Amorphous Cores in Transformers
Patrick Turner
may be some benefits with the use of amorphous core material for output
transformers instead of the usual laminations in the form of C-cores , E
& I , or Toroidal or ( Uni-cores , from AEM in Sth Aust . )
I wind all my own OPTs , Power trannies and Chokes for my amps and I
know only too well that the output transformer quality is one of the
most important items affecting fidelity in any tube amp .
So far , my transformers measure well and sound well but there's always
room for improvement .
So , has anyone used amorphous core material ?
From Patrick Turner http://www.turneraudio.com.au
GRANT G 10
(Tamura with IE core and Tribute with C-Cores) along with several other
high-end trannies. Amorphous are my favorites. Permalloys also sound great
(Tamura, Nature Sound, Ulbrich). The Amorphous and Nickel-Permalloy are far
ahead of iron in terms of delicacy, detail, and life.
-Grant
Mark Harriss
Neosid dealer in Australia and winding some 50W transformers to
suit EL34 valves, if i remember rightly it's called C5 material.
Peter Roovers
apparantly checked these things out.
Maybe you could mail him. (great website to !)
Peter R.
JHPage
in...@turneraudio.com.au says...
I have built a pair of output transformers (parallel feed, no DC) for 50
type triodes using Phillips 3C85 core material and they sound quite good.
I used 2-1 inch square cores stacked (2" x 1") in a dual C type
configuration and wound the coils on standard 2" x 1" EI nylon bobbins.
They have about 100 hours on them and I will do some hard measurements in
the near future. They do have 2 main drawbacks: 1.) the core material is
very hard and brittle (looks like black glass) so it chips easily and
leaves very sharp edges 2.) in small quant. they are fairly expensive,
about $180US for 8 "C" and 4 "I" sections including bobbins.
Joe Page
joe...@enteract.com
-----= Posted via Newsfeeds.Com, Uncensored Usenet News =-----
http://www.newsfeeds.com - The #1 Newsgroup Service in the World!
-----== Over 80,000 Newsgroups - 16 Different Servers! =-----
Chris Merren
Amorphous material has been around for a while and there are different
types of amorphous materials...
Amorphous means that the material is non-crystal in atomic
structure...but can recrystalize if the temps are brought up in
manufactureing....
Not all amorphous materials are suited for Audio OPT's....
Amorphous materials vary greatly ....some do not share any similarities
to each other, not even slightly..
As for the shape the material is in...this is not the issue... whether
it comes in C-core or E-I lams ect.. is not what makes it better or
worse or different....
Here are some types..
Amorphous-B: IRON-BASED ...Sat Flux density about 15.5 K guass
Amorphous-G: IRON-BASED.... 14K gauss sat.
Amorphous-E: COBALT BASED... 5.75K gauss sat.
So I don't wnat to hear this Amorphous sounds better than Iron
crap....because IT IS IRON-based...
It's the losses that are lower with the amorphous-B that help out in
sounding better....
The choice would be AMORPHOUS-B for Audio OPT's.... I will explain just
ahead..
The standard stuff for OPT's is M6 grain oriented also known as
MAGNESIL....
Now lets compare this AMORPHOUS-B...
Amorphous-B can be due to the squarness of the B-H loop ..can be used up
to 12K gauss nicely in Audio since it is fairly linear up t this
point..( % core-distortion).... Magnesil on the other handis only
usefull up to 12Kgauss as well in audio..although it sats about 18K
gauss...it's usefull range in audio is limited to 12K gauss for
linearity reasons.( % core distortion)...
Amorphous Alloy-B core loss at 10K gauss at 1Khz is about 3
watts/pound.... while magnesil M6 is about 12 watts per/pound at the
same given conditions..
As for permeabilty...the Amorphous Alloy-B is just about exactly double
the permeability of M6 magnesil for sma e given conditions....
All in all if you are designing an OPT...you do so by limiting the peak
flux density excursion at the lowest frequency of interest at full
output power....this way setting it to say 12K gauss.... then you check
to see if the inductance is good after the gapping ect....You DO-NOT
design a OPT with a inductance as a design target...then the core
distortion would be random..not good..
So the comparing an OPT of Amorphous Alloy-B with one of M6 Magnesil of
the same area...the turns would need to be the same... ..BUT the
Amorphous Alloy-B would have 3/4 less core loss and twice the
inductance...
This would be VERY desirable in SE applications.....
As for C-core vs. E-I lams..... The C- core would have a longer
inductance curve.. i.e the volume needs to be turned up to get the
inductance up...C-core incremental inductance would peak at close to
full power output...while the E-I lams would have a quicker incremental
inductance curve.. i.e the inductance would peak at about 1/2 the power
output the taper off... SO you can see the advantage of E-I lams in SE
applications as well....
As for Colonel Willy McLyman and his Inductor book....the guy knows his
shit.. I also worked for NASA at The Jet Propulsion Laboratory in
Pasadena, CA. doing power supply and magnetics design...
CHEERS
CM
Dave Slagle
> As for C-core vs. E-I lams..... The C- core would have a longer
> inductance curve.. i.e the volume needs to be turned up to get the
> inductance up...C-core incremental inductance would peak at close to
> full power output...while the E-I lams would have a quicker incremental
> inductance curve.. i.e the inductance would peak at about 1/2 the power
> output the taper off... SO you can see the advantage of E-I lams in SE
> applications as well....
are you referring to the differences of the core vs shell type of assembly?
the way i see it the shell type config with c-cores is nearly identical to
the scrapless lam config.
dave
Patrick Turner
For push-pull amps here is my basic design logic for a 50 watt PP amp :
Work out maximum voltage RMS across primary at 1Db below clipping .
So for 50 watts into 7 K ohms I get 591 V . So round this up to 600 .
22.5 X V X
10,000
No of primary turns = _______________________
B X Afe
X F
Where 22.5 and 10,000 are constants for the equation to work in all cases ,
X is multiplied by ,
V is volts across primary ,
B is flux density in Tesla ,
Afe is cross sectional area of the central leg of the core ,
in square millimetres
F is the frequency of the sign wave voltage .
So we have 600 V . Choose the frequency to be 14 Hz . Choose the core to be
62.50 mm stack of 44 mm E&I laminations of non grain oriented 3% silicon steel
. The maximum B at saturation is about 1.8 Tesla .
If I insert these values into the above equation I get 1,948 primary turns .
16 layers of 0.4 mm cu dia wire neatly layer wound give 2,064 turns . From
the above equation I can work out what B is at 50 Hz at 600 V . B will be =
0.47 Tesla .
At about 12.5 watts , V = 300 V and B = 0.235 Tesla .
This compares favourably with the original Williamson design .
From what I've read the distortion in the signal due to the iron is a lot less
than the valve distortion , even if 4 X KT88 tubes are used in triode mode .
[What more could you want ?] . The distortion can be calculated from the
formulas in the Radiotron Designer's Handbook . Pentodes give lots of D ,
ultralinear and triode give about 3 times less , due to the impedance driving
the transformer being low .
From what I've seen on my CRO the above formula seems to actually work .
The high freqency losses do not seem to be of serious concern . There is not a
huge increase in plate currents due to any losses at 10 Khz .
Of course the above does not take into consideration the interleaving of P and
S sections to get the OPT to have low leakage inductance and hence transfer
energy efficiently up to 70 Khz . The high freqency geometry is a seperate
issue to core material choice .
I have found the basis of 2,000 primary turns on a big block of iron to be a
good recipe for fine music .
The maximum mu of the material in this example is about 4,000 and the
inductance , Lp , at low V is around 50 to 100H . Maximum Lp is about 250H and
seems quite sufficient when compared to the load value and the needs for
stability in my designs .
I've used C-cores and GOSS E&I lams and got fabulous Lp figures but the Fs ,
saturation frequency , was about the same .
From what I've read from 1950 books the use of GOSS halves the 50 Hz
distortion over using non-GOSS , which is a third the price of GOSS . The
saturation behaviour of non-GOSS is less sudden than GOSS . And for my SE
designs I prefer to use GOSS C-cores as the mu is higher when the air gap has
been optimised for non saturation .
So Chris perhaps you would like to comment on the foregoing and let me know
what size of type B amorphous cores I'd need to do no worse than in the above
calculations .
I would maybe find it hard to find Willy McLyman's book in the library at the
local uni . Feel free to email scanned pages .
Patrick Turner http:www. in...@turneraudio.com.au
__________________
Chris Merren
>
Choose the frequency to be 14 Hz . Choose the core to be
> 62.50 mm stack of 44 mm E&I laminations of non grain oriented 3% silicon steel
> . The maximum B at saturation is about 1.8 Tesla .
Two problems here..
One...only power transformers can be driven at the full excursion of 1.8
T.... because the core distortions can be filtered...but still power
trannies need some headroom..so they usually go 1.6T with M6...
OPT's need to be limited to 12,000 to 13,000 gauss for core distortion
reasons...12,000 gauss is about .5% core distortion in M6... when
gapping you can go a little beyond due to the linearity the gap
introduces...how far beyond is a matter of the gap size ..
Second... You need only plug in 20Hz for the turns equation...since it's
not like choosing inductance where you have tpo worry about phase
shifting below the frequency of interest...
It sound like you probably were doing OK..since because you were using
1.8T and were using 14Hz ...instead of 1.2T and 20Hz...which is only off
from each other by a tiny ammount...
> From what I've read from 1950 books the use of GOSS halves the 50 Hz
> distortion over using non-GOSS , which is a third the price of GOSS . The
> saturation behaviour of non-GOSS is less sudden than GOSS . And for my SE
> designs I prefer to use GOSS C-cores as the mu is higher when the air gap has
> been optimised for non saturation .
Forget these 1950's books...they are good for theory BUT most of the
materials and charts do not pertain to materials used today...
M6 grain-orientedis also known as Hypersil in the old books made by
Westinghouse..today ity is refered to as Magnesil...
M6 hits it's peak permeability at about 13,000 gauss...meaning that it's
inductance hits peak when driven to 13,000 gauss...well here's the
kicker... That just so happens to also be the max flux density point to
use with audio transformers beacuae the M6 core distortion is about .66%
at this point.. I would figure this at 20Hz at full output.. Now if you
build a C-core OPT you would not get full inductance till you drive the
OPT to full power output...( incremental inductance )..... Now replace
this C-core with E-I lams and we have a whole other story.... The E's
will first pass the magentic pathlength and the I's will have no
involvement...The gap between the E's is mucho smaller than the gap
between the Es and I's...this would be for 1:1 alternate stacking...SO
the reluctance is less between the E's...now the E's will start to peak
at approx half the peak flux density(6500 gauss) ...then start
saturating ....because it's like having half the volume of material
doing the work..at half the power point of 13,000 gauss...NOW the
relauctance builds causing the magnetic path lenth to now jump into the
I's to complete the loop...since this new path is lower reluctance....
Now each E and paired I share the burden of passing the magnetic path
length.....but now the inductance starts to deline after the mid point
of (6500) ,....this is due to the BIGGER gap between the E's and the
I's.... So the incremental inductance curve of E I lams is
parabolic...with it's peak somewhere around half the peak flux density
point..while the C-core incremental curve is also parabolic but peaking
at the full 13,000 gauss..
One has advantages over the other and vice versa....especially when
listening to music and lower listening levels..
CHEERS
CM
Patrick Turner
The man who has no problems makes nothing . So let's see if we could address the
problems .
I don't know what you mean by " only power Xfrms can be driven at the full excursion
of 1.8 T....because the core distortions can be filtered ....."
Sure some power Xfrms do run at 1.8 T ; none of mine do ; I run at 0.9 T max for
quietness . I use an oversize core and sure it's heavy but nothing gets hot like I
see in other amps . I don't know what is meant by ... core distortions can be
filtered ....
Filtered by what ? By the rectifier to get clean DC ?
The Transformer Equation is my guide to an adequate number of primary turns and it
seems we agree that my method
gives a correct starting figure even though you find it odd that I'd use 14 Hz and
1.8 T rather than 20 Hz and 1.2 T .
My design , for this post anyhow , has 2,000 turns on Afe = 44mm X 62.5mm , of non
oriented Si Fe .
I have compared my designs with Williamson's which had 4.400 turns on Afe = 32mm X
44mm . At 12 watts into a 10 Kohm RL the B at 50 Hz is less than 0.3 T and
distortion is low . My design gives just lower B at the same power and hence same
low distortion as Williamson's . Well that's what my calculations from the the RDH
indicate . But the design allows power to go up to 50 watts easy . The maximum
inductance of mine is half the Williamson's but it is sufficient compared with the
RL of 5 Kohms . The initial inductance at low Va-a is ok for LF stability at low
levels . The benefit of a bigger core is less P turns and much less leakage
inductance , LL , as LL varies as Np squared .
I use my design for 2 or 4 tubes , [4 tubes are best] and have waist free impedance
matching secondaries .
To be honest I find your discussion of the way reluctance skips about between Es and
Is somewhat baffling . Even if I felt I could follow you completely I still must
decide on how many turns . All the testing and listening have not yielded any
unpleasant surprises and all my OPTs have at least tested better than QUAD , Leak ,
Dynaco , Copeland , Jolida ,Etc .
I've not compared with Tango , Tamura or Hammond . And I'm not into making entirely
silly sonic claims .I find that if the OPT measures well the valves get a chance to
give you good music . And how far do you have to bend over to the laws of
diminishing returns ?
Sure , over the last 50 years there must have been some changes to materials
mentioned in 1950's books [RDH for example] .
Well , would not most changes be improvements ? ie , if you substitute today's GOSS
material in a 50's design you maybe will get wider usefull bandwidth , less losses
and less distortion . So I must be doing that anyhow as the RDH has been my closest
design friend . However not all the magnetic Si Fe that you buy today is of equal
quality and it's not unusual to be lumbered with modern product that is barely up to
the best 50's materials . But I must say the best GOSS E&I lams I've had came from
dear old England and gave me fabulous inductance figures but the saturation
frequency behaviour was only marginally different to NON- GOSS at one third the
price . No free lunches in electronics .
So as a start point I find 2,000 turns on a big block of iron to be ok , and non
oriented Si Fe to be ok for PP amps at least .
Do you see any problem using GOSS C-core material for SE OPTs and chokes ? So far I
think chokes I make with C-cores
are much better than using E&I lams .
I've tried air-gapping PP C-core OPTs but the LF instability tended to rise ,and the
Lp was less , and so LF phase shift increased and I didn't like it much . They say
Partridge did it but that doesn't mean I would .
So which is better listening at low level [in theory at least] the E&I or C core ? I
couldn't quite make out which type you meant was best .
Phil talks about low frequency caused phase distortion of high frequencies in normal
iron cores , I guess due to the changeing Lp that is across RL . Gee , I dunno about
this one . I guess there would be some IMD caused by the inductance change changeing
the total RL but Lp to RL ratio is so healthy that the phase Dn or IMD would be low
.
I've tried Bi amping with appropriate pre-amp filters and heard no improvement .
Perhaps someone on the Joe-list which Phil refers to might like to prove his point .
I've never seen phase distortion of HF caused by LF on my CRO at any time . But then
you can't say it doesn't exist ; and I wasn't using a phase shift detector during
two tone tests . And there would be those who would say the tiny artifacts make a
huge difference . Well what of phase distortion or doppler effect in speakers , and
musical instruments ? Where ever it is largest would be the first place to try to
reduce it , would it not ?
Phil also talks of parafeed amps . I assume he means seperate choke feed to each PP
plate and a PP OPT slung between each plate to collect the plate power , maybe with
a CT connected to ground through a capacitance . Gee , what a lot of trouble to go
to . But exactly what was meant by parafeed amps ?
The question still remains in my mind , exactly what amorphous cores of what size
would I need at least for the same overall performance ? And if I used the same
wind up on a bobbin and shoved in some am'cores what difference would it make ?
The man who questions nothing , knows nothing .
Patrick Turner http://www.turneraudio.com.au
CM
excursion
> of 1.8 T....because the core distortions can be filtered ....."
> Sure some power Xfrms do run at 1.8 T ; none of mine do ; I run at 0.9 T
max for
I have 1.6T power trannies that are extremly quiet...and nice and
small...but run hot..but running hot is fine since it is designed to do so..
. I don't know what is meant by ... core distortions can be
> filtered ....
It means that the filter section of the power supply that comes after the
pwer transformer will do the filtering of the harmonics generated by the
core distortion due to pushing a power trannie close to 1.8T .......this is
perfectly fine...provided you have assembled the trannie with the high temp.
wire coatings and interleaving material to deal with the higher
tempts....this is done when space is a major factor...also gapping the power
trannie is a very valid option when trying to extend the flux density
excursion beyond 1.8T for M6 core material....since inductance is not an
issue like in OPT's.....
> The Transformer Equation is my guide to an adequate number of primary
turns and it
> seems we agree that my method
> gives a correct starting figure even though you find it odd that I'd use
14 Hz and
> 1.8 T rather than 20 Hz and 1.2 T .
I highly doubt you have much signal below 20hz making it's way through the
power amp circuit caps..(poles) and to the primary side of the OPT
windings.... using 1.8T and 14Hz does not make any sense...unless you can
explain it to me mathamatically...It's like falling the down the stairs and
landing on your feet...since the equation come close to an optimum
number...At a flux density of 1.8T with M6 grain oriented core
material...you have close to 2.6% core distortion, not usefull in audio
applications....and at 14Hz you have basically no significant signal in the
OPT...even if your amp is capable.. the program material does not contain
these signals...
1.2T is used because it produces about .4% core distortion....and is
targeted by the designer as being the most % distortion tolerable by choice
of the designer....Now since the flux density increases with lower
frequencies and power output...the choice of 20Hz is used since this will be
lowest realistic worse case lowest frequency .... So @ 20Hz and at maximum
output power the peak flux density excursion is limited to 1.2T being .4%
core distortion.....this mean that the OPT can be cranked to full power
output with a 20Hz signal and only make .4% distortion at worse case...with
M6 core material..
Since frequency and flux density are both in the denominator of the turns
equation....ww can multiply them to simplify using a single constant....
1.2T x 20Hz)= 24 and (1.8T x 14Hz)=25.2 ........ very close.... 25.2 is
equivalent to running the flux density at 1.26T @ 20Hz which is about .5%
core distortion....
As far as the equations go...The 20Hz is basically a fixed number in
audio....but the choice of where the designer chooses to limit the peak flux
density is really a matter of what % core distortion you want in your OPT...
remember that gapping can extend this flux density for the same given core
distortion figure..
> My design , for this post anyhow , has 2,000 turns on Afe = 44mm X 62.5mm
, of non
> oriented Si Fe ...
This is totally true...since the original Williamson OPT's were
NON-grain-oriented cores..and thus makes a fair comparison.
> I have compared my designs with Williamson's which had 4.400 turns on Afe
= 32mm X
> 44mm . At 12 watts into a 10 Kohm RL the B at 50 Hz is less than 0.3 T and
> distortion is low .
Williamson gave the specs out as 15 Watts @ a peak flux densit of .725T @
20Hz....
My design gives just lower B at the same power and hence same
> low distortion as Williamson's . Well that's what my calculations from the
the RDH
> indicate . But the design allows power to go up to 50 watts easy . The
maximum
> inductance of mine is half the Williamson's but it is sufficient compared
with the
> RL of 5 Kohms . The initial inductance at low Va-a is ok for LF stability
at low
> levels . The benefit of a bigger core is less P turns and much less
leakage
> inductance , LL , as LL varies as Np squared .
Using a bigger core(double) would definetly help lower the turns as you have
done...But this makes mucho higher winding capacitance...and capacitance is
of major concernb especially with the bigger plate loads..like 10K...what
you do to reduce the leakage also increases the capacitance.... what total
capacitance did you calculate???
> To be honest I find your discussion of the way reluctance skips about
between Es and
> Is somewhat baffling .
It might be baffling to some..but it is the way E and I's work ...
> Sure , over the last 50 years there must have been some changes to
materials
> mentioned in 1950's books [RDH for example] .
> Well , would not most changes be improvements ? ie , if you substitute
today's GOSS
> material in a 50's design you maybe will get wider usefull bandwidth ,
less losses
> and less distortion .
Westinghouse in 1946 was making grain-oriented silicon steel under the name
HYPERSIL....this was both available in C-cores and E-I lams..... It has the
same characteristics as todays GOSS M6 magnesil..This was the C-cores that
Frank McIntosh was using when he was making his first unity-coupled OPT's in
1949... The stuff was VERY expensive and typically used in aero-space
applications and some Hi-End stuff....then in the mid 50's the cobalt based
Supermendur $$$$was the choice..and still is..
> So as a start point I find 2,000 turns on a big block of iron to be ok ,
and non
> oriented Si Fe to be ok for PP amps at least .
> Do you see any problem using GOSS C-core material for SE OPTs and chokes ?
So far I
> think chokes I make with C-cores
> are much better than using E&I lams .
> I've tried air-gapping PP C-core OPTs but the LF instability tended to
rise ,and the
> Lp was less , and so LF phase shift increased and I didn't like it much .
They say
> Partridge did it but that doesn't mean I would .
Partridge used the GOSS material and the air-gap was done by increasing the
number of n : n alternate stacking of the laminations... as you increase the
grouping size of the alternate stacking lams...this increases the effective
air-gap and thus lowering the inductance, but with GOSS OPT's operating at
1.2T at 20Hz in a P-P application you usually wind up with more inductance
then needed..so the gapping is not a problem with low frequency problems...
> So which is better listening at low level [in theory at least] the E&I or
C core ? I
> couldn't quite make out which type you meant was best .
In theory and in practice the E I lams WILL exhibit a quicker incremental
inductance curve and peak at half power...thus giving better low-end
performance at lower listening levels than C-core of the SAME inductance..
BUT since the losses are lower with C-cores they can have better detail in
the highs and sound cleaner..since the grain is rolled in the same direction
of the C-core.....so I don't know which is better..it depends on your
preference..
> The question still remains in my mind , exactly what amorphous cores of
what size
> would I need at least for the same overall performance ? And if I used
the same
> wind up on a bobbin and shoved in some am'cores what difference would it
make ?
Well...As I stated in another thread...amorphous alloy-B would produce twice
the inductance and 3/4 less core losses than GOSS M6....for the same give #
of windings and same core area and same flux density of 1.2T....BUT
Amorphous-B is limited to 1.2T for linearity sake.....
If you play your cards right with amorphous-B.....you can gap to increase
the flux density excursion thus reducing the number of turns to achieve a
higher flux density of equal core distortion....while still maintaining a
better L than GOSS equivelent..
CHEERS
Chris
Patrick Turner
Thanks for your detailed reply ; things are becoming a lot clearer .
Obviously we have a different attitude to heat . So many amps that I have to
repair have failed due to heat related stress .
For every 10 degrees C rise in operating temperature there is a reduction of 10%
in reliability . OK , that's a bit arbitary but it's my rule of thumb . If a
power Xfrmr runs at 50 dgrees C in winter then on a 37 degree C day the temp
will escalate to scorching and electrolytics , resistors and all the other bits
get pushed up in temperature and its at a time like this that something will
fail . Tube amps are heavy things anyway ; you may as well stack on the iron and
permit no more than 20 degrees C rise . And drill lots of holes through the
aluminium chassis so all the bits are cooled by air rising quickly through the
amp . OK , so I run power Xfrms at 0.9 T ; it's a bit wasteful but all the hot
running mass produced amps are inferior . Let's agree to disagree .
I have encountered power and OPTs with the E&I lams alternately stacked in
bunches of say 4XE then 4XI , then 4XI then 4XE . So there is an attempt to
effectively add an air gap in the core . I've always thought this bunched
stacking was just stinginess on the part of the manufacturer who couldn't pay
the workers to go E I , I E , E I , and so on . I have not tried gapping cores
like this . But I have tried playing with the gap on C-cores . When the gap is
increased in a power trans the L goes lower and the idle current goes higher and
losses and temperature rise . The higher the Mu the lower the idle current . The
C-cores I've bought locally from AEM have only fair flatness on the cut faces .
When I gave them a hand rub on a flat oil stone a few times I was able to reduce
idle losses by 30% , because I'd reduced the gap . AEM of South Australia make
their unicore GOSS material so that if you want the cores can have about half
the idle current of the C-cores , ie the Lp is very high . I personally don't
like the way it's so difficult to secure the unicore material prior to
varnishing . And once they are wound and varnished they can't be pulled apart
and rewound . The unicore gives about the same performance as an uncut toroidal
GOSS core but can be used as direct replacements for E&I lams . Potted in a good
type of wax the unicored OPT would be ok .But 2000 turns on a big core would
still be required I think .
My obsevations on OPT distortion in valve amps co-relates to your figures which
suggest Dn at 2.6% at 14 Hz and 0.4% at 20Hz , ie there is a big increase in
distortion as you approach Bs . But it isn't the full story . The generater
impedance , Rg , which is R a-a in parallel with RL , affects the distortion .
Suppose we used KT66 valves . So if we have a 10 Kohm RL and 3.2 Kohm triode R
a-a then Rg is 2.42 Kohms . The Dn might be much more than you say . If we used
pentodes the R a-a is 60 Kohms , RL is 5 Kohms then Rg is 4.6 Kohms . The Dn
will be double or at least a lot more . The equations in the RDH don't tell any
fibs , nor does my CRO .
Added to the last consideration is the fact that speakers don't offer a
completely resistive load at low F . My own speakers have Fbox tuned to 27 Hz
and there are two impedance peaks at 15 Hz and 33 Hz . So around the frequency
where most other amps could saturate or react badly to reactive loads , mine are
not troubled .
I allways design for the widest possible bandwidth . I like my amps to go down
to low F at full power . I agree there isn't much signal below 20 Hz but the
clients , friends , all the audio club members I know like amps that have 10 Hz
to 100 Khz of bandwidth if possible . I am never going to reproduce the garbage
that was made in the 1950's . Now Bs is a function of voltage so if I design for
14 Hz at full power then at half Vo or a quarter of full power Bs will occur at
7 Hz . Since most listening is done at well below 50 watts the bottom end in
amps I build is exceptional .
I have two monobloc amps I recently built which use a 75mm stack of 44mm tounge
iron and have 2,400 primary turns .
They go lower than the example in my last post . Each amp has six EL34 output
valves at present , but the amps are set up so six KT90 tubes culd be used for a
maximum of around 200 watts . But with humble EL34s I get 100 watts class Ab1
with a decent class A portion . I have some recordings of some mainly percussion
instruments and huge bongo drums . It is very easy to get the amps to start
clipping and I don't want the low frequency behaviour to be anything short of
fabulous . At the local community music school the drum/percussion group can
produce the most terrifically musical din ; can be heard miles away , like the
Lost Tribe Of Subburbia .When I replay recordings of this sort of stuff , I
want to rattle windows , just like at the live performance . So I like good bass
and so I design for lower frequencies than engineers in the past would have .
And modern electronic music has low frequencies which wobble your kidneys . This
is the modern world . The other reason for very low F capability is stability
with NFB .
The problem of shunt capacitance , Csh , is not a problem . If we look at the
equations in the RDH we find that there is no point in interleaving beyond a
certain number of times as the law of diminishing returns applies . So with the
2,000-turns-on-a-big-block approach the turn length is going to increase and
hence Csh too . But we can afford to keep the primary away from the secondary if
we use 4 pri sections and 5 sec sections . One can work through the multi step
graphs in the RDH to find out how altering the P to S gap will affect the
leakage inductance . Crikey it's a hard slog ; almost as bad as trying to
decipher the capacitance tables for OPTs . But I've got a formula which is
easier to use . Any way I find you never need to go to more than 4P and 5S
sections . 5P and 4S work similarly . I use plastic insulation of 0.75mm thick .
I don't use 3 layers of "empire tape" as Williamson says . The dielectric of the
material is around 2.0 .
I definitely don't use paper of any sort . I use 0.05mm of plastic between
neatly wound primary layers . Wax impregnation is used . The typical completed
OPT gives less than 5 Mh of LL refered to the primary .The capacitance gives
different response poles depending on RL and the tube selection ; the response
is determined by Rg [like at low F] . So if Rg is 2.4 Kohms then the Csh can be
twice that when Rg is 4.8 Kohms , for the same HF pole . I think there is about
600 Pf of Csh measured at each annode with the OPT in an amp , and with one side
of the sec grounded .
What I have found is there are minimal shunt or series resonance problems and
the dominant roll off is caused by capacitance .
The wide open loop response in my best amp permits a total of 38 Db of loop plus
local output valve cathode feedback to be applied with a resistive load before
instability prevents more NFB being applied . I normally only use 16Db of NFB
and set up for 16 Hz to 65 Khz of unconditionally stable bandwidth . Read my
other postings on square wave testing for more about tuning tube amps if you
want to . All amps are bandpass filters and if we have NFB then the response
slopes and poles are important considerations .Wide open loop response is
required for stability , and fidelity .
I doubt I'll get too many people knocking on my door demanding that I use cobalt
OPT material . I would then ask them to go shopping at Mr High End and they can
supply the OPTs around which I would build them a superlative amp . I'll do
anything to keep people happy providing the price is right . I suspect some of
the modern emphasis on amplifier materials quality is a bit of an obsession . I
think other priorities come way before using cobalt stuff .
So , after all that , do I use amorphous ? It would be more sensible than cobalt
, surely ?
Well , from what you say let's just replace the 44mm X 62.5mm GOSS core with
amorphous B and keep all the turns the same .
If ungapped it would saturate at about 1.5 times higher F . But Lp max could be
1000H , easy .
So then I gap the cores so Lp max is say 250H and I find that the initial Lp is
high enough for stability .
The more I gap the more Lp stays the same .
And the gapping stops saturation [like gapping does in a SE design when DC is
present].
So LF capability is as good or slightly better than GOSS .
Sure , I know about how to gap for best wanted all round performance . My SEUL
amps using 13E1 beam tetrodes required it .
Now I have a dilema .The Melbourne Audio Club allerted me about the possibility
of am'core use .
I was a bit skeptical but it seems now to be quite feasible . I have to tell
them . I can't afford the material at present .
I guess I'll have to ask them to provide me so I can try out the experiment to
validate the theory .
One other thing , I looked at a Russian website where am'core material was
alledgedly available ; they use it big time in electric trams and all sorts of
things . But their life expectancy for the material was a lousy 30 years only .
With all due respect to Russians , who make great tubes this "expectancy" could
be in fact a lot shorter . How long do you reckon this stuff will last even if
made by Americans ? Does it suddenly crumble into powder at a use-by date ?
Patrick Turner http://www.turneraudio.com.au
Chris Merren
> Obviously we have a different attitude to heat .
Yea your right...but I meant that it is do-able when space is a
problem...I also mentioned that gapping the power trannie has many
benefits..since L is not needed..
> When the gap is
> increased in a power trans the L goes lower and the idle current goes higher and
> losses and temperature rise . The higher the Mu the lower the idle current .
You mean the DC idle current????? This whould have nothing to do with
the L ....since L is a function of AC excitation...
> Added to the last consideration is the fact that speakers don't offer a
> completely resistive load at low F .
As long as your OPT is doing it's job ...don't introduce problems that
other devices are causing.. My concern with the OPT is that it is
flat....
> I allways design for the widest possible bandwidth . I like my amps to go down
> to low F at full power . I agree there isn't much signal below 20 Hz but the
> clients , friends , all the audio club members I know like amps that have 10 Hz
> to 100 Khz of bandwidth if possible .
Bandwidth has nothing to do with Bs...... I was refering to % distortion
in my last thread....
I also agree with using wide bandwidth...but that is a function of the L
inductance for the low frequencies..not the Bsat....
I use a -3dB point of no greater than 2Hz in my P-P designs....In most
of my P-P designs I have my -3dB point roughly at 1Hz....this way I have
vertually no phase shift by 20Hz...
> The problem of shunt capacitance , Csh , is not a problem .
Not so...
> But we can afford to keep the primary away from the secondary if
> we use 4 pri sections and 5 sec sections .
You are negelecting primary self-winding capacitance...since you are
refering to interwinding capacitance..
One can work through the multi step
> graphs in the RDH to find out how altering the P to S gap will affect the
> leakage inductance . Crikey it's a hard slog ; almost as bad as trying to
> decipher the capacitance tables for OPTs . But I've got a formula which is
> easier to use . Any way I find you never need to go to more than 4P and 5S
> sections . 5P and 4S work similarly .
Forget RDH.... You need to model the system for yourself and do the
mathamatical derivations for the frequency response, i.e. the COMPLETE
transfer functions...then you will see whats happening...
Why do you think Williamson used split bobbins wound in opposition ???
this was to control the capacitance while maintaining constant leakage
for a given geometry...
> I doubt I'll get too many people knocking on my door demanding that I use cobalt
> OPT material . I would then ask them to go shopping at Mr High End and they can
> supply the OPTs around which I would build them a superlative amp . I'll do
> anything to keep people happy providing the price is right . I suspect some of
> the modern emphasis on amplifier materials quality is a bit of an obsession . I
> think other priorities come way before using cobalt stuff .
Sepurmendor solves many problems...with it's Bsat at 2.4T ...except the
wallet problem :(
CHEERS
CM
JHPage
> > I doubt I'll get too many people knocking on my door demanding that I use cobalt
> > OPT material . I would then ask them to go shopping at Mr High End and they can
> > supply the OPTs around which I would build them a superlative amp . I'll do
> > anything to keep people happy providing the price is right . I suspect some of
> > the modern emphasis on amplifier materials quality is a bit of an obsession . I
> > think other priorities come way before using cobalt stuff .
> Sepurmendor solves many problems...with it's Bsat at 2.4T ...except the
> wallet problem :(
>
> CHEERS
> CM
>
>
curious, so I called the company I buy my laminations from (Magnetic
Metals) and ask for a quote for 1.5 inch EI (150 EI) cores of Supermendur
(49%Co 49%Fe 2%V) in a 50lb amount. I was told 15 - 18 weeks for
delivery, and the cost would be $75-$95 per pound depending on their cost
for the stock metal, they would let me know the exact price after I
placed my order. For reference, a 1.5 inch square stack of Supermendur
laminations is 5.93 lbs. Ouch!
Joe
Patrick Turner
Gee , So If I use 5 pounds of Supermendur in one OPT it will would cost me around $1,500
Australian allowing for freight and insurance and the exchange rate . To do a nice job
I'd have use silver foil or wire . And arrange exotic insulation , potting ,
terminations , labels , and mounting bolts and paint finnish , not to mention a month's
R&D .
Can I wind one for you at a quoted all in price of US $2,800 ?
Delivery in 6 months .
Well , is it worth it ? anyone producing such wonders will scream "yes ! , of course it
is" ,
but are they puulling our leg ?
If cobalt reduces distortion from 0.1% to 0.095% then don't you think it would never be
worth it . Just what would be the measurable benefits ?
Even if the core distortion was zero , you are still going to get valve distortion .
Patrick Turner http://www.turneraudio.com.au
Patrick Turner
Yes , a lot of older amps like the QUAD 2 monoblocks were compact and they allowed a
certain temperature rise . And we get some real hot days in Oz , much hotter than in
dear old England .
QUAD were producing 1950s "life style" products , which were very chic for the time .
Smallness and hotness are derived from accountants' control of quality . Big deal .
Space isn't a problem in specialist audio gear made to a specification , not a price ,
unless it's a mobile phone . And think of solid state class A amps . Gee , the heat to
reliability ratio becomes very accute , which is why there are so few in the shops .
Shops don't like selling smoke producers .
Perhaps you misunderstood about my core gapping experience . I've assumed that E&I
lams have a large effective gap due to the alternate grain direction and stacking
method . Well , much larger than a toroid of same Afe . Toroidal cores run cooler as
the iron losses are lower than E&I lams , ie the idle AC [magnetizing current] is less
. In a power tranny I thought as much Lp as possible would be a good thing to lower
the iron loss . I don't set out to get high Lp , but nice when it's there when C-cores
or GOSS is used . I still design for B = 0.9 Tesla . I find that is just a nice way to
go , along with a big core . The iron loss is nearly the same as the copper loss .
Anyway , none of my trannies run hot , and they stay cool , give good regulation and
overall efficiency , and can sustain big overloads ; what more would you want ?
OK , so next time I wind a power tranny I'll put a bigger gap in the core and I'll bet
it runs hotter ; if not , I'll buy you a beer .
The only time I deliberately gap a core is when I do SE amps that have DC in the
primary as well as AC . I won't go into details , but it takes a bit of juggling to
get it just right so Lp is adequate and the amp doesn't saturate or show huge
distortion at full power above 20 Hz . I like GOSS in C-cores for SE amps .
Gee , I thought bandwidth has got a lot to do with Bs . If Bs is as low as 10 Hz at
full power then the amp can go down to 10 Hz . Sure , Dn will be high , but such an
amp can be used for a sub woofer . In fact at a quarter power it'll go down to 5 Hz .
But if you measure the Lp even at very low V a-a then the LF pole is generally at very
low F , and the inductance to RL a-a ratio seems very healthy , so , as you say ,
there's not much phase shift or distortion at 20 Hz due to the OPT .
People are saying that using amorphous cores will stop phase modulation of HFs by the
LFs . Gee , let them explain all , so we'll all know . They say the non linear change
in Lp causes it . But the lowest Lp is still ok ; I think the effect they allege
occurs is very small . Let them provide some numbers .
I recall saying capacitance in my typical OPT measured about 600 Pf at a plate
connection with CT grounded and one side of secondary grounded . This includes the
total shunt C seen by the valves and comprises winding self C and P to S C . Say you
connect it up to two KT66 in tetrode mode with no RL . If R a-a is around 60 Kohms
then expect response to be down 3 Db due to C shunting at about 8.88 Khz . This roll
off occurs at a higher F when RL is connected as the more favourable total R to C
ratio reduces the effect of C . It's all in the RDH . In the case of KT66 triodes with
no load then the 3 Db point is over 160 Khz . But resonance of Csh and leakage L will
occur at such HF but it isn't a big bother . The top end wriggles in OPT response are
only troublesome if they occur too close to 20 Khz , which may make NFB use difficult
.The various stray Cs and LLs would comprise a very complex filter system .
All I know is that some Csh is unavoidable and in my OPTs it seems to be low enough ,
along with the leakage L . All the mass produced amps I've tested are worse . Valve
amps still sound good despite their percieved problems .
The complex mathematical approach is all very well . I do use some maths . But to
quantisize all the bits of C and L all over the place in some kind of extended model
is a pain and it varies between every OPT design . I've seen such extensive models in
old Wireless World magazine articles and in some older books and the equations stretch
across the damn page . Not for me . I've got a common sense appreciation of why the
OPT won't go to 1 Mhz , and why things have to be so . The simpler equivalent circuits
are enough to understand .
There are some engineers who could not build a child's swing without a week's time
spent on equations using calculus and a lot of vague notions and arm waving . I just
get some rope and a few planks and by lunchtime I've got a smiling kid . After awhile
,OPTs get like that ; easy .
I tried winding a split bobbin Williamson . I was not impressed . The overall
performance was not as good as my simpler design . The trouble of 4,400 turns means a
lot of leakage and stray C . In my design the need for a split bobbin is isn't there ;
the differences in DC resistances Csh and LL for each half primary are negligible .
One must remember that in 1947 the price of wire and iron was huge compared to now .
Williamson used used thin wire on the smallest core possible to get the copper losses
just under 10% . Other unsung adventurous DIY folks at the time probably got good
results by spending more on copper and iron and using less turns .
Did you see Joe's post . The costs of exotic OPT materials would make a grown man cry
.
Patrick Turner http://www.turneraudio.com.au
Jack Crenshaw
Clearly,
since vacuum-tube amps (almost) always use OPT's, the quality of the OPT is
bound to have a lot to do with the sound.
I'm bumfuzzled, though, by your comment,
>Forget these 1950's books...they are good for theory BUT most of the
>materials and charts do not pertain to materials used today...
>M6 grain-orientedis also known as Hypersil in the old books made by
>Westinghouse..today ity is refered to as Magnesil...
It's not that I don't agree with you; I don't have enough knowledge to do that,
though what you say certainly sounds reasonable. What's got me bumfuzzled
is that I talked to a fellow by phone who winds trannies for a living. Someone
here put me onto him -- sorry, can't recall his name; maybe the someone can
remind me.
Anyhow, I was a bit set back when he told me that he doesn't actually build
transformers, only rewinds them onto existing cores. I asked him if this was
a good idea. I specifically said, "Surely there must be better magnetic materials
today." He said no, that in the end, iron was iron, and the old iron is just as
good as the new stuff.
Was he talking through his hat, or merely selling his expertise, or what?
Jack
Patrick Turner
Did you see my queries to everyone on the subject ?
God invented electro-magnetic theory to keep us occupied , lest we sin , and to prove
His majesty .
Meanwhile a few souls seem to understand a bit . So amorphous cores came to be . There
are a lot of different grades and types . It seems it is very expensive . But it seems
as though there is a main benefit of a more constant Mu . And these cores would be of
benefit in SE amps . It seems that in a conventional PP design there is a large change
in Mu from say 1,000 to say 10,000 during the voltage cycle from low V a-a to high V
a-a respectively . Therefore there is a change of primary inductance which is in
parallel to the RL a-a . Now the claims I've heard are that this dynamic change at say
50 Hz causes phase modulation of the higher freqencies , and would/might sound awful .
The Lp on SE amps is more constant as the core is gapped to stop the combined AC an DC
saturating the core . Although difficult to follow , the RDH sheds some ligtht on this
behaviour . Anyway the same claiments say that since Lp on SE amps is more constant ,
then there is less phase modulation than PP , and it is one reason why they sound so
well , when all the other measurements indicate they should be utterly inferior .
Well I have not seen any talk of phase modulation in any OPT-behaviour descriptions in
the RDH or any old or new textbook .
I havn't noticed the phenomenon whilst looking at a CRO with two tones present . And
nobody has posted measurements to prove that the problem is a serious concern . My own
view is that the lowest amount of Lp in a good PP OPT is still a lot compared to RL so
the load change due to Lp change won't change the total nature of the RL much for
frequencies above 50 Hz .
One could measure for phase modulation if one used a detector sensitive to phase
changes . Phase modulation is one way to modulate RF frequencies . The basic
principles can be applied at 10 Khz .
Perhaps the phase modulation is similar to the doppler effect produced in loudspeakers
and in musical instruments . It may be of little concern ; I'm sure that I'm not sure
. Perhaps someone else might like to provide accurate measurements of the differences
between an optimised amorphous cored OPT and an optimised GOSS transformer in the same
amp .
I really don't have the resources to puchase hundreds of dollars' worth of cores to
experiment so that the claims by high-end manufacturers could be verified , eg , that
"our OPTs sound better because we use amorphous cores" .
I would think some manufacturers would only go to am'cores because it was cheaper to
achieve the same result , but be able to market it at the higher price .
Then there are the cobalt alloys if you really want to drain the budget . Again , we
have no measured samples . And no listening tests , I suspect .
Most of my clients know nothing of core materials ; but it is still in my interest to
know enough to establish what is the absolute best if it is possible to do so , and
what the costs and pros and cons may be .
The OPT in any hi-fi amp should not colour the sound .
Over the last weekend I was able to demonstrate a pile of speakers and SE amps to the
Audioplile Society Of New South Wales and they were pleased with the effort . I used
special interconnects , speaker cables , power line filters . These clearly improved
the sound . There is not much logic or measurement to support such a claim . But then
why does good valve gear stay in such demand ? Why doesn't vinyl dissapear ?
I try to build gear that measures well but I don't listen to measurements .
So where does core material rate in these considerations ?
And Jack , don't forget those old books .
Phil
Sorry to get in on this conversation so late, I just noticed the thread tonight!
My main thought in our emails on the subject was to find something with a small ratio
of maximum permeability to intial permeability, and the cobalt based amorphous cores
definitely win that contest. Unfortunately, they appear to be (1) non-existent, and
(2) seriously expensive. There are a couple of "ordinary" materials that, while
nowhere near the cobalt materials, nevertheless have max-to-min ratios that are a lot
smaller than GOSS. Listing GOSS first, with min, max, ratio, and saturation, they are;
GOSS 1,500 40,000 26.7 20.0
Sendust 30,000 120,000 4.0 10.0
1040 40,000 100,000 2.5 6.0
GOSS is 97% iron, 3% silicon
Sendust is 85% iron, 9% silicon, and 5% aluminum
1040 is 72% nickel, 14% copper, and 3% molybdenum
Sendust is so brittle that it is easily powdered, so it is cast rather than wound. I don't
even know if it available in cores. 1040 is nickel, so it is expensive, but not
as bad as the cobalt alloys. Put a decent sized gap in series with either of them, and
the fluctuation in inductance should be pretty small. Plus, the high initial mu means
fewer turns -- if the gap does not have to be so big that it wipes out this advantage.
I have no idea how the gap size should and will change when going from one material to
another, but it hopefully requires a similar percent reduction in mu.
I have no idea if the constant variation in mu does indeed cause phase-shift at high
frequencies, but that is what someone on the Joe-net said -- I think. My natural
pessimism makes me think that it is true, and that the unavailable, super expensive
cobalt cores will indeed sound best of all.
Parafeed is an idea that was "brought back to life" by Mike LeFevre, the idea being
to get the DC currrent out of the tranny. The plate current is simply sent on to the
power supply through a current source, usually an inductor optimized to work at high
as well as low frequencies (though I should think that an ordinary choke in series
with a high frequency inductor would work just as well, and cost less). The tranny
has one end connected to the plate, and the other connected to a "power supply",
which is usually just good capacitors. The caps can be connected to ground or to the
cathode, which has the advantage that the current through the cathode resistor
becomes DC, allowing the use on no or poor bypass caps -- in either case it takes the
bypass cap out of the sound, which is good. It is mainly used in SE amps since, as you
say, it is a lot of work for a PP amp, which is largely DC balanced anyway. In
parafeed, you can use the exotic cores without fear of saturation from DC current. You
can also avoid the need for an extra air gap, or at least use a gap only to reduce
variations in mu, without worry about saturation.
Phil
Patrick Turner
Patrick Turner wrote:
> Hi Phil, nice to hear from you again.
> I'll try to comment on your post,
> But who makes these claims?Where is their financial interest?I am not the world's ultimate
> pessimist, but I am right up there withleading skeptics, even cynics. Ho hum, I need my
> attitude to survive
> without emptying the bank too fast.
>
> IMHO, the turns on the primary are governed by the distortion at low
> frequencies becoming greater as those frequencies approach the
> saturation frequency. The fact that the innitial, or minimum mu is
> higher does not mean that fewer turns can be used; saturation
> will still occur at a certain tesla level, regardless of the mu.
>
> Take a typical PP design with 2,000 turns around a central core area
> of 36 mm by 55 mm, say we use double C-cores of GOSS.
> When the cut cores are carefully polished and closely fitted
> together you get a mu of about 7,000+ max at 1.2 tesla which might be at
> 400 volts rms across the primary, at some lowish F.
> And you may get a mu minimum of about 1,500 at very low tesla, say at 1 volt rms
> across the primary.
> Now if you gap the core with say a layer of cigarette paper
> between the cores, the minimum mu might be reduced say by 4 times.
> So the minimum inductance will be reduced by 4 times.
> But the inductance will be substantially be made constant;
> the L variation can be reduced ten times, and we can satisfy
> both conditions, ie that the core won't saturate at full power
> at 14 Hz and there is enough L which is strapped across the primary load
> so that the load seen by the tubes is mainly resistive even at 14 Hz.
>
> I doubt such an approach would lead to significant sonic gains in fidelity.
> However, I will try it on the next amp I make.
> Should anyone want to prove it to be different then I would welcome
> such effort; gee they would have to get busy in their worshop,
> and present me with their observations.
> I won't accept just theory.
>
> >
> >
> > Parafeed is an idea that was "brought back to life" by Mike LeFevre, the idea being
> > to get the DC currrent out of the tranny. The plate current is simply sent on to the
> > power supply through a current source, usually an inductor optimized to work at high
> > as well as low frequencies (though I should think that an ordinary choke in series
> > with a high frequency inductor would work just as well, and cost less). The tranny
> > has one end connected to the plate, and the other connected to a "power supply",
> > which is usually just good capacitors. The caps can be connected to ground or to the
> > cathode, which has the advantage that the current through the cathode resistor
> > becomes DC, allowing the use on no or poor bypass caps -- in either case it takes the
> > bypass cap out of the sound, which is good. It is mainly used in SE amps since, as you
> > say, it is a lot of work for a PP amp, which is largely DC balanced anyway. In
> > parafeed, you can use the exotic cores without fear of saturation from DC current. You
> > can also avoid the need for an extra air gap, or at least use a gap only to reduce
> > variations in mu, without worry about saturation.
>
> Shunt feed, or Parafeed, might even have been used commercially,but there are precious few
> examples.
> To keep the "iron caused interferance" at low levels in SE amps,
> the choke/chokes feeding the DC to the plate must be works of art.
> A decent 100 Uf cap off the plate can be used to couple an ungapped
> OPT which has it's other primary end grounded.
> Two valves can be used in PP with a center tapped PP OPT, with the CT connected to ground.
> But the primary inductance of both the feed choke and the OPT are in
> parallel so there is a need to keep both Ls high lest the total L be too low.
>
> My experience tells me it is wiser to just do a decent OPT and no need
> for parafeed, at all.
>
> Of course there is the third way, and that is to use a SRPP output stage,
> ie, two seriesed valves with the "top" valve's cathode connected
> to the OPT via a cap. So thus the current source would have no
> inductance or iron problems.
> A PP version would drive each half of the primary at all times.
>
> The possibilities are endless; there are more ways than one
> to minimise DC saturation effects.
> Luckily the most obvious deleterious iron effects seem to occur at
> large signal volts and low F, quite different to solid mistake amps
> whose cross over distortion makes them as bad at 1 watt as
> a poor valve amp at 10 watts.
> But then most stupid silicon is almost intolerable, at any level.
> Regards, Patrick Turner
Patrick Turner wrote:
> Hi Phil, nice to hear from you again.
> I'll try to comment on your post,
> But who makes these claims?Where is their financial interest?I am not the world's ultimate
> pessimist, but I am right up there withleading skeptics, even cynics. Ho hum, I need my
> attitude to survive
> without emptying the bank too fast.
>
> IMHO, the turns on the primary are governed by the distortion at low
> frequencies becoming greater as those frequencies approach the
> saturation frequency. The fact that the innitial, or minimum mu is
> higher does not mean that fewer turns can be used; saturation
> will still occur at a certain tesla level, regardless of the mu.
>
> Take a typical PP design with 2,000 turns around a central core area
> of 36 mm by 55 mm, say we use double C-cores of GOSS.
> When the cut cores are carefully polished and closely fitted
> together you get a mu of about 7,000+ max at 1.2 tesla which might be at
> 400 volts rms across the primary, at some lowish F.
> And you may get a mu minimum of about 1,500 at very low tesla, say at 1 volt rms
> across the primary.
> Now if you gap the core with say a layer of cigarette paper
> between the cores, the minimum mu might be reduced say by 4 times.
> So the minimum inductance will be reduced by 4 times.
> But the inductance will be substantially be made constant;
> the L variation can be reduced ten times, and we can satisfy
> both conditions, ie that the core won't saturate at full power
> at 14 Hz and there is enough L which is strapped across the primary load
> so that the load seen by the tubes is mainly resistive even at 14 Hz.
>
> I doubt such an approach would lead to significant sonic gains in fidelity.
> However, I will try it on the next amp I make.
> Should anyone want to prove it to be different then I would welcome
> such effort; gee they would have to get busy in their worshop,
> and present me with their observations.
> I won't accept just theory.
>
> >
> >
> > Parafeed is an idea that was "brought back to life" by Mike LeFevre, the idea being
> > to get the DC currrent out of the tranny. The plate current is simply sent on to the
> > power supply through a current source, usually an inductor optimized to work at high
> > as well as low frequencies (though I should think that an ordinary choke in series
> > with a high frequency inductor would work just as well, and cost less). The tranny
> > has one end connected to the plate, and the other connected to a "power supply",
> > which is usually just good capacitors. The caps can be connected to ground or to the
> > cathode, which has the advantage that the current through the cathode resistor
> > becomes DC, allowing the use on no or poor bypass caps -- in either case it takes the
> > bypass cap out of the sound, which is good. It is mainly used in SE amps since, as you
> > say, it is a lot of work for a PP amp, which is largely DC balanced anyway. In
> > parafeed, you can use the exotic cores without fear of saturation from DC current. You
> > can also avoid the need for an extra air gap, or at least use a gap only to reduce
> > variations in mu, without worry about saturation.
>
> Shunt feed, or Parafeed, might even have been used commercially,but there are precious few
> examples.
> To keep the "iron caused interferance" at low levels in SE amps,
> the choke/chokes feeding the DC to the plate must be works of art.
> A decent 100 Uf cap off the plate can be used to couple an ungapped
> OPT which has it's other primary end grounded.
> Two valves can be used in PP with a center tapped PP OPT, with the CT connected to ground.
> But the primary inductance of both the feed choke and the OPT are in
> parallel so there is a need to keep both Ls high lest the total L be too low.
>
> My experience tells me it is wiser to just do a decent OPT and no need
> for parafeed, at all.
>
> Of course there is the third way, and that is to use a SRPP output stage,
> ie, two seriesed valves with the "top" valve's cathode connected
> to the OPT via a cap. So thus the current source would have no
> inductance or iron problems.
> A PP version would drive each half of the primary at all times.
>
> The possibilities are endless; there are more ways than one
> to minimise DC saturation effects.
> Luckily the most obvious deleterious iron effects seem to occur at
> large signal volts and low F, quite different to solid mistake amps
> whose cross over distortion makes them as bad at 1 watt as
> a poor valve amp at 10 watts.
> But then most stupid silicon is almost intolerable, at any level.
> Regards, Patrick Turner