<<previous : next >>
Contact

From stephen.delf–(at)–elcom.gen.nz Sat Jan 6 20:34:32 CST 1996
Article: 4960 of rec.audio.tubes
Path: geraldo.cc.utexas.edu!cs.utexas.edu!math.ohio-state.edu!uwm.edu!vixen.cso.uiuc.edu!newsfeed.internetmci.com!in2.uu.net!munnari.OZ.AU!comp.vuw.ac.nz!welcom!stephen.delft
From: stephen.delf–(at)–elcom.gen.nz (Stephen Delft)
Newsgroups: rec.audio.tubes
Subject: Re: AB crossover dis 1/2
Date: Fri, 5 Jan 1996 23:24:00 GMT
Message-ID: <960107130936366--(at)--elcom.gen.nz>
Organization: WELCOM BBS
Distribution: world
References: <4ci4ak$4g--(at)--loner2.ix.netcom.com> <960105102622355--(at)--elcom.gen.nz>
Lines: 145

(Re: apparent DC bias shift at grid, in overdriven Class AB p-p output
stage, due to grid current in part of cycle: and use of large value
resistors in series with (output) grids to minimise this
“rectified-input-signal” bias shift.)

(SD) (snip) …if you then reduce the grid-to-ground resistors to bring
(SD)back the _total_ grid resistance << for the output tubes >> even to
(SD)the over-spec figure regularly used in guitar amps… you are also
(SD)changing the loading on the phase-splitter/driver tube, and you may
(SD)change the level and nature of this stage’s distortion contribution
(SD)to the amp’s overall sound.

RA>(snip) _preamp stages_ (snip) Cathode biasing is self-correcting to a
RA>certain extent (snip) much larger grid resistors than for fixed bias.
RA>(snip) resistor in series with the grid _after_ the grid to ground
RA>resistor, which won’t change the loading on the previous stage.

Hi Randall….no disagreements here…except your last paragraph, where
I think we have our lines crossed. If I put the question poorly, let
me now try to be more exact – and so please forgive the long post – and
any typing errors.

You refer to fitting grid resistors to (cathode biased) preamp stages.
Certainly…I agree…this is the classic “grid stopper” position, also
sometimes used to modify instability, or unwanted RF input.

I was referring to the difficulty of putting (say) 100k in series with
the grids of the _output tubes_ (either fixed or cathode biased)…
….in circuits where the present total resistance from grid-to-ground
is _already_ higher than recommended by the tube manufacturer…..and
where reducing the present grid resistor value to compensate, would also
change the load on the previous stage.

Typically – 6L6 pair in guitar amplifier: DC path from output grid to
ground = 220k plus a bit for bias resistors – which would be fine for
a cathode-biased o/p stage (Rg max=500k), but is higher than recommended
(Rg max=100k) for the (many) guitar amps with fixed bias.
Personally, I like the tonal effect in some amps of putting 47k or
similar resistors in the output tube grid leads. Unfortunately, with
typical fixed bias guitar amps, this mod may be encouraging other
problems which can lead to unstable overheating of tubes ( and greater
amp damage if the circuit is not properly fused.)

Placing (say) 100k in series with the output grid, makes Rg-total about
330k (compare: recommended 100k)….which brings one closer to a
situation where some samples of tubes may become thermally unstable when
overdriven – not a good idea in a guitar amp. Inserting resistors of 1k
(or even 10k) at the output tube grids, for reasons of stability or
tonal adjustment, makes insignificant increase in total DC resistance
>from grid to ground. But inserting something like 100k (or up to 470k as
suggested by Thiele) _does_ increase Rg-total significantly.

Reducing the presently fitted grid resistors to about 120k for each grid
will restore Rg-total back to its original value of 220k plus a bit.
Unfortunately, this change also reduces the AC load on the plates of the
previous phase-splitter/driver stage. This can alter the subtle
coordination of “which stage, distorts in what way, and at what level”
which is what makes some guitar amplifiers particularly expressive
musical instruments. It is also one of the factors which allows
musicans to notice the “feel” of one amplifier, different from the
feel of another amp which uses the same basic components but with a
few slightly different values.

Alternatives include EL34’s (or 7591’s for lower outputs) either of
which should be happy with fixed bias and around 330k Rg. Both
are relatively easy to drive, can make very nice sounding guitar amps,
but not exactly the same as a similar circuit with 6L6’s.

Some early 30 to 40W p-p 6L6 guitar amps used cathode bias on the
output pair. This will tolerate higher DC path grid resistance. Max
output is a little lower than with fixed bias (not very signifiant in
view of the ear’s log. sensitivity), and some people say it sounds
sweeter and a little more compressed. A cath-bias output stage with
chassis-mount cathode resistor, a solid aliminium type bypass cap, and
adequate fusing should be a least as reliable as the common fixed bias
arrangement. Personally, I like the sound of small cath-bias p-p amps.

Putting small iron-cored chokes from output grids to the fixed
negative bias supply would allow one to specify DC Rg separately from
signal impedances….But quite apart from cost, it is hard enough on a
small guitar amp chassis, to align power and O/P transformers, PS choke
and reverb transformer to minimise stray field interaction….without
throwing _two more_ chokes into the recipe!

**** So here is one of the basic tube guitar amp interlocks:

Many old amp designs are just too good musically, to be believably the
result of paper calculations alone; I suspect they were adjusted
empirically until experienced musicians said “OK, that sounds good”.

Often this has not corresponded with tube manufacturers’recommendations
and while Leo Fender’s original customers did not have to worry too much
about wearing out 1950’s and 1960’s tubes, we should not be wasteful of
the remaining NOS stocks. Even where current tube makers produce good
replicas, it is unlikely they would ever reproduce _all_ flavours of 6V6
or 6L6 for example, and there will always be some people will prefer to
use varieties only available from diminishing old stock.

The grid resistance in many nice old Fender (and other) amps is already
higher than the tube makers recommend for that tube with fixed bias.
Most of the time, this (roughly x2) variation doesn’t usually cause any
problems: raising it _any further_ is probably inviting thermal
instability, runaway tubes, and maybe other kinds of amplifier damage
unless it is adequately fused.
Rearranging to put more resistance in series with the grid, without
increasing Rg to ground (if one wishes to minimise input DC offset and
crossover distortion when mildly overdriven) also lowers the load seen
by the previous stage, and this subtly changes the amplifier’s
performance in other respects.

In other words, Leo and his crew had probably found a pretty good
engineering/performance compromise, pushing Rg as high as could be
risked in practice, and squeezing the maximum gain/headroom compromise
>from the previous stage.

A HiFi amp design would probably either run the phase-splitter at higher
current and with lower value plate resistors….or place a buffer stage
between p/split and output stages. And either of these mods will _also_
subtly change the performance of a guitar amp – probably in a way which
many guitar players would not appreciate.

The fascination – and the challenge – of the best old tube guitar
amp circuits, is that they include _many_ distortion mechanisms, often
the result of complex interaction between one stage and the next. (A
close analysis would be tedious, and might require a “fuzzy logic”
computer system to give useful answers in a reasonable time.)

As with fine acoustic guitars, if you change one part of the system, you
also change its interactions with other parts. It is certainly possible
to modify one aspect of a classic tube guitar amp’s performance (and
make whatever adjustments to ensure the whole system works reliably
again). The result may be a very nice guitar amp, but probably _not_
your original amp performance, with just that one aspect changed.

Incidentally, it was this parallel with the interactions between
components in a fine acoustic instrument which got me interested in
making amps, rather than just repairing them.
I have not found any way of removing crossover kinks from the signal
just beyond clipping (without changing the old-guitar-amp tone and
sig.level/distortion response)….except by using a fairly high
no-signal current in the p-p output stage, keeping B+ down to voltages
which result in acceptable plate dissipation in use at this bias setting
….and accepting an output limit of about 35 to 40 watts per pair of
6L6’s.
>>> Continued to next message

From stephen.delf–(at)–elcom.gen.nz Sat Jan 6 20:34:40 CST 1996
Article: 4961 of rec.audio.tubes
Path: geraldo.cc.utexas.edu!cs.utexas.edu!usc!math.ohio-state.edu!uwm.edu!vixen.cso.uiuc.edu!newsfeed.internetmci.com!in2.uu.net!munnari.OZ.AU!comp.vuw.ac.nz!welcom!stephen.delft
From: stephen.delf–(at)–elcom.gen.nz (Stephen Delft)
Newsgroups: rec.audio.tubes
Subject: Re: AB crossover dis 2/2
Date: Fri, 5 Jan 1996 23:24:00 GMT
Message-ID: <960107130937366--(at)--elcom.gen.nz>
Organization: WELCOM BBS
Distribution: world
References: <4ci4ak$4g--(at)--loner2.ix.netcom.com> <960105102622355--(at)--elcom.gen.nz>
Lines: 8

>>> Continued from previous message
This roughly corresponds with postings from other guitar amp people on
a.g.a. and r.a.tubes, but I would be interested to hear any other
thoughts from anyone, on minimising crossover distortion in simple tube
guitar amps, in the region just beyond the beginning of output stage
clipping.

Regards, Stephen Delft.

From reaike–(at)–x.netcom.com Sun Jan 7 09:30:05 CST 1996
Article: 4969 of rec.audio.tubes
Path: geraldo.cc.utexas.edu!cs.utexas.edu!usc!howland.reston.ans.net!ix.netcom.com!netnews
From: reaike–(at)–x.netcom.com(Randall Aiken )
Newsgroups: rec.audio.tubes
Subject: class AB crossover distortion
Date: 7 Jan 1996 06:37:25 GMT
Organization: Netcom
Lines: 161
Message-ID: <4cnpn5$rj--(at)--loner2.ix.netcom.com>
NNTP-Posting-Host: atl-ga28-17.ix.netcom.com
X-NETCOM-Date: Sat Jan 06 10:37:25 PM PST 1996

In <960107130936366--(at)--elcom.gen.nz> stephen.delf–(at)–elcom.gen.nz
(Stephen Delft) writes:

SD>(Re: apparent DC bias shift at grid, in overdriven Class AB p-p
SD>output stage, due to grid current in part of cycle: and use of large
SD>value resistors in series with (output) grids to minimise this
SD>”rectified-input-signal” bias shift.)

>RA>(snip) _preamp stages_ (snip) Cathode biasing is self-correcting to
>a certain extent (snip) much larger grid resistors than for fixed
>bias. (snip) resistor in series with the grid _after_ the grid to
>ground resistor, which won’t change the loading on the previous stage.

SD>Hi Randall….no disagreements here…except your last paragraph,
SD>where I think we have our lines crossed.
SD>You refer to fitting grid resistors to (cathode biased) preamp
SD>stages. Certainly…I agree…this is the classic “grid stopper”
SD>position, also sometimes used to modify instability, or unwanted RF
SD>input.
SD>I was referring to the difficulty of putting (say) 100k in series
SD>with the grids of the _output tubes_ (either fixed or cathode
SD>biased)…in circuits where the present total resistance from
SD>grid-to-ground is _already_ higher than recommended by the tube
SD>manufacturer…..and where reducing the present grid resistor value
SD>to compensate, would also change the load on the previous stage.

Hi, Stephen…I understood we were originally talking about push-pull
output stages, where all the discussion and conclusions about
resistances were accurate and appropriate, if fact, if you refer to my
original response, you will note that I was talking exclusively about
output stages; however, this exact same technique that can be applied
in reducing crossover distortion in overdriven class AB amps also
applies to reducing a particularly bad form of _preamp_ distortion in a
cathode-biased (or fixed-bias) RC-coupled class A amplifying stage. I
thought you made reference to using the same technique of series grid
resistance for this problem, which was the reason for my comment. More
about this later…

But first, let me clarify one point. In the summary at the top, you
make reference to apparent DC bias shift because of “rectified input
signal”. Actually, there is no rectification of the input signal
(unless you clip the phase inverter asymmetrically); it is more a
_clamping_ action caused by driving the output tube grid positive.
This clamping holds the positive peaks of the input waveform at a
relatively constant level, which, in conjunction with the AC coupling
and high source resistance of the bias voltage and input signal,
results in the waveform shifting downward as the amplitude is
increased. I am sure you understood what I was talking about, I just
don’t want to confuse anyone else reading this who might have
interpreted this wrong.

Secondly, your comments about lowering bias resistances to accomodate
large series grid resistors were spot on; that is the same thing I was
referring to. Many older amplifiers (Marshall and Fender, among
others) _did_ use 220K grid resistors; however, most changed to 100K at
some point. I suspect this was to prevent tube failure, but might have
been to reduce distortion in the output stage by lowering the drive a
bit. Interestingly, I have read comments on these newsgroups
questioning whether anyone has ever seen a case of tube failure caused
by a 220K grid resistor in place of a 100K. I suspect the tube
manufacturers conservatively rated that value, but I personally
wouldn’t push it too far. This is why I recommended the cathode
follower as the best method of obtaining low source impedance.

SD>Putting small iron-cored chokes from output grids to the fixed
SD>negative bias supply would allow one to specify DC Rg separately
SD>from signal impedances….it is hard enough to align

You could get around the radiation problem with using toroidal cores,
but the problem is the required inductance to get a 100K load at even
50Hz is around 300 henries, which is a rather large inductor…and the
inductive reactance would rise as the frequency went up, changing the
load on the phase inverter and further changing the frequency response
and gain.

SD>In other words, Leo and his crew had probably found a pretty good
SD>engineering/performance compromise, pushing Rg as high as could be
SD>risked in practice, and squeezing the maximum gain/headroom
SD>compromise from the previous stage.

Don’t forget that in those days the amplifier was never expected to be
run into clipping; clean power was what the amp was designed for. I
suspect Leo and friends never worried about the crossover distortion
increase at hard overdrive.

SD>I have not found any way of removing crossover kinks from the signal
SD>just beyond clipping (without changing the old-guitar-amp tone and
SD>sig.level/distortion response)….except by using a fairly high
SD>no-signal current in the p-p output stage, keeping B+ down to
SD>voltages which result in acceptable plate dissipation in use at this
SD>bias setting….and accepting an output limit of about 35 to 40
SD>watts per pair of 6L6’s.

I prefer cathode followers in my own designs to remedy the crossover
distortion problem; however, I have used EL34’s with between 50K and
100K grid resistors and 100K bias resistors with no problems, and find
the sound to be quite good. Remember, though, even changing the bias
current changes the sound; I personally don’t care to modify classic
old Fender and Marshall amps to remove crossover distortion, I feel
that that stand on their own. My comments are directed primarily at new
designs.

One interesting note about what led me to research this problem: in
comparing my fantastic-sounding old 50 watt non-master Marshall with a
JCM800 50 watt that didn’t sound particularly good cranked up, I found
the crossover distortion on the JCM800 to be much worse at any bias
setting within reason, even though the plate voltages weren’t far off.
I suspect the transformer primary impedances were very different. I
also believe that this form of distortion is audibly bad even in low
amounts.

Now, in regards to my comments about preamp series grid resistors: The
same mechanism that causes the excessive crossover distortion in an
overdriven push-pull stage can cause a very bad sounding distortion
known as “blocking” in a preamp stage.

If you drive an AC coupled preamp stage into clipping, the input grid
will _also_ tend to clamp the input signal to a relatively fixed level,
and the waveform will also shift downward. If you drive the tube hard
enough, the tube will be cut off for a significant portion of each
cycle; in particular, if you have large coupling capacitors and
resistors, the high RC time constant of the circuit will cause the tube
to shut completely off on large transient waveforms until the average
level drifts back into conduction range. This causes a nasty “cutting
in and out” sounding distortion which is particularly bad if the stage
in question is within a negative feedback loop. In addition, this
blocking can also occur in output stages as well.

There are a few of ways of fixing this; one is to not try and get too
much distortion from a single tube section. Use resistive voltage
dividers to reduce the level between stages.

The second is to use diode clippers to limit the input drive like
Marshall does in their JCM900 series. I would avoid this like the
plague, as it generally sounds thin, buzzy, and lifeless (some will
undoubtedly disagree, but I noticed that even Marshall abandoned this
approach on the new amps starting with the 30th anniversary series…).
Refer to the alt.guitar.amps newsgroup for several good discussions and
opinions of these diode circuits.

The other method is to use large value series grid resistors as I
mentioned earlier and in the previous post. In this case the resistors
are not acting as “grid stoppers” or RFI prevention resistors, but
instead, as current limiting resistors to prevent the clamping action
that occurs during heavy overdrive. As I said in the previous post,
most, if not all, preamp stages are cathode-biased, allowing the use of
large series resistors (typically 100K to 470K) _after_ the normal
grid-to-ground resistor, which doesn’t affect the loading of the
previous stage.

By the way, Stephen, thanks for starting such an interesting technical
discussion! It probably should be on alt.guitar.amps, though; I hope
we don’t offend the audiophiles on here by talking about intentional
distortion , but the same problems can apply to a hi-fi amp on
transient signals; particularly the blocking problem, as these
amplifiers often use global negative feedback which aggravates the
problem.

Randall Aiken
reaike–(at)–x.netcom.com

From reaike–(at)–x.netcom.com Mon Jan 8 22:18:31 CST 1996
Article: 5036 of rec.audio.tubes
Path: geraldo.cc.utexas.edu!cs.utexas.edu!bcm.tmc.edu!news.msfc.nasa.gov!newsfeed.internetmci.com!howland.reston.ans.net!ix.netcom.com!netnews
From: reaike–(at)–x.netcom.com(Randall Aiken )
Newsgroups: rec.audio.tubes
Subject: AB crossover dis
Date: 8 Jan 1996 23:57:57 GMT
Organization: Netcom
Lines: 73
Message-ID: <4csb25$gg--(at)--xnews2.ix.netcom.com>
NNTP-Posting-Host: atl-ga6-04.ix.netcom.com
X-NETCOM-Date: Mon Jan 08 3:57:57 PM PST 1996

From: reaike–(at)–x.netcom.com(Randall Aiken )
Newsgroups: rec.audio.tubes
Subject: Re: AB crossover dis
References: <4cqcqj$h7--(at)--inet02.li.net>

In
ATEC15–(at)–mail.mot.com (John Byrns) writes:
>
>In article <4cqcqj$h7--(at)--inet02.li.net>, bbruhn–(at)–ewshost.li.net (Bob
>Bruhns) wrote:
>
>> As an alternative to resistors in series with the output tube
>>control grids, you could drive the grids with DC-coupled cathode
>>followers. This would allow you to actually drive the output tubes
>>into class AB2 with a capacitively coupled phase splitter, without
>>rectified drive voltage charging the coupling caps and messing up
>>the bias.
>>
>> Also, I suggest some kind of limit on the cathode bias, if used.
>>In a class AB or B stage with a bypassed cathode resistor, the DC
>>bias is a function of average cathode current, and therefore the DC
>>bias increases with the audio level. A certain amount of this may
>>produce a pleasing effect, but it can easily wind up in class C
>>unless a limit is placed on the voltage that can be developed.
>>

>
>At this point why use cathode bias at all? Since you already have the
>negative supply for the cathode follower, all you need to do, is add a
>voltage divider to provide the necessary negative bias on the cathode
>follower grid, which will inturn bias the output tubes. It would seem
>like this would be cheaper, and easier than putting all that junk in
>the cathode circuit of the output tubes. Fixed bias would also
>eliminate the bias shift problem of cathode biased class AB
>amplifiers.
>
>
>Regards,
>
>John Byrns

John, you have a good suggestion here, but I think you missed the point
of the discussion. In my original answer to the question (see previous
posts on this thread), I detailed the cathode follower bias method you
and Bob suggest for fixed-bias operation of a class AB amplifier. It is
the preferred method of eliminating the bias shift caused by clamping
of the positive peaks of the waveform in the output grids in
conjunction with a high impedance signal and bias source.

However, cathode biasing is sometimes preferred in guitar amplifiers
because it gives a “warmer” sound with additional harmonic distortion
that is considered to sound good. In this case, the cathode follower
is still the preferred method for eliminating the harsh crossover
distortion _above_ clipping, but the smaller bias shift that varies
with signal in a class AB amplifier (due to the average cathode
waveform not being zero) both above and below clipping, will still
occur, allowing the amplifier to retain its desirable harmonic
distortion.

As to limiting the cathode voltage in a class AB amplifer: in general,
the output stages of guitar amplifiers do not develop enough drive
voltage at the output tube control grids to shift the output stage too
far past class B operation to start sounding bad. The objective in
adding the cathode follower or series grid resistors is to limit the
amplifier to class AB mode to avoid the harsh-sounding crossover
distortion of class B, while allowing the normal clipping to occur.

Randall Aiken
reaike–(at)–x.netcom.com