KT88 PA build

[shooter]

Ok, starting my “build/mod”, the links below describes how I got here;
http://el34world.com/Forum/index.php?topic=21311.0
http://el34world.com/Forum/index.php?topic=21424.0

Working schematic below along with screen shots.
The data in the graph is “relative”, scaled for easier display
Vin and V1aPlate do what I expected, V1 b however!;

Is the output of V1b the result of Impedance/loading mis-match? (yellow)
And the kt88 is just amplifying what’s coming in?

The sin waves look clean, symmetrical throughout.  The spectrum is pretty clean, higher HD at lower freq, but I’m guessing that has to do with the amplitude drop-off at higher freq.

Once I get a flatter freq response, I’ll start playing with KT88 plate load to “optimize” output power.
 



 

[shooter]

I pulled the 88, no change in freq response, swapped 6SN7s, no change, Changed KT88 G1 R from 220K to 470k, no change.  back to head scratchin

[PRR]

There's no scale on the curve.

But I would look for a 10Meg snuck in where you think you have 10K grid stoppers.

[shooter]

I checked for strays, found none.  Did find if you look at input to KT I had the 15k stopper in the wrong place, fixed.

I "re-scaled" all the values to get them to plot reasonably
Here's the raw data;

               800hz   2.8K     8K       15K
IN             1.00    .96       1.02     .96
V1aPlate   8.80    8.64      9.28     8.24
V1bPlate   44.20  38.80   31.60   18.80
KT88Plate   400     368      312      208

All values Volts pp

I changed the CC from V1b to 88 grid just because, no real change

[PRR]

> V1aPlate 8.8
> V1bPlate 44.2

V1a plate to V1b plate shows V1b gain of 5

We expect a gain of 15.

You have the pot turned down, don't you??

With wiper at 0.3 of total resistance, wiper impedance is 210K. -3dB at 8KHz. 95pFd against 210K is 8KHz.

But low-gain 6SN7 should not have that much pFd in its grid.

Is there some shielded wire here?

In General-- you can get away with 1Meg volume pots in guitar but not in wide-band gear. 250K is borderline for >15KHz flatness. Your V1a can easily drive 250K.

I'm not inclined to look at KT88 plate because this will depend directly on load impedance vs frequency. It may be near-flat on dummy load; it won't be close with a loudspeaker, even in UL. I don't know what your goal is (just re-invent a wheel?), but most "good" amps run NFB from OT back to the driver.

[shooter]

V1a plate to V1b plate shows V1b gain of 5

The 1st stage is un-bypassed because I don't need the gain

You have the pot turned down, don't you??

Yessir, I adjusted it for max clean symmetrical sine wave out of 2nd gain stage, n you're close, it's about 80/20 split on the 1M
I will swap it for a 250K.

Is there some shielded wire here?

Nope, if it gets to build from breadboard will be

(just re-invent a wheel?),

I don't think there are any new wheels to be invented
I just have parts ageing like me, I want a "blackbox" audio amp that you set your streaming whatever on and music comes out the speakers.  I also want to sharpen "you get this because of that" skills

I will/should have NFB in the final, but for now I'm just trying to get a flat response PRE.  The KT is just on the bench for now since it didn't seem to matter in or out.

 

[shooter]

I checked 80Hz just because, and I suspected was a smidge better than 800hz.

I ||'ed a 220K across the pot, nothing really changed relative, all volts did get a bump.

I'm gonna change the bypass cap 1st, then if nothing changes,  re-config the 1st gain stage to fully bypassed, 2nd gain stage to something closer to a PI, probably start with true CF, then start splitting Rk and Rp till I get to 50/50 and see.

[jjasilli]

I"m shooting in the dark.  But if there's trouble I would first simplify the circuit:


*  temporarily remove both grid stoppers
*  temporarily remove the bypass cap from V1b
*  The 100uF filter cap to the KT88 screen seems excessive
*  The 47uf & 100uF filter caps to V1b&a seem excessive.  try 8 - 20uF
*  See the tube Chart p. 3: http://www.mif.pg.gda.pl/homepages/frank/sheets/093/6/6SN7GTB.pdf  The chart calls for a Supply Voltage (E_bb) as low as 90V - 300V.  You show 201V & 370V.  You might want to tame supply voltage.
*  Also note the Chart value for R_s (your 1M pot) is also 100K, etc., but much lower than 1M.

[Ed_Chambley]

I"m shooting ion the dark.  But if there's trouble I would first simplify the circuit:


*  temporarily remove both grid stoppers
*  temporarily remove the bypass cap from V1b
*  The 100uF filter cap to the KT88 screen seems excessive
*  The 47uf & 100uF filter caps to V1b&a seem excessive.  try 8 - 20uF
*  See the tube Chart p. 3: http://www.mif.pg.gda.pl/homepages/frank/sheets/093/6/6SN7GTB.pdf  The chart calls for a Supply Voltage (E_bb) as low as 90V - 300V.  You show 201V & 370V.  You might want to tame supply voltage.
*  Also note the Chart value for R_s (your 1M pot) is also 100K, etc., but much lower than 1M.

Following along trying to learn.  I am amazed I understand most of this.

Would you mind sharing why you consider the 100uf excessive for the KT88?  And even though every design I have looked at has the 6sn7 caps around 8-20uf and I simply follow, is there a reason to reduce farads for the 6sn7?

[jjasilli]

I too wish I knew more about theory. So I try to stick to established circuit design, then mod from there.  (E.g., one of my "brilliant" mod ideas caused motorboating, so I just restored the circuit.  If I had built the circuit wrong to begin with, tracing out the motorboating issue would have been a major PIA.)


Shooter's signal strength is falling off.  I believe (from my memory which is always faulty) that excessive gain can actually cause volume drop.  Here, signal strength is also falling-off as frequency increases. So there might be a frequency related issue.  That makes me think caps.  My suggestions have one or both of these things in mind. 


Also excessive B+ filtering can lead to signal issues.  As HotBlue reminds us in a current, nearby thread, B+ filter caps are actually IN the signal path. 


Hope that helps.  Can explore further if desired.

[shooter]

Thanks for the suggestions;

Currently the KT88 is out and has zero effect on sig loss/freq.
I've looked both in time and frequency out to 250khz and besides HD, which is at least in the "normal" guitar range, there is nothing above the "grass", which is 20 - 50mV

I added a 2nd filter stage to separate 1st gain from 2nd, that didn't help.
the gain stage volts at plate are 123 and 136.  I was actually thinking of boosting the tubes current, IIRC I is only 1mA or maybe less, thinking about getting it up to 2 - 3 in stage 2, since 1st gain stage works just fine.

Hopefully I will get the bypass cap swapped but it was a long work day and I hear my green beer n painting calling!

[PRR]

> I ||'ed a 220K across the pot, nothing really changed

That does not change the impedance *at the wiper*.

But the voltage should not "bump" either.

[jjasilli]

the gain stage volts at plate are 123 and 136.


With a supply voltage of 300V and 100K plate resistors, the tube chart calls for:
* cathode resistors of 1K8, not 3K2 -- you should check your load lines -- and
* 100K grid leak resistors in the following stages.  You have 1M (pot) and 220K (KT88), which deviate from the recommended design.

BTW: I forgot -- always check tubes first.  Maybe the tube is faulty.

[shooter]

That does not change the impedance *at the wiper*. 

Ah, I thought the pot was in || with the plate load n the plate load seen whatever R was dialed on the pot, 180k vs 1M

And "the bump" came from not having the kt88 in, B+ at the pre-plates went up 20ish volts.

220K (KT88),

The data sheet I have shows 470K max for under 32W cathode biased

1K8, not 3K2

The datasheet I have, shows with Ebb between 90 and 180, Rp at 100k, Rk works from about 2.8k to 3.4k
The grid leak at 100k.  I choose them based on my "expected" gain needs, then observed gain and distortion and found, it worked!!!  the bypassed stage gave me a gain of 13ish, the un-bypassed around 6ish, clean, symmetrical sine

I swapped the bypass cap, no love, but noticed later, to late to play with HV, I only swapped 1 of the CouplingC's
doing that next.  I have ran 4 tubes through, all about same. 

[jjasilli]

The data sheet I have shows 470K max for under 32W cathode biased
That's not the point.  The point is that 220K resistor affects the AC operation of V1b, and that resistor, among other things, is out of spec.


A 6SN7 can drive a KT88:  https://www.youtube.com/watch?v=VRsDAzpOqnc 

I think there is an undo attachment to a circuit design that isn't working.

[shooter]

A 6SN7 can drive a KT88:

absolutely, I've done 2 successful builds with 6sn7 n kt88, but they were guitar based and if they fell off >8K, I didn't check, and the 3 owners so far didn't care/ notice.

this one is going to get a re-boot, probably Friday.
I started with just 1 gain stage, got it clean, got it flat across the spectrum, just not enough gain, so I "mirrored" the 1st gain stage and that's when things went south.

The point is that 220K resistor affects the AC operation of V1b

Yup, I originally had a 100k, found that by going to 220k, I got a gain boost from just the single gain stage.
left it there when I added the 2nd stage, changed it to 470k simply as a "what if I do this", but it didn't effect the freq drop-off, or gain much 220k or 470k, (didn't go back to 100k).

I'm betting on the CC, or B+ tap change,  and if that doesn't get it, i'll simply borrow someone's working design, (I have 3-4 schematics), once that works, I'll go back and find why mine didn't, maybe .

[Willabe]

For all the new/young guys;

The utube vid in reply #14 and below is a VERY BAD and DANGEROUS idea!!!!!! DON'T DO THIS!!!!!!!!

http://www.youtube.com/watch?v=VRsDAzpOqnc

[drgonzonm]

To me, this amp is being rubegoldberged. 
Just curious, what circuit was the basis for the amp design, and could you post the basis circuit?

1.  I would get the amp operating at 400 Hz, (range of the human voice, also typical range for PA systems, see Crowhurst's work), then determine what's happening at the other frequencies. 
2.  The bypass cap on the KT88, may be high at 100 microfarads, most of the SE's I"ve seen have 50 microfarads.
3.  I would consider evaluating each triode, and the pentode/triode using RDH4 approaches, and GE design approach documents.   Valvewizard, may also provide some evaluation methods. 

I found a discussion using Lundahl, O/T's in a some what similar power amp setup.  The design used 50 microfarad caps on the KT88, cathode. 

I'll be the first to admit, the guys providing answers to your questions like you have a ton or megatons more experience than I do.   

[Ed_Chambley]

For all the new/young guys;

The utube vid in reply #14 and below is a VERY BAD and DANGEROUS idea!!!!!! DON'T DO THIS!!!!!!!!

http://www.youtube.com/watch?v=VRsDAzpOqnc

Absolutely, I do not think this is a very safe setup, but it does prove the 6sn7 is driving the output.  If you were to reach and pick it up by the 2 buss wires, you would be the new fuse to ground.

[Ed_Chambley]

I too wish I knew more about theory. So I try to stick to established circuit design, then mod from there.  (E.g., one of my "brilliant" mod ideas caused motorboating, so I just restored the circuit.  If I had built the circuit wrong to begin with, tracing out the motorboating issue would have been a major PIA.)


Shooter's signal strength is falling off.  I believe (from my memory which is always faulty) that excessive gain can actually cause volume drop.  Here, signal strength is also falling-off as frequency increases. So there might be a frequency related issue.  That makes me think caps.  My suggestions have one or both of these things in mind. 


Also excessive B+ filtering can lead to signal issues.  As HotBlue reminds us in a current, nearby thread, B+ filter caps are actually IN the signal path. 


Hope that helps.  Can explore further if desired.

Thanks JJ, that is all for now.  It did help, thanks.  I do recall reading HBP comments and have been reading about this topic.  Seems the more I read, the more important tube data becomes.


I will watch Shooter and PRR.  I to come up with brilliant ideas.  
I am reminded of a build where put in a mod I do all the time.  That amp is now completely apart and ready for a rebuild.  This time I think I will stick to the good advice I did not listen to which is build it, then mod it.

[Willabe]

If you were to reach and pick it up by the 2 buss wires, you would be the new fuse to ground.

 

[PRR]

I think you are riding off without direction.

Volume pots for wide-range audio can NOT be 1Meg. Loading the input side does not change the wiper impedance. You need to get 100K-250K pot.

You say "should have NFB in the final". I know about making the amp good BEFORE adding NFB. But it is kinda-good (no major fault) already. And you have to trade-off a lot of gain to make it better with NFB. And you do not have a lot of gain there.

How are you going to apply NFB to this thing? The most obvious way is a voltage-divider from speaker back to V1b cathode. But you have a cap here! There's ways around this (see AA-Champ). But simplest is to lose that cap and work with V1b gain near 8.

If '88 needs 24Vpk drive, then V1b needs 3Vpk before NFB. With a mere 12dB (4:1) NFB you need 12Vpk at V1b grid. OK, V1a can do that. But V1a flat-out won't deliver much more than 20Vpk from a 1Vrms input. So the most NFB available with these tubes is about 7:1 or 17dB. OK, that's maybe more than you can handle with a non-fancy OT. And you probably WILL need some treble roll-off inside the NFB loop to avoid trouble with OT top-resonance.

[shooter]

But it is kinda-good (no major fault) already

With the exception of a near linear drop in volts per frequency, kneeing at about 4k.

wide-range audio can NOT be 1Meg

I have both 100, 250k, I'll wire up a 100k.

rubegoldberged.

Yup!  here's the "engineering:
had Iron, tubes, desire
decided on pre tube, looked at available B+, looked at datasheet, calc'd required gain, selected the parts from the sheet, build the pre and got what I expected - gain wise.  I was happy.
Added a pot (wrong one for "gain control')
decided to check freq spectrum since it's an audio amp, and here we are.

I have plenty of other ppl's builds as fall back, hell, I have a Marantz 7 preamp board and a 60W SS PA, but where's the fun it that

 

[shooter]

 
It's bedtime, here's what I have
Did the following
1.   Changed, 1st stage output CC, Changed gain pot to 100k, moved both gain stages B+ to the end of the rail.

Results, Happy Place!!, back on track!

Quick data;
800hz  G1 V1b  2.48Vac pp,  At Kt88 plate  404Vac pp   Gain  162.9

15Khz      “   “    2.1Vac pp.       “        “         396Vac pp   Gain  188.6

Plate volts  110

2 on the gain pot is where the clean signal stops outta the KT.  I’ll “tweak” that

[PRR]

> 2 on the gain pot is where the clean signal stops outta the KT.  I’ll “tweak” that

For an assumed (artificial) input signal, and withOUT the NFB that you will probably want.

I'd move on to the NFB. Taking 4:1 or so of gain reduction in NFB will get your input sensitivity near nominal.

[shooter]

I'd move on to the NFB. Taking 4:1 or so of gain reduction in NFB

Thank you for getting me to gain reduction

I would like to "tweak" gain, max clean, by experimenting with "optimal" Plate load(resistive DL) @KT88. Then;

I'm guessing true global NFB at V1a Cathode since it's un-bypassed, I can split Rk for an injection point.  That puts the gain knob inside, so not sure on that.

[PRR]

> NFB at V1a Cathode

Taking NFB around 3 bass-cuts (V1a-V1b, V1b-V2, OT) asks for instability.

Also gain control inside a NFB loop screws things up.

Look at AA-Champ. NFB to your V1b. Either unbypass that cathode, or use AA-Champ's scheme of bypassed bias stacked on a smaller NFB resistor.

[shooter]

NFB to your V1b

Than you kind sir
It was boring at beer-bait, n gas, so my brain re-arranged the gain stages, so I will probably bypass V1a, un-bypass n tap for NFB.  I will look at the AA champ

[shooter]

I updated the schematic (post #1) with NFB sketched in

Would I be better off controlling gain at input instead of ‘tween stages.  The reason, I can only get to 2 on gain knob

I wired up the NFB, tried a couple different “series” values, nothing did much for “extending” the max clean sine out of the KT88 (started with 15k, ending with 4.7k)

Taking 4:1 or so of gain reduction in NFB

Are you talking signal, or Resistor voltage divider?

From what I’ve seen of distorted happy guitar sound, image S2A won’t sound very happy? 

Just observations;
 @ 1k 10ohms gets me a smidge more gain, 9.6 @ 4ohm, 11.5 @ 10ohm,  need to see the extremes to confirm
Vk drops from 24.3V S1 image, to 19.3V S2 image.

[PRR]

> controlling gain at input instead of ‘tween

Same either way. Get a hot iron and try it.

> NFB..., nothing did much ...(started with 15k, ending with 4.7k)

Why did you stop before you got anywhere?

You should have gone down until the gain *dropped*.

You have "no" NFB.

Math? V1b has gain near 15. KT88 let's pencil gain of 10 to the plate. OT is say 625:1 impedance or 25:1 voltage. Total gain from V1b input to loudspeaker, about 6.

To get "a little" NFB, the resistor ratio should be about 6. This works for guitars where we want "slight" speaker damping.

"Good" NFB would be a gain-drop of about 10. So to 0.6! You can't get there with a simple divider. (You could be close if you started from a 16 Ohm tap and strapped it back to V1b cathode circuit.)

Gain reduction of 10 is really pretty heavy for tube circuits without special transformers.

I would change your 4.7K to 100 Ohms. Overall gain of 2 is 3:1 of NFB and will be "some". I might even go to direct connect for near unity-gain around V1b and NFB factor near 6.

Then your vol-pot will have to be turned 3 to 6 times higher, which may put you at a happy spot.

[PRR]

> Vk drops from 24.3V S1 image, to 19.3V S2 image.

From clean to clipped?

Then your idle current is too high or your load impedance is too high. Ideally there is no change of Ik from clean to clipped. In reality because of tube curvature we tend to arrive at a small rise from clean to clipped. Say 10%-10%.

[shooter]

Why did you stop before you got anywhere?

Can I still plead ignorance or have I been here to long for that excuse

I'll try a straight wire, then work up if need be.

From clean to clipped?

Yessir, I kinda thought it was more than I've observed on other builds.  It stays about the same either 4 or 10ohm, (will double-check though)

Won't Ik drop though as I drive G1 to cut-off?

 

[shooter]

  I left off the ground wire from spkr side

Moving forward;
  , I have a 4:1 NFB, my knob will not distort the 88's output on 9!!!!
I'll get No#s soon, but appears from quick glimpse, same Vacpp at plate as fully distorted earlier, but clean!
also noticed a minor distortion in pre at 10, so I think it's moving from Beta to Version 1

Again, thank you kind Sirs.
 

[shooter]

I plugged in a guitar/organ speaker, an MP3 player and NICE!, I still need to plot out Freq response and hook up a full range speaker.  appears the NFB eats up some highs, so I'll be checking freq response with, without  NFB and see.

[shooter]

For those with insomnia, I uploaded an .XLS file of data, otherwise;
(NFB in, All #'s in rms)

The amp is flat from 100hz to 15Khz, check.
The Audio power is 4Wrms at DL!  (32Wdc Idle), not checked!

I’m at 156Vac rms on plate,  plate DC at 303Vdc, Vk 23.4Vdc
PreOut is 12.5Vac rms with NFB, without NFB the 88 still clips with about the same drive out of Pre.  Put in a 6550 and get about the same results.
Input signal at .3Vacrms (1Vac pp)

I changed DL from 10ohm to 4ohm, I lose about 8 – 10Vac at plate at 4ohms.

Gonna raise input signal and verify the pre has more clean than the PA can use.

Unless there is some circuit piece I’m missing, I’m gonna change PS and shoot for the same 32W dc Idle, at 400ish VDC plate and see If AC power goes up.
But the breadboard is set up perfect for testing a pile of small bottle PA tubes, so……