Small Line Amp (Tube)

[NewYorkDave]

Way too much playing around today... I tried 12AV7 + 6SN7, and 12AV7 +6CG7/6FQ7. The results were very close--as befits the tube manufacturers' claim that the SN7 and CG7 have the same characteristics.

In both cases, THD+N was 0.1% at +4dBM and 0.5% at +14dBM. But it started rising rapidly from there (2.5% at +18dBM and 4.7% at +20dBM). This turned out to be distortion from V1A at higher input levels. Removing the bypass cap from V1A improved matters greatly (1% at +18dBM, 2% at +20dBM), at the cost of some gain, of course. The distortion figures are a little higher overall than what I was getting with the 6SL7 as V1, but at least now there's no problem with HF rolloff from a high-Z source. With a 50K source, frequency response is 20-20, -0.13, +1.01. Again, that 1dB bump at the bottom is due to the output transformer.

IMD+N is 0.26% (unweighted) or 0.18% (A-weighted) at +14dBM

Graphs from today's tests can be found in this album. I did not take screenshots of the FFT, however. The album just contains images from RMAA automated tests. The THD+N numbers I'm quoting here are from FFT, since RMAA does its THD+N test at 3dB below full scale and therefore the results tend to be slightly optimistic.

Dig the little bumps at 60Hz, 120Hz, etc... Power line frequencies creeping in. That happens in a breadboard :wink:

By the way, in case you're wondering why I seem determined to incorporate 12AV7 into this design somehow, it's because I have a sh*tload of them. But now it looks like I need to pick up some more 6CG7 as well.

So, in conclusion, it seems what we have here is a small line amp (or beefy preamp, depending on your perspective) that will work well with a few different tube types, both miniature and octal. I'll build a proper prototype someday, but it's time for me to call the "development work" done on this one and move on to the many other projects (both in and outside the "day" job) that are piling up...

 

[tk@halmi]

Just so the torch keeps burning... It looks like we are getting a replacement transformer from Edcor sometime next week. I will try to block out some time to test it that weekend. Finally I have Dave's one triode output circuit on a turret board so it should go quick.

 

[NewYorkDave]

Are they going with gapped M6 this time?

 

[NewYorkDave]

Well, despite what I said earlier, I did find just a little more time to putz around with this circuit. I had to leave work early yesterday and wasn't going to be there long enough to get involved in any "real" work, so I spent the last hour or so of the abbreviated workday looking at the 12AV7/12BH7 combo in more detail. I'm inclined toward this pairing because I happen to have several of these tubes in stock.

Here's an updated schemo.

I opened the feedback loop and measured the output resistance of V2 at 2.2K, which means the plate resistance is around 2.7K. But the output impedance figure given on the schematic is the closed loop output Z measured at the transformer secondary. I know it seems improbably low, but I triple-checked and sure enough, it takes about 65 ohms shunted across the secondary to drop a very lightly-loaded +12dBU output to +6dBU. That's feedback for ya. Of course, that all goes out the window when you push the amp to maximum output, but I digress...

I also ran a number of tests with the transformer secondary bridged (by about 7-8K) instead of terminated. This particular circuit/output transformer combination didn't seem to mind the "mismatch"; it didn't cause any weird peaks in the frequency response and a 1kHz square came through mostly nice and clean with only the slightest ringing.

So, I think I'm gonna go with this version, since it performs well and I'm already stocked on the tubes. I want to build four to six of these amps for my mixer. Of course, with an input transformer added, it'd also make for a nice high-gain mic pre.

 

[bcarso]

What's the gain go to without any feedback now?

 

[NewYorkDave]

Brad,

The gain of V1A is 20dB with cathode unbypassed, 26dB bypassed. The composite OLG of V1B and V2 is 46dB, or 32dB with 100K for Rfb, or 26dB with 50K Rfb. Subtract 12dB from the total figure for the output transformer, of course.

 

[bcarso]

[quote author="NewYorkDave"]Brad,

The gain of V1A is 20dB with cathode unbypassed, 26dB bypassed. The composite OLG of V1B and V2 is 46dB, or 32dB with 100K for Rfb, or 26dB with 50K Rfb. Subtract 12dB from the total figure for the output transformer, of course.[/quote]

Thanks. I just wondered what kind of open loop gain you had available and how close to the edge of the cliff things were getting when negative feedback was applied.

 

[pstamler]

Dave, have you compared the distortion performance, especially the harmonic spectrum, between the various versions? The 12BH7 is a good deal less linear open-loop than the 6SN7, and I've found it has a nastier distortion spectrum. YMMV, of course. And have you tried a 12AV7 as output tube?

The 12AV7 and the 6SL7 should have virtually identical noise performance in this circuit. Dunno about their distortion performance in the first tube section, though; there aren't a lot of things as linear as a 6SL7 open-loop.

Peace,
Paul

 

[pstamler]

Oh, and another thing: you were mystified why you got identical clipping performance when you paralleled two sections of 12AV7. If you're in voltage clipping that makes sense; volts is volts, and putting two sections in parallel won't change the point at which they clip.

On the other hand, if you parallel them, you can then run into a transformer with half the primary impedance and get equivalent outputs from the tubes; keep the same secondary impedance and you're now using an output transformer with 3dB less stepdown, so now your clipping point will be 3dB higher.

Peace,
Paul

 

[NewYorkDave]

THD and IMD comparison of 12BH7 and 6CG7 in V2 position, same circuit, same B+, same tube (12AV7) for V1, same output level:

http://groupdiy.twin-x.com/thumbnails.php?album=124


Ignore the "Right Channel" display on all graphs because nothing was connected.

 

[NewYorkDave]

For "those what care", here are some tests with the amp in medium-gain/low-ish distortion mode (with AC heaters):

Webpage

I don't bother with testing at +4dBM because single-tone tests at that level will always give results that are much better than you'll get with real program material. So I test 10dB higher. I included tests at +22dBM to give an idea of how it does near the threshold of clipping--and yes, the spectrum does become a bit "rich" at that level :wink:

 

[alk509]

Cool, Dave! How about some soundbites?

Peace,
Al.

 

[NewYorkDave]

Soundbites of what? Me going "duuhhhhh...?"

:razz:

The amp only lives as a breadboard on my bench at work, so recording musical samples isn't really an option till I build a proper prototype and take it home.

Anyone who thinks DC heaters are always a must for a tube preamp should look at those noise figures. They ain't bad for an open breadboard running at a gain of 46dB, sitting on a bench, mostly unshielded, in a room full of fluorescent lighting and various power supplies humming away. With a net gain of +34dB (including the output transformer), that would make the EIN something on the order of -132dBU. Figure a weak mic through a 1:10 input transformer would be stepped up to about -40dBU at V1A grid, so the SNR would be quite acceptable.

 

[NewYorkDave]

Here's a three-position "range" switch:

 

[NewYorkDave]

A better view at +14dBM output, courtesy of a real-time FFT (and a good friend who hooked me up :thumb:

Spectrum

This is less optimistic than the results in RMAA (posted earlier) because RMAA tests for THD+N at 3dB below whatever operating level you set.

 

[tk@halmi]

[quote author="NewYorkDave"]Are they going with gapped M6 this time?[/quote]

Yes, it is gapped. It is on the regular XSE1 core so it uses the M6 lams, and is bigger than the nickel core part. More later...

 

[pstamler]

[quote author="NewYorkDave"]A better view at +14dBM output, courtesy of a real-time FFT (and a good friend who hooked me up :thumb:[/quote]

Hmm. On the one hand, the spectral distribution is good, and there seems to be nothing over the 5th harmonic. On the other hand, the overall level of the distortion is high for only 10dB over nominal level.

Is the bulk of the distortion coming from the output stage?

Peace,
Paul

 

[NewYorkDave]

All of the following were measured at +14dBM output, 1kHz signal from a 50K source impedance.

Here's the spectrum with the signal going into the grid of V1B, with V1A disconnected.

Image

So, a little bit--but not much--of the distortion seen earlier is due to the input stage. The difference seems to be mostly in noise level.

As a point of interest, here's a comparison between the 12BH7 and 6CG7 in the V2 position, running open-loop, signal input to V2 grid (V1 not connected):

12BH7

6CG7

The interesting thing to note is that although the BH7 shows an overall higher percentage (due to a couple dB more second, and 10dB more fourth harmonic), the CG7 has much greater levels of nasty upper harmonics--which, given the published plate curves of the 6CG7 and its big brother, the 6SN7, is not what you'd expect. The outcome would no doubt be different if the circuit values (especially load impedance) were adjusted to suit the 6CG7 better.

Unless there was some flaw in the test setup that skewed the results, the 12BH7 seems the superior choice in this particular circuit despite giving higher THD at normal operating levels.

 

[pstamler]

[quote author="NewYorkDave"]As a point of interest, here's a comparison between the 12BH7 and 6CG7 in the V2 position, running open-loop, signal input to V2 grid (V1 not connected):

12BH7

6CG7

The interesting thing to note is that although the BH7 shows an overall higher percentage (due to a couple dB more second, and 10dB more fourth harmonic), the CG7 has much greater levels of nasty upper harmonics--which, given the published plate curves of the 6CG7 and its big brother, the 6SN7, is not what you'd expect. The outcome would no doubt be different if the circuit values (especially load impedance) were adjusted to suit the 6CG7 better.

Unless there was some flaw in the test setup that skewed the results, the 12BH7 seems the superior choice in this particular circuit despite giving higher THD at normal operating levels.[/quote]

Wow, that 6CG7 looks genuinely nasty. Is it possible this is a bum tube? Do you have any other specimens handy? I confess I've never seen a member of this family act quite like that. (Never? Well, hardly ever.)

Peace,
Paul

 

[NewYorkDave]

The tube was a new-in-box GE JAN 6CG7 from '75. I do have some others, but they're the same type, might even have the same production date. If I have time, I'll check another sample or two before I dismantle the breadboard.

Looking at inconsistencies in my results over the course of these experiments--even with the same tubes, components and test setup--I think there are still some parasitics in play with this breadboard. And the lack of shielding certainly doesn't help the "N" part of the THD+N figure. So now that I've tweaked the circuit about as far as I care to do, the next step is to construct a proper prototype and actually listen to the thing :wink:. And I think it'll probably measure out better as well. It'll be a while before I have time to build it, but I'll report back after I've gotten around to it.