MOSFET output stage

[Eeyore]

Mackie HR 824 s are almost an industry standard and are active.

Berhinger 2031a are favored by a number of my pickier friends - also
powered.

No disrespect to the above two, I'm sure they're quite competent, but
'serious' studios won't be using them.

KRKs are very popular at the moment with high end studios for
mini-monitors.
http://www.krksys.com/

Graham

 

[Eeyore]

All that yields is your post.

Okay, sillyness aside, without the quotes it does indeed return some
results. But I couldn't find a definitive answer either way. I do remember
reading about a BJT that had an ft of 500 GHz a couple of years ago, which
was a new record at the time. It's indeed possible that things have changed
since then, although I'd bet that BJTs can't be far behind.

Of course this is all a whole different game than audio, but frankly I think
both BJTs and MOSFETs offer sufficient bandwith in audio applications these
days.

BJTs designed for audio vary very widely in stated fT.

Graham

 

[Eeyore]

True, none of this is relevant for audio. Anything will work for
audio.

Almost anything. RCA's 2N3055s weren't exactly the fastest in the world. 800kHz
fT IIRC. I hate to think what their germanium predecessors like the OC35 were
like.

Graham

 

[Jan Panteltje]

The real 802.11G throughput is 2.8MB/s at the best. An uncompressed
audio channel takes roughly 100KB/s.

That number you give is 100% correct (tested with scp, it reports the speed).

But using UDP (not TCP/IP) we can go faster:

original file:
-rw-r--r-- 1 root root 2473920576 2008-08-23 19:00 goldfinger.ts

Main PC:
date;cat goldfinger.ts | netcat -q 0 -x Maximize-Throughput -u 10.0.0.155 1234;date
Mon Sep 22 21:26:19 CEST 2008
cat goldfinger.ts 3.20s user 32.41s system 12% cpu 4:36.56 total
netcat -q 0 -x Maximize-Throughput -u 10.0.0.155 1234 3.67s user 215.37s system 79% cpu 4:36.56 total
Mon Sep 22 21:30:56 CEST 2008

Other side (eeePC)
netcat -u -l -p 1234 > /dev/zero

Completed, file length / total time gives 8.9MB/s
x 8 makes about 70 Mbits / second, clearly something went wrong..... my math?

However, I could not play it, when I tried to write to a RAM cache for speed,
it was all garbled.

UDP = no error correction if packet not received, packet may arrive in any order too.

Anyways, I repeated the test the other way around, now for a mp3 file:
-rw-r--r-- 1 user user 103760023 2008-09-09 19:03 instrumental.mp3

home/user> date;cat instrumental.mp3 | netcat -x Maximize-Throughput -u -q 0 10.0.0.150 1234;date
Tue Sep 23 00:17:16 CEST 2008
Tue Sep 23 00:17:40 CEST 2008

That makes 24 seconds for 103760023 bytes, makes 4.32 MB/second, that is
34.58 Mbps.

There is some data loss, as the received file length is: 103579799
But it seems to play very well, the end is also there, but for course
data integrity is bad, there must be damage somewhere, but I listened to part of it,
and that sounded great, but it is too long to listen to all of it now, maybe tomorrow.

 

[krw]

He said kBps.

Don't know what I was looking at. ...ahead at WiFi??

44.1k samples / sec x 2 bytes each = 88.2kBps. Allow overhead for collisions etc. Actually if
you transmit the full AES/EBU SPDIF data it's THREE bytes.

100kBps sounds about right for one mono channel. Or 150 kBps with flags etc.

Ok, now put three or four of these together and it's still easily
manageable over WiFi.

 

[krw]

Jan Panteltje write:
I mean I try to avoid wireless whenever possible. So when I'm at home, I
watch cable TV and listen cable radio and use my landline phone...

Why? I spent yesterday installing my router into some built-in
cabinets in the great room. The builder put the power outlet in the
top cabinet and all of the LV stuff (phone, Ethernet, and speaker
wires) in the lower cabinet, with a granite slab in between. I
wanted the router on top of the cabinet, where it's out in the open
but can't be seen (from anywhere but the upstairs hall).

 

[Eeyore]

[email protected] says...

Don't know what I was looking at. ...ahead at WiFi??


Ok, now put three or four of these together and it's still easily
manageable over WiFi.

With Quality of Service and no latency ?

Graham

 

[Eeyore]

But they were tough triple-diffused parts. Some other peoples' 3055's,
like the Fairchilds, were epitaxials, very fast but very fragile, and
about half the silicon area.

Were they triple diffused ? RCA called them hometaxial.

The Motorola part was faster but less rugged. Many hobbyists got caught out with
exploding amps by using the wrong brand because JEDEC hadn't specified the devices
adequately and the authors of the articles hadn't thought to look into it.

Eventually, I think Motorola made a 2N3055H to indicate RCA compatability.

Graham

 

[RichD]

An unusual design but the copper thingy is very likely
heatsink, in which case you're referring to the actual output
devices.

The copper thingy is flexible, about 15' long,
attaches to the box with the magnets at one end.

It would indeed be unusual to use that
as heatsink, but perhaps novel and efficient.

They in turn usually have devices called 'drivers'
which precede them, although it's
less necessary with mosfets, only for ultimate performance..

So the output driver feeds the output stage?
Is that the standard jargon?

 

[Eeyore]

The copper thingy is flexible, about 15' long,
attaches to the box with the magnets at one end.

I have never seen anything like it. What is the brand ?

It would indeed be unusual to use that
as heatsink, but perhaps novel and efficient.


So the output driver feeds the output stage?
Is that the standard jargon?

Yes. Somtimes there may be more than one one stage of drivers too. This
is where expertise in amp design comes into play.

Graham

 

[Phil Allison]

"John Larkin"

But they were tough triple-diffused parts. Some other peoples' 3055's,
like the Fairchilds, were epitaxials, very fast but very fragile, and
about half the silicon area.

** Fairchild were making silicon transistors here in Australia in the early
1970s including their "fast" 2N3055. There were also several imported 3055s
around at the time like " Power Physics " which were bog standard spec as
well as the similar Philips type BDY20.

One of my jobs back in 1973 was to "de-bug" kits built by customers of an
electronics parts store and one of the most popular kits was a basic " CDI
Ignition " module based on a magazine article. This consisted of a self
oscillating inverter with two 3055s driving a pot core transformer whose
secondary produced around 300 volts DC after rectification which then
charged a 2 uF Mylar cap. The energy from the cap was periodically dumped
into the ignition coil primary by firing an SCR that simply shorted out the
300 volt rail.

The idea was that self oscillation ( at an audible 5 kHz ) of the 3055s
would automatically cease due to the short on the secondary when the SCR was
triggered and so allow it to commutate off ready for the next triggering.

Unfortunately, the store owner has purchased a large stock of locally made *
Fairchild 2N3055s * and filled the kits with them.Those damn Fairchilds had
enough HF gain to keep on oscillating despite the short ( at about 35 kHz as
seen on a scope ) when the SCR fired, so preventing it turning off.

The 3055s then just sat there in total silence getting hotter and hotter.




....... Phil

 

[krw]

With Quality of Service and no latency ?

Send it a dozen times. Timing shouldn't be too much of a problem.
The clock can be updated periodically. This stuff isn't all *that*
hard, at least good enough for audio.

 

[Eeyore]

[email protected] says...

Send it a dozen times. Timing shouldn't be too much of a problem.
The clock can be updated periodically. This stuff isn't all *that*
hard, at least good enough for audio.

Your phrase "good enough for audio" does not inspire confidence. I work in the professional /
production area of audio. 'Good enough' usually isn't for us. Indeed, for a host of reasons I'd also
probably want to transmit 24 bit audio. Some customers might want 96 kHz sampling too.

Go on. Ask why !

Graham

 

[Eeyore]

probably want to transmit 24 bit audio. Some customers might want 96 kHz sampling too.

Buffer the data. Include a GPS receiver in each speaker to synchronize
the outputs to sub-ns accuracy. That will eliminate those awful cable
prop delays and perfectly coordinate the granularity of the layering
of the holographic soundstaging, or something. The critics will love
it.

John

That's certainly an interesting concept. ;~)

Graham

 

[krw]

Your phrase "good enough for audio" does not inspire confidence. I work in the professional /
production area of audio. 'Good enough' usually isn't for us. Indeed, for a host of reasons I'd also
probably want to transmit 24 bit audio. Some customers might want 96 kHz sampling too.

Yes, I know you're a "professional" audiophool. If you can't hear
it, it's good enough. Fer instance, millisecond is good enough as
long as all channels are together. They all get the same signal.

Go on. Ask why !

Why? I could care less about "professional" audiophoolery. BTW, we
were talking about powered speakers, not "professional" grade audio.
96kHz? What nonsense.

 

[Don Klipstein]

Yes, I know you're a "professional" audiophool. If you can't hear
it, it's good enough. Fer instance, millisecond is good enough as
long as all channels are together. They all get the same signal.


Why? I could care less about "professional" audiophoolery. BTW, we
were talking about powered speakers, not "professional" grade audio.
96kHz? What nonsense.

I occaisionally hear artifacts in 16 bit 44.1 KHz, in music.

It is easy to make a test signal turn up severe artifacts with 44.1 KHz
sample - see what happens with a sinewave at a higher audio frequency that
is several Hz off a frequency that the sample frequency is a multiple of.

Since I only occaisionally hear artifacts in music with 44.1 KHz 16 bit,
and when I do I usually find them minor, I would expect a sample rate
twice as high as that to be OK.

- Don Klipstein ([email protected])

 

[Eeyore]

[email protected] says...
probably want to transmit 24 bit audio. Some customers might want 96 kHz sampling too.

Yes, I know you're a "professional" audiophool.

Just a professional.

If you can't hear it, it's good enough.

Some people's ears are sharper than others. I expect the phone would be good enough for you ?

Fer instance, millisecond is good enough as
long as all channels are together. They all get the same signal.

It wasn't specifically the timing I thought you had in mind.

Why? I could care less about "professional" audiophoolery. BTW, we
were talking about powered speakers, not "professional" grade audio.
96kHz? What nonsense.

Many top studios are now mastering at 24 bit 192 kHz. Using products like this, reckoned to be the best in
the world and manufactured by another company I used to work for.
http://prismsound.com/music_recording/products_subs/ada8xr/ada8xr_home.php

Graham

 

[Eeyore]

probably want to transmit 24 bit audio. Some customers might want 96 kHz sampling too.

I occaisionally hear artifacts in 16 bit 44.1 KHz, in music.

It is easy to make a test signal turn up severe artifacts with 44.1 KHz
sample - see what happens with a sinewave at a higher audio frequency that
is several Hz off a frequency that the sample frequency is a multiple of.

Since I only occaisionally hear artifacts in music with 44.1 KHz 16 bit,
and when I do I usually find them minor, I would expect a sample rate
twice as high as that to be OK.

For some reason 88.2 kHz never caught on. It's always been 96, double the 48kHz sampling used in some other
digital audio products, as far as I can remember.

Graham

 

[Eeyore]

BTW, we were talking about powered speakers, not "professional" grade audio.
96kHz? What nonsense.

Do you think professionals don't use powered speakers ?

Graham

 

[RichD]

LMAO ! Yes, Kevin had indeed had as have I.
Amps that sold commercially into the
pro-audio market.

Did you use MOSFET on the output stage, and why?