This is in regard to my custom home-built ST-70. After reading a lot about this, I am considering removing the Zobel stability network (series RC shunting to ground) between the input stage and the phase splitter. And in addition removing the cap in the GNFB loop. As I understand it, these are meant to reduce the possibility of HF oscillation, but may not really be necessary. My GNFB is approximately 15db (which I believe is fairly moderate). The OPT's are Hammond 1650K.
So I really have two questions, 1) do you think this is a wise mod (ie. will it really improve the sound?), 2) If I do this, how can I tell if it has introduced any adverse behavior? Please keep in mind that I don't have any test equipment, other than my old ears.
Thanks,
Dave
So I really have two questions, 1) do you think this is a wise mod (ie. will it really improve the sound?), 2) If I do this, how can I tell if it has introduced any adverse behavior? Please keep in mind that I don't have any test equipment, other than my old ears.
Thanks,
Dave
Depends on the circuit and output transformer. In the original ST70 circuit, removing the RC on the input pentode's plate will cause strong ringing. The sound will have a hard, pinging quality. A scope is necessary to adjust the capacitor value to be smaller but still not cause ringing, but don't change the resistor's value. The speaker load can affect the tuning as well.
Last edited:
spot on.....Don't do it. If you tinker with the values, one could end up destroying a precious tweeter without knowing one has done it....and a mystery orange glowing anode from an output tube. One of the tell-tale signs of amp instability in reactive loudspeaker loads is momentary squealing when the amp is turned on/off. It may even pick up an AM station too.
rich
It is hard to tell how much different the phase response versus frequency, and the frequency response of the Hammond 1650K transformers are, versus the original Dyna A470 output transformers.
It may turn out that you need to find a person in your area (and highly suggested for you to do so) who has the following:
An Oscilloscope
A square wave generator
A non inductive 8 Ohm power resistor load The knowledge to know how to adjust the feedback loop(s).
There is a dominant RC pole in the pentode plate circuit. There is a feedback capacitor from the (correct) ultra linear taps back to the pentode circuit. There is the global feedback network. All of these, plus the A470 Output transformer are what makes the Dyna Stereo 70 as stable as it is with different loudspeakers.
If I remember, there was about 26 dB of global feedback in the ST70.
If done and adjusted properly, it should be possible to use 10 dB or so less global feedback. That actually should be easier to optimize than 26dB. But the frequency response, distortion, and damping factor will likely be slightly worse. However, it might make the amp better with some types of loudspeakers that have more reactive loads.
In some cases, it may be better to use one tap lower, i.e. 8 ohm speaker on the 4 Ohm tap.
It may turn out that you need to find a person in your area (and highly suggested for you to do so) who has the following:
An Oscilloscope
A square wave generator
A non inductive 8 Ohm power resistor load The knowledge to know how to adjust the feedback loop(s).
There is a dominant RC pole in the pentode plate circuit. There is a feedback capacitor from the (correct) ultra linear taps back to the pentode circuit. There is the global feedback network. All of these, plus the A470 Output transformer are what makes the Dyna Stereo 70 as stable as it is with different loudspeakers.
If I remember, there was about 26 dB of global feedback in the ST70.
If done and adjusted properly, it should be possible to use 10 dB or so less global feedback. That actually should be easier to optimize than 26dB. But the frequency response, distortion, and damping factor will likely be slightly worse. However, it might make the amp better with some types of loudspeakers that have more reactive loads.
In some cases, it may be better to use one tap lower, i.e. 8 ohm speaker on the 4 Ohm tap.
Try any feedback scheme you want. Just not on my speakers.
I will have to use the scope, generator, and resistive load first.
Only after that, when I connect to loudspeakers I still watch the scope (my hearing does not go to as high of a frequency as my loudspeakers, but I do not want to burn out the tweeters).
I will have to use the scope, generator, and resistive load first.
Only after that, when I connect to loudspeakers I still watch the scope (my hearing does not go to as high of a frequency as my loudspeakers, but I do not want to burn out the tweeters).
Last edited:
The problem with the ST-70 circuit is one shared with many circuits: the designer has decided to include a laggy band limited piece of iron inside a GNFB loop - which needs to avoid and laggy band limited device in order to function correctly. This is a hangover from the faulty 'Williamson' method of GNFB.
The phase lag can turn your ST-70 into a nice oscillator to blow out your tweeters so that's what the RC is for - to damp down HF in the loop.
While it's not sabotaging your GNFB loop the OPT is band limiting it so that the bass and treble don't get as much feedback as the midrange, which is unfortunate because they are the areas that generally need it most.
Then if you increase the feedback to compensate the lag turns your ST-70 into an oscillator again. All of these problems are caused by the OPT being in the loop.
Therefore the only way to avoid these issues is to remove the laggy, band limited device from your feedback loop in the same manner that a SET amplifier does. However you'll still be wanting the same power as the original ST-70 had originally, so just put the GNFB loop around the driver and power tube - or all the tubes - instead. Do this to turn your multiple tubes into a 'superTriode' - i.e. a collection of tubes with GNFB around them to create a single linear, low drive impedance unit - like a perfect triode in fact.
This will then drive the OPT correctly with no lag or bandwidth issues so oodles of NFB can be used, and finally the OPT can be driven properly and will sound the best it can with full stability.
The phase lag can turn your ST-70 into a nice oscillator to blow out your tweeters so that's what the RC is for - to damp down HF in the loop.
While it's not sabotaging your GNFB loop the OPT is band limiting it so that the bass and treble don't get as much feedback as the midrange, which is unfortunate because they are the areas that generally need it most.
Then if you increase the feedback to compensate the lag turns your ST-70 into an oscillator again. All of these problems are caused by the OPT being in the loop.
Therefore the only way to avoid these issues is to remove the laggy, band limited device from your feedback loop in the same manner that a SET amplifier does. However you'll still be wanting the same power as the original ST-70 had originally, so just put the GNFB loop around the driver and power tube - or all the tubes - instead. Do this to turn your multiple tubes into a 'superTriode' - i.e. a collection of tubes with GNFB around them to create a single linear, low drive impedance unit - like a perfect triode in fact.
This will then drive the OPT correctly with no lag or bandwidth issues so oodles of NFB can be used, and finally the OPT can be driven properly and will sound the best it can with full stability.
Dyna said this is to help balance the phase inverter at high frequencies. "An internal capacitive feedback loop balances the phase inverter at the highest frequencies, and the arrangement provides accurately balanced driving signals to the output tubes" See p.1 here: http://www.thetubestore.com/lib/the...promocode=&gc=clear&promocodeaction=overwrite
Last edited:
The phase inverter arms need to see the same balanced loading conditions. That condition can start to break down for a number of reasons - and all to do with the next stage (the output stage), not with the inverter stage itself in the Dynaco Mk 4.
If the output stage operates in class A, with lowish signal conditions, then the changing plate and screen voltages for both sides are balanced, and so the effective EL34 grid capacitance is balanced, assuming the output transformer winding section shunt capacitances are similar. If one side starts to move in to grid conduction, or its anode voltage starts to compress, or the common cathode voltage changes, then imbalance starts to appear.
Its worth noting that the Williamson configuration inserted a driver stage between the inverter and output stage, for the very reason of maintaining balanced conditions on the inverter to very high frequencies. Cost cutting by amp manufacturers easily led to the driver stage being removed, and if needed some new means to alleviate the collateral damage.
Clarification please. I am specifically referring to the capacitor in the global negative feedback loop, not the one in the input stage Zobel network. The former (as I understand it) has to do with phase compensation in the feedback loop, whereas the latter has to do with limiting high frequency bandwidth, both for stability, but operate differently.
My question specifically has to do with the connection of the feedback capacitor to the OPT. Hafler connected it to the primary (I assume in order to bypass the phase anomalies caused by the OPT). However all "modern" amps parallel the capacitor with the feedback resistor connected to the output. It seems to me that his approach was the correct one, but it sure hasn't passed the test of time. Your thoughts?
My question specifically has to do with the connection of the feedback capacitor to the OPT. Hafler connected it to the primary (I assume in order to bypass the phase anomalies caused by the OPT). However all "modern" amps parallel the capacitor with the feedback resistor connected to the output. It seems to me that his approach was the correct one, but it sure hasn't passed the test of time. Your thoughts?
Negative feedback method 1. (M1)
The small cap from the Ultra Linear tap back to the negative feedback node is high frequency compensation. The resistor from the secondary back to that same negative feedback node is for low and mid frequencies.
Negative feedback method 2. (M2)
A parallel resistor and capacitor from the secondary, back to the negative feedback node.
The difference is:
Method 1. The feedback has high frequency lags due to the leakage reactance 'Across' the primary windings: plate 1, UL tap 1, UL tap 2, and plate 2.
Method 2. has high frequency lag due to the leakage reactance 'From' the primary 'To' the secondary.
The loudspeaker load may be capacitive at high frequencies, and most speaker wires are capacitive at high frequencies. It will reflect that back from secondary to primary. But since there is leakage (inductive) reactance from the secondary to the primary, there is less effect on the high frequency negative feedback when it is taken from the UL tap (M1), than when it is taken from the secondary (M2).
3. The series RC that loads the pentode plate is the dominant pole frequency, and rolls off the gain at very high frequencies.
At a high enough frequency, the phase lag in the amplifier causes the negative feedback to become positive feedback.
The dominant pole reduces the gain at high frequencies, so that the gain and positive feedback does not cause the amp to be an oscillator.
One definition of an oscillator: positive feedback plus more than unity gain.
Does that help?
The small cap from the Ultra Linear tap back to the negative feedback node is high frequency compensation. The resistor from the secondary back to that same negative feedback node is for low and mid frequencies.
Negative feedback method 2. (M2)
A parallel resistor and capacitor from the secondary, back to the negative feedback node.
The difference is:
Method 1. The feedback has high frequency lags due to the leakage reactance 'Across' the primary windings: plate 1, UL tap 1, UL tap 2, and plate 2.
Method 2. has high frequency lag due to the leakage reactance 'From' the primary 'To' the secondary.
The loudspeaker load may be capacitive at high frequencies, and most speaker wires are capacitive at high frequencies. It will reflect that back from secondary to primary. But since there is leakage (inductive) reactance from the secondary to the primary, there is less effect on the high frequency negative feedback when it is taken from the UL tap (M1), than when it is taken from the secondary (M2).
3. The series RC that loads the pentode plate is the dominant pole frequency, and rolls off the gain at very high frequencies.
At a high enough frequency, the phase lag in the amplifier causes the negative feedback to become positive feedback.
The dominant pole reduces the gain at high frequencies, so that the gain and positive feedback does not cause the amp to be an oscillator.
One definition of an oscillator: positive feedback plus more than unity gain.
Does that help?
Last edited:
Both methods are 'correct', but each is a compromise. One helps stability, but compromises on distortion by not taking all the feedback from the output. The other compromises stability but minimises distortion. Which one to choose depends on various things, like how stable is the circuit and how high in quality is the OPT - the better the OPT the less need to compromise on distortion in order to achieve stability.
The issue is really cost of manufacturing the output tranny. Our perfectionist well matched and well balanced multisectioned isn't visible here, as it appears the secondary is simply one winding with taps and the global nfb taken from the outer. Any seasoned engineer will immediately spot that the leakage inductance esp with the 4 ohm winding has a poorer coupling with the whole primary with a view of misbalancing the leakage components. So a leakage inductance spike will be noticed with a square wave with that section of winding with poor physical coupling, as the gnfb is taken from the 16 ohm complete winding.
There is alot of hit and miss when it comes to dealing with stray parasitic components with many designers adopting strategies applicable to the culprit windings. The small cap on that section of screen is a sure indication of a slight misbalance in the parasitics and worse when it comes to correcting any phase differences between the other halves. This last bit becomes immediately obvious that one is taking on the devil here, so we know a good well balanced sectionalised transformer is an expensive time-made item as an attempt has been made to neutralize the parasitics. As for messing with small capacitors that can effect the response above audibility ? Again experts will maintain esp when it comes to snubbers across transformers, many can detect an audible differences even though their values are minimal. Input transformers are the easiest to play with. It has been well known that correcting a particular transformer response at 50kHz, that effect can be hear downstream in audio frequencies in how the harmonics play out.
One of the easiest lab tests I do to check nasty phase issues is to set up the Lissajous test with the XY on the scope that has a coupled timebase and notice when the ellipse really starts. This relatively easy test seems to have vanished over the decades and the important knowledge it provides; in the 1950's I used it a lot as it was part of the radio engineers curriculum, but now ?
Keep at it !
rich
Some amplifiers are designed and tested on an 8 Ohm load resistor. Then they are connected to loudspeakers and listened to. It works. Sometimes, someone finds a speaker and speaker cable where it does not work.
One tradeoff of negative feedback is open loop gain to closed loop gain ratio. If the ratio is large, the distortion is reduced more than with a small ratio. A large ratio also gives a better frequency response. A large ratio also gives a better damping factor. This all tends to be good for an 8 Ohm load resistor. But all other things being equal, the stability can get worse, especially on some loads (speakers). Speakers vary, and some can challenge large feedback ratios. What is a 'large ratio' depends on the amp topology, etc., and particularly on the output transformer.
This problem is not limited to tube amps, it also applies to solid state amps that do not even have an output transformer. There was a whiz-bang super low distortion solid state amp that a company produced and sold. It tested great on a resistive load. Unfortunately it used a very large open to closed loop ratio; it was not stable with loudspeakers. All of the amps (or most all) were returned for repair. Some tweeters also may have needed to be replaced.
An additional problem of amplifiers is Transient Intermodulation Distortion (TMD). This was written about by Matti Otala; and by a second person from Finland. Suddenly, lots of solid state amplifier manufacturers were claiming they were using lower amounts of negative feedback to keep the TMD low.
Build an amplifier that intrinsically has fairly low distortion, and that has a fairly good frequency response, and if possible that has a fairly good damping factor. That is easier said than done. But then you will only need to apply a small amount of negative feedback (or you may even prefer not to use negative feedback at all).
One tradeoff of negative feedback is open loop gain to closed loop gain ratio. If the ratio is large, the distortion is reduced more than with a small ratio. A large ratio also gives a better frequency response. A large ratio also gives a better damping factor. This all tends to be good for an 8 Ohm load resistor. But all other things being equal, the stability can get worse, especially on some loads (speakers). Speakers vary, and some can challenge large feedback ratios. What is a 'large ratio' depends on the amp topology, etc., and particularly on the output transformer.
This problem is not limited to tube amps, it also applies to solid state amps that do not even have an output transformer. There was a whiz-bang super low distortion solid state amp that a company produced and sold. It tested great on a resistive load. Unfortunately it used a very large open to closed loop ratio; it was not stable with loudspeakers. All of the amps (or most all) were returned for repair. Some tweeters also may have needed to be replaced.
An additional problem of amplifiers is Transient Intermodulation Distortion (TMD). This was written about by Matti Otala; and by a second person from Finland. Suddenly, lots of solid state amplifier manufacturers were claiming they were using lower amounts of negative feedback to keep the TMD low.
Build an amplifier that intrinsically has fairly low distortion, and that has a fairly good frequency response, and if possible that has a fairly good damping factor. That is easier said than done. But then you will only need to apply a small amount of negative feedback (or you may even prefer not to use negative feedback at all).
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.