harmonic distortion figures in triode stage

[5v333]

i clearly has some reading up to do!!! thanks for the links!

back to the main question.

figures!

do you guys aim for any specific harmonic distortion figures when designing on paper? are there any general thresholds? what is tolerable, what is painfull what is clean? im talking about studio stuff now, not guitar amps and such..

 

[ruffrecords]

do you guys aim for any specific harmonic distortion figures when designing on paper? are there any general thresholds? what is tolerable, what is painfull what is clean? im talking about studio stuff now, not guitar amps and such..

The first thing to remember is that distortion in triodes is directly proportional to signal level. So if you have, say,  40dB of open loop gain and a +20dBu output, the signal at the grid/cathode of the first stage is -20dBu or a bout 75mV. Bottom line is you can usually ignore the distortion in the first stage as it is likely to be alt least an order of magnitude below that of the output stage. So, rule one, concentrate on output stage distortion.

In my early days of designing tube mic pres I avoided NFB for all the usual reasons and looked for topologies that gave cleanish gain without it. I settled on the mu follower and with something like a 6SN,7 or its B9A clone the 6CG7, you could get 20V rms into a 10K load at about 0.4% distortion, nearly all second harmonic. The only trouble with this design was it could not drive a 600 ohm load.

Later I moved to the SRPP stage which has four times the drive capability of the mu follower for the same quiescent current but the price you pay is about 5 times the distortion; typically about 2% at 20V rms but still  mostly second harmonic . However, 20dB of NFB drops this to 0.2% and lowers the output impedance so through a 2:1 transformer wit will drive over +20dBu into 600 ohms.

I would suggest your push pull design could aim for similar figures. If you look at the specs for ancient tube mic pres and consoles they often spec the maximum output level for 1% distortion or 0.5% but in the absence of NFB expect it to be significantly higher.

Cheers

Ian

 

[5v333]

now we´r talking!!!

your giving me some specific ballpark figures here! just what i needed!
i also like the style of finding low distortion in openloop. NFB seems tricky but i think ive read that NFB is good for stability and noise suppression so a little bit of that seems desirable aswell but perhaps only under good conditions such as low distortion in openloop.

id like to hear more personal views like ians on figures when designing!

this is just on paper but with the HD figures in the output stage i get from my ac loadline at full amplitude class a(p-p 90V ) is about 2nd 2,54% 3d -0.44% and 4th 0.02%.
if i take thoose figures and do a thd sum i get about 0.017%... and then the pushpull transformer cancels the even harmonics. this, i guess, is tolarable..! but maybe im doing something wrong...

for now im throwing away about 10dbs with global NFB. more feedback seems to degrade the bandwith of the output severly and i get bumps in the very low end. band width is about 3hz -115khz. theese lundahl tube transfomers are not to mess with 

OT voltage ratio is 9:1 right now.

doing alot of loadlines and compare distortion figures seems to pay off!

 

[ruffrecords]

9:1 seems a little on the high side for the OP transformer ratio. A 600 ohm load will look like 48K on the primary - no load at all for the tubes and you will need 18V rms to get 2V rms out.  4:1 would probably be a better bet.

Cheers

Ian

 

[5v333]

you are right about that!

im thinking about terminating the output with a ≈200R resistor. the output will mostly see 20K to 50K inputs (AD converters or other stuff)

ill have the resistor hooked up to a realy so i can disconect it if low impedance input would be present.

i had it 4.5 : 1 at first but i sit here with too much gain. and the feedback approach didnt satisfy me so i reconfigured it. completly different sound! less tuby color more swift and transparent like a still ocean.

im gonna see if i can down configure the input transformer cause im about 6-8db over unity.

thanks again!

 

[ruffrecords]

Why do you want to load the otuput with 200 ohms? After output level, the most significant factor in the amount of distortion you get is the value of the load. I can see no reason to load it with less than 600 ohms.

Cheers

Ian

 

[5v333]

hhmmmmmm...

maybe im mixing up stuff again... reflected load/ plate impedance...

when doing resistive coupled stages we think of loads as 2-5 times the previous resistor or Ri...
is there a similar relationship with transformer coupled stages. what are we aiming for?

will a higher load resistor affect the output impedance of the plates?

 

[ruffrecords]

hhmmmmmm...

maybe im mixing up stuff again... reflected load/ plate impedance...

when doing resistive coupled stages we think of loads as 2-5 times the previous resistor or Ri...
is there a similar relationship with transformer coupled stages. what are we aiming for?

will a higher load resistor affect the output impedance of the plates?

In the early stages of an amplifier you are trying to get as much voltage gain as possible. You are not interested in power so you maximise gain by making the load as light as possible. In the output stage you are interested in delivering power to a load. The power delivered to the load is V^2/R where V is the output stage swing and R is the load seen by that stage. Clearly the smaller R is the bigger the power will be. Just to put this into context, +26dBm into a 600 ohm load is 400mW. By contrast, your first stage swinging 5V into its 300K load is just  83 microwatts.

The maximum power transfer theorem states that a source with a given internal resistance of R will deliver the maximum power into a load equal to R. For a triode, the internal resistance is the plate resistance and for a 6SN7 it is about 8000 ohms at the current you are running. To match a 600 ohm load the turns ratio needs to be around (8000/600)^0.5 = 3.65. 4:1 is close enough.

There is a lot more to it that that but the above is enough to get you in the right ball park.

Cheers

Ian

 

[5v333]

this is how i looked at it.

plate R is 8000, two plates = 16000

OT ratio 9:1

16000/9^2 = 197R or about 200R

 

[PRR]

> The maximum power transfer theorem....

That works for DC, and for AC, but DC+AC leads to four different situations.

For transformer loading:

If gain is short, RL=Rp is maximum output for limited drive. If you now find more drive, THD tends to approach 10% just below clipping.

If drive is available, RL=2*Rp is approximately maximum power. THD tends to approach 5% just below clipping.

If power is ample, RL= 3 to 5 times Rp leads to lower THD.

Given free choice you "never" want to add load, it hurts power and distortion. Even if you want distortion, the reduced max power means no real advantage.

(I forget the 4th case; probably an odd corner we never go to.)

 

[ruffrecords]

this is how i looked at it.

plate R is 8000, two plates = 16000

OT ratio 9:1

16000/9^2 = 197R or about 200R

I think each plate sees 4.5:1 so it is 8000/4.5^2 or about 400R

As PRR has demonstrated, there is a lot more to it.

Cheers

Ian

 

[5v333]

i have been figuring about the 8000/4.5^2.

what i did... was to use the total ratio and divide by 2.
while you are dividing the impedance ratio by two instead as we are talking about resistance, not voltage.

my way was more like a voltage-ratio-approach while you are prob stating the more correct impedance-ratio-approach!
is this correct?

also we are talking calss a here right?

the main thing must be to know ex what each plate sees so that harmonic distortion and impedance (∆V/∆mA) and gain can be derived from a loadline plot. yes/no?

loadlines are cool!

 

[ruffrecords]

The important thing, as you say, is to know what each individual anode sees. It is easy to get confused with push pull transformers; I know I do. Just remember that as far as the signal is concerned, the output transformer centre tap is at 0V,

Your design is more a balanced class A amplifier than a push pull. One pair of tubes does the hot and the other pair the cold.. You could remove the top or bottom pair completely and it would still work (with half the gain).

Cheers

Ian

 

[5v333]

hehe yeah trying to understand theese stuff is like solving a rubix cube for me. turn one row and suddenly other stuff doesnt add up but im getting there. slowly...

but to be sure now. if i put a load resistor of 400R on my 9:1  OT.
each plate will see an ac load of 400*4.5^2 = 8100R


another thing about harmonic figures. i know even harmonics cancels out in the transformer.
but the stated 2% 2nd harmonics from my plot, can it - even if canceled - contribute to intermodulation with other harmonics for ex?
if so i guess i should try to find a plot that reduces even harmonics.

 

[5v333]

just noticed on the output that the lowend -3db point is around 23hz when the secondary winding is loaded with about 330R. and not about 12.5Hz as was expected...

OT ratio = 9:1
prim inductance = 200H
Lf -3dB = 23Hz

this suggests the impedance from the plates are 28900R and not around 15K.

wouldnt this say that the AC load of each tube is (330 x 9^2)/2 = 13365

im not sure what route to go...

either i keep the load on the secondary with about 330R, but then i have to change the interstage cap between the 12ay7 and 6sn7 from 33nF to 10nF. (i get low freq resonance inside the amp when the feedback resistor is reduced till a certain point, not good for the ears)

or i reduce the 330R resistor so it looks like 15K on the primary and the lowend point is abot 12.5Hz again.


any more free advice and explaination about the load figures are very appreciated.

 

[DaveP]

Best advice now, is to stop the paper work and start the bench work.

You will need a good sig gen and a scope, then you can see the waveform you've designed, listen to it if possible.

I have read that most people cannot hear distortion until it reaches ~ 2%, below that is where people hear "Tone".

Best
DaveP

 

[5v333]

thanks Dave!!

guess what. this morning i turned on my build projekt, oscilloscope and funk generator, and it turns out that my scope doesnt wanna play anymore... crap!!!
im on the hunt for a new and better.

i have a tektronix 500 mainframe with a generator plugin that goes to 40mhz and does some trix if needed. its not very modern but i like it and theres schematics and calibration proceedures avaliable.

also i have two mic cables to and from my ADDA converters and speakers so i can just plug it in and flip some mute buttons on my computer and the sound from the computer goes through the circuit and back for further routing and listening.

i found some in RDH about distortions and it talks about tolerable, objectional and percievable amounts of distortions. also that it depends on the bandwidth. more bandwidth calls for less distortions. i guess its the same with a lot of stuff, like jitter for ex.