• 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.

SE sound from a P-P amplifier

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
next piece of Tube modified spectrum.
 

Attachments

  • tubefdbk2.gif
    tubefdbk2.gif
    8.7 KB · Views: 1,019
SA,

It appears to me that third is close to the same with the tube fb, fourth is almost identical to the Optimos alone, and H2 is much increased, contributing principally to the much higher THD of 0.3%.

The characteristic you might be uncovering is known in psychoacoustics as 'masking', whereby largish low order distortion tends to cover over higher order distortion.

This would give the amp the sonics of a SET, but not quite........ (If you want it to sound like a tube, use a tube!)

It also opens up the prospect of distortion cancellation with a tube preamp of the same kind, with similar operating point........

Have you had a bit of a listen? Impressions?

Thank you for your impressive work on this. It's fascinating!

Cheers,

Hugh
 
OK, here is the Tube Feedback version schematic. ( I'll post the CCS schematic next post.) The component values shown will change some depending on the tube used.

For higher current operation, the tube load resistor needs to be higher wattage and the CCS adjusted to that current, and RLL might be a lower value if you don't want to lose too many tube volts from the B+. However, going lower on RLL likely will require a larger coupling capacitor than the 25 uF shown, depends on the lowest frequency you want to handle.

The 5K gain adjust pot is used to set the closed loop gain of the amplifier. Different tubes used in the feedback will have different Mu factors (and consequent 1/Mu attenuation), so this provides a way to adjust the gain. The original OptiMos had a feedback attenuation factor of 10,000/270 = 37, so 1 volt input signal gave about 38 Volts output (37 + 1 for a non-inverting input). The 6LE8 I used had a Mu of about 18, so the 5K gain adjust pot was adjusted to about its middle to restore 37X attenuation. This means that tubes with a Mu of 37 or less can be used easily by adjusting the gain adjust. (this controls input sensitivity of the amplifier as well)

For tubes with a Mu of greater than 37 you will get more amplifier closed loop gain than normal. This case may be fixable by feeding a small amount of output signal to the grid, I haven't tried it yet. (Use a pot and resistor, say 100K pot and 10K resistor in series, from grid to Optimos NFB tap/output point and a 1K resistor say from grid to ground.)

The 220K resistor from output tap to - input is used to control gain when the tube is warming up and it also keeps DC offset under some control. It is over-ridden by the tube, AC feedback wise, when the tube warms up. I still needed a DC offset control to fix input offset current (at first I thought the 50uF cap was leaking, but its not from that, its unbalance in the Beta or current gain of the NPN versus the PNP input differential pairs) at the - input. Thats why the 100K Ohm resistor to the 5K Offset Control is there.

(Its possible that the Offset pot might have to be connected to +12V instead of -12V. The Optimos uses dual complementary differential pairs for its input and it just depends on which pair has less Beta as to which way input offset current will occur.) The 50 uF cap from cathode to - input is to keep tube DC off the input, it just needs to be big enough to pass the lowest frequency against the 220K and 100K resistor loads it sees on the - input. 50 uF is overkill, I just used what I had handy.

The CCS should be made with a variable current setting pot if you intend to try different tube types in the circuit or just to optimise the operating point of a given tube. A high impedance CCS using a cascode like design will give the best results sonics wise. Be careful not to exceed the dissipation ratings of the tube.

OH, I forgot to draw in the RC and RL output networks between the output and load resistor. They are still there, as shown in the original OptiMos schematic. Negative feedback gets taken before these networks, as the original 10K NFB resistor is shown.

Some other notes:

The Optimos amplifier I am testing with has Analog Devices/PMI Mat02 and Mat03 matched transistors for the input differential devices, instead of the stock MPS8599 and MPS 8099 transistors. These give better input DC offset voltage results due to matched characteristics, but do not necessarily fix input offset current, which depends on the match between the pairs.

Some amplifiers will go unstable at near unity gain (without additional high freq. miller compensation) which would occur here due to the 220K DC feedback resistor when the tube is not warmed up yet. I didn't run into this problem, but other amps may be different.

Use a suitable fuse in the output when connecting to a real speaker, too many chances here for the output to go to the rails if something goes awry.

It would be a good idea to shield the tube feedback circuitry in a grounded metal box to keep out noise, hum and extraneous feedback. And use feedthru caps and large local electrolytic caps to ground and the shield for the tube related supplies. The box is also a good safety shield for the unexpected High Voltage present in this solid state amplifier.

Don
 

Attachments

  • optimostubeschematic.gif
    optimostubeschematic.gif
    9.8 KB · Views: 1,129
Here is the CCS ( constant current source) circuit I used for my test rig. Many other CCS circuits are around, like the ones using an LED for a voltage reference. I just used some parts that were handy. I think a cascode type CCS, with its high output impedance, will give the least interference with the tube sonics.

There is only around + - 2 Volts variation on the - NFB input for normal gain setup, so not much voltage drop or wattage dissipation is required of the CCS unless you decide to run with some high current tube.

OH, I should have mentioned that the 6LE8 tube I used has a gm of around 11000. This makes for 1/gm = 91 Ohms for the cathode output impedance I think. So it clearly takes control from the 100K type resistors connected to the - input. Maybe one could lower the 220K feedback resistor some more to fix the DC offset, buts its nice to have the DC offset control to fine tune the output offset.

http://www.mif.pg.gda.pl/homepages/frank/sheets/135/6/6LE8.pdf

I haven't had the opportunity to do much critical listening with this circuit due to my recent move and the fact that my speakers need some repair work. (cones rubbing in the mid ranges, also I left my CD player at a friends in Conn., I have to get back there this Spring to pull a bunch of stuff I forgot and left in storage)

I have to get the speakers fixed up soon, living in silence here, well... I have some darn dog always barking next door. I need some 2nd harmonic masking badly here!

Don
 

Attachments

  • testrigccs.gif
    testrigccs.gif
    3.9 KB · Views: 857
AKSA said:

It appears to me that third is close to the same with the tube fb, fourth is almost identical to the Optimos alone, and H2 is much increased, contributing principally to the much higher THD of 0.3%.

The characteristic you might be uncovering is known in psychoacoustics as 'masking', whereby largish low order distortion tends to cover over higher order distortion.

How come? As the second and third harmonics should already be (nearly) masked by fundamental according to masking theory.
I would think it is rather psychology than psychoacoustics. The simple mechanism of psychological masking one error with another.
If you read a 'flawless' book or listen to 'flawless' music than you will subconciously seek some flaws until you find it is generally flawed or dull or cliche. Best musicians/virtuosos do some subtle mistakes while playing here and there from time to time, you know why?
Sorry, Hugh, I don't get your point, it drives me mad to read that low order distortion should mask higher order distortion and this situation should be belived better. Similarly I don't find unpretty woman with make-up any prettier.

regards
Adam
 
New and Improved Schematic Version

It just occurred to me that the DC offset problem can be fixed much more elegantly by just using the standard Op Amp configuration.
Improved schematic attached.

I have made the DC input resistance to both the + and - amp. inputs the same, 100K here. The input currents should be matched with matched input diff. amp. stage transistors, and will track better with temperature change this way. This should work well for Gainclones and other IC chip Amps too.

Only drawback is that before the tube warms up, AC gain will be at exactly unity, so high freq. compensation will be stressed during warmup. If you get oscillation during warmup, then change the 100K feedback resistor to 220K and put a 100K from - input to ground, and change the DC offset pot to 50K.

Don
 

Attachments

  • tubefdbk3.gif
    tubefdbk3.gif
    9.3 KB · Views: 808
Warm up transient

Just a note on warmup transients. The amp will put out some big transient pulse to the rail during tube warmup, due to the electrolytic caps charging up to operating conditions. This needs to be fixed before connecting a speaker to it. Several methods using a timed relay are possible, I'm still thinking about the best way to handle this.

I hope to try out some other triode tubes in the feedback this weekend to get more distortion spectrums. I would like to know if this essentially 2nd harmonic only boost is generic with all triodes.

The 6LE8 tube is an interesting tube by the way, it works quite well in vac. tube current mirrors for the output tube, and with +12 volts on the suppressors its pentode characteristics square up just like a beam tetrode.

Don
 
I could definately see this helping out an all-solid state system, but I can't figure out how it would help in my case where I already have a SE choke-loaded 12B4A linestage feeding my gainclone. Maybe I'm missing something, but in my case, it could end up being too much of a good thing?
 
Adam,

I don't really know enough about masking to disagree with you, but I'm not sure that your point is anything other than a semantic interpretation. The sound would be perceived differently because of increased H2; that is certain, and this is clearly a psychoacoustic phenomenon, so perhaps we agree to disagree?

I'm amused by 'the unpretty woman done up with makeup'. Perhaps you are young enough to insist on the best? :angel:

Don,

I admire your schematic. However, best to set a constant gain and avoid pots, particularly as power amps do not like very low gain settings and compensation should be left fixed!! Further, by using a current source (very clever, too, BTW) on the cathode, you run the tube in constant current and thus could remove some of the lovely tube sound. Why not try it with a simple resistor - it might sound better!

Cheers,

Hugh
 
Hi Hugh
I wanted to say that these distortions have very little to do with each other. Typical odd-order crossover ditortion of solid state amplifier is mainly a low level signal phenomenon, whereas 2nd (and third) harmonic gets vivid with higher level, so one more reason for no masking.
Low order harmonic distortion may be nice. It enhances high frequency contents for higher levels, which can make an illusion of better dynamics, because high freqs are better hearable. Also low orders are mainly musical intervals adding octave, octave+fifth, 2 octaves... These colorations one can like or not and no attitude is moraly better. Just if you like this 'make-up' do 'make-up; because you like 'make-up'. Don't state "my spots are not visible anymore".

regards
 
Update

I tested a half dozen types of triode with this inverse feedback scheme this weekend and it appears that all triodes have fairly similar distortion spectrums, mainly 2nd harmonic added.

I also found that putting some load resistance on the cathode will LOWER the 2nd harmonic (surprisingly), significantly in many cases, 10 dB or more (around .5K to 3 K load thru a large cap). Too low a load resistance makes the 2nd harmonic AND all higher harmonics go up again, so there is some kind of optimum point for each tube type.

I also found that putting the resistive attenuation pot ( to normalize closed loop amplifier gain back to that of the original amplifier resistive attenuation network) on the cathode end rather than the plate end lowers the 2nd harmonic significantly too. Looks like the tubes like to have the maximum plate voltage swing for lowest distorton. (also a surprise)

I changed the DC idle current ( by 2x) thru the tubes and found that increased idle current in some cases increased 2nd harmonic and for other tubes it decreased 2nd harmonic dist. Looks like I need to do a sweep thru a DC idle current range to see if there is some optimum idle current level for minimum distortion for each tube type. Clearly, at some high idle current, one will run out of voltage headroom on the cathode, as the cathode to grid DC bias voltage becomes too low for the input voltage swing of the amplifier.

More work to do!

Don
 
Further Update

I tested distortion versus DC idle current thru a tube (6S4A) in the inverse tube feedback amplifier scheme and found that distortion generally increases with increasing tube idle current and spreads out into the higher harmonics too. This is expected since the cathode to grid bias drops at higher currents, limiting the voltage headroom for the input signal swing.

Optimum cathode load resistance for minimum distortion has a bigger effect at low idle current (drops 2nd harmonic by 20 dB at 5 mA idle on a 6S4A tube using an optimum 2K load versus CCS only)

The optimum load resistance drops with increasing current, but has less effect on dist. (0.775K optimum load for 20 mA idle drops 2nd harmonic by 4 dB versus CCS only)

My previous post about better results using resistive attenuation on the cathode end, rather than the plate end, may be incorrect, I found my test setup for that was changing the input resistance. I will re-test this effect again correctly.

I noticed an interesting effect during testing, The tube 2nd harmonic distortion and this amplifier's internal 2nd harmonic distortion appear to be in opposite phase. By adjusting the tube 2nd harmonic level it is possible to get a near null of 2nd harmonic in the output distortion. This will only work however for amps with this phase of 2nd harmonic. Probably is easier to just change the gain of the N channel versus the P channel outputs in the amp for this nulling effect.

Don
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.