KT120 vs. 6550 shoot out in my Conrad Johnson Premier 11

I found this very interesting so I thought I would share. I've been wanting to compare 6550s to KT120s for some time in my faithful friend the CJ Premier 11. I wanted to compare bias (will the KT120s bias properly), max power, power bandwidth, and total harmonic distortion. I found some expected results and some very surprising results. This is not a perfect test bed environment in that I don't have the power resistors to load both channels simultaneously, so I used only the left channel for all tests described here.

The Premier 11 has received good press over the years, most notably in Stereophile, where in 1994 it received a full review with published performance tests (using GE 6550A's I assume, which was standard issue back then for a new Premier 11 owner). Since this is a well regarded amplifier, I figured this is as good of a test bed as any to compare the KT120 to the 6550.

Here's a picture of my amp. This shot shows KT120s plugged in for the right channel and Winged C 6550Cs plugged in for the left channel. For all tests I used only the left channel and swapped tubes accordingly.



The KT120 tubes biased up perfectly in the amp--the bias trim pots had ample adjustment room on either side. The 10KHz square wave looks very similar to the Stereophile published measurement. Still, I am surprised there is so much initial overshoot and less damped ringing than I would have expected. Nonetheless the amp is very easy to listen to and not fatiguing at all.



Other measurements I took with the KT120 tubes:

• Max power (across 8 ohm resistive load): 49 watts @ 1KHz. This is very surprising in that CJ published 70 watts/channel.
• Power bandwidth: 27Hz - 33KHz, +0.6dB, -0dB (the +0.6dB is surprising, and as I found out later, this is an artifact of the KT120 tubes)
• THD at 1KHz and max power: 0.3%
• Unconditionally stable as near as I can tell. I tried no load, 8 ohm resistive load, 2 uF cap only load, 20 uF cap only load, and 8 ohm resistive load in parallel with 20 uF cap.
• I had no way of measuring stability margin.

Now for the Winged C 6550C's. Here's a picture of its 10KHz square wave at 1 watt output across an 8 ohm resistive load. You will notice it is virtually indistinguishable from the wave generated by the KT120 tubes.



Other measurements I took with the 6550 tubes:

• Max power (across 8 ohm resistive load): 45 watts @ 1KHz. Still very surprising.
• Power bandwidth: 24Hz - 33KHz, +0dB, -0dB (excellent)
• THD at 1KHz and max power: 0.15% (Half that of the KT120s!!)
• Unconditionally stable as near as I can tell with same loading parameters as described above for the KT120s
• I had no way of measuring stability margin.

Sound-wise, the 6550's are just perfectly sweet for this amp. The KT120s are "not bad", especially in the mid range where they are a bit fuller which is very pleasing to my ears at least, but there is evidence to me that this amp was built around the 6550 tube, and it measures to prove it--especially producing HALF the distortion, and producing an extremely flat response within the amp's power bandwidth.

Also I should mention that these 6550's were purchased straight from Conrad Johnson. I know that CJ takes great care in selecting only premium manufactured tubes. It is possible that the KT120s would measure as well if they were culled by CJ.

And while this thread isn't about the CJ amp itself, I gotta say I am impressed with these output transformers. They measure fantastic. Observe the nearly vertical side walls on the square waves and the straightness of the edges of the square waves right up to the top of the overshoot. IM-pressive.

kind of surprising you're only getting 45 watts out of that. 6550's are certainly capable of doing more. I guess you've verified that all the voltages and whatnot are where they ought to be, all that fun stuff?

Looks like I have a mystery on my hands. These amps aren't that easy to work on being 100% PCB based as I believe they are. I don't think I've ever even popped the bottom off of this one.

I wonder if changing some component values in the feedback circuit would correct the overshoot.

K -- One thing you might try with the differing distortion levels is to try the test again between both tubes, but this time, reverse the tubes in their sockets from the way they were installed for the first test.

Without resetting the DC balance and AC balance for each tube, it leaves the balance of the tubes as a variable besides just the change in tubes themselves.

Also would be interesting to know what the frequency response was. The undamped ringing implies it will likely be quite high.

As for the power output, that is basically a product of B+, and loading conditions used. Stereo Review did a rather (in)famous performance review of Fisher 500C some years ago, and while praising it, the review was also grossly misleading in some ways, and quite lacking in others. I have grown to have a weary eye regarding their published results.

Dave

The 70W rated power seems to be only available with the 2-4 Ohm load, as shown in the Stereophile test result, even though the specs say 70W for 2-4-8 Ohm loads.

Okay, things are starting to make sense now. I popped the hood off of the Premier 11 and took some voltage measurements on the output tubes:

Here's what it looks like without the top cover:


Voltage measurements:

Plate: 508V
Cathode: 1.15V (properly biased according to manual)
Grid to cathode voltage: -60V
Cathode resistance: 20 ohms

From this, I can calculate:
57.5mA quiescent current per output tube
29 Watts plate dissipation, which for a 6550 is about 70% of max.

The measurement of 49 watts max output power through an 8 ohm load now makes sense with a 20 ohm cathode resistor at 500V plate. It's interesting that if CJ would have used a 10 ohm cathode resistor, they would have obtained maybe 10 watts additional output power, but I can see why they didn't--they needed at least a 0.7V margin to run the bias detection LED circuit.

A quick glance at the 6550 chart, and I can see that CJ biases this amp a good amount into AB1 at max power. It appears there's about 10 to 15 watts of class A power before the tubes cut over to class AB1. (A guess really, since I don't know what the primary impedance of the transformer is.)

sony6060 said:

I wonder if changing some component values in the feedback circuit would correct the overshoot.

Click to expand...

Undoubtedly it would. But the amp doesn't sound fatiguing or bright, so I guess I'll leave it be.

dcgillespie said:

K -- One thing you might try with the differing distortion levels is to try the test again between both tubes, but this time, reverse the tubes in their sockets from the way they were installed for the first test.

Without resetting the DC balance and AC balance for each tube, it leaves the balance of the tubes as a variable besides just the change in tubes themselves.

Also would be interesting to know what the frequency response was. The undamped ringing implies it will likely be quite high.

Click to expand...

With the hood popped off, I readjusted the AC balance pots for minimum distortion with the KT120s and 6550's at 1KHz at max power. Each tube type was readjusted for bias and AC balance.

Distortion readings at 1KHz at max power are now:
KT120: 0.24% (a 20% reduction from previous measurement)
6550: 0.12% (a 20% reduction from previous measurement)

Because distortion at max power was reduced by 20% in both cases, I speculate Conrad Johnson tunes these amps for lowest distortion at something like 60% of max power, either that or they tune for distortion at max power at a different frequency.

Frequency response at reference of 1KHz at max power through 8 ohm load (with KT120 tubes) is:
-0dB: 33KHz
-1dB: 62KHz
-3dB: 90KHz
-6dB: 101KHz

Yeah, looks like this amp has plenty of high frequency extension. Doesn't make a lot of sense to me (but what do I know?) why they would tune it for only -1dB down at 60KHz. All I can say is the amp does not sound harsh, edgy, or fatiguing at all, even with source material I know to be bright to begin with. Probably speaks to the output transformers being of excellent quality and winding workmanship.

jazbo8 said:

The 70W rated power seems to be only available with the 2-4 Ohm load, as shown in the Stereophile test result, even though the specs say 70W for 2-4-8 Ohm loads.

Click to expand...

Ah ha. So my measured results are right on for an 8 ohm load. Somewhat disappointing as I thought I had a 70 watt amp with 8 ohm speakers. Rereading that Stereophile article, I noticed they do not reference or discuss "table 1" anywhere in the text, but do display it at the very end of the article. Hmmm. :thumbsdn:

kward said:

Because distortion at max power was reduced by 20% in both cases, I speculate Conrad Johnson tunes these amps for lowest distortion at something like 60% of max power, either that or they tune for distortion at max power at a different frequency.

Click to expand...

I say this because I sent the amp back to CJ about 5 years ago for repairs and tune up, including new set of tubes all around. Since then the amp has had 100 to 150 hours of play time, so relatively speaking its like I got the amp back last month.

Sooooo, there are no impedance taps on the secondary of the OPT? Just a secondary winding with a Com and Hot lead coming off of it?

Frequency response tests are typically run at 1 watt power levels (as opposed to your maximum power tests), so that the OPT's power bandwidth doesn't limit it. Response curves run at maximum power determine the power bandwidth of an amplifier, which will always be narrower than the frequency response curve is.

Dave

OPT Wiring

It seems you can re-wire the OPT taps, so better matching is available for 8 or 16 Ohms, it is probably pre-wired for 4 Ohms at the factory.

output transformer

dcgillespie said:

Sooooo, there are no impedance taps on the secondary of the OPT? Just a secondary winding with a Com and Hot lead coming off of it?

Click to expand...

Premier 11 output transformer can be configured for 2,4,8 or 16 ohm output matching. http://www.conradjohnsonowners.com/viewtopic.php?f=11&t=361

As far as I remember, CJ used to ship Premier´s configured for 4 ohm output unles some other matching was asked. My Premier 11A does have 8 ohm handwritten clearly on shipping carton, but no remark for 8 ohm match on amplifier. Unmatched output can be resason for less than specified output power figures.

Best Regards

Kimmo

Just a note: You can bias the 120's hotter and obtain better distortion figures, and better sound.

That said, I still prefer the sound of 6550's to KT120's in my designs.

Thanks for the information. I have no experience with these CJ amplifiers, so I didn't know if alternate output impedance settings were possible. Testing the amplifier without the proper load applied invalidates all of the previous test results provided.

Dave

Sigh... On my amp I've got a yellow wire and a white wire connected to positive, and a black wire and a gray wire connected to negative. (There's also a green wire connected to positive which I assume is for the feedback and another black wire connected to negative which I assume goes to the ground plane of the amp. ) So, it does appear mine is wired for 4 ohms. Up until yesterday, I'd never even looked inside this amp...so I'm under a learning curve too.

dcgillespie said:

Frequency response tests are typically run at 1 watt power levels (as opposed to your maximum power tests), so that the OPT's power bandwidth doesn't limit it. Response curves run at maximum power determine the power bandwidth of an amplifier, which will always be narrower than the frequency response curve is.

Click to expand...

I will measure this aspect of an amp's performance this way from here on out. Thanks for the info. And thanks to everyone else for chiming in.

Power bandwidth is properly measured at 1/2 of the manufacturers rated power (-3db from rated power).

Interesting. Seems to be somewhat of a misleading specification to me then since many very average performing units will measure much, much better at only 1/2 of rated power output. I know that never would have passed muster back when the FTC was on the warpath about the ratings for consumer stereo equipment.

Before that time, high quality units often had the power bandwidth information combined with distortion information, such that a caliber unit would typically be rated at "Less Than X% THD within 1 db of XX Watts, 20 Hz to 20 kHz.

This is a much more meaningful specification, since not only does it represent a greater portion of the available power output, but also stipulates the maximum distortion produced within the specified power bandwidth. Even this however was not acceptable to the FTC, which ultimately mandated preconditioning of a unit, and no reduction of power to specify the power bandwidth at. Whatever the power capability was within the power bandwidth of a given unit, was the power the unit had to be advertised as capable of. Power bandwidth ratings at so many db below rated power output were gone overnight. The FTC also stipulated that the impedance be provided that this performance was produced at as well.

Dave

Watts per channel, # channels driver, Frequency response with less than % distortion is the FTC requirement for power specification starting in November 1974 for stereo amplification.

Power bandwidth is a half-power indication of the bandwidth of the unit under test.

These are two different specs.

Agreed! My original statement was a general one primarily intent on showing that frequency response measurements are not made under full power conditions, but under low level power conditions, and that frequency response testing under rated power conditions more generally relate to the power bandwidth response of a given unit. There is no question however, that power bandwidth by strict definition is defined by the -3 db power points. For the highest quality audio equipment however, these points should occur far enough outside the audio bandwidth, defined as 20 Hz to 20 kHz, such that the power delivered within this bandwidth has an essentially flat response.

Dave

I do not understand power and distortion figures. If distortion is below 1%, you can push it further up to 1%, this will be your max power point. With KT120, I would rise idle current close to 100mA four each tube. Should get better distortion numbers at moderate power that way.

Updated measurements

Understanding now that the output transformers are wired for 4 ohms, I thought I better rerun the measurements. I still do not have a way to drive both channels simultaneously (I need another BNC to RCA adapter), so all measurements are taken on the left channel only. This makes the test less than ideal, but it's the best I can do at this time.

Both sets of tubes and all my equipment were warmed up for 30 minutes, output tube bias was set according to the instructions in the manual, and AC balance was tuned to minimize distortion at 1Khz at just under max power. Half power distortion measurements for both sets of tubes required no additional adjustments of the amp's AC balance.

I used a 4 ohm resistive and non inductive load. The 10KHz square waves exhibited identical shapes as posted previously under the 8 ohm load, so I will not repeat them here.

6550 tubes:

• Max power 72 watts @ 1KHz
• Frequency response @ 1 watt: On the low end: 5 Hz at -0.6dB (the lowest my equipment can reliably measure). On the high end: 42KHz at +1dB, 81 KHz at +3.3dB, and 106KHz at -1dB. In other words the amp is fairly flat till about 32KHz, where it takes a sharp rise, peaking at +3.3dB at 81KHz, then falls off quickly after that. (note this matches the Stereophile article very closely).
• Power bandwidth: 22Hz (-0dB, but onset of notch distortion) to 24KHz (-0dB, but onset of phase distortion)
• Distortion (THD): 0.15% at 1KHz at just under max power, 0.035% at 1KHz at half power.

KT120 tubes:

• Max power 72 watts @ 1KHz
• Frequency response: same as for 6550, above
• Power bandwidth: 24 Hz (-0dB, but onset of notch distortion) to 24KHz (-0dB, but onset of phase distortion)
• Distortion (THD): 0.13% at 1KHz at just under max power, 0.12% @ 1KHz at half power.

I was curious about the half power THD measurements so I performed them twice for both sets of tubes, and I'm confident they are correct. So this is a virtual dead heat now between the tube types. The only significant difference is that at half power, the 6550 tubes have 1/5th the distortion of the KT120 tubes. Since I doubt anyone can hear the difference between .03% and .15% THD, sonically I would call these tubes equal relative to measured performance. Technically speaking however with respect to distortion, it does appear the Winged C 6550C has the slight edge at lower power levels, where most people will probably do most of their listening.

I'm sure there are sonic differences between these tube types, for example I can hear more bass with the KT120s and the mid range is a bit fuller with the KT120s in my amp.

Now I can rest easy knowing that I can swap tube types without fear of loosing performance, and I really can pick the tube I want based on sonic merits.