Build Thread:VECA MS (VCA comp)

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
tata said:
Gustav,

i must say im havig soo much fun with this VECA now. the MS mode really makes a change, no more negative compression on vocal buss, i can hit it really hard - needle will break soon (dont care, its a cheap 1mA DC meter) - unless i change the resistor and call it a 40dB GR meter or so..

im happy that i have R101 socketed in the 1st build too - so it will be an easy "mod".
thanks for marking it on the PCB (and on the schemo)!
(the asterisk and MikeClev's post in the "SSL Clone compressor ratio measurements" thread was a big help.)

Yes, theres a reason it is marked on the schematic and the board, but I can't make the connection with what you are presenting  with the ramping sine. I guess its a visualisation thing, and I prefer to make sure nothing major is at play before noodling with details. (like meter calibration, level set-up to know what I am actually seeing, if I care enough to take measurements of something).

Just closing down on a night shift, so I won't be checking up on it today - If its needed, Ill put it in the errata.

Gustav
 
im very curious of your findings w/ the measurement

in the meantime i think i found the right word, instead of "negative" i actually meant "inverted" compression.
sorry for the confusion.
 
My measurements.

3Khz sine fed from a tone generator.

Comp set in 2:1 (The proposed mod does not affect the internal relation between ratios, so just this ratio is enough for the test)

Setting a reference

-2 dBu to unit, threshold fully out. Adjust make-up gain to match output to -2dBu

Dial in threshold, so output reads -4dBu.

We will now test the assumption that our threshold is set at -6 (4 less than our -2dBu input), which corresponds to a -4dBu (2 less) output at a 2:1 (4:2) ratio.

Test 1.

Adjust test tone to 0dBu.

Output measured at - 3.1dBu.

- 6 relative to the 0dBu input (-6), and a -3.1dBu (0-3 = -3) output 6:3 = 2:1

This holds up to the reference setting with a .1dBu discrepancy.

Test 2.

Adjust test tone to -4dBu

Output reads -4.9dBu.

- 6 relative to the -4dBU input (-2), and a -4.9dBU relative to the -4dBu (-1)  is 2:1

This holds up to the reference setting with a .1dBu discrepancy.

Conclusion

The 0.1dBu discrepancy is so low, it can easily be ignored (maybe just due to the meter).

Ratio functions effectively at 2:1 with no modification to resistor values at normal operating levels.


So what is happening in Tata's test?


As the test under normal operating levels shows, more or less input will generate more or less output, relative to the ratio setting, when signal is above threshold. Test also shows the ratio to be correct in the stock build.

Looking at Tat's test again, it could be the bump at first is the nose of the sine peeking in, slams the signal down, then shows negative ratio type behaviour due to crapping out completely.

It could  be an error in the build, but the resistor change should not be necessary, as my findings show. The ratio is linked to the threshold in this build, and changing  the resistor to a value as high as suggested probably had the side effect of moving the threshold out of the way.

Making up a random tests can be like kicking tires just to see what happens, but for actual trouble shooting, and making comments on a design, it is advisable to measure the voltage of the signal and follow a more scientifically minded approach (especially when making corrections based on the tire kicking). If further testing is needed, I would advice starting out at known, standard operating levels. We need some physical measure, and dBFS is a digital reference which makes no sense as a reference in the physical domain, unless we know, and are 100% sure about the converter calibration level. (And even if we are, measure the voltage to make sure!)

Hope that helps!


Gustav
 
Gustav,

i copy-pasted the dBU / dBFS values from the RME manual, so i guess thats correct.
i suck at maths and i dont really have instruments to measure dBU, but i have the beautyful RME meters.
as i mentioned the Ratio does matter, the inverted compression happens on 4:1 and 10:1.

let me show you with a practical, real life test:

VECA_R101_120k_vs_200k_10-1.png


top is the RME out / VECA input signal, red is the VECA w/ R101=120k, and green is the VECA R101=w/200k. all other settings are the same.
Ratio 10:1, Threshold and Recovery at middle, MS at M, Blend OFF, HP SC OFF.

can you see the inverted compression kicking in on the red item and not on the green? this is a vocal example. (showing just one channel of a "stereo" wav, where both channels are the very same)

with the 200k i can just use the VECA as i expected, unlike w/the 120k.
yeah, at "rockandroll" levels.. :)

dbFS is often used in this forum, for eg in the "SSL Clone compressor ratio measurements" thread / MikeClev's calibration.. without mentioning which interface he uses (dBFS / dBU may vary, right?..)
 
tata said:
let me show you with a practical, real life test:

1. I am glad you are enjoying the unit with the change you made.
2. I will not be implementing your resistor change in the errata, not even for rock n roll.

I think anything else I could write would be a repeat of this last post.

http://groupdiy.com/index.php?topic=59576.msg792895#msg792895

Gustav
 
if anybody could run the test i originally showed (ramping sine, especially at 10:1 setting) i would be thankful.
that way i could learn if both of these builds have some (and same) problems - or its by design.

are you gonna update the BOM btw -not that i need it anymore, but new builders could find it useful?

thank you.
 
I wish I had the wisdom to give some sort of meaningful reply, but I took a glance at the thread again, and I just dont.

If you still think it makes sense to ask someone to replicate what you're doing, theres nothing I can say to change your mind.

Gustav
 
tata said:
if anybody could run the test i originally showed (ramping sine, especially at 10:1 setting) i would be thankful.
that way i could learn if both of these builds have some (and same) problems - or its by design.

are you gonna update the BOM btw -not that i need it anymore, but new builders could find it useful?

thank you.

I had to give up helping, since you are reverting to a request  for replication of a non-replicable test (as it stands, your test did not rule out problems with levels, possibly pushing the side chain headroom, or the converters coming back too hard. This could easily be ruled out by the simple test I posted).

I tried to help you by doing a controlled test to verify for you that the design is correct, so your problem will boil down to either the test method or an error in your build. Ruling out the test error is impossible if you ignore the need for real testing.

Even so, I really want people to have great, working units, so I have something to share that could point you the right direction.

I am working on a split version of the basic GssL design. In the first prototype, the internal meter is still showing expected GR, but I am getting a lot more compression when measuring the signal path, and even negative ratios.

I thought this could be of help to you, since it could be a symptom of a build error on your VECA. CV through the 1K going to the signal path VCAs is way too high in this scenario.

Doing these tests, I have done comparisons with the VECA, and I (once again) guarantee you, your problem is not the result of a design problem with the VECA, but it could quite possibly be related to what I am seeing in the split version. Ill work on adjusting the values accordingly on the prototype, but that is a result of the single-channel CV load.

Wether its the signal you are pushing into the unit, or a build error, you are going into the side chain too hot. 1200 or 1206 on the input? Should be 1206, for example...

Hope that helps in some way.

Gustav
 
1206 / 1646 / 2180B

the "SSL Clone compressor ratio measurements" thread has some reports of 10:1 having inverse compression in some builds, so i guess im not the only one experiencing it.

i have the front control panel redesigned to fit in 1U, and added 1.5:1 ratio - so far so good, no inverse compression (still having the 200k resistor instead of 120k), but i dont call it done yet. more experiments soon - but in the meantime if i could have 22 seconds of your life and youd be kind to record the signal for me in stereo / 2ch, id still like to see how a properly built VECA is crapping out.
maybe you dont understand it, but those waveforms tell me a lot. i can see the "knee", and such.
10:1 is giving me 1dB change when input changes by 10dB - i guess thats good.

3 builds - all the same. i wonder what i messed up, if....

i dont see why i should not push the gears "hot", i could link you lots of discussions about "hot being the new standard". (or at least no problem with it, since the digital / hybrid era)

btw 2180B - no need to disconnect pin4 in this circuit?

IMG_7584_vca_1U_front+.jpg

two Ratio controls on the photo - the standalone is having the 1.5:1 mod.
attack and release pots arent the perfect ones, but this is what i had at hand for now.
(the DIY PCBs / components dont touch the PCB under, i have that PCB in the foil / bag.)

thank you
 
tata said:
i dont see why i should not push the gears "hot"

You should do what you want - you seem confident and content in your approach, and I realise now that my last post was misplaced.

tata said:
btw 2180B - no need to disconnect pin4 in this circuit?


- You have to ask how to reverse a bypass switch.
- You think dBFS is an absolute level reference, and keep clinging to the notion theres some validity to your position instead of accepting the info given to you.
- You are "not good enough at math" (by your own account) to follow the test I shared, but you are confident enough in your ability to completely ignore it and my requests for a valid measurement.
- You insist that pushing a sine wave into the unit at completely unknown levels is somehow a meaningful test.
- You think other builders can replicate your test without any reference to levels.
- You do not care (or know how to?) measure, if you are pushing 20dBU into the unit
- The unit should be able to take any level going into it (Rock n Roll)
- You can't read a schematic (No, theres no need to lift pin 4, since its not connected).

I think there is a discrepancy between what you know, which conclusions you have drawn, and what you aim to achieve thats hard to overcome. I am used to helping people who have no basic knowledge. It seems I dont know how to help people who lack basic understanding, but somehow still knows everything.

If you want to get serious about trouble shooting, I would ask you to bring the unit back to stock configuration, manufacture a steady sine wave at around 1-3KHz at a level of 0dBU (0.775V), or maybe even just measure the voltage on your output with aforementioned sine  wave, so we can at least get some idea of your converter levels. That way,  we can possibly rule out flawed testing and start looking for the build error.

If not, just keep going, as long as you are happy with what you get out of it.


Gustav
 
- i didnt ask how to reverse the bypass switch.  it wasnt a question, i knew how to do it.
(i wrote : "one more thing that i must change : the direction of the Bypass switch is the opposite of what i prefer. i think its always a good idea to leave extra pins for users to choose (there is room on the PCB) - but i know its your design, your decisions.")
by "extra pins" i meant pins for connectors. i think thats an elegant solution for user options.

and i think the unit is not true bypass 'by default' - "What a silly mistake!  I crossed the connections for the hard bypass on the board layout." (quoted from your Errata) when you say it IS, noting the Errata would be helpful, i think.

- i copy-pasted the dBFS / dBU values of my RME for reference.

- unknown levels? no reference to levels? i thought any of you can do that math easily (see above)

- pin 4 is not connected? what is R74 connected to then?  yes, i see the other two VCAs dont have pin4 connected . yes, i dont know exactly how VCAs / VECA works, but i think that question wasnt as stupid as you think. i cant read schematics.. well, sort of true. but if you remember our conversation about the MS76 / blend..  and for eg. i spotted a drawing error in a hairball schemo, corrected some MCI schemo..
some schematics are harder to read, especially if they have the labels and values overlaid ;)

and for a serious build guide, i wouldnt let "pifoFarad" printed in - especially after a stupid noob (me) did let you know about it..

Gustav said:
It seems I dont know how to help people who lack basic understanding, but somehow still knows everything.
Gustav

i hope you mean someone else, because i dont remember stating that i know everything..
i came to this forum to ask, learn and share my experience / ideas.

ill stop here - please forgive if i did any bad for your business, i wish you good luck and many happy customers in the future.

VECA rocks anyways.
and it may fit into a 1U - if anyone cares. (we do)





 
tata said:
i dont know exactly how VCAs / VECA works, but i think that question wasnt as stupid as you think.

Usually, when I am helping someone who is still developing the very basics, they work with me, not against me. When they encounter problems, they usually dont make the conclusions that they are ready to do circuit redesign.

I dont think asking the questions is stupid - not at all.

In your case, I took the time to check your issue. I would have included it in the project errata if there was one, but the extend of the problem you are having, in light of the measurements I made, makes it obvious you have either a build error or a measurement error. A simple test with controlled levels could rule out the measurement error. You wouldn't even have to know exactly what the reference level is, just be able to abandon the idea that 0dBFS in the DAW = 0dBU, and that that type of level is just what the cool kids send to their outboard.  Most people can adjust to something sensible without knowing the absolute level.

I am sorry if you feel attacked, but it is frustrating to deal with. You may not have noticed, but I did lay out a way to return to serious trouble shooting, and I would be happy to do so! Theres no telling if a measured approach would not reveal something I missed, but I can't add errata based on the random data you present, when a controlled experiment validates the design.

And confusion about levels is not uncommon. Theres a company that makes and sells a pre-assembled GssL, and in their video demo, the unit compressed with the treshold dialed out. It takes 12dBU to trigger compression under these conditions (too hot - not rock n roll).

Gustav
 
i wish id know where i stated that 0dBFS = 0dBU..
you took the time to do something different that i was asking for, only feeding / recording one side of a 2ch signal.

once you wrote : "This will vary wildly with stereo content, level and relative to M/S setting, and lastly, it may not be hitting the 1:2 ratio exactly, but something just beneath "  - 1:2 ratio?..

you dont think that asking the question is stupid, but you dont reply to pin4 being connected on one VCA  - and trying to tell me i cant read schematics in basic level.

i work against you? do you have problem with me being able to redraw schematics and making my own PCB designs form them at home? (just 1 sided PCBs - as i got no cool PCB factory at hand..)

you feel frustrated (and now i do feel attacked, yes), so dont waste your precious time on me anymore.
just focus on the happy customers and ones who suggest that i should populate all the filter caps in the HP/SC to do proper testing.. (w/ HP SC being set to OFF)


 
tata said:
i wish id know where i stated that 0dBFS = 0dBU..

You didn't, but you referred to dBFS levels into/out of the compressor, which is not helpfull. I do not feel you responded to this problem in a constructive manner (the rock n roll mantra in this thread).

I understand you feel attacked, and I apologise for venting frustration at first. If you feel like it at some point, I would still love to move forward with trouble shooting. It would be so simple to eliminate measurement error on your part, then take the next step forward. To me, re-designing is a few steps after verification.

Checking one channel first is also quite sensible. The side chain has been redesigning to vary the M/S content weighting, but if you push mono content into it in stereo, you are bound to encounter "aarhus mode" type behaviour. If this is actually whats going on, and you redesigning to counter-act, you may be surprised to see the result on program material containing actual, stereo content.

One of many theories we could investigate, but first - it would be good to make sure theres no measurement error.

I wish you the best.

Gustav
 
Hello forum!

Been reading and re-reading the schematics and these forum posts while waiting for the opportunity to order some kits. Found this little bit of info and just wanted to do a little bit of nitpicking. It's a general concept, not particular for this build or compressor design. Apologies if it's exceedingly obvious, too pedantic and so on. ;)

Majestic12 said:
The basic GSSL circuit is a real stereo compressor that cannot be "unlinked" to use both channels of it seperately. Both audio channels have their own VCAs that are responsible for the compression, but they are getting the exact same control voltage from the sidechain circuit.

In case of a stereo signal this totaly makes sense since you usually want to have the same amount of GR on both sides to not change the stereo image.

There is actually no way to avoid altering the stereo field in dynamic stereo processing. The exception is the rare cases of 100% mono or 100% stereo signals(dual mono), where of course, full link or unlink will ensure there is no change in stereo width. If the signal actually is stereo, with partial correlation between channels, processing will alter the stereo field.

Using fully linked dynamic processing on a partially correlated signal will increase the correlation between the two signals. In practical lingua, this means that a stereo signal will be somewhat pushed towards the center/mono. Of course depending on the amount of change. For the exact same reason that introducing an unlinked (totally uncorrelated) process will decrease correlation, pushing the signal towards the stereo sides.

The effect is somewhat subtle, yet obvious enough that I constantly tweak the amount of linking between channels when applying limiter to final masters to match the original signal. Or to push the signal in the desired direction, if a stereo field alteration is desired. Haven't applied that logic to compression though as I use no compressor with such a feature.

This got me thinking about building it as a dual compressor with link/unlink rotary fader. Even though it will probably only used for the first few percent of travel from linked towards unlinked, it may increase the utility and fidelity.

As I said: nitpicking.. Though I did have a big aha! moment when I discovered this aspect of audio processing years ago. So wanted to share in case someone wants to expand their knowledge base. And perhaps someone wants to implement the idea before I get around to it. Please report if you do! :)


Edit for clarity: this have nothing to do with the mid/side sidechain processing in the veca. That is, as far as I know, an unique feature for this compressor which is very clever and very attractive! :)


Cheers from 10+ year lurker,

Andreas Nordenstam
 
What are the SC Filter modes? M, S and B are Mid, Side and Both? What about L and 0?
I don't find this info anywhere...
 
dubful said:
What are the SC Filter modes? M, S and B are Mid, Side and Both? What about L and 0?
I don't find this info anywhere...

The schematic is on the site. If you check it, you'll see that switching the side chain filter clockwise just puts another capacitor in series. The board is set up for the use of 11 steps in total - I use less...

The side chain filter switch does not alter the M/S balance. It just raises the high pass frequency. The lettering was done on a panel prototype for personal fun and enjoyment, then kept.

L = low, M = mid, S = Speech, B = Brass.

The neighbouring MS Balance switch alters the Mid/Side balance.

Gustav
 
Just to follow up on this, I finally got my VECA front panel sorted out and - well, wanted to show off! I'd originally wanted to do something with a laser-engraved walnut veneer sheet, but that eventually got too complex. I'm actually really glad it did, as Frank at frontpanels.de has done a top job and I'm really happy with how it's turned out. Huge shout-out for his NRG enclosures as well, they're orders of magnitude more robust than the Modushop ones I'd been using (disclosure: I've only used the slimline Modushops, so the others may be tougher). The internal mounting rail is also really useful.

The VECA is the top unit in the photo, named "White Peak" after the Peak District in England (where I'm from). The bottom unit is Gustav's MS76, with various frills. The VECA is still front and centre of my 2-bus, super versatile and sounding great as a bus compressor - but I find myself using the 'Stall' mode (which I've labelled 'Level') for tracking as well, as it switches in time constants that seem to my ears to behave more like a classic levelling amp than an SSL.

All in all, it's a cracking piece of kit.
 

Attachments

  • TTA - White Peak & 1176.jpg
    TTA - White Peak & 1176.jpg
    668.7 KB · Views: 74
Back
Top