Group DIY Vari-mu

[gemini86]

I remember a forum member speaking of a burst generator being useful for sidechain testing. I'm at work so I'll let others do a search. I imagine a white noise burst at random volumes pre-recorded on a looping wave file would help you see what's going on as well...

Also, I don't believe in limiting yourself in terms of "I don't do ICs, I'm an old tube guy!"... While I agree there is a beauty in all tube gear, I would limit that beauty to a vintage original, or a clone thereof. I know this isn't the "poor mans" compressor, so I'm not looking for the cheapest solution, but using glass in a sidechain seems wasteful to me. That's my opinion, YMMV.

 

[DaveP]

Gemini86,

What I mean is I have no IC expertise and I would be outside of my comfort zone.

Didn't mean to sound snooty.

I think I'll stick with what I know, if someone cleverer than me can modify it later to an IC equivalent, then they are free to do it.

best
Davep

 

[gemini86]

Sorry for my misunderstanding. I get where you're coming from.

 

[DaveP]

Right I've been doing some maths on the spec for the sidechain and I need someone to check it.

This is all about the current and voltage needed to charge the timing capacitor.

I have surveyed all the vintage compressor specs and presented them in a table, (I don't think I can post a spreadsheet).

Based on the compressor tube spec to get to 10umos, the column has the negative voltage which could be -15 to -35 depending on the tube.  The timing spec is based on one time unit (tau) which is 0.63 of that voltage.  Where there is no attack resistor I back-calculated the output Z of the driving circuit, and used that.

These are the formulae used:-  1amp= 1coulomb/1 sec

1coulomb= Capacitance in Farads x Voltage on the Capacitor (the 0.63 figure)

Substituting: Amps= Capacitance x Voltage/time in seconds.

For example:- WE 1126, Amps= 0.000,0001 x 12.566/0.001
=0.00126 Amps. This is the charging current.

The voltage drop across the 10k output Z is 12.56V so the charging voltage is 12.566 + 12.566 = 25.2V.

The Watts needed to do this is therefore 25.2 x 0.00126 = 0.0317W

The very high output from the Collins 26U is because it has a dedicated secondary on its transformer.

These levels all seem reasonable to me but there maybe some who disagree, I'll be interested to get some feedback on this as it will help me design the new sidechain.

By the way, I can never get the attachment button to work, can someone explain please?

best
DaveP

 

[Michael Tibes]

I remember a forum member speaking of a burst generator being useful for sidechain testing. I'm at work so I'll let others do a search. I imagine a white noise burst at random volumes pre-recorded on a looping wave file would help you see what's going on as well...

I believe the trick is to use a short loud sine burst at 15-20 kHz. The effect is that you trigger compression with a more or less inaudible signal, so all you'll hear is the thumping of the compressor. This makes it easy to set the balance between the 2 tubes. For the U73(b) there is a detailed description (frequency / burst length and level) in the description (probably only in german?), but the concept doesn't need too much science if it is only about the tube balance.

Michael

 

[PRR]

> I can never get the attachment button to work, can someone explain please?

Maybe you could say what it does or does-not do for you? Perhaps your O/S, and where you keep your files?

Does it accept the file-name then choke on the size or file-type?

> I don't think I can post a spreadsheet

"Allowed file types: doc, gif, jpg, jpeg, pdf, png, txt
Restrictions: 1 per post, maximum individual size 1000KB"

You "can" attach simple spreadshets. Save As... CSV or TXT. Both will lose all fancy-frills, but will keep numbers in line. CSV may be a wee bit better. *.csv is not an allowed filetype; after saving re-name it to foobar.csv.txt and tell users to remove the .txt before asking their spreadsheet app to open it.

Sometimes we jam the storage-quota so I'll attach any darn thing to check....

 

[DaveP]

Thanks PRR, I'll try again next time.
best
DaveP

 

[alexc]

Hi Dave

Very interesting table you have calculated there!

Don't have any useful information to add, unfortunately.
Maybe the more experienced among us can add some comment.

It seems to make sense, based on the sidechain circuits time constant calculations of the basic cap/resistances

Some of the calculated attack times seem quite fast. Sub ms and so on.

A bit over my head, but look forward to the results!

I added an attachment using the "Attachments and other options" item
Browses to a file location fine, adds the file fine but in my browser, Firefox 8.0.1 does not open properly.
Noticed this on other postings too.
Seems to want to open everything as an 'index.php' file
I think it is aFirefox issue.

Cheers

 

[DaveP]

Hi Alex,

Thanks for your comments.

Its not hard to back calculate the output resistance when there is no attack resistor.

Taking the WE1126 as an example, the published spec is 1mS and the capacitor 0.1uF.

As the attack time is  Time = C x R, then R = T/C  so 0.001/0.000,0001 = 10,000.

I needed to have a ballpark feel for the voltages and currents involved in order to design the sidechain amp

best
DaveP

 

[DaveP]

I've been doing some work on the bench so I don't know if this still qualifies for the drawing board.  No doubt someone will tell me!

I've been checking out the side-chain circuit for the Federal AM864/U.  Its nice and neat but as PRR pointed out, the 6SQ7 is a "lame tube", having less poke than an 12AX7.

I subbed a 6BF6 which has 12AU7 characteristics.  I found that the plate cap has a bearing on the frequency response and has to be 100nF to have a flat response and also the attack resistor has to be reduced to 12K.  The CV set-up is 2M/100nF.  Plate resistor 22K, supply 250V, plate volts 150, cathode resistor 1.6k and cathode volts 6V.

This set-up gave me -22V across the frequency spectrum, but if I used an external diode down to earth I got -28V (extra 6V from cathode).

OK so I wanted to measure the attack time.  Putting a square wave through it and one channel of the scope on the input and the other channel inversed on the output, I got a straight rise time of 20uS then it curved off and joined the flat top of the square wave after a total of 60uS.  This seems quite promising to me, so I thought I'd report on this before going further.

I propose to move to the LAB once I start the construction in the New Year.

best
DaveP

 

[emrr]

I'll post some librarian corrections for you, at least.

1126 1941
86-A 1941
SA-39 1947
26W 1945
26J/356E, same thing, could leave one off. 
Collins LA-1D (?)  Not a Collins unit.   

The very high output from the Collins 26U is because it has a dedicated secondary on its transformer.

Not sure what you mean here; it has a tertiary feedback winding.  The first SA-39's also used tertiary winding before switching to cap coupled feedback from the plates. 

I have yet to use a vari-mu limiter with low-Z sidechain drive.  Plenty of the Hi-Z variety can certainly erase a leading transient.

 

[DaveP]

Thanks Doug,

I think I got the dates off the drawing and manual printing data so they were probably later editions.  I only wanted to put them in date order to track the tube changes and trends.

This from Vintage Audio site

"
Western Electric 1126 limiting amplifier (compressor)

This is Western Electric's second limiter after the rare WE110.

Introduced in the late 30s and produced into late 40s it mechanically consists of three separate chassis each with separate faceplates. The three units are held together in one integral unit with a steel binder on each side of the three rack mount chassis. On top is the 126 limiting amplifier, center is the 298 control panel with controls and meter, bottom unit is the 20B regulated power supply. The three units together comprise the 1126 system."

I think it must be that old because it uses those weird spring chokes!

best
Davep

 

[emrr]

Updated 1126 date, found an ad.  1126C is dated 1950, and changes to a pair of 6sn7 output tubes, additional stages as cathode followers feeding the side chain.

 

[DaveP]

Doug,
Can you post that circuit or PM it to me ?

Or any others I may have missed out.

best
DaveP

 

[emrr]

Need to get it scanned, putting it on the list. 

 

[DaveP]

I have now made a prototype side-chain and tested it.

They red Voltages are DC and the blue AC signal.  The output for 0.65 input is -24V CV but that was just a convenient level, the maximum output before clipping is about -70V CV, not that all of that would be needed for later tubes.

I tested the output resistance to be ~250 ohms, that would make the attack about 25uS with a 0.1uF cap theoretically.

With the component values shown, the sidechain provides a steady -24V from an input signal from 30Hz up to 1Mhz!!! so something will have to be done to tame the HF response of these TV tubes.

best
DaveP

 

[leadbreath]

hey dave,

very interesting post, good too see some inovation and exploring new ideas!!

in keeping with the originl spirit of this post i will be donating £10.00 for the use of your intellectual property.

mick

ps: go dave go!! go dave go!!!

 

[DaveP]

Hi Mick,

That's very generous of you, and you don't even know if it works yet!

Anyway, thanks for the support.  PRR's 12AU7 compressor was an inspiration to many,  so we are all standing on the shoulders of giants here.

That's about it for me now as I'm on holiday til the new year.  I'll be back on the case in January.

all the best for Christmas

DaveP

 

[PRR]

> I tested the output resistance to be ~250 ohms

For resistive load.

Put cap and decay-resistor at CV. My gut says the cap-loaded impedance is very much higher, because the time-cap charges-up which charges-down your 470nFd.

. . . . .  ah, the charging effects make the attack time very short but it won't hold the initial peak on a sustained signal. 17V peak in with 0.1u hold-cap and decay resistor equal to your 39K ground-reference resistor gives -14V on the first peak but sustained 17V pk in decays to to 8V in 40mS. Musical result: organ attack is dipped but then level rises 4db-5db, which is not right.

While 39K seems to be a small decay resistor, "decay" should happen after the sound stops, not while sound is steady.

Your "39K" needs to be much less than your lowest decay resistor.

Depending on GR grid power requirements, this may be tough.

 

[DaveP]

PRR,

I see my circuit was incomplete and misleading.
When I did the test, the CV point was in fact connected to a 0.1uF cap and 2M release resistor.

I originally tried a diode bridge but it worked better with those two 39k.  The value of those resistors affects the frequency response, higher or lower and it gets spoilt.  Most of the old 40's circuits used a simple triode with a load resistor and diode like that.

All my tests so far have been on a continuous sine wave for the purpose of providing a high CV from a low output resistance.  I think you are saying that I need to also examine it under single peak conditions?  Are you also saying that 0.1uF is not really big enough to keep the voltage down long enough?

best
DaveP