ST70 7199 / 6GH8A amplitude vs frequency
The network you've identified is absolutely one of the main reasons that distortion rises at the higher frequencies -- but also absolutely necessary to keep the amplifier stable into more typical loads (i.e. a speaker) than a simple resistive load represents. All NFB amplifiers need some element to control the closed loop response -- otherwise, the NFB would try to make the response go to the moon. The problem with that is the OPT, where at supersonic frequencies, the phase shift created by the combination of the transformer and the impedance of a real world load connected to it can cause the negative feedback taken from the transformer to become positive feedback. That can then turn a power amplifier into a power oscillator, destroying tweeters and other components. To prevent this from happening, the response of the amplifier with the loop closed is intentionally controlled (tapered off), so that the loop gain is below a value of 1 by the time the frequency reaches the resonant frequency of the OPT, where the problems normally begin. It is the purpose of the step network you've identified to provide the requisite controlled roll off. The best OPTs will have a resonant frequency so high, that the required roll off does not dip down so low as to affect the upper portions of the audio band. Even if it does however, the power required in this region is so low, that the elevated distortion in this portion of the band is really a moot point.
Dave
Dave
I didn't look up the specs, but I would expect that 1v input signal should be enough or nearly enough to drive a stock ST70 to full output power, or at least certainly higher than 1/8 of a watt.
jazbo8
Well-Known Member
The extra zero was no mistake, when I said a well-calibrated 339A, I left out the bit that the unit has been modified for minimum distortion, so it's not spec - but ultra-spec. For more on the modifications, you need to go through the rather long thread over at diyaudio.
Hummmm -- unless they're using some outboard indicator, a distortion level that low is hard if nearly impossible to even resolve with the standard 339A meter and range offerings. Maybe they've been able to squeeze an additional lower distortion range into the design with their modification. I could see a resolution that low for SS work. But it would be well wasted on traditional vacuum tube gear. Impressive results though none the less!
Dave
Dave
jazbo8
Well-Known Member
As shown in the picture, the reading was done by the ShibaSoku automatic distortion analyzer. You are right, the guys that worked on the mods for the 339A are primarily working on SS designs, since that level of THD is just wasted on tube gear. But I posted the picture to show that vintage lab test equipment are often better performers than the typical low-end DSP-based test gear.
As it turns out the iphone output is not 1VRMS max... and too bad, because its sine wave distortion is lower than any equipment I have.
That being said, the iphone provides a maximum of power such that the output if the ST70 measured
(5.0vRMS) 20hz 3.125w
(5.7vRMS) 30hz 4.06w
slow upswing to 7vRMS (biggest output power change from iphone)
(7.0vRMS) 100hz 6.12w
(6.4vRMS) 300hz to 200hz 5.12w
(6.2vRMS) 400hz to 9K 4.8w
(6.0vRMS) 10K to 18K 4.5w
(5.8vRMS) @20K 4.2w
Even with this you can clearly see the launch starting at 1000hz to 20k the same exact findings at 24 watts on the last test.
The question is what slight changes are to be made to push the feedback circuit out a bit, maybe to 5000hz
A further observation shows that the amp held up at 20hz on the iphone, though it is not determined if that is becuase of the iphone or the 3.125w output of the amp, so ill have to test with the function generator at the same output rms as the iphone
Also on a different note...... though the voltage sag on the bias was greatly reduced by pulling it off of the isolation transformer it was still noticeable by .10 to .15 volts measuring from the 1.56v testpoint
On a whim, I replaced the GZ34, sire enough the sag was gone..... while the GZ34 has nothing to do with the bias, it has much to do with the EL34's. This leads me to believe that the bias want drifting.
the current outputs of the EL34's were... I may test this theory at the power supply for the grid bias
That being said, the iphone provides a maximum of power such that the output if the ST70 measured
(5.0vRMS) 20hz 3.125w
(5.7vRMS) 30hz 4.06w
slow upswing to 7vRMS (biggest output power change from iphone)
(7.0vRMS) 100hz 6.12w
(6.4vRMS) 300hz to 200hz 5.12w
(6.2vRMS) 400hz to 9K 4.8w
(6.0vRMS) 10K to 18K 4.5w
(5.8vRMS) @20K 4.2w
Even with this you can clearly see the launch starting at 1000hz to 20k the same exact findings at 24 watts on the last test.
The question is what slight changes are to be made to push the feedback circuit out a bit, maybe to 5000hz
A further observation shows that the amp held up at 20hz on the iphone, though it is not determined if that is becuase of the iphone or the 3.125w output of the amp, so ill have to test with the function generator at the same output rms as the iphone
Also on a different note...... though the voltage sag on the bias was greatly reduced by pulling it off of the isolation transformer it was still noticeable by .10 to .15 volts measuring from the 1.56v testpoint
On a whim, I replaced the GZ34, sire enough the sag was gone..... while the GZ34 has nothing to do with the bias, it has much to do with the EL34's. This leads me to believe that the bias want drifting.
the current outputs of the EL34's were... I may test this theory at the power supply for the grid bias
heavy loading on the power transformer will drop the voltage output on all of the secondaries. Probably not really an effect of the rectifier so much as a limitation of the power supply as a whole.
Also, just a mention but you won't see a steady 1.56 across the cathode resistor when the amplifier is passing signal. You'll get AC+DC on there and it will skew the readings. If you want to see if the bias supply is sagging, you'd have to measure voltage right at the output of the bias supply.
Also, just a mention but you won't see a steady 1.56 across the cathode resistor when the amplifier is passing signal. You'll get AC+DC on there and it will skew the readings. If you want to see if the bias supply is sagging, you'd have to measure voltage right at the output of the bias supply.
Yes the 1.56 is a product. Im going to be measuring the bias itself independently from the 1.56 test port.
All of the bias testing mentioned had no signal on the input with a shorting plug on the rca jack.
if the test ports change and the bias remains stable, then what im looking at is a change in the total current of the EL34's on the channel.
That being said a current change can be caused by the transformer as well as the GZ34.
Filament voltage, plate voltage etc.
However , even though all transformers drop their output to some degree under load. When the load is gone it should swing back. and not after 10 minutes.... UNLESS... heat is a factor
however less doubtfull than the other possibilities
That being said..........
At constant wall voltage...... with a constant bias........ with a constant AC input on the GZ34.... the transformer cannot be an issue.
This would be the next logical test.
As the curve still has a takeoff with the same characteristics, once this voltage issues is solved
the iphone test will be done again in the exact same way using a 7199 vice 6gh8a
If the curve remains yet the same again, i will assume that the problem lies within the feedback circuit and will make a minute pf adjustment to see if the curve moves. If it does ill proceed from there and find the optimum value.
Also as an afterthought , i want to remeasure the rms voltage going to the el34's after the phase inverter and then flip the probes. at 10k to 20khz If the flipping of the probes matches the flip on the oscope display ill know that it is capacitance or loading in the probes causing the issue. If it does not change ill know that it is in the circuit itself
Last edited:
How much is it changing? I've never seen bias voltage remain absolutely rock solid, but large amounts of drift might indicate a problem. Most of my stuff will stay solid easily within 2%. Also keep in mind that component values can shift as things warm up, including that cathode resistor, the power supply filters, and the windings within the transformer itself, plus minor fluctuations in line voltage. All of those things can contribute to slightly shifting operating parameters.
JRUBIN said:
JRUBIN,
Not sure that I interpret correctly, but one doesn't go about it this way. Referring you back to my post #20 and the subsequent explanation of Dave, #22. The 82pF value is (or should be) primarily selected to give stable output with NFB, not a desirable frequency response. No problem in fiddling with that, but the yardstick should be the appearance of, say a 5 kHz square wave at the output, regarding overshoot/undershoot (a quick-and-easy frequency response check). If the cap needs to be 82pF for such stability, there is little you can do - one cannot risk 'better' loop response at the cost of possible momentary instability at the output.
[Many amplifiers (even s.s. ones) show the typical D-graph you found and worse! The problem is to design the amp inherently such that loop gain remains constant to some 10kHz. Many do not succeed, at the cost of some stridency at h.f. ... But such a discussion falls outside the purpose of this thread.]
JRUBIN,
Not sure that I interpret correctly, but one doesn't go about it this way. Referring you back to my post #20 and the subsequent explanation of Dave, #22. The 82pF value is (or should be) primarily selected to give stable output with NFB, not a desirable frequency response. No problem in fiddling with that, but the yardstick should be the appearance of, say a 5 kHz square wave at the output, regarding overshoot/undershoot (a quick-and-easy frequency response check). If the cap needs to be 82pF for such stability, there is little you can do - one cannot risk 'better' loop response at the cost of possible momentary instability at the output.
[Many amplifiers (even s.s. ones) show the typical D-graph you found and worse! The problem is to design the amp inherently such that loop gain remains constant to some 10kHz. Many do not succeed, at the cost of some stridency at h.f. ... But such a discussion falls outside the purpose of this thread.]
Follow -
Even though somewhat OT, I need also mention that the output stability depends on the NFB resistor cap bypass. There appear to be several schematics on internet; mine shows these as C12 = 390pF (with R16 = 1K as feedback resistor). C12 will thus also need to be calculated/chosen to obtain correct stability. One may be able to decrease the 82pF and compensate with C12 to restore desired frequency response; this might be possible/necessary because of different OPT characteristics.
But again warning: One cannot do this randomly; it requires knowledge of NFB basics.
Even though somewhat OT, I need also mention that the output stability depends on the NFB resistor cap bypass. There appear to be several schematics on internet; mine shows these as C12 = 390pF (with R16 = 1K as feedback resistor). C12 will thus also need to be calculated/chosen to obtain correct stability. One may be able to decrease the 82pF and compensate with C12 to restore desired frequency response; this might be possible/necessary because of different OPT characteristics.
But again warning: One cannot do this randomly; it requires knowledge of NFB basics.
Please note that I am reading all of the responses, but can only do so much testing in one sitting....
Lets dispense with the low hanging fruit.
Phase inverter output - uneven signal issue:
Sure enough, swapping the scope probes showed that the larger waveform followed the probes and is not an issue with the AMP!!!!!
1000K @ 9vrms 30 second runs
Bias voltage after filter, before circuit board.
BIAS TESTPORT
-39.4 1.56
-39.3 1.55
-39.3 1.548
-39.1 1.53
-39.2 1.54
-39.1 1.53
-39.0 1.52
-38.9 1.51
-38.8 1.51
This shows that the current read as voltage at 1.56 test port is directly related to the bias voltage which was suspected.
Suspecting that this was a fluctuation of input ac voltage to the selenium rectifier, i decided to test the power transformer. To get a better resolution I chose
one of the AC wires to the rectifier, as there would be a greater fluctuation with higher voltage
First starting at @ 9VRMS on output 30 second runs, bias readjusted to 1.56 slashes denote fluctuation
RECT in TESTPORT RECT during load
365 1.56 362/3
364/5 1.548 361/2
364/5 1.548 363/2
364/5 1.544 361/2
cranked it to 10.8VRMS on output no readjust
363/2 1.52 359/60
363/2 1.52 359/8
362/1 1.51 358/7
363/2 1.52 357/8
363/2 1.52 357/8
363/4 1.53 359/60
The power transformer is dropping voltage over time under load, and this is cascading through the rest of the unit. Is this symptomatic of a tired transformer?
Back to the output of the phase inverters.
So now i'm measuring with the same probe first with the o-scope measuring P/PV (F & R denote the EL34 in the channel front/rear)
HZ
500 F 6.92
500 R 7.00 +.08
1000 F 6.88
1000 R 6.92 +.04
5000 F 6.80
5000 R 6.84 +.04
10000F 6.68
10000R 6.80 +.12
15000F 6.52
15000R 6.68 +.16
20000F 6.44
20000R 6.56 +.12
I did it once again with my fluke RMS meter for good measure
500 F 5.97
500 R 6.00 +.03
1000F 5.96
1000R 5.99 +.03
5000F 5.85
5000R 5.91 +.06
10000F 5.66
10000R 5.77 +.11
15000F 5.37
15000R 5.56 +.19
20000F 5.13
20000R 5.33 +.20
So obviously, the rear output is higher than the front on the phase inverter. I imagine the POT recommendation of the 1% resistor would fix that
notice though that as the frequency increases, there still is a growing difference in amplitude. I could make a vague assumption that if this were better balanced
the differences at high frequency would be smaller, thats just a guess though
Lets dispense with the low hanging fruit.
Phase inverter output - uneven signal issue:
Sure enough, swapping the scope probes showed that the larger waveform followed the probes and is not an issue with the AMP!!!!!
1000K @ 9vrms 30 second runs
Bias voltage after filter, before circuit board.
BIAS TESTPORT
-39.4 1.56
-39.3 1.55
-39.3 1.548
-39.1 1.53
-39.2 1.54
-39.1 1.53
-39.0 1.52
-38.9 1.51
-38.8 1.51
This shows that the current read as voltage at 1.56 test port is directly related to the bias voltage which was suspected.
Suspecting that this was a fluctuation of input ac voltage to the selenium rectifier, i decided to test the power transformer. To get a better resolution I chose
one of the AC wires to the rectifier, as there would be a greater fluctuation with higher voltage
First starting at @ 9VRMS on output 30 second runs, bias readjusted to 1.56 slashes denote fluctuation
RECT in TESTPORT RECT during load
365 1.56 362/3
364/5 1.548 361/2
364/5 1.548 363/2
364/5 1.544 361/2
cranked it to 10.8VRMS on output no readjust
363/2 1.52 359/60
363/2 1.52 359/8
362/1 1.51 358/7
363/2 1.52 357/8
363/2 1.52 357/8
363/4 1.53 359/60
The power transformer is dropping voltage over time under load, and this is cascading through the rest of the unit. Is this symptomatic of a tired transformer?
Back to the output of the phase inverters.
So now i'm measuring with the same probe first with the o-scope measuring P/PV (F & R denote the EL34 in the channel front/rear)
HZ
500 F 6.92
500 R 7.00 +.08
1000 F 6.88
1000 R 6.92 +.04
5000 F 6.80
5000 R 6.84 +.04
10000F 6.68
10000R 6.80 +.12
15000F 6.52
15000R 6.68 +.16
20000F 6.44
20000R 6.56 +.12
I did it once again with my fluke RMS meter for good measure
500 F 5.97
500 R 6.00 +.03
1000F 5.96
1000R 5.99 +.03
5000F 5.85
5000R 5.91 +.06
10000F 5.66
10000R 5.77 +.11
15000F 5.37
15000R 5.56 +.19
20000F 5.13
20000R 5.33 +.20
So obviously, the rear output is higher than the front on the phase inverter. I imagine the POT recommendation of the 1% resistor would fix that
notice though that as the frequency increases, there still is a growing difference in amplitude. I could make a vague assumption that if this were better balanced
the differences at high frequency would be smaller, thats just a guess though
transformers don't really "get tired". They will get hot with use, which changes the DC resistance of the windings, which changes the output voltage. I would expect it to change for a bit as things warm up, but eventually it should reach a steady-state temperature and things ought to be reasonably stable at that point.
I'm curious, what iPhone app are you using? This one has settings for 0.7, 1.0 and 1.4 V:
http://sgenerator.scorpionzzz.com/en/index.html
I have played with it a bit, and the settings seem to match what I read on a scope.
Also, where does the distortion level you quoted come from?
David
The app is called signal gen in the app store amplitude is measured by gain from -48db to 0 db max. I downloaded sgenerator and will give it a look.... thanx
HMMMM, given the very small value for the capacitor....... can the resistor be adjusted for the capacitor given an equation. or is this not possible and the capacitor must be adjusted.
I know that measuring caps in circuit is an almost worthless endeavor, however i noticed, with the tubes removed from the driver board that
Left .0980nf Right .040nf
Left .076nf Right .065nf