• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

FET Source Follower Distortion

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I'm testing in triode mode at 8Vp-p as that is the max into my sound card before clipping.

The gyrator represents a low DC impedance, and a high AC impedance to the tube, so gain is not reduced significantly over a CCS.

Yes, my sound-card is a 10K input (I expected 47K for some reason) and was the initial cause of the high distortion.

The LF353 buffer distortion is quite low at this level when driving the sound card (2nd below-120dB, 3rd below -110dB, 4th on below -130db), so it is acceptable as is. I have an OPA2134 but they didn't like the proto-board. They generated a lot of noise, probably due to poor power-supply bypassing. I'll eventually build a breadboard with ground plane for the OPA2134. For now I'll stick with the LF353 until it becomes the dominant distortion source.

I think testing in pentode mode will pose some problems. As you point out, Gain with either a CCS or Gyrator is going to be high.
 
The LF353 is an old design. Its open-loop output-impedance is not symmetrical,and so it may distort more when driving a capacitive load - either the follower's input capacitance, or the cable's.

How about trying a more capable op-amp - especially one design for capacitive loads, eg TI LM8261, or a LT "C-load" model - may well improve it.


The LM8261 can drive any capacitance, and has specified performance for a 2K load, so you won't need a follower on this one's output.
 
Part of the original question, I believe, was "before I measured worse performance with a similar circuit, now it's cleaned up. Why?"

Looking at the two FFT plots attached to Post #1, it's fairly clear that the one that shows poor performance contains a whole school of 60 Hz harmonics. The actual circuit performance is buried under all the 60 Hz garbage. With the new layout and ripple filter, most of the 60 Hz is gone, leaving a good noise/hum floor to make measurements.

I bet that's the cause of the better performance measured in the second plot with the new circuit and layout.

As for FET source followers driving tubes like the 300B. I've played with that quite a bit. I used an AOT1N60 NMOS for the source follower. It has very low gate-drain cap, hence, lends itself well to source follower applications. I found the THD of the complete amp to be better at low FET currents, actually. Not dramatically better, but measurably better. Same was true when using a cathode follower. I suspect this is due to distortion cancellation between the input tube (grounded cathode) and source/cathode follower. Morgan Jones writes about this in VA4. I measured and listened to both circuits and ended up preferring the cathode follower. YMMV.

~Tom
 
Spectrum of tubes and FETs will be the same: in both cases are used non-linear devices with deep feedback. But distortions of FETs will be as the rule much lower due to higher transconductance. Of course, distortions caused by capacitance modulation will play higher role with FETs, so as usual the trick is to find optimal solution.
 
Rod, per your recommendation I've ordered a couple of LM8261s to try. However in light of the performance of the low distortion oscillator with the LM353 buffer, I don't expect to see a great deal of difference.

If the LM353 were the culprit, wouldn't there be a high level of 2nd order harmonic content when only driven by the low distortion oscillator?
 
Steven, the LM353 will probably work fine for many duties, especially where the capacitance of the load is minimal - but one of the biggest improvements in modern op-amps is the ability to drive difficult loads with comfort.

Knowing the actual value of load capacitance will help ensure the design is optimised.

The LF353 shows high open-loop output impedance (>30 ohms) which can be a recipe for trouble where more than a handful of pF are expected.

The problems include instability (which can be fixed with 100 ohms of resistance in the output), and added distortion.

The supply voltage also plays a role, and if you are not working with +/- 15V, the LM8261 will also improve performance. +5V single rail will suffice for many test environments.
 
I built a buffer using the LM353 while I'm waiting for better op-amps on order.

I used a LM317 pre-regulator with two TL431 shunt regulators for +/-15Vsupplies.

The first FFT is of the low distortion oscillator only.

Second FFT is of the LDO through the buffer.

The buffer adds a slight degradation of S/N, the second harmonic is up about 5dB and the third harmonic is up about 3dB, and there is some garbage beyond 10KHz but otherwise looks pretty good.

Third FFT is of the output from the source follower (IRFI820) without the buffer.
Fourth FFT is of the output from the source follower with the buffer.

Here there is a significant improvement in S/N, and a lot of the EMI is eliminated. I suspect the lower impedance output of the buffer is reducing the EMI suscpetance of the wiring.

Again there is about a 5dB increase in 2nd Harmonic distortion, but 3rd, 4th 5th, and 6th are down significantly, or they were being masked by the EMI.

Possibly the lower load of the buffer compared to the Sound card input is resulting in greater signal levels of the harmonics, but there is only a 0.24 dB increase in the level of the fundamental compared to a 5dB increase in the second harmonic. Shouldn't all signals be effected proportionally?

Am I still seeing an erroneously high distortion level, or is this actually an accurate measure of the tube? (6J9P, NOS)
 

Attachments

  • LNO Only.JPG
    LNO Only.JPG
    173.8 KB · Views: 349
  • Buffered-LNO.JPG
    Buffered-LNO.JPG
    178.8 KB · Views: 338
  • Tube-w-SF.JPG
    Tube-w-SF.JPG
    180.9 KB · Views: 332
  • Buffered-Tube-w-SF.JPG
    Buffered-Tube-w-SF.JPG
    174.3 KB · Views: 331
To answer my own question, this is probably still not an accurate measure of tube performance.

The high levels of distortion are probably due to my driving the input to near 0dB ref, and although I'm not clipping I suspect that I am over-loading the sound card, in which case I would expect to get increased distortion.

I'm working on a buffered Twin"T" filter to reduce the fundamental and thus prevent me from overloading the input.
 
By using a Twin T filter I am no longer overloading the front end of the sound card, and I see a drop in 2nd harmonic of about 10dB. The rest of the harmonics are about the same.

I expect this is representative of the tube performance.

I measured the response of the twin T filter with a 8.24V p-p signal at various frequencies, and used the values to make a correction table (1).

The mean offset (excluding values close to the notch) is 2.44dB so I add that as a correction factor to measurements I make to get the true value of the harmonics(2).

I think I can improve the noise level slightly when I make a housing for the Low Distortion Oscillator and Twin T filter as currently everything is spread out on the bench.

Here is a typical 6JC6A (NOS RCA).
 

Attachments

  • Corrected THD Spreadsheet.JPG
    Corrected THD Spreadsheet.JPG
    46.7 KB · Views: 293
  • 6JC6A-No5_Thd.JPG
    6JC6A-No5_Thd.JPG
    60.2 KB · Views: 287
  • 6JC6A_No-5.JPG
    6JC6A_No-5.JPG
    108 KB · Views: 281
I tried 8 6J9P-E last night. These were from two suppliers, but all the same manufacturer varying from 1977 to 1987. All were in the same blue/yellow/black marked boxes with data sheets wrapped around the tubes.

All 6J9P were triode strapped with Gyrator set to 220V at the anode.

Eight were tried at 3mA, with an average 2nd harmonic of -83.2dB, worst -77dB, best -93dB. All show the same trend in harmonics. The best is shown as FFT 1.

At 1mA both 2nd Harmonic distortion and noise floor were worse (2), but harmonics above 2nd were better.

At 5mA the noise floor improved, but higher harmonics are increasing (3).

At 7.5mA, the trend continues, but second harmonic is starting to decrease. This is probably a tipping point. (4)

At 10mA, third harmonic exceeds the level of second harmonic at 3mA. (5)

Back to 3mA and 8Vrms out. The signal is starting to overload the op amp input to the twin-T filter. Above 8Vrms harmonics look like clipping. (6)

Finally I tested a 6CJ6 at 3ma for comparison. (7) It is interesting in that second harmonic is much greater than the 6J9P, but beyond that harmonics are much lower. This is probably why the 6CJ6 is preferred even though the overall thd measurement is higher.

At low current (1mA) the 6J9P looks similar, but has a higher noise floor. This is where I've been running 6J9Ps in the past. A good compromise that is worth listening to might be between 1 and 3mA.
 

Attachments

  • 6J9P_3mA.JPG
    6J9P_3mA.JPG
    151.4 KB · Views: 260
  • 6CJ6_3mA.JPG
    6CJ6_3mA.JPG
    152.8 KB · Views: 53
  • 6J9P_3mA_8vrms.JPG
    6J9P_3mA_8vrms.JPG
    153.5 KB · Views: 32
  • 6J9P_10mA.JPG
    6J9P_10mA.JPG
    158.4 KB · Views: 38
  • 6J9P_7-5mA.JPG
    6J9P_7-5mA.JPG
    152.2 KB · Views: 37
  • 6J9P_5mA.JPG
    6J9P_5mA.JPG
    153.6 KB · Views: 40
  • 6J9P_1mA.JPG
    6J9P_1mA.JPG
    154 KB · Views: 249
I'm putting the two LD Oscillators and a the Twin T filter in a chassis to help shield them and clean up the wiring.

In order to isolate the boards from the magnetic field of the transformers I've oriented them as as shown in the attached photos.

It occurred to me that since the two transformers are nearly identical (same part number from the same mfg), it might be possible to cancel some of the B field by wiring the primaries anti-phase. If so, which transformer orientation will help reduce the field the most?
 

Attachments

  • TwinT_Layout_a.JPG
    TwinT_Layout_a.JPG
    151.7 KB · Views: 84
  • TwinT_Layout_b.JPG
    TwinT_Layout_b.JPG
    159.2 KB · Views: 91
The block diagram shows the test set I’ve build. It has the two Low Distortion Oscillators selected by a three pole double throw switch (have to switch both power connections plus the output), followed by an adjustable attenuator and a buffer. The buffered output drives the UUT.

The output of the UUD goes through a buffered step attenuator with selections for 0dB, -10dB, and -30dB.

This drives the Twin T filter, which is followed by a divide by two and buffered by a differential output buffer. The divide by two compensates for the differential driver to the sound card.

FFT 1 is a 6J9P at 2.5mA and 220Va driven to 2.8Vrms out (8.12Vp-p). From it we can see that the measured fundamental is at -0.32dBfs ref to 2.8Vrms. So I have a measurement error of 0.32dB that will need to be added to the measured value.

Does this also get added to all harmonics as a correction factor?

Comparing FFT 1 and FFT 2 we see about a 12dB difference in second harmonic and about a 10dB difference in subsequent harmonics. This is attributed to overloading the input of the sound card, and is one of the main reasons for building the test-set. The greater difference in second and subsequent harmonics is in part due to the Twin –T filter which was measured previously at -0.26dB.

FFT3 is the same 2.8Vrms but with a -10dB attenuator switched in.

FFT4 is the same 2.8Vrms out with the -30dB attenuator switched in, and bias point adjusted to 180V.

To calculate distortion of FFT5 (45Vrms) it is necessary to measure all harmonics and adjust them for the attenuator (-30dB), test-set and Sound-Card errors (-0.26dB).

Taking the first five harmonics I get -75, -96, -108, -120, -128dB
Correcting for the attenuator I get -75+31=-44, -96+30=-66, -108+30=-78, -120+30=-90, -128+30=-96.

The soundcard error of -0.26dB should be added, however I have not expanded the plots enough to get that much resolution so I ‘m leaving it off for now but to nit-pick it should be added in.

The fundamental is calculated by adding the gain over 2.8Vrms in dB. 45/2.8=16.07, converting to dB I get 24.12dB..

So now the harmonics need to be adjusted by 24.12 to get their relative value to the fundamental, giving -68.12, -90.12, -102.12, 114.12, -120.12

From this I get 0.039%thd.

Is this correct? It seems low to me from what I expected at 45Vrms from a triode strapped 6J9P.
 

Attachments

  • Twin-T Block Diagram.JPG
    Twin-T Block Diagram.JPG
    19.3 KB · Views: 75
  • 6J9P 2-8vpp.JPG
    6J9P 2-8vpp.JPG
    159.2 KB · Views: 68
  • 6J9P Twin-T-0atten.JPG
    6J9P Twin-T-0atten.JPG
    159.2 KB · Views: 34
  • 6J9P Twin-T-10dB.JPG
    6J9P Twin-T-10dB.JPG
    152.7 KB · Views: 28
  • 6J9P Twin-T-30dB.JPG
    6J9P Twin-T-30dB.JPG
    155.5 KB · Views: 29
  • 6J9P Twin-T-30dB_45Vrms.JPG
    6J9P Twin-T-30dB_45Vrms.JPG
    153 KB · Views: 31
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.