harmonic distortion figures in triode stage

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

5v333

Well-known member
Joined
Jun 30, 2013
Messages
631
Location
Gothenburg
hi!

im back with my makeup/output pushpull amp for my NYDeq now that summers live gigs are over and i have repaired my tektronix tm503 mainframe unit containing my funktion generator and freq counter.

ive been re drawn the amps stages many times from doing versions of the loadlines and calculating the 2nd and 3d harmonic distortions on paper.

almost everytime ive seem to come up with improvments, sometimes just a fraktion sometimes x2 or x4 lower harmonics.

figures on paper and real life figures might not be the same i guess but the only distortion analyser ive got for the moment is my ears so getting good stuff on paper first seems like a good goose to follow.

initialy doing 2-4 different graphs get me in the ballpark, then 2-5 versions  more will finetune the amp to very low figures.

however, what is low here?

for now i have,  on paper, a transformer coupled pushpull line output stage with about Hd2 = 2.5% and Hd3 0.13% (even harmonics cancels out)

and a symetric 12ay7 pre stage with Hd2 -0.68% and Hd3 -0.45%.

theres also a couple db global feedback here.

so what is to much, tolerable, good and really good?
 
Distortion is triodes is directly proportional to output level. It also depends on the total load seen by the plate. If you can add that information it will help us to asses if what you are achieving is reasonable or not.

Cheers

ian
 
hi Ian!

im calculating the harmonic distortion from AC load lines and ive noticed latly that working backwards (ac line before dc loadline...kind of atleast) is a cool way of finding your spot.  i then work out the harmonics from the highest amplitude (where AC line touches the zero volt grid line at one end)
http://diyaudioprojects.com/mirror/members.aol.com/sbench102/po-dis.html
this method is what i use.

first stage unbypassed:

B+ 235V
aiming for bias point VG -1V Vq 75V Iq 1.35mA

plate resistor is 120K grid/load resistor 330K into 6sn7 grid. (theese values are picked also with concern of lowend rolloff with the type of cap i wanna use)

AC line starts on the voltagescale from the right at:
(0.00135 x 120K|330K) + 75V = 193.8V
and through the bias point.

 
alright.

last stage is the pushpull 6sn7 output.

B+ 215V

aiming for Iq 14mA Vq 200V  Vg 4V

transformer is coupled 9+9 : 1+1+1+1
transformer primary DCR is 655 x 2
aiming for a secoundary load of 740R
if im right should reflect as (740 x 4.5ˇ2)/2 + 655 at one of the plates. about 8150r.

ac loadline starts from about 315V and crosses the bias point.

now this way of looking at the AC loadline and the reflected load in pushpull stage is something that i have come up with on my own kind of so im not 100% that this is true...

this is open loop figures aswell. i guess AC loads will change when i feedback resistor from plate to the cathode of the first stage is added. right now that resistor is around 120k hanging in aligator wires.
even if i wanted it to be lower and throw some more gain away, it sounds pretty good as it is and squarewaves look less strange (ramp is tilting the other way from the output around 400hz) inside the loop.

 
just noticed that in real life, the voltages around the first stage are not spot on. could it be beacause of the unbypassed cathode resistor?

the distortion figures are probobly something else from my loadline calculations...

the voltages in the last stage are perfect.
 
There is something wrong with the arithmetic on your schematic. If the cathode resistor is 820 ohms and it has 1V across it the the current through it is 1.22mA. The plate current cannot be bigger than this.

Cheers

Ian
 
Ian you are right!

I had choosen a cathode r of 750 at first but measuring with a dmm i got 0.85v at cathode and around 90v at plate...

So i threw in a slightly larger r to
get up t 1v.

I havnt figured out whats going on there...
 
5v333 said:
Ian you are right!

I had choosen a cathode r of 750 at first but measuring with a dmm i got 0.85v at cathode and around 90v at plate...

So i threw in a slightly larger r to
get up t 1v.

I havnt figured out whats going on there...

The data sheets for tubes are for what used to be called a 'bogey tube' which I guess means an average or median one. It is not unusual to see a 10% variation between datasheet voltages and measured voltages. Also, tubes are very high impedance devices so it is only too easy to shift the operating point owing to the the impedance of the measuring instrument.

Cheers

Ian
 
In case it helps, I've found this site very helping for spit-balling triode biasing/distortion calculations:

Triode Loadline Calculator

You pick: supply voltage, set maximum current (to define a load line), then slide the Q point and input swing, and it will tell you the plate resistor, cathode resistor, the gain, and the 2nd, 3rd, and 4th harmonic distortion, and estimated THD.
 
RtSpect is a free spectrum analyser for PC and though very simple you'll be able to see what harmonics your triode stage actually has. Obviously you'll need to make up a "probe" to plug into your PC with a HV cap to stop the HT going into your PC's input with some form of attenuation too.

Andy.
 
Ian
Seems that what you are saying leeds us to the discussions about reliability in tubes new and old etc.

Hmmm.. need to considder this and see what could be done.

Am i right with the freq knee point in the feedback network by the way? Im aming for sub 1hz.

 
5v333 said:
Ian
Seems that what you are saying leeds us to the discussions about reliability in tubes new and old etc.

Hmmm.. need to considder this and see what could be done.

Am i right with the freq knee point in the feedback network by the way? Im aiming for sub 1hz.

Personally I do not use NOS tubes. I have found many to be more inconsistent than current production. For small signal tubes I have seen no evidence that current production are any less reliable than that the original designs were when they were made. However, I have had lots of leaky and noisy NOS tubes.

Power tubes are a whole different ball game.

Your feedback network pole is well below 1Hz which should work well with the 50Hz one between the stages.

Cheers

Ian

Edit: if you have a fixed gain you might like to think about closing the loop at dc.
 
50hz??? i thought i made that 13hz..!  :eek:
in open loop i think the 12ay7 out impedance is about 37.8K.  grid to ground resistor is 330K so, isnt the pole at:
1/(2*pi*(37.8K+330K) * 33nF) =13.1hz

close it at dc? does that mean omitting the blocking cap in the feedback?
theres a thought in my head to insert a fine gain calibration like +- 3db for left/right balancing, but for now im glad if i can pull it together and use it for some mastering duties.
 
5v333 said:
50hz??? i thought i made that 13hz..!  :eek:
in open loop i think the 12ay7 out impedance is about 37.8K.  grid to ground resistor is 330K so, isnt the pole at:
1/(2*pi*(37.8K+330K) * 33nF) =13.1hz
I am sure you are right; I just did it quickly in my head 10^9/6*330K*33 and got 50!
close it at dc? does that mean omitting the blocking cap in the feedback?
theres a thought in my head to insert a fine gain calibration like +- 3db for left/right balancing, but for now im glad if i can pull it together and use it for some mastering duties.

It is important to get the two low frequency poles at least a decade apart in order to maintain stability. That is one reason why it is hard to make a tube preamp with a wide range of gains by altering the NFB because you alter one of the the poles at the same time. Closing the loop at dc eliminates one pole and makes it unconditionally stable at LF. It does mean omitting the blocking cap and also rejigging the bias to account for the dc flowing in the NFB network.  Here is how I do it in the EZTubeMixer project:

https://drive.google.com/file/d/0B_n67A1hN3qtelg2R21Xa0dIbVk/view?usp=sharing

Cheers

Ian
 
oh thats great  ;D phuw...

trying to understand your circuit:

the top triode in the output doesnt let any HT voltage to get in contact with the feedback junction. the only voltage there is the bias voltage of the output stage. right?
so i would guess if you are aiming for 1v - 1.5v bias voltage at the first stages cathode and what looks like a 50% ratio voltage divider across feedback and ground, the output tubes yields a bias voltage of 2-4V across the 680r resistors.
and there by using part of the biasvoltage from the output stage in the first stage. no need for a cap.
 
The output stage is an SRPP. The output dc voltage is almost exactly half the supply volts (¬300V)/2)

http://www.valvewizard.co.uk/SRPP_Blencowe.pdf

The 47k feedback resistor to the 12AX7 cathode takes a small current because the 12AX7 cathode is at about 90V. This current plus the 12AX7 cathode current flows through the 470R resistor that sets the 12AX7 bias. The 47K to ground carries the current (just under 2mA) and sets the nominal closed loop gain to 6dB. Gain is increased by bypassing this at ac.

This circuit was derived from an earlier design where I first tried to close the loop at dc in a fixed gain headphones amp:

https://drive.google.com/file/d/0B_n67A1hN3qtTnFBRlFrWXFnQ1k/view?usp=sharing

Probably easier to understand. Again, the output should sit at half the supply volts but the 47K feedback resistor pulls it down somewhat (hence raising the 12AX7 cathode to 90V in the later circuit).

The basic idea was inspired by the Pultec MB-1 mic pre:

http://www.geocities.ws/rafafreddy/DIY/pulteq-mb1/schem.jpg

Cheers

Ian

Cheers

Ian
 

Latest posts

Back
Top