U73b Translation

I'm in the process of translating the TAB U73b compressor circuit description from German to English.  Most of the circuit is straightforward but I would like to check with someone more knowledgeable about the reason for cutting the top end above 15kHz.  Is this because it was designed for FM or TV stations and there would have been problems with the 19khz stereo marker?  Or LOPT whistles from monitors?

When its finished I will add it to the archive.  Some of our German contributors can do the rest of of the operating manual easier than I can!

best
DaveP

DaveP said:

I'm in the process of translating the TAB U73b compressor circuit description from German to English.  Most of the circuit is straightforward but I would like to check with someone more knowledgeable about the reason for cutting the top end above 15kHz.  Is this because it was designed for FM or TV stations and there would have been problems with the 19khz stereo marker?  Or LOPT whistles from monitors?

Click to expand...

Definitely for FM broadcast. The maximum audio frequency is 15kHz, with a steep filter. Limiting audio BW to 15k instead of 20k limits the actual Rf spectrum to about 180kHz, which in turns allows 200kHz separation. With 20k audio BW, the spectrum would be about 250kHz, which would impose 300kHz separation, so a given channel would contain 30% less usable stations.

Definitely for FM broadcast. The maximum audio frequency is 15kHz, with a steep filter. Limiting audio BW to 15k instead of 20k limits the actual Rf spectrum to about 180kHz, which in turns allows 200kHz separation. With 20k audio BW, the spectrum would be about 250kHz, which would impose 300kHz separation, so a given channel would contain 30% less usable stations.
Thanks Abbey,

I had my suspicions the filters were for FM radio.  I would not have thought that a recording studio would want to lose the "air" above 15kHz, so if I make one. I'll leave out the filters.

Best
DaveP

I have finished the most relevant details on the U73b (attached)

These need to be read with the basic U73 circuit description on which I'm still working!  Still something to be getting on with.
You can find these manuals at Kubi's site:-
http://audio.kubarth.com/rundfunk/index.cgi

This is an interesting circuit as it's Feed Forward, not too many of those around.  Apparently practically every European recording went through this limiter during its heyday.

Back to the U73 Schaltung!

Best
DaveP

This is what Abbey was referring to, Thanks to Engineer's corner:-

CARSON'S RULE Calculating FM Modulation Bandwidth


A formula is used to determine the RF Bandwidth or occupancy for an FM signal. If you are designing an FM system on microwave or satellite, you will need to take care that your signal does not cross-talk into other signals on the system.

Frequency Modulation creates modulation sidebands that theoretically extend to infinite bandwidth. These sidebands consist of Bessel Functions of any order. From a practical standpoint the band occupancy of an FM modulated carrier only needs to count the Bessel Function sidebands of significant amplitude. The formula that calculates this bandwidth is called CARSON'S RULE.
CARSON'S RULE requires knowing the modulating frequency and the maximum frequency deviation of the transmitted carrier. As an example, a monaural RF band modulator will have a peak deviation of 75KHz and the highest audio frequency is 15KHz. To calculate the CARSON'S RULE bandwidth occupancy of this signal, add the highest audio frequency to the peak deviation (15KHz + 75KHz = 90KHz), then multiply by two to include both the upper and lower sideband (90KHz X 2 = 180KHz). The CARSON'S BANDWIDTH for this signal is 180KHz. Since there are many Bessel Function sidebands beyond 180KHz, FM channels must be spaced considerably farther apart than 180KHz. The FCC has determined that a spacing of 400KHz provides sufficient "Guard Band" to effectively prevent inter-channel cross-talk, but that 180KHz is sufficient bandwidth to receive the original modulation with less than 1% distortion. The distortion is due to a failure to receive all of the modulation energy.
Similar CARSON RULE calculations can be used for other modulation bandwidths and peak deviations, with similar considerations for Guard Bands between channels.
Amplitude Modulation bandwidth can be considered exactly two times the highest frequency of modulation, while Frequency Modulation bandwidth is described by Bessel Functions that extend much higher than those of Amplitude Modulation. In fact the "FM Advantage" in signal-to-noise ratio stems exactly from spreading the modulation over a greater bandwidth than Amplitude Modulation.
CARSON'S RULE
BANDWIDTH = 2 X (PEAK DEVIATION + HIGHEST MODULATING FREQUENCY)

There you go.  You never stop learning on this forum!
best
DaveP

OK, it's done. :
This is the standard U73 circuit description attached below.  Some of its features are modified by the U73b circuit changes that I've  given earlier, so they need to be read together.

They have used an E88CC to save space in their cassette format.  To beef up the output level they have used a high inductance choke with a DCR of 15k, I wonder how CJ would wind one of those?

It would be possible to use a triode wired pentode like a 6HB6 (which has the same performance) with parallel feed from a 15k resistor in the same way as a REDD47 circuit,  you could then use a parallel 6BZ7 for the first tube....just thinking out loud,mmmm

best
DaveP

DaveP said:

I had my suspicions the filters were for FM radio.  I would not have thought that a recording studio would want to lose the "air" above 15kHz, so if I make one. I'll leave out the filters.

Click to expand...

Don't leave them out.

ALL properly conducted DBLTs on bandlimiting show a distinct preference for sharp filtering.

My tests circa 1980 were with multiplex filters screwed up to allow 20kHz.  I was quite surprised to find this Millenium, tests with 15kHz filters showed the same result including some that our own Jan Didden was involved in.

I thought this century, with all music sources already heavily band-limited cos EVIL digital, that this preference would be less.

Ricardo,
Would you adjust them to 20kHz then?
best
DaveP

ricardo said:

Don't leave them out.

ALL properly conducted DBLTs on bandlimiting show a distinct preference for sharp filtering.

My tests circa 1980 were with multiplex filters screwed up to allow 20kHz.  I was quite surprised to find this Millenium, tests with 15kHz filters showed the same result including some that our own Jan Didden was involved in.

I thought this century, with all music sources already heavily band-limited cos EVIL digital, that this preference would be less.

Click to expand...

I'm not understanding this at all.

Assuming we're not trying to keep out something nasty (aliasing, carrier signal etc) I've never heard a sharp band limiting filter I like as much as a gentle one? 

In this case the reason to band limit is gone? If so why would we leave the filter in?  And if we did why would we want it sharper? (other than to preserve as much below 15k as possible).

Genuine questions, I must be missing something.

ruairioflaherty said:

I'm not understanding this at all.

Assuming we're not trying to keep out something nasty (aliasing, carrier signal etc) I've never heard a sharp band limiting filter I like as much as a gentle one? 

In this case the reason to band limit is gone? If so why would we leave the filter in?  And if we did why would we want it sharper? (other than to preserve as much below 15k as possible).

Click to expand...

When you conduct proper DBLTs (MUCH more expensive than buying the latest AP) your whole view of what is important and what is trivial changes AS IT SHOULD.

The correct reaction to DBLTs is to modify your theory/prejudices so they fit the EXPERIMENTAL RESULTS.  Not claim the results must be rubbish cos your theory says so.

I don't KNOW why in the 21st century results with mostly bandlimited material gives the same results as my 1980's test with mostly vinyl & tape stuff which is pretty yucky at HF.

I could pontificate but it would be from the wrong orifice 

In the meantime, if you want to put something in your power amp which WILL make it SOUND BETTER, try a sharp cut filter between 15-20kHz.

But you'd better not let the deaf Golden Pinnae look inside or dem objectivists measure it cos the MAGIC BETTER SOUND will disappear in a puff of smoke as soon as they KNOW what you are doing.  ;D
_______________________

I would use 20kHz and sharp cutoff cos that's what I tested in Jurassic times.  You wanna try something else .. go ahead but check it against something that is known to give good results.

Just be aware its not easy or cheap to do DBLT properly.

Ricardo,

I like to understand the reason for things.

Carbon comp resistors have voltage dependent characteristics (as well as the noise).

Capacitor" sound" is related to lack of piezo-electric effect.

Feedback affects the way higher harmonics are treated, etc, etc.

The translation I have just laboriously carried out specifically says the filter is to reduce the bandwidth for FM broadcast.

While I respect your experience, I can't see how restricting the bandwidth would make a better recording (MP3 excepted).

People all over the world try to make their Hi-Fi systems, tweeters in particular, extend to 20kHz or beyond.  My understanding is that it this helps the transients to sound more faithful to the original sound (the square wave multiple harmonics effect).

In fact, I would be worried that phase shift associated with any filter would be undesirable, I am a disciple of the less is more philosophy  as far as electronics is concerned, one of the reasons why I prefer tube circuits to chips.

Please take my comments in the constructive spirit in which they are intended, any more light you can throw on the subject would be greatly appreciated, as would any comments from other members.

I don't start a build until I understand all the angles.

best
DaveP

Dave, just after the war, Harry Olson conducted a famous http://ejjamps.com/PDF/experiment-that-saved-high-fidelity.pdf

My 1980 results and that of other researchers this century may be the equivalent of the Chinn & Eisenberg report.  It may be that the remaining distortions today are still less obtrusive with sharp cut filtering.  But I don't think so.

The important point is not to ignore carefully controlled listening tests.  The carefully controlled is important.  Sighted tests are worse than useless.

Some of these tests were carried out to see if zillion MHz sampling was worthwhile.  All the reliable results show that anything beyond CD quality (properly dithered 16b 44.1kHz) was a waste of time.  All the reliable results show either no benefit or else an advantage to the lower sampling (and hence greater bandwidth limited) signal.

But if you are building something, why not install a switch so you can test it for yourself and increase the store of human knowledge?  To be really useful, the switch has to be capable of being integrated into a sensible DBLT but even a crude switch with a little care over levels will tell you a lot.

The pseudo Golden Pinnae install far more useless stuff that either have no sonic affect or often degrade sound .. eg the capacitors carved from solid Unobtainium by virgins.  Why not install something that has proven better sound.

A listening test is probably the most important test we can apply to an audio chain.  Why not do it properly (DBLT) and actually take action based on the results?

Ricardo,

Thank you for the article, it was very interesting, I often wondered why 1940's radios had a 10nF cap from the plate of the output tube to earth!  It was because they preferred the treble cut I guess, just what they were used to.

I will take your suggestion on-board and try to make the filter switch-able.  I doubt that I could conduct meaningful listening tests at my age as I listened to too many bands in the 60's in the confines of the Marquee club in London.  I can just about hear/feel a 15kHz sine wave if my ear is a few inches away from the tweeter!

best
DaveP

DaveP said:

I doubt that I could conduct meaningful listening tests at my age as I listened to too many bands in the 60's in the confines of the Marquee club in London. 

Click to expand...

Well then, you should know what music *really* sounds like!!!

Can you tell us more about your tests Riccardo? 

How were they configured?  Source? etc.

ruairioflaherty said:

Can you tell us more about your tests Riccardo? 

How were they configured?  Source? etc.

Click to expand...

The bandwidth limiting was quite early on in our Blind Listening Test Journey.

Some important landmarks along the way include

http://www.aes.org/e-lib/browse.cfm?elib=2476  Intermodulation Distortion Listening Tests (which was about MUCH more than IMD and includes some of the most important stuff on Audible Speaker Distortions.
http://www.aes.org/e-lib/browse.cfm?elib=2859  Loudspeakers: An Approach to Objective Listening
http://www.aes.org/e-lib/browse.cfm?elib=3798  Absolute Listening Tests-Further Progress
http://www.aes.org/e-lib/browse.cfm?elib=10251 Is Linear Phase Worthwhile?

We got one of Lipsh*tz & Vanderkooy's ABX boxes but the final method we used for nearly 2 decades was an ABC method which allows statistical significance much more quickly.  We quickly found out that complicated forms and scoring was counter productive. 

Our methods differed in many ways from the false prophets Floyd & Olive.  eg our victims chose THEIR own music and set the levels they liked.

When you take care of these and other 'minor'  details you find, unlike Floyd & Olive, that the man (and particularly) the woman in the street as perceptive as 'trained listeners'.  But trained listeners usually came to a conclusion more quickly.

The Bandwidth stuff was relatively easy to detect compared to the Phase & Dither stuff.

We never published anything on Bandwidth Limiting but others have.

I'm sorry to be so brief.  I could write books on DBLTs but probably never shall.  I've provided more details on this and other forums.

I'm reading the braunbuch and the excellent translating of daveP for the U73 ( not the b version )but still I don't understand the purpose of the RC network at the interstage transformer secondary . point 38 and 13.
Can someone explain ?

38 and 13 comprise a filter with the -3dB point at 16Hz, this starts working from 40Hz.  This is apparently there to compensate for the difference in the ears perception of sound when the volume is changed (see Fletcher Munson Curves).  The volume changes when switching from limiting to compressing with switch 78.  The central position on 78 is a straight through so you can hear the effect of in/out.

This compressor is a feed-forward design that relies on the side chain being a perfect mirror of the control amp, so it has to be very carefully adjusted to match.  If the side chain is more potent at high inputs than the control amp, then the output level will fall rather than staying  flat as a limiter should.  So it's not a project for the faint hearted 

Thanks for your interest
DaveP

aha, that's an understandable explanation but it seems they left it out on the U73b version.

I guess it's because the U73 was for FM broadcast use, where fidelity to the input programme was important, whereas the U73b was for recording studio use.  If the sound needed adjusting then that was done in the mix.
You made a good point

DaveP