emrr

Re: the Poor Man 660 support thread
« Reply #3240 on: June 02, 2017, 05:58:57 PM »
Spend the money for some real Daven/Shallco and they won't fail.  Elma shouldn't either, can't say why they did.  Every old broadcast service limiter used rotary switches in path, and those were not allowed to fail.   I wouldn't use the cheapo's like Ural, but I have no experience with those failing either.   alexc appears to have no problem with them. 
Best,

Doug Williams
Electromagnetic Radiation Recorders

"I think this can be better. Some kind of control that's intuitive, not complicated like a single knob" - Crusty

"Back when everything sounded g


Kingston

Re: the Poor Man 660 support thread
« Reply #3241 on: June 03, 2017, 11:48:00 AM »
I had 4X of those Elmas on a build and all of them failed with the exact same fault: certain range was consistently losing contact. Pretty "awesome" stereo glitch and bubbling fx.

Similarly I have a quad Elma rotary as a stereo passive balanced attenuator and its just horribly glitchy. Never ever going to use them again and I've made it a point to remove all Elmas from within 100 meters of me for peace of mind and safety.

http://www.michaelkingston.fi/files/correct-and-accurate-rotary-switch-placement-and-handling.jpg
« Last Edit: June 09, 2017, 03:49:30 PM by Kingston »

alexc

Re: the Poor Man 660 support thread
« Reply #3242 on: June 07, 2017, 10:53:06 PM »
If I wanted to keep the stock attack time, but mod the release times, I could wire the TC network in parallel with the stock 10uf and just vary the resistors, is this correct?

For the background on the pm660 time constants mods .... https://groupdiy.com/index.php?topic=29056.20 describes the 'Bluebird mods'  which I think was the earliest ....

Basically the story is this :

- the pm660 is loosely modelled on the original fc660 which has a 'time constants' circuit consisting of a 'fixed part' and a 'variable part'  - they are parallel'd with each other across the rectifier bridge

- the pm660 'tc fixed part' consists of r17 (221K), c3 (0.22uF)  and r18 (120R)
- the fc660 'tc fixed part' consists of 220K and 2uF

- the pm660 'tc variable part' consists of 'attack pot' 10K rv7 and c4 10uF
- 'the fc660 'tc variable part' consists of a 2 deck rotary wired in parallel with 6 selections available

It can be said that the actual 'attack' and 'release' times are quite inter-related, the time circuit having both a 'fixed part' and a 'variable part';  so trying to seperate out the 'attack' and 'release' into two distinct controls is not really valid

 ie.  a single control is probably warranted, as in the case of the original fc660.

NOW, the pm660 'attack' pot doesn't do all that much, and the 'single 10uF cap' is wayyyy long,  so many have gotton rid of all that and moved more towards the fc660 by using a single 'time constant' rotary switch of 6 positions and 2 decks.

[one can also simplify the original fc660 2 deck 6 position into a single deck switch]

BUT the question is 'do I really need the 10uF ?' (which is super long and not usable in some peoples opinion) was mostly asked *after* people (like me) had already soldered C4 into the pcbs. 

SO, I kept it in circuit -  they were expensive at the time!

Note that if one does desire the 10uF super-extra-long release time, it is too large to easily fit to the back of a lorlin; and further, there's not much space left in a pm670 to lay it down elsewhere, so best to use C4   :)

...

THUS my choice was to keep the on-board 10uF as 'selection 1' of a 6 position lorlin 2 deck switch with two decks : one for the caps, and the other for the resistors - basically the 'Bluebird' mods.

....

NOW  there's a number of ways to 'patch in' the 'time constant' rotary switch :

option 1

- one can short out the C4 cap if 10uF is not going to be used, and patch in the lorlin in place of the 'attack'  pot 
    ie.  rv7 connector pins 1,2  (pins 2,3 are linked on the pcb)

or

option 2 (my way)

- one can populate C4 cap and use it as the first selection ie. 'send' it to the lorlin -  this requires

      i) cut to the 'attack' rv7 connector pins 2,3 which are linked at the pcb   
     ii) link from rv7 connector pin 3 trace to 'top' of C4 trace ...   at back of pcb near the connector
     iii) wire rv7 connector pin 3 to position1 selection terminal of the lorlin cap deck
     iii) patch in the lorlin at rv7 connector pins 1,2

....

The easiest thing is option 1 - short C4, ditch 'attack pot' rv7 and use connector rv7 pins 1,2 to patch in a lorlin

Here's Bluebirds drawing  :)  with a 2deck 5 positon rotary and his choices of the cap and/or resistors

..

Now the Lorlin switch has 2 decks on the back laid out as 2 half-circles of selection terminals with a pole terminal in the middle of each.

The pole terminals are to be connected in parallel ...  and that goes back to the pcb. (rv7 pin 2)
The selection terminals are soldered .. to 'one end' of  ...  the caps on one deck   ....  and resistors on the other deck

The 'remaining ends' of the caps *and* resistors are all wired together into a ring shaped 'cage' assembly ..
.  that 'common' ring is soldered up strong, and then a wire goes back to the pcb (rv7 pin1)

The lorlin layout is out there somewhere .... 

... insert pic ...
..

As to the  actual values :

- the cap can go from 10uF (very long release time) down to 0.22uF (very short), around 200V dc rated.
- the resistors generally from nil or 25K .. 500K ..  can be 0.5W

Now we know this 'time constant network' is also a 'filter' circuit for the output of the recifier bridge, it's a bit like a 'psu ripple filter'

The 'control signal' we derive from the rectifier bridge has a strongly and quickly varying 'dc' waveform (negative going) plus some 'ripple'.

The time constant network affects this ripple as well as dictating the 'timing' of movement of this negative going control signal (the desired end result of the sidechain amp).

Anyway, at the lower valued cap'n'resistor choices in play (fastest attack+release), it  is the 'least filtered' with  more ripple  - so one can use some additional resistance, in series and/or parallel to 'clean things up a bit'    ..   wrt to the phenomenon of 'control signal leaking into the audio signal'

It's hard to describe and easier to hear and see on a cro  :) Its a balancing act :)

Hence the sorts of choices made are .. like Bluebird has on his - the deck resistor starts low K, the deck cap starts high C  and each progress on the switch to high K and low C.

Actually determining the real 'timing' values in seconds or milliseconds and even microseconds is not easy .. it's a bit hit and miss and requires a more 'pro' level of skill and knowledge.

So, getting hung up on the specific times isn't appropriate until one has the measuring skill to verify.

One can do some rough calculations to best decide on the values and how they relate to reality.

Which is why trying to intimately map the pm660 time constants to the fc660 is a little apples-to-oranges. At the detailed level, the pm660 is only loosely based on the original, including in the control amp and thus the timing circuits.

One can go ahead and use the fc660  time circuit exactly if desired but probably makes sense to simplify a little.

In fact, I believe to correctly map fc660 times to pm660 requires more sidechain power than is available in in the mighty pm660 with its thunderous wastage of energy! Some people went on to make a pcb module to drive more power with transistors - they became known as the 'scamp' ers  :)

....

Fortunately, it all works out pretty well using the 'Bluebird' type mods.

I used 6pos dual deck lorlin, the decks in parallel  (fast to slow)

0.22uF, 0.5uF, 1.0uF, 2.2uF, 4.5uf and onboard 10uF on deck 1  //  in parallel with
      26K,    68K, 150K,330K, 470K and  Open Circuit   on deck2

which is, of course' also  in parallel with the on-pcb 'fixed timing' part as I described earlier.

Go ahead and calculate   ;D   

In practice, it will probably take a few attempts to get the timing that makes you happy. I used a resistance substitution box to find the rough R value for each of the C values while testing ....  that worked reasonably well :)

...

I find the setting I use most in middle of the pack - something like 1uF // 150K
Its fast attack (some ms) and a mid level 'release'something like 0.5s  ;) at generous GR of -12dB or so.


It can release very fast but that does begin to get 'challenging' in the area of distortions of various kinds....

It has a very fast attack but that does begin to get 'challenging' in more subtle ways ....

----

Such is the magic  of big tube limiter design and construction  !!
« Last Edit: June 10, 2017, 01:34:11 AM by alexc »
I ping therefore I am

Holger

Re: the Poor Man 660 support thread
« Reply #3243 on: July 20, 2017, 11:51:18 AM »
I had 4X of those Elmas on a build and all of them failed with the exact same fault: certain range was consistently losing contact. Pretty "awesome" stereo glitch and bubbling fx.

Similarly I have a quad Elma rotary as a stereo passive balanced attenuator and its just horribly glitchy. Never ever going to use them again and I've made it a point to remove all Elmas from within 100 meters of me for peace of mind and safety.

http://www.michaelkingston.fi/files/correct-and-accurate-rotary-switch-placement-and-handling.jpg

Michael, just saw this, excellent...
Any other reliable 24 pos switches on the market except for the chinese ones?