especially considering the Lundahls I use are (too) transparent already. 

The other channel where I use a nice old Amperex has sweeter mids and trebles. The 99 reminds me of my solidstate pres.

The big downsides are: 

huge current draw on the filament (a hair short of double the original) .  Don't try this unless you are prepared to add an extra 12V supply with its own regulator and heatsink.  You probably need an extra trafo too unless you have something really big to begin with. 

The tube is about 16mm taller than a regular 82.  I'm not sure it has enough space in the 2U cases everyone uses. 

At the end after 2 days of countless recording tests (voice and guitar) I decided to keep to the original design.  The extra gain and lower distorsion sound great on paper but I have solid state that can do that better.  The original is sweet and I like that.  I will likely use the 99s for Ruffrecords' headphone amp. 

Several points that should be mentioned.


You never said why you wanted to change the original tube to ECC99. Compared to 12AU7 it is very linear with very soft breaking point. That explains your mentions of "solid state sound". I've used 6N6P a lot and it behaves exactly the same. They can be overdriven in a very controlled way but you need extreme levels for that (or extremely low plate voltage way below 100V).

You accidentally made a broadcast preamp like grandma used to make. In my opinion it is the highest level of tube amp engineering to create something where no one can tell tubes were even used.

ECC99 has a slightly higher mu-value which explains the minor gain increase. The winning factor in any output stage compared to 12AU7 is the far lower plate resistance (2.3k vs. 6.6k of 12AU7). It might happily drive even 600:600 (1:1) output transformer where 12AU7 is struggling with even 2:1 to 600ohms.

Which brings us to the bass cut and the bass bump of the cap. One thing is for sure, ECC99 easily drives the Lundahl with no bass cut. But you cannot remove the bass bump. It's a physical constant from the output cap and transformer interacting together (inductance related resonant frequency). You can only move the bass bump to your choice of tradeoff frequency. Move it too low and the bump peak increases with potential loss of headroom.

I can't tell why high frequencies would change with any of your mods. Are you actually measuring these with FFT plots?

RMAA plot attached. 

To the ear the difference is subtle, the original sounds a bit more open and light, sometimes sweeter.    The bass cut is not a problem for what I do (acoustic guitar).  With the solidstate pres I used before i would often engage the basscut filter.  On the G9 I simply omitted it (i have seen RMAA plots before starting the build)

Anyway now I am thinking to mod just one channel and match with the microphones.  A deep sounding mic (say a tube LDC) could go to the original and an SDC to the ECC99?  Anyway it is nice to have options. 

I've stacked a few more caps (a total of about 13uF) on the ouput and the bump moved from 33Hz to 20.  So maybe a 15uF will be perfect, or probably it doesn't even matter at all. 

The 1:1  transformer could be very interesting, thanks for the suggestion.  With a 1:10 on input it might happily work on ribbons?!

Alexandru marian said:

The 1:1  transformer could be very interesting, thanks for the suggestion.

Click to expand...

For this you have to do the calculations. I only said it might drive. You would only win 6dB of gain with this.

I think it is worth a try.  6db might not be wow for a loud source, but with a low power instrument (nylon guitar) with low output mics (I now use DPA 4011s) those 6db are wow.  That's why i build it from the first with 1:7 on the input( 1530) and now I'm thinking I should have tried 1578 (1:10)

http://www.mh-audio.nl/tubecalculator.asp

Looks like ECC99 can do 600:600 easily. Note, LL5402 is not designed for this and won't work.

Thanks, nice link, I'll play with that!  Sowter should have some nice 1:1...

But looking at the 5402, it is 2+2:1+1.  In G9 both sides are in series so 4:2.  With the input windings in parallel, would it not be 2:2?  Or maybe with one input winding omitted?  What I am missing?  Bad impedances?  Sorry I was a total noob before starting the G9 and that hasn't changed too much

Yes indeed "bad impedance". The input winding is designed to be driven by very low impedance source. It acts somewhat like a 600ohm impedance when configured in series. Paralleled it will act like 300ohms (in practice you will need more to drive it) and ECC99 ran out of juice. There would be some stray capacitance related effects when configured parallel as well (read: a bit more unpredictable frequency response).

That said, there is an abundance of 600:600 line level transformers out there. They are easy to find and cheap. And easy to have perform really well, even the cheap ones. You will need a much bigger output cap.

Unfortunately I can't get past the reduced treble response.  I wouldn't care about the graph but I can hear a veil or heaviness over the recordings as well.    It must be some incompatibility between stages.  If the gain is set to near max, and output very very low, the graph is almost ruler flat.  The bass bump goes away completely.  Same for setting the gain to 2 or 3 and output to max.  But anything in between, the bump is high and treble rolled off.  I've seen the same roll off on your ECC88 moded G9 with OEP output. 

While I can't understand a word of those descriptions of gains and their differences, you should get used to the idea of placebo while looking at graphs and listening to the music at the same time. I'm absolutely positive your mind is playing tricks.  Humans haven't got the physical capability to hear a 0.5dB difference at 20khz. This includes four year olds with no hearing damage from concerts. Unfortunately the mind is far more powerful and it will happily let you hear whatever you imagine you hear.

By gain I mean gain switch, by output -the pot.  So low switch setting + max pot, or viceversa, high gain + low pot, these two make the linear treble graphs.  In-between settings, like 50-50, come with the treble cut.  The original config has some of this too, a very low pot setting produces ringing on the trafo.  The 10k output helps reduce it (tried with and without the resistor)

I might be imagining things, but I have also sent  recorded samples to a recording-enthusiast friend. He did not have a clue what is what, just that I am testing some preamp settings, and he had the same impression as me.  It is a subtle difference but I think i can hear it.  In a way the original config is like a bright microphone (you know many have a treble hump). ECC99 sounds darker and neutral.  Nothing wrong in theory but it is a bit too much,  leaving that "veiled" impression in the end.  Perhaps I need to push the tube more.  I've seen it used even with 120 ohm cathode resistors (i now have 500 up and 310 down) 

You are looking at the normal G9 behaviour with those somewhat changing frequencies and this "ringing" you mention. Nothing to do with any transformer. The first stages are in a negative feedback loop and the input gain is changed by varying the amount of feedback. This has side-effects like the frequency response changes you are seeing, including the resonance pole.

The output stage will not change its frequency response with any input gain rotary switch or output (passive attenuator) pot configuration. No matter what tube or output transformer there is. The varying responses should all be from the input gain feedback configuration.

Also, did you at any point measure THD with these tube and bias changes? It would be helpful if instead of "veiled" you would say "significantly more distorted" with the graph along side to back it up. Now it seems you are shooting from the hip in complete darkness. Yes, ears are completely unreliable as a measurement device.

And again, what was the motivation for these changes in the first place?

I said it above, THD (and IM I will add now) are at least 2-3 times lower with the 99 according to RMAA. 

I do not say "veiled" as in distorted, it is definitely not more distorted. It sounds like a different EQ curve.  And it matches the RMAA curve. 

I got the "ringing trafo" idea somewhere around this forum.  I don't know if it is really the output trafo, but with the 10k resistor strapped across the output, with ECC82 there is a +1db bump in the very high treble, without the resistor it is close to +3db.  Only happens if the pot is at a very very low position. 

In the same scenario, 99 is linear.  But in most common usage scenarios (gain and pot in the 50-70% range) the treble is reduced as can be seen in the graph above.  For my taste this is maybe more aggravating than the reduced bass of the original. 

I've tried the 99 out of curiosity, see if it makes it sound "better".   

At this point it seems to me that while the ECC82 resistance is a bit too high, the ECC99 could be too low.  I guess next I will try with a small resistor in between output cathode and cap.