DC Restoration & Blocking Distortion: Simulation and Analysis

[gadget73]

thanks for the writeup. I'll give it a second read when I have the time to really try and process it. I'll also try tacking in a couple diodes to one of my amps to see what really happens here. I've got a fixed bias amp that is reasonably easy to operate on, so that should serve. I don't have digital test gear or anything with interfaces to the computer, so the best I'll be able to do is post pics of the scope and record THD measurements.

 

[Retrovert]

thanks for the writeup. I'll give it a second read when I have the time to really try and process it. I'll also try tacking in a couple diodes to one of my amps to see what really happens here. I've got a fixed bias amp that is reasonably easy to operate on, so that should serve. I don't have digital test gear or anything with interfaces to the computer, so the best I'll be able to do is post pics of the scope and record THD measurements.

The low frequency response is the major beneficiary. It would be helpful if you could measure at, say, 20 Hz, 30, 40 Hz, 50 Hz and 100 Hz instead of at 1 kHz (or thereabouts), which is the standard for many harmonic distortion analyzers.

Note: using only a diode without a bias source causes the baseline to be 0.7 V above ground. That can cause blocking distortion because it allows a DC voltage at the grid. So any experiments may not demonstrate the best possible results. The figures show what happens at that 0 V crossing.

 

[Retrovert]

Another issue, which I will annotate into the diode section as soon as I have the chance.

A diode has a non-linear voltage drop which varies with current. So it can be anywhere from 0 to 0.7 VDC, or whatever the drop is for the diode. (The actual voltage drop varies with the technology. LEDs can have a drop of roughly 2 to 5 volts depending upon the materials used. It is not clear to me how much current is involved, but it is likely very small. This may cause the diode to behave more like an ideal device with 0 V drop. At this point I do not know and I lack the test equipment to measure it.

Any insights?

 

[nerdorama]

signing up. Thanks for sharing all your work.
John

 

[dcgillespie]

Carlyle took the concept far enough to say that properly adjusted, it would even cause R/C coupling to generate class AB2 operation of the output stage -- although I question how effective it might be at doing this. It would be interesting to see what the real world performance improvements would be, when an actual amplifier like the Dynaco ST-70 (your example) had its performance measured both before, and after the clamps were installed.

Very nice work and presentation!

Dave

 

[dcgillespie]

Retro -- In your application with the ST-70 -- how are the diodes doing anything more than the control grids are already doing in the first place? The control grids inherently clamp R/C coupled positive (above ground) signals, which is what causes the blocking distortion in the first place: Since the coupling cap cannot drive the grid positive, the cap then charges on (attempted) positive going (above ground) peaks, which generates the basic problem. Are not the diodes shown only reinforcing what the control grids are already inherently doing?

Dave

 

[Retrovert]

Carlyle took the concept far enough to say that properly adjusted, it would even cause R/C coupling to generate class AB2 operation of the output stage -- although I question how effective it might be at doing this.

Carlyle, it should be noted, was using a fixed-bias scheme at –12 VDC and he was eliminating the negative peak, not the positive.

I believe the bias change and focus on the negative peak led him to comment about driving the grid into AB2. But, having said that, he also said he hadn't tried it.

For those of you unfamiliar with this, aamplifier classes like "A", "B', "AB", have two subclasses "1" and "2". Subclass 1 is no grid current flow and 2 has grid current flow. Subclass 2 is generally avoided because grid current heats the grid, potentially damaging it (this is why transmitter tubes which are specifically designed for grid current have very heavy grid wires and heat sinks), and creating distortion.

It would be interesting to see what the real world performance improvements would be, when an actual amplifier like the Dynaco ST-70 (your example) had its performance measured both before, and after the clamps were installed.

You are correct. I will be doing this when I rebuild my ST-70, but the real world conspires to keep me away from my HiFi projects. This winter it will happen.

Very nice work and presentation!

Thank you. Creating all of this was entirely too much work. Next time I think I'll stick to proving things to myself.

I still must finish that monograph and get it out there. I have an entire analysis of DC restoration in televisions that wasn't included.

 

[primosounds]

Holy cow retro man, an impressive work in progress. Kudos to you for an indepth analysis of the problem. Do you know if blocking distorting is also a problem in single ended amps? I am working on SEP amps to be precise.

 

[Retrovert]

Retro -- In your application with the ST-70 -- how are the diodes doing anything more than the control grids are already doing in the first place? The control grids inherently clamp R/C coupled positive (above ground) signals, which is what causes the blocking distortion in the first place: Since the coupling cap cannot drive the grid positive, the cap then charges on (attempted) positive going (above ground) peaks, which generates the basic problem. Are not the diodes shown only reinforcing what the control grids are already inherently doing?

No, the diodes do two things: prevent the grids from being driven positive AND prevent the coupling capacitor from charging to a DC value which both creates a positive bias and asymmetrically alters the waveforms on the positive excursion.

Explanation for others about positive grid issues.

When the grid is driven positive it makes the grid highly attractive to the nearby cathode electrons than the more-distant plate. (On the scale of the tube microstructure, of course.) This attractiveness causes current to preferentially flow through the grid back to the coupling capacitor and to the previous stage's plate and through the grid leak resistor to ground. This current flow, like all current flow in non-superconducting wires, causes heating. That heating alters the grid characteristics, causing damage. This is why red-plating a tube can melt bits of the grid.

Beyond the heating, the positive grid also attracts ions, typically residual gas which emerged from the metal structures, mica, and cathode coating, either because of natural diffusion or being baked out of oxides. That makes the grid even more attractive. Most tubes are not designed for positive grids (see my above comments about transmitting tubes and heat sinks), and the tube's characteristics become non-linear.

So a positive grid and current both alter the grid's behavior and cause distortion. This is why one of the rules of HiFi is: Thou shalt not drive thine grids positive, lest the Tube Gods destroy thine clean output signal and visit upon thee the plague of distortion and also potentially destroy thine tubes before their normally appointed hours, for the Tube Gods are mighty and fickle, and may do that to spite thee for not honoring the Holy Dictates of the Datasheet, hallowed be its name, and, by the way, thou shalt also honor the Holy Dictates of Screen Dissipation, also in the Datasheet, just in case thou felt like blaspheming there as well, thou hast been warned.

So that's a different issue than the coupling capacitor charging.

When the coupling capacitor charges it temporarily maintains the grid at a constant DC level until that charge dissipates. That prevents the AC signal from the previous stage from controlling the plate. Instead we have the plate at a fixed DC voltage until that capacitor discharges. The only way that can happen is back to the previous stage during the AC reversal (it is at the plate voltage so this is unlikely) and through the grid leak resistor (primary means). That's why the blocking distortion has a bias excursion time controlled by the RC constant. Bigger capacitors extend longer, bigger resistors slow the discharge rate.

So we have both (a) positive grid and (b) DC bias with an RC decay time.

While a larger coupling capacitor better passes lower frequencies and compresses them less, when the coupling capacitor is too large (this depends on both the capacitor and the grid leak resistor) the bass response suffers, becoming flabby and muddy. A smaller capacitor better passes the highs at the expense of bass, which is less compressed but is attenuated. So that's why the first response to increase capacitance (from, say, 0.22 µF to 0.47 µF) can cause significant bass issues if the grid leak resistor is not accordingly adjusted. This is the reason why we don't see 100 µF capacitors being used for interstage coupling, and why it is common to see 0.1 µF and rare to see anything larger than 0.22 µF or 0.47 µF. Hence the values I chose for simulation.

That make sense? (Question for both Dave and anyone who hasn't studied this issue nearly as much.)

 

[Retrovert]

Holy cow retro man, an impressive work in progress. Kudos to you for an indepth analysis of the problem..

Thanks. Science is our friend, but doesn't always easily communicate its secrets.

Do you know if blocking distorting is also a problem in single ended amps? I am working on SEP amps to be precise.

Yes, it is. Think about it. A Class A (a single-ended amplifier is always Class A because the single tube is on for both halves of the waveform) still has a coupling capacitor removing DC from the gain stage output before that amplified signal is being applied to the output tube's grid. So any time you have a capacitor removing a DC component blocking distortion can arise.

So a SEP or SET amplifier should benefit from a negative clamp, just as any other amplifier would.

Please add it and report back with some numbers.

 

[trobbins]

The coupling cap still gets DC bias charged due to the diode (rather than through the grid), so blocking distortion still occurs with same time constant for coupling cap bias voltage to return to average.

 

[Retrovert]

The coupling cap still gets DC bias charged due to the diode (rather than through the grid), so blocking distortion still occurs with same time constant for coupling cap bias voltage to return to average.

Ahhh, but blocking distortion happens at the 0 V crossing. It does not, to my knowledge, happen at the negative peak. So if the diode prevents charging to a positive peak, it prevents blocking distortion.

Think of this another way. Positive grid current makes the output tube into a dual diode as current (abnormally) flows cathode to grid instead of (normally) as cathode to plate. That's a cause of crossover distortion. So by reducing the charge to the positive peak there's no sustained current flow to create the problem.

 

[dcgillespie]

That was my thought. My understanding has always been that on positive going signal swings presented to the output tube grid (and assuming a fixed bias design), once the peak signal reaches Eg=0, that effectively prevents the grid from being driven any further positive, because the coupling cap cannot develop the power required in the grid circuit to drive the grid positive, due to the current it will draw in that conditon. Therefore, the grid end of the coupling cap is effectively locked (for the purposes of this discussion) at basically a ground level. But as the driver stage is driven to a level that overdrives the output stage, then the continued positive swing of the driver stage, combined with the other end of the coupling cap being effectively grounded by the output tube grid, does the dirty deed of then charging the coupling cap to greater potential. Now the stage is set, as now on the ensuing negative swing from the driver, the increased charge across the coupling cap drives the output tube further into cutoff (assuming a Class AB1 design), making it late for the hand-off on the next positive swing. Hello blocking distortion.

Retro -- are you suggesting that the diode (1) protects the output tube grid? (very understandable), and (2) also locks the coupling cap closer to ground level than the effective cathode/grid diode action does? If so, then I would think that would only make the actual blocking distortion worse.

Dave

 

[Retrovert]

That was my thought. My understanding has always been that on positive going signal swings presented to the output tube grid (and assuming a fixed bias design), once the peak signal reaches Eg=0, that effectively prevents the grid from being driven any further positive, because the coupling cap cannot develop the power required in the grid circuit to drive the grid positive, due to the current it will draw in that conditon. Therefore, the grid end of the coupling cap is effectively locked (for the purposes of this discussion) at basically a ground level. But as the driver stage is driven to a level that overdrives the output stage, then the continued positive swing of the driver stage, combined with the other end of the coupling cap being effectively grounded by the output tube grid, does the dirty deed of then charging the coupling cap to greater potential. Now the stage is set, as now on the ensuing negative swing from the driver, the increased charge across the coupling cap drives the output tube further into cutoff (assuming a Class AB1 design), making it late for the hand-off on the next positive swing. Hello blocking distortion.

Yes, I believe this to be a substantial part of what is happening. Thank you for contributing that wording and explanation. The output of the coupling capacitor, being directly connected to ground through the diode, cannot rise above ground. (Taking into account the forward voltage drop. Which, as I previously noted, is a non-linear and current-dependent phenomenon, so it may not be as much an issue in practice.) So, yes, the positive swing and positive climb is totally eliminated no matter what the tube would otherwise do, particularly when hard driven. The output of a negative clamp is perfectly below ground, and this is where the output tube best operates.

Blocking distortion is not well studied, and I don't have a lot of hard data on what transpires at that 0 V crossing. Simulation is only going to be possible when one adds in accurate tube models to either side (driving and driven) and that's a lot of work. I believe LTSpice can do it, witness the extensive simulation done on the PAS and other amplifiers, but it's a messy job to get right. Empirical testing is going to be fare easier.

A related issue is that as the grid becomes more positive it attracts ions which cause it to become more positive, and this (low-level) runaway also locks it into blocking distortion. Modeling that problem is likely very difficult. I think I would drop down into real world experiments for data and forgo the simulation.

Retro -- are you suggesting that the diode (1) protects the output tube grid? (very understandable), and (2) also locks the coupling cap closer to ground level than the effective cathode/grid diode action does? If so, then I would think that would only make the actual blocking distortion worse.

(1) Yes, it protects the grid from grid current. Depending upon the circumstances (current flow, voltage, duration, tube construction, etc.) grid heating causes damage to the tube to varying degrees. It is never totally harmless, but it may not cause meaningful damage, i.e. shortening lifespan or altering characteristics. The grid should never be driven positive, except in tubes designed for this. Screens fail for similar reasons. The dissipation limits are clearly stated, but these are honored more in the breach than the observance.

(2) I think I can clarify this. The diode prevents the flow of cathode-to-grid current, because the grid is not driven positive, and it creates a maximum value for the grid. The grid is not locked (i.e. maintained) at ground, it cannot move above ground. It can, however, greatly move below ground. That's how the AC input signal now controls the grid. It is purely negative (again, taking into account voltage drop and diode bias) and functions as it always did. If the input signal is purely negative the grid remains negative, same as before. It is only with the positive half of the cycle that further change is stopped (clamped) at 0 V. So the problem cannot be worse, because the entirety of the problem for positive swing is now gone.

That's why it makes things better.

 

[Retrovert]

One key thing to remember in all of this debate is that the DC restoration modification being proposed is trivial to implement and has no meaningful expense.

Adding a diode to ground is a very simple task. Try it and see. Measure it with instruments, measure it with one's ears. If you like it, leave it. If you don't like it, remove it.

The expense is tiny, the risk to the amplifier should be non-existent (if properly installed, if you don't know how, get a technician who does, again, no warranty, YMMV), and the time involved is minimal. As there's no great cost or great risk no reason exists to not try it and see if what is being proposed actually works for you in practice.

 

[gadget73]

The low frequency response is the major beneficiary. It would be helpful if you could measure at, say, 20 Hz, 30, 40 Hz, 50 Hz and 100 Hz instead of at 1 kHz (or thereabouts), which is the standard for many harmonic distortion analyzers.

Note: using only a diode without a bias source causes the baseline to be 0.7 V above ground. That can cause blocking distortion because it allows a DC voltage at the grid. So any experiments may not demonstrate the best possible results. The figures show what happens at that 0 V crossing.

The THD meter I have is also my audio generator. It will do THD measurements basically anywhere I ask it to. I think it will bottom out at 1 hz, and top out right around 100 khz.

I can't really bias the diode positive on this with a negative DC grid supply. Initially I was planning to pull the NFB and see if the grid can even be driven positive before the output stage craps out. I've never taken any measurements on it to see what exactly is giving up first. I had considered trying this on the Fisher 600 which has a positive voltage on the grid, but then I realized that fitting a diode to ground would cap that at a fair bit below the actual DC voltage it sits at normally, and that won't work. Wouldn't be a big deal on a typical cathode bias amp to do this though. Tickle a half volt or so of positive voltage into the grid, compensate the cathode resistor as needed, off you go.

 

[Retrovert]

The THD meter I have is also my audio generator. It will do THD measurements basically anywhere I ask it to. I think it will bottom out at 1 hz, and top out right around 100 khz.

Nice! Sounds like it should do plenty of bass measurements. A lot of them are fixed frequency.

I can't really bias the diode positive on this with a negative DC grid supply.

1.5 V batteries and resistors suitable for a 50% divider. Presto: instant 0.75 V source! Tinker with upper value for exact forward drop cancellation. Remember, this is just like the old C battery used for radio bias. No current, just a voltage

Initially I was planning to pull the NFB and see if the grid can even be driven positive before the output stage craps out. I've never taken any measurements on it to see what exactly is giving up first.

NFB should not be affecting the grid voltage. It's the plate voltage which causes the positive grid voltage.

 

[trobbins]

Dave, yes my appreciation is that the coupling cap average DC bias voltage shifts to a higher level as the grid (or protection diode if used) conducts current. That situation won't really be any different if the protection diode is used. That was described in a few vintage articles (Crowhurst Audio Nov 1959, plus Merlin and Randall Aiken have done articles).

The Paul Ruby mod was the first use of a neat means to supress that blocking distortion effect, by clipping off the negative going grid voltage peak, and so assisting the recovery of the coupling cap to its nominal DC bias level. This was discussed in a few threads over at music-electronics-forum , along with a simpler negative clipper using a single zener and 2 diodes. Without any such remedial clipping circuit, the RC time constant of coupling cap and grid leak and anode load on driver stage determines the recovery timing. And if a cathode biased output stage with bypass cap is used, then there is the extra effect of the time-constant of that bypass cap due to anode/screen current.

 

[gadget73]

1.5 V batteries and resistors suitable for a 50% divider. Presto: instant 0.75 V source! Tinker with upper value for exact forward drop cancellation. Remember, this is just like the old C battery used for radio bias. No current, just a voltage

yes, but if its got -15v currently, its a bit difficult to positive bias the diode while still holding the grid some fair bit negative. If i put a big resistor divider rig in there to allow both to happen, the voltage drop across the resistors to the diode would defeat the purpose.

NFB should not be affecting the grid voltage. It's the plate voltage which causes the positive grid voltage.

Thats more to see what stage craps out first. If the output stage starts showing distortion, it shows up at the grids too via the feedback loop. Pulling that out of the chain will let me monitor both the output and the grids to see whether the grids drive positive before the output stage distorts without the feedback making things look funny.[/QUOTE]

 

[crispycircuit]

How would you apply it a Magnavox amp? Or Eico circuit?