ST70 7199 / 6GH8A amplitude vs frequency

[JRUBIN]

Ive posted a 1 off video showing my issue of the amplitude change of the 7199 outputs in which
the change increases with the frequency. I found something, however even though it corrects the amplitude, the phase does not align, tested with both a 7199 and a 6GH8A

Surely this is masking another imbalance somewhere that needs correction.

The 47K resistors are matched since this video

from an older video

with mod

 

[gadget73]

am I looking at output from the speaker terminals comparing the two channels, or is this on the phase inverter itself? If its unbalanced at 40 khz, does that really matter much? You're well beyond the audio frequency range. By design, the output response rolls off beyond 20khz to keep the amplifier from oscillating. You don't want response out to infinity.

something you can do is to change the cathode side 47k to a 39k, and put a 10k trimmer in series with it. That would let you tweak in the AC balance more precisely if needed.

Sounds like you're introducing some local feedback on the stage. The .05 is on the grid of the 7199. Connecting from that point to the output of the phase inverter would either give some local positive or negative feedback at very high frequencies, depending which side you're on. Basically it would either bump or drop the very high frequency output. Have to be extremely careful if you bump it. Audio amplifiers will actually turn into fairly hot RF oscillators if you get the output back to the input at the right rate. They do a nice job of baking tubes when that happens too.

 

[JRUBIN]

am I looking at output from the speaker terminals comparing the two channels, or is this on the phase inverter itself? If its unbalanced at 40 khz, does that really matter much? You're well beyond the audio frequency range. By design, the output response rolls off beyond 20khz to keep the amplifier from oscillating. You don't want response out to infinity.

something you can do is to change the cathode side 47k to a 39k, and put a 10k trimmer in series with it. That would let you tweak in the AC balance more precisely if needed.

Sounds like you're introducing some local feedback on the stage. The .05 is on the grid of the 7199. Connecting from that point to the output of the phase inverter would either give some local positive or negative feedback at very high frequencies, depending which side you're on. Basically it would either bump or drop the very high frequency output. Have to be extremely careful if you bump it. Audio amplifiers will actually turn into fairly hot RF oscillators if you get the output back to the input at the right rate. They do a nice job of baking tubes when that happens too.

The output you are seeing are those off the phase inverter. (LEFT CHANNEL) I understand that at some point frequency response will roll off and 40K may be way off in the deep end, but I chose 40 as the top end because it matched specification. at on or below 1% THD
indeed the decade resistor on the 47k helped and a trimmer would be a more permanent solution. I cant figure why the amplitude would change on just one of the phase inverted signals. id imagine the response would fall evenly.
Ive used several tubes as the phase inverter to rule out a bad tube as a possibility. all outputs have an 8ohm 100W dummyload, which the 331A feeds off of


In testing further, just to be sure I tied a decade capacitor to both .1mfd capacitors and saw no difference when increasing total capacitance
the .05 also showed little effect with the decade capacitor.
The remaining two caps 82mmf and 390mmf value changes with the decade capacitor did have a noted effect, however it effected both evenly , not providing a remedy to the balance issue at hand.
The 270K resistors both read 300, however are withing 1% of each other
everything before pin 9, effects pin 1 and 8 the same so I imagine this would not provide the solution i would be looking for
the 1K resistors to the EL34's are not precision, but are in spec, i imagine this is not an avenue either
the feedback from the EL34's tie to pin 6 and 3 so it is before phase inversion, again a dead end
the 270K's are matched to 1% or less
aside from the 47K's, which are matched, i see nothing that would seperately effect the seperate phases coming from the 7199. In consideration of a pot at the 47K's , I wouldnt know the balance as it is different at each frequency. my assumption would be to balance at the top end.

what did I miss?

 

[gadget73]

wonder if some of it might be stray lead capacitance from the decade boxes. You're high enough in frequency that it might just be enough to play in here. Might be worth putting in some parts with normal lead length to see if that doesn't change stuff.

Which side is the stronger side, the plate or the cathode output? The cathode side should be lower impedance, so I would expect it to be slightly less affected by any stray capacitance in the leads and whatnot.

Ever swapped the two EL34's? Differences in internal capacitance there could possibly come in to play, though that probably ought to be low enough in value as to not affect anything.

AC balance is one of those things you just have to find a happy place for. I'd probably set it somewhere towards the upper part of the normal audible range and make sure you don't wreck the bottom end performance. I ended up tweaking things back and forth on the Bogen to settle at a happy compromise.

 

[gadget73]

also, just out of curiosity, how does yours stack up relative to these numbers?

Dynaco ST-70: Base Line Testing

 

[JRUBIN]

A bit of foolishness on my part. Lets weed out the errors first.
1. I did not read the specification correctly, power response is measured from 20 cycles to 20KC !!!!!!! LESS THAN 1% measured at 1db below maximum power output (24 to 27 watts MAX)
In a quote from DCGILLESPIE
<quote>
Both Channels -- @ 20 Hz = 31.6W RMS @ 1 db down, THD = 1.9%
Operating @ 1 kHz = 37.5W RMS @ 1 db down, THD = .32%
@ 20 kHz = 34.1W RMS @ 1 db down, THD = 3.3%
</quote>

my 1K values are 16W (.165%) and at the incorrectly measured 32W (.308%)
I dont ahve a 20K value yet but
my 30K values are 16W (2.3%) and at the incorrectly measured 32W (3.2%)

This also resembles the figures from ELPASO TUBE AMPS, and he has a lot more equipment for testing than Id ever hope to have



The other issue is humming of the test frequency coming from the amp. It was very hard to find but I used my iphone spectrum analyzer and found it to be in the forward facing shells of the output transformers. Here is the readout with the iphone sitting on the shell when the frequency output was 3KC

 

[JRUBIN]

I was just replying to that while you had posted

Heres a photo from ELPASO Tube amps testing of a channel

 

[jazbo8]

Thanks for sharing these data, would either of you have the harmonic distortion composition (taken from the spectrum analyser) with the amplifier near clipping?

 

[gadget73]

not I. Don't own a spectrum analyzer or an ST-70.

 

[dcgillespie]

In my base line testing, I was driving both channels at once to get a more accurate picture of how the amplifier performs in actual use, as we generally don't listen to just one channel at a time. With that element factored in, your single channel results are not unlike those of mine.

I think it would be safe to say that what you are really experiencing is a measurement error, rather than an actual error. It typically would not be possible to achieve the low level of distortion you are obtaining if the imbalance were as real as implied. Measurement of a cathodyne phase inverter balance is a tricky thing to do -- particularly at higher frequencies, as the loading on each output must be kept EXACTLY identical, or imbalance will be indicated that is likely not there in actual practice. Check to see that your connecting cables and the input impedance of both channels of your scope are identical relative to both resistive and capacitive values, so that their loading effects will cancel each other out when hung on the inverter.

Dave

 

[JRUBIN]

Hello Dave,

Aside from ensuring that the 47K's matched up I did a test where I took a THD at 20KHZ both with and without the o-scope connected to the .1 MFD capacitors.
The THD was remarkably higher with the Oscope connected which leads me to believe that the scope may be playing a part in what im seeing. This is somewhat of a paradox as the
measurement device is ruining the measurement under test. I should mention now that for the THD tests Ive posted, none of them had the probes connected to the .1uf capacitors and
therefore didn't ruin the results.

Youve noticed that im only testing one channel. I agree both should be under load to represent a more accurate real life measurement. There are two reasons:

1. My 47K's are quite mismatched on the right channel (yes this would stop me from measuring, but not for putting under load
2. And this is the big issue : running at 24WATTS ive noticed, especially at low frequencies, that the bias doesnt hold, it drops. I give the AMP a rest between frequencies to collect itself
but find that im making adjustments at almost each interval. Ive seen selenium rectifiers die, but never saw them exhibit this. I can only surmise one of two things.

A) The power transformer cant handle the load after all these years
B) My isolation transformer cannot operate the whole workbench and the AMP tandem.

I will be testing the B possibility tomorrow with an inline voltmeter from mains to the AMP

Be that as it may, in keeping the bias at the right intervals, ive completed testing at 24 Watts, so I have somewhat of a baseline

Note that at 20 hertz the amp output completely falls apart, ive not added 20hz to the chart. Further note that Im running 6GH8A tubes and have heard that there may be an adjustment needed for them
with regard to CAPS and resistors


Though not optimal, the characteristic curve that ive seen in other results says im in the ballpark
Note that 10K to 20K was NOT done logarithmically and changes the look of the chart, giving the top end a linear appearance, i wanted to get more detail on the high end

Here is the conventional chart with a best effort on 20hz with an average of a few samples... but may in fact be total rubbish. all other values are the same as previous but logarithmic

Also, your document on the ST70 is fantastic..

Regards

Jordan

 

[dcgillespie]

Jordan -- What are you using as a signal source, and to measure THD with? A couple of points if I may --

1. Generally selenium rectifiers -- as used in the ST-70 -- will not fail, this because they don't generate any internal heat due to the extremely low current flow. For all the cry against selenium units (with good cause in high current applications), I've asked for anybody to post of an ST-70 where the selenium unit actually failed. No takers yet, but tons of ST-70s have had them changed out for fear of their existance damaging the space/time continuum. Generally, the bias should hold relatively steady from quiescent, to full power conditions if the AC power is holding steady. Your isolated bench source may in fact be an issue.......

2. The 20 Hz performance is problematic and should be investigated to determine the cause. If AC balance issues have all been eliminated, then consider that Dynaco output transformers are wonderful units, but the original devices do have their issues relative to performance consistency from unit to unit. It would also be a good idea to close couple your source and THD meter, so as to measure the baseline THD of your measurement system at 20 Hz so you know what where you're starting from.

Let us know!

Dave

 

[JRUBIN]

Hello Dave,
From my post in EEVBLOG the following baselines were made. The signal generator is an FY3224S.
I may need to retest 40hz and below based on my own comment. Seems I could not achieve 5.5VRMS for whatever reason. My assumption is that below 40hz the P/P voltage of 20V is exceeded to reach 5.5vrms

The THD meter is a freshly overhauled HP 331A that I spent weeks fixing and calibrating.

The Thread on my efforts is located here including the associated movies http://www.eevblog.com/forum/testgear/my-hp-331a-distortion-analyzer-thread/

The thread focusing on the 331A and the FY3224S is located here http://www.eevblog.com/forum/testge...unction-signal-generator/msg823997/#msg823997

All of my movies on the work for this amplifier can be found at https://www.youtube.com/playlist?list=PLK_zQ6_j4TF13qvrS3i2fw-p9trdjUhVm

Im still curious about the changes, if any, are required to support the 6GH8A tubes vice the 7199's

Im going to offload the test equipment off of the isolation transformer and reserve it only for the ST70, with that ill check the stability of the bias

[from the thread in eevblog]
Further work on my 331A as well as using a scope of the output to tune in the absolute lowest distortion

Input voltage 5.5VRMS except for 40hz and below (around 3.7VRMS MAX) NO LOAD

I REPEAT NO LOAD MEASUREMENT!!!!

10 .205
20 .195
30 .205
40 .205
50 .195
60 .205
70 .205
80 .205
90 .205
100 .21
200 .21
300 .21
400 .205
500 .21
600 .21
700 .208
800 .208
900 .208
1000 .205
2000 .195
3000 .195
4000 .192
5000 .192
6000 .2
7000 .2
8000 .2
9000 .198
10000 .2
15000 .198
20000 .192
25000 .192
30000 .192
35000 .183
40000 .182
45000 .180
50000 .176

 

[dcgillespie]

Very neat restoration thread on your 331! I've had my share of time under the hood of my HP gear as well. For distortion measurements, I use a couple of 339As, which are just stunning pieces of equipment from back in the day. HP is the best.

As for your results however, with a typical straight through mid-band distortion of about .2%, what do you attribute this too? The generator, or the analyzer? If the distortion of the generator is sufficiently low (for vacuum tube work being defined as at least below .1% throughout the audio band, and much lower at mid frequencies), then it would seem that your indication of .2% is high, and certainly so for HP gear. Does your restored unit meet HP specifications now? I am not familiar with the generator you are using, as for distortion measurements, I use the internal oscillator in the 339As.

I'm hardly trying to imply that this is the problem with your ST-70, and I really don't want to get your thread off target from your work on that amp here. I am just keenly aware of how important the capability of the measurement set is when using it to be judge, trial and jury for a piece of gear being worked on.

Dave

 

[jazbo8]

As for your results however, with a typical straight through mid-band distortion of about .2%, what do you attribute this too? The generator, or the analyzer?

According to the posts at EEVblog, isn't the 0.2% about what the FY3224S is able to do? That's pretty far from the HP, below is shot of a well-calibrated HP339A, the THD is just 0.00006%!

 

[JRUBIN]

Yes the FY3224s is nothing amazing. The chart i provided shows the no load direct connection from the generator directly into the 331A.
loading up the 3224 even at 50ohms significantly increases the P/P voltage required to achieve the desired RMS output. BY P/P voltage im implying
the voltage setting on the 3224S and NOT the product. I may be in need of a better function generator that would be up to the task. or do my testing at lower wattage

I am made to understand that the iphone function generator app pushes less than .0001%THD at 20khz at its max 1VRMS. but that only leaves me with .125 watts output

 

[JRUBIN]

No doubt the isolation transformer was the culprit

With a maximum rating of 300 WATTS, I was way overboard. I use it for antique radios and small electronics, not power amps

Plugged into the isolation transformer alone the amp pulled 163 watts

Consider that the AMP plugged into mains at 120.2 volts draws : slightly higher [somehow the isolation transformer bucks it to 117-118v]

179 watts at rest
196 watts at 9VRMS ( 10 watts)
add in the oscilloscope/function generator/ 331A/ and a cooling fan and you see where this is going.........

However even without the other items, im still left with

179 watts at rest
196 watts at 9VRMS ( 10 watts) [yes , now measuring with both channels in to separate 8 ohm dummy loads, switch set to mono]

Thats already 2/3 of the maximum output. let alone 24 watts output

Unfortunately, even with everything else removed from the isolation transformer, the ST70 simply pulls too much current for such a small device to power.

Plugging into mains saw the amp stabilize. There are fluctuations, but it is due to devices within the rest of the house slightly effecting the mains voltage.
but those changes are reflected in my mains volt meter, and i take into account what the average house voltage is.

It stands to reason that the voltage to the 331A also changed, but it should be able to handle such fluctuations with its voltage regulator and well within spec.
the ST70 has no such technology

 

[dcgillespie]

Glad you're making progress!

Jaz -- I think you've got an extra zero in there. A properly operating 339A strapped to itself will typically measure about .0006%, for about -104 db.

Dave

 

[mfaz28]

How about the (almost forgotten) Sylvania 7687 ?

 

[Audiovet]

One small matter I find missing (hope I have not missed relevant commentary!), is the action of the 82pF phase capacitor over the 270K pentode load resistor (on my schematic drawing, in series with 18K). With the ST70 this plus other Cs start to diminish loop gain upward of some 7kHz, and thus NFB, ergo: A rise in distortion. This is what I see reflected in the distortion graphs.