Bob Cordell's Power amplifier book

[Kindhornman]

Thanks Kgrlee.

I would think those golden eared reviewers would have made it fairly easy to identify the speaker that they obviously have a personal reason to select and could use location as a clue to identify that speaker. I have read enough of those types of reviews to see they often don't align in the least with the technical testing and reviews of those results.

 

[ostripper]

Do you end up with the same number of turns or do you need to increase the turns to get the same inductance? What if you reduce the diameter of the turns with the toroidal winding, would it really make any measurable difference? I don't think we would have to worry about saturation in an open air core coil winding like that would we with so few windings?

Inductance stays about the same , capacitance changes. The big thing is the field's much smaller. I'm going for it - I'll wind a pair for my new OPS's , check it out at 200W.
I have an L meter , I'll see any differences from a standard coil. An inductor like this
would be especially useful on a "mini-ops" .

That why they use huge air cores on real good speaker X-overs , no ferrite - no saturation.

OS

 

[Kindhornman]

Thanks OS,
It does seem though that you will need a bit more real estate to fit the coil because of the diameter of a toroid, would you turn it 90 degrees or make it flat on the board?

 

[Bob Cordell]

You will be very disappointed, the distortion is going to be high. Only 33ohm degeneration resistor, no cascode on LTP, no darlington VAS and 2EF. That's the hallmark of a cheap amp.

To me, distortion means nothing. It's the sound. I don't care to even debate tube or SS. It's the sound and that's the only thing that matter.



Current is push pull between P and N. You have P on one side and N on the other side. Current travel the whole length of the board. With P N alternate, the current loop is only on each pair. This is local current loop. This is particular important if you have thin trace.

I worked with MOSFET for years, not on audiophile. If it is done correctly, they are very rugged. But if you are not careful and leaving parasitic at the drain and gate path, they tend to oscillate and burn the gate very fast. They die a very quiet death. You turn on, it just die.....cold. You measure, everything shorted. Seen this too often already. Layout is very important and you have to treat it as RF device.......as they are. They are being used in RF power amp by making use of their Cgs capacitance for tuning. That's the reason I decided against using MOSFET myself even with all the supposedly advantage and stay with BJT. The last one I burn was used in my power scaling design for a tube guitar amp last year. You leave the lead a little long, you might get into trouble. By long, I mean 1", not a few inches.

You cannot use zener, transorbs to clamp the gate, it's internal. we tried it all. Only cure is short thick wires. You treat it as RF device, they last and last.

This is exactly right about vertical MOSFETs, mush less so with laterals because they are slower. The equivalent ft of a vertical power MOSFET can be in the 100's of MHz depending on operating current. I discovered this and discussed in in my JAES papaer on my MOSFET power amplifier with error correction back in 1983. This is why I use gate Zobel networks; one has to damp the device resonances to prevent oscillation under some conditions. One mechanism for burnout is gate punch-through due to excessive gate-source voltage caused during oscillation. More than about 20V peak inside the device can destroy the gate. See the paper on my website at CordellAudio.com - Home.

Cheers,
Bob

 

[kgrlee]

Still very interesting .... Richard. I know it's not Dadod's amp , but my CFA (ND) transitions from sub to single digit near 7-8W @70ma bias (below). All 3 of my CFA's do this.

Is this SPICE world or real?

Making a clamp that is truly transparent below its intended operating point is a lot harder than it looks. The non-linear diode capacitance is a real problem, especially if it's in parallel with Cdom as in the Baker clamp. Measure the 20 kHz THD before and afte connecting the clamp.

This is certainly the case.

But what is your policy on clamps? Your earlier posts suggest you prefer to let the amp saturate.

 

[DouglasSelf]

Is this SPICE world or real?

This is certainly the case.

But what is your policy on clamps? Your earlier posts suggest you prefer to let the amp saturate.

In general, yes.

 

[ostripper]

Is this SPICE world or real?

This is certainly the case.

But what is your policy on clamps? Your earlier posts suggest you prefer to let the amp saturate.

Correlating the RTA results at >50W/8R of 5 members that have built 100's of amps -
(Just sold out the whole first group buy of the new slewmaster) with my
cordell/keentoken BJT models ....BOTH worlds converge quite nicely.
Down to the uA and PPM.

Doug is right on with the comment on capacitance and performance "hit"
using the diode as clamp. No 1n914's or 4148's here ! BAV21 is 5pf ,
raises THD20K by 2-5ppm. Member 5thElement RTA'ed the blameless
at 3-5ppm with no clamp ... but still <10ppm using the BAV.
A high capacitance diode even "erodes" phase margin on a TMC compensated
design. (at clip).
The transistor clamp (like in some examples) works well if the beta enhancer
and clamp is the same device.
With a dissimilar BJT pair as beta/clamp , a TMC design will produce a short burst of
oscillation right as the clamping takes place. Using two similar ksc1845's (on my
"wolverine") , this "chirp" goes away ??
Perhaps someone could elaborate on why this happens (only with TMC - not
with a standard miller compensation).


Either technique , real world <10ppm on a TMC blameless. It's a simple choice
of trading 5ppm for a more reliable amp.
Must be doing something right , Badger builders quite often sell their last amp
because it's not even in the same league with the blameless , SQ wise (most use the clamp).
The new EF3 blameless is quite unreal - I'm listening now.
Edit - I had no choice but to solve this errata , the builders did not like the saturated waveforms (complaints !!)

OS

 

[kgrlee]

Is this SPICE world or real?

Correlating the RTA results at >50W/8R of 5 members that have built 100's of amps

What is RTA? Are these numbers THD1k or THD20k

1n914's or 4148's here ! BAV21 is 5pf raises THD20K by 2-5ppm. Member 5thElement RTA'ed the blameless at 3-5ppm with no clamp ... but still <10ppm using the BAV.
...
The transistor clamp (like in some examples) works well if the beta enhancer and clamp is the same device.
With a dissimilar BJT pair as beta/clamp , a TMC design will produce a short burst of oscillation right as the clamping takes place. Using two similar ksc1845's (on my "wolverine") , this "chirp" goes away ??

Perhaps someone could elaborate on why this happens (only with TMC - not
with a standard miller compensation).

Either technique , real world <10ppm on a TMC blameless. It's a simple choice of trading 5ppm for a more reliable amp.

Is 'wolverine' with this scheme on the WWW somewhere?

 

[kgrlee]

But what is your policy on clamps? Your earlier posts suggest you prefer to let the amp saturate.

In general, yes.

Very wise. DBLTs, including mine, show overload & recovery behaviour is probably the biggest audible difference in amps around 50W 8R.

Gotta have something to audibly distinguish your stuff from the opposition

I'm discounting the Golden Pinnae amps which are marginally unstable on real speakers as I'm sure your designs aren't in that exalted category

 

[kgrlee]

I would think those golden eared reviewers would have made it fairly easy to identify the speaker that they obviously have a personal reason to select and could use location as a clue to identify that speaker.

Well in the tests where they provided a pair of speakers & set them up, the speaker would be allocated to A, B or C and then swapped out for unknown pairs in the same position.

In a serious test, the victim is NEVER told what he is listening to or for ... whether its speakers, amps, cables bla bla.

It's remarkable how the sound changes for these Golden Pinnae (note capitals) when they can't see the speaker

 

[Kindhornman]

kgrlee,
I stopped reading Stereophile many years ago as I found the articles discussing the listening tests and the setups to be so far from the measured results to be incomprehensible. Special cables, little blocks to raise the cables of the floor, special power cable and so many other things make the testing useless to me. I didn't have much problem with some of the actual measured results but then they would come up with crazy reasons why the listening tests and measurements didn't correlate in the least. It seemed the higher the costs the more the excuses many times.

RTA Real time analysis?

When I worked in a medical lab doing actual testing on patient samples we always used double standards. The standards at the beginning of the run must match the standard at the end of the run or all tests would have to be repeated. I look at dblt as the same type of testing though not really, if you can't identify certain things at least two times correctly I would question the results for sure.

 

[ostripper]

What is RTA? Are these numbers THD1k or THD20k




Is 'wolverine' with this scheme on the WWW somewhere?

"poor mans AP" = RTA - TrueRTA Real Time Analyzer FAQ Page ... 20K.

A blameless is real hard to measure with a mid level sound card - but a 300$ one
will give reliable results. Blameless has lower THD than most sound cards !
Second one one I made for the slewmaster - post #599
http://www.diyaudio.com/forums/solid-state/248105-slewmaster-cfa-vs-vfa-rumble.html
Huge thread , the matching "top" CFA is post 6782 - read to 6863 and see two
abusive squarewave and overload tests.

All the other amps have been used and abused - never has one shown any errata.
Most are 20-30ppm 20K .
By counting the group buys - >300 output stages , toner etched .... I keep finding
the design in remote places on the web.

OS

 

[31697B]

Hi,

The RTA being used, I have doubts about its accuracy below -100dB. Everything I have measured and compared against the best SOTA instruments show this to be true. Have you access to better test equipment?

I use mainly three instruments to measure distortion: The Audio-Precision 2722A; The ShibaSoku 725D; and the Panasonic VP-7722A. Any one of these, new, cost what a new car does. I started with sound cards and modified old HP339A's...... they cant do it either.

Not saying a 100-300 dollar tester cant do as good... but I just doubt it. Pls check so we can have confidence in your numbers. If you like, i can measure an amp for you.


THx-RNMarsh

 

[kgrlee]

When I worked in a medical lab doing actual testing on patient samples we always used double standards. The standards at the beginning of the run must match the standard at the end of the run or all tests would have to be repeated. I look at dblt as the same type of testing though not really, if you can't identify certain things at least two times correctly I would question the results for sure.

There's a similar principle when measuring speakers and also doing DBLTs.

A DBLT is a measurement. Your DBLT panel is your instrument. Like all instruments, it has an accuracy and you need to check it regularly. You make sure your panel ONLY has true golden pinnae and no deaf Golden Pinnae.

I don't do ABX tests but ABC tests cos you get statistical significance faster.

And you NEVER tell your panellist you are just checking his accuracy either.

 

[Kindhornman]

When doing an abc test would the person doing the test know the abc selection of would that person be also blind? I imagine the answer is yes and that is the basic difference between the abx and the abc test, in the abx it would be random unknown selection.

I'll email you some questions I have about spiders if you have the time to answer questions. .

 

[Bob Cordell]

you can wind a air core toroid on a insulating former without any mag material - the numbers can be found in many online calculators

https://www.google.com/#q=toroidal+inductor+calculator

just enter u_r = 1


still some fringing field close in - but much reduced at distance vs cylindrical solenoid

I used air-core toroidal output inductors on my Super Gain Clone. I wound them on a pencil, then removed the pencil and formed them into a toroid. Got about 3uH. However, they took up quite a bit of space.

Cheers,
Bob

 

[ostripper]

Hi,

The RTA being used, I have doubts about its accuracy below -100dB. Everything I have measured and compared against the best SOTA instruments show this to be true. Have you access to better test equipment?

I use mainly three instruments to measure distortion: The Audio-Precision 2722A; The ShibaSoku 725D; and the Panasonic VP-7722A. Any one of these, new, cost what a new car does. I started with sound cards and modified old HP339A's...... they cant do it either.

Not saying a 100-300 dollar tester cant do as good... but I just doubt it. Pls check so we can have confidence in your numbers. If you like, i can measure an amp for you.


THx-RNMarsh

That's what I have planned.

As far as the RTA , it all comes down to what digital/analog environment it has.
Use this - (below) and trueRTA will have it's "grass" at -120db. It should
have the capability to accurately measure sub PPM.

I agree that most forum members are using typical 100$ internal cards. Not
really precision with a lot of variables that can not be avoided.
PS - check out pete millett's external "box" and -140db "grass" here -
Soundcard Interface
OS

 

[31697B]

It isnt the noise floor of FFT that is not real (it isnt as far as the real noise of the DUT) but the data shown below -100 is not real (IF it is based on ADC/DAC chip inside).

Excellent interface though to safely handle higher input voltages. Though it add distortion (small but measurable) as well. WE should compare them to be sure.

THx-RNMarsh

 

[31697B]

kgrlee,
I stopped reading Stereophile many years ago as I found the articles discussing the listening tests and the setups to be so far from the measured results to be incomprehensible. .

I had a chance to measure the freq response of a 2-way speaker which had been reviewed. What I found did correlate with the listener's review BUT......

he was hearing the effects of floor reflection causing a broad hole in the lower mid. I used the same distance to measure and height of the speaker stand as the reviewer. Unfortunately, when speaker tests are made, it is a 1 m and no reflection from a room involved. So naturally the review and measurement do not correlate well.


Thx-RNMarsh

 

[AKN]

That's what I have planned.

As far as the RTA , it all comes down to what digital/analog environment it has.
Use this - (below) and trueRTA will have it's "grass" at -120db. It should
have the capability to accurately measure sub PPM.

I agree that most forum members are using typical 100$ internal cards. Not
really precision with a lot of variables that can not be avoided.
PS - check out pete millett's external "box" and -140db "grass" here -
Soundcard Interface
OS

Looks nice. Have you seen any real measurements of your "planned" sound card?

Below noise floor of my old trusty Asus D2 internal PCI card.