Small Line Amp (Tube)

[edanderson]

cap coupling into a 10k resistor, the THD plot is similar, but higher, with 100Hz to 10KHz at about 15%. below 100Hz it rises as before. above 10KHz there is a smooth rise without the peak around 40KHz seen in the transformer output configuration. there is a small dip centered around 400Hz.

for reference, this is what i have built up:

i notice that my voltages are a bit higher than indicated on the schematic.

ed

 

[NewYorkDave]

...and what's the residual THD+N of the signal generator itself?

 

[edanderson]

[quote author="newyorkdave"]...and what's the residual THD+N of the signal generator itself?[/quote]

at this bench, a loopback cable gives 0.0029%.

guess what?

bad tube.

the one i had been using i pulled from the newyorkdave 12av7/12bh7 lineamp, which worked when i put it away. goodbye international servicemaster, and hello GE. now the THD is below 2% in the resistor loaded configuration, and the distortion is all clean looking 2nd harmonic.

thanks for all the diagnostic pointers guys. i'm sure they'll come in handy some other time when i've actually tested the tube i'm using before plugging it in.

ed

 

[NewYorkDave]

Excellent. That's the amount/type of distortion you should be getting at "normal" operating levels and no feedback.

It was an interesting troubleshooting exercise, anyway.

Something interesting to try: increase the cathode resistor to 510 ohms and series-feed the transformer. I know the xfmr is only rated for 8mA, but it might not suck too bad run a couple of mA above that. See if the LF response/distortion goes into the toilet or not, check saturation level, etc. If it sucks, back to shunt feed...

 

[edanderson]

[quote author="NewYorkDave"]Something interesting to try: increase the cathode resistor to 510 ohms and series-feed the transformer.[/quote]

only too happy to oblige. it does work; THD is now below 1% across most of the audio band, and i get about another 1db of gain. however, below 200Hz THD starts to rise, and by 30Hz it has crossed over the distortion figure for the shunt fed configuration. down at 10Hz, you end up with 4% THD in series fed instead of 3% with the shunt fed configuration. the output level also rolls off starting at 200Hz, and is -1db at 40Hz and -3db at about 20Hz. so, yes, it is "worse" in the low end, but maybe for some people the benefit across most of the audio band would be worth the extra distortion below 30Hz and weaker low end. the droop may be more pronounced at higher signal levels (this was at -10dbu input as before).

while i have all these plots up on the screen, i should also mention that there is an odd dip in the frequency response at 40KHz. the smooth looking rolloff is interrupted by a dip that juts down about 4db and then pops back up nearly 3db to continue the smooth slope downward. any ideas on what could be causing that?

using one 12AV7 this way makes about 12db of gain (with a 5:1 output transformer). i think i'll try building up the "one bottle" 12AV7 pre next, and see if i can stand what i assume is reduced output drive capability in return for a bit more gain.

thanks for providing an interesting circuit to experiment with dave.

ed

 

[PRR]

> down at 10Hz, you end up with 4% THD

There should not be any 10Hz sources in a classic studio; if there were, the distortion and smoke in transmitter or disc-cutter would mask any distortion in the studio. (Heck, we didn't have any 10Hz LAB-sources when those transformers were introduced.)

> output level also rolls off starting at 200Hz, and is -1db at 40Hz and -3db at about 20Hz.

Perfectly acceptable in many classic studios. Not much music or voice has any real strength below 40Hz. If you really wanted 20Hz, there are better choices than the utility UTC cans.

However, I do think you are soaking the core. Do you need to?

Think. +20dBm is ~60V and 4mA peak on the 15K side. You "could" run the tube as cold as 4mA idle, swing from 8mA to 0mA, and make +20dBm. Swinging to zero current is hard and gives large distortion; but 8mA idle swinging 12mA and 4mA is a pretty clean swing. In light of the 8mA rating, I'd try idle currents like 5mA or 7mA and see what it does to midband THD at +20dBm and what it does to bass flatness.

Yes, +20dBm. +4dBm is a nominal average. Peaks will be 16dB higher. For good sound, THD should be very low at +4dBm and not-large (say several percent) at +20dBm. If it soft-clips at, say, +12dBm, you can't pass a "+4dBm" signal cleanly.

> an odd dip in the frequency response at 40KHz

There is an internal resonance in the winding. That's the drawback for using windings higher than roughly 10K. A lower plate resistance would reduce that, also improve bass response, but lower plate resistance normally comes with higher current, which is sure to droop the bass. NFB from the plate might help, but you have no place to put it. I'm sure the designer felt that -4dB an octave beyond the Audio Band was not a serious fault, or not worth the expense of a much more complicated winding technique.

 

[NewYorkDave]

Tamas kindly sent his custom transformer to me, since he didn't have time to play with it. I've just done some bench tests and I have some results to share.

First, the bad news: this transformer is not suitable for series feed with more than a couple of milliamps of DC through the primary. The primary inductance starts out huge at 0mADC and plummets from there. You could probably get away with up to 2 or 3mADC if the source feeding the primary were fairly low-Z.

Now here's the good news: this transformer tests superbly with 0mADC. Here are some measurements...

Turns ratio: 5:1
Primary DCR: 222 Ohms
Secondary DCR: 9.2 Ohms
Primary inductance (measured at 20Hz, +10dBU): 795H (!)
Frequency response (measured at +15dBU input level, source impedance 7K, load impedance 620 ohms):
0dB (and I mean flat) from 10Hz to 10kHz, -1dB at 18kHz, -2dB at 27kHz, -3dB at 35kHz.
Self-resonant frequency, measured as above: 200kHz.
Saturation level (20Hz), measured as above: I can't say. The generator I have here can only produce up to +17dBU at 20Hz, and that wasn't enough to saturate the transformer.

Initial impression: this transformer should kick some serious ass for a shunt-fed, AC-coupled application like the line amp posted earlier, or the One-Bottle pre, or a Pultec MB-1 or Altec 1566A (and so on). But you really need to go with a gapped transformer for series feed.

 

[tk@halmi]

Hi Dave,

Thanks for taking the time to test it out. For sixty bucks it is a bit expensive for a cap coupled transformer. The XSM series fits the bill for a fraction of the price. I am a little pissed because they ignored the requested spec.

Cheers,
Tamas

 

[PRR]

> not suitable for series feed with more than a couple of milliamps

Can you gap it? Pry off the mounting, pry the lams out, put all the Es in one side and put the Is on the other end? Tap it with a rubber mallet or trashy hardcover book to get the lams butt-flush, and put a few rubber bands around to hold it for testing. Try again with a sheet of copier paper in the gap. You have "too much" AC inductance and no DC stiffness; a small air-gap should put it in a better place for series work.

 

[NewYorkDave]

ignored the requested spec

I hope it was just an honest misunderstanding.

For what it's worth, I measured the primary inductance at 20Hz and 50Hz. Looking back at the sketch I posted, I probably should have specified the frequency at which the inductance would be measured. I don't know if it would have helped in this case but it's a good practice nonetheless.

There could also be a flaw in their testing method... or mine! Maybe it would be best to lash this up in an actual tube circuit before passing final judgement. I'll try to get to that in the coming week. I don't expect a big difference in results (since the xfmr is not gapped) but if there is, I'll be man enough to admit it :thumb:. But if it still flunks and you call Edcor about it, I'm sure they'll make good on it. (Hey, 60 bux is 60 bux! :wink.

Here's a thread with some detail about the instrument used to make the measurements:
http://www.groupdiy.com/index.php?topic=10493

 

[edanderson]

my experience with edcor has been very very positive, but in order to get what you want, you often have to be very specific. as nydave mentioned in the edcor transformer thread over in the lab, transformers for installed sound is their primary gig. getting them to deviate from their standard practices can be tricky, but in my experience, they always will try to make you happy.

if you test the transformer in circuit and discover that at the DC current required, saturation causes the primary inductance to fall too low, i'm willing to bet that returning the transformer and asking them to re-stack the lams will not be a problem.

ed

 

[NewYorkDave]

Still not lookin' too righteous...

Test setup:

 

[NewYorkDave]

OK, back to the schematic posted at the top of the thread. Over the past couple of days, during coffee breaks, I've lashed up a breadboard of this circuit. I finished it and powered it up this afternoon... Of course, it's misbehaving. But I'll be back with results once the debugging is done.

 

[NewYorkDave]

So, to continue my monologue... It's not unusual to have some weirdness when powering up a breadboard for the first time, usually due to sloppy layout, connections you forgot to make, etc. For instance, I'm getting a max undistorted output at 3 or 4dB less than it should be. There's probably a rational explanation for that--the tubes are "pulls" and one might be weak--but the other problem is truly weird.

With the gain pot turned up more than halfway and any kind of load--even a 10X scope probe or voltmeter probe--connected to the dry side of output coupling cap, the circuit would oscillate. It would do the same with no load and a finger touching the body of the cap, but not with a plastic probe touching the cap. With nothing touching the cap, and nothing connected to the dry side, the oscillation wouldn't happen even with the pot turned up full. Finally, I discovered that I could prevent the oscillation by adding a miniscule amount of capacitance (in the form of a short clip-lead) at the grid of V1A.

Yeah, this sounds like a typical case of some HF parasitic going on, but here's the weird part: it's not an HF oscillation as far as I can see on my 100MHz scope--it's motorboating!

Oh well... I'll get back to it tomorrow if I have the time. That damn "job" thing keeps taking time away from my tinkering.

 

[Larrchild]

Is it possible that the 470 uf is so large that it's charging slower than desired?

"Cathode Bypass Caps:
Larger values (22mf to 330mf) = More gain and bass. If ya got too much (oscillation and "motorboating",) try smaller (2.2uf, 1uf, .68uf, .1uf)= less gain and bass. "

http://www.sonicdeli.com/ThunderTweakWeb/tweaks.htm

Guitar Amp wisdom, but germaine.

 

[NewYorkDave]

Since the output stage is inside a feedback loop, I don't want any LF rolloff outside of the pole I deliberately provided with the .022uF/1Meg coupling (-3dB at 7.2Hz). The idea with the big Ck was to make the output stage flat down to the bottom of the bandwidth. But if the problem still eludes me after checking the more conventional causes, I'm not averse to trying different values of Ck just to see if that does anything.

Still, I have a feeling the cause is a bit more exotic, since adding a couple of pF to the grid of V1A "cures" it. (Still wondering WTF is up with that, but we'll see). The large value of Ck wouldn't seem to account for that.

 

[PRR]

Reduce R11 to 470 ohms.

Increase C7 to 50 or 100uFd.

I'm shooting in the dark, and I'm not happy with the cost of these changes, but my guess is power rail oscillation.

> miniscule amount of capacitance... at the grid of V1A.

But wait. Isn't V1A grid connected to a low-Z source?

 

[Larrchild]

How does one prevent phase shift in a feedback network thru a large R and C value from becoming non 180 degrees and regenerative?

I agree too much Ck wouldnt make the grid act like a tuning circuit.

 

[NewYorkDave]

But wait. Isn't V1A grid connected to a low-Z source?

No, it's being driven from a source Z of 50K. I did this to approximate the response with a step-up input transformer.

I'll short the 50K build-out and see what happens. That could give a clue. I'll also try tacking extra filter capacitance across the rail(s).

 

[NewYorkDave]

Shorting out the 50K didn't help. But the VLF oscillation went away when I used a shielded grid lead for V1A. I'm still a little baffled by the whole thing but I don't intend to investigate it further.

Still only getting max out of +18dBM with a 12@V7 for V2. This is a whole 3-4dB worse than predicted. I tried different AV7s (including NOS) with no substantial difference. Interestingly, it does a whole lot better (+21dBM) with a 12AU7 for V2. But then again, this particular AU7--an Amperex "Bugle Boy"--has big plates, twice the size of the AV7's. I don't know if other varieties of AU7, such as cheap Chinese, would work as well. When using a 12BH7 for V2, the output can reach +22dBM max. (I'm testing into a 600-ohm load, hence "dBM", but just in case ya missed it...).

By the way, I define "max output" as the point at which the waveform is just beginning to show the first signs of squishing. Real clipping usually happens at least a couple of dB above this.

In addition to AV7, I tried AY7 and AX7 for V1 and both worked well. So there is some latitude in tube selection depending on availability and personal taste.

I have to get back to earnin' the paycheck now, but I'll take some measurements sometime soon.

EDIT: Oh, for anyone who's keeping track, I changed the plate and cathode resistors on V2 to 12K and 270 ohms, respectively. This nets a little more output.