D-AOC PCBs - the building and help Thread

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
One other good thing about this project - I've started writing a test tool which generates the transfer function graph for a compressor/limiter function.

Input dBu vs Output dBu with amplitude sweep  with some knobs for adjusting min/max input levels, sweep time and profile. That sort of thing.  Maybe even some freq domain stuff if I get really motivated.

Doing it as a Labview application using windows sound card interface and hoping to target the Win7 runtime environment so it can run without any special whatchamcallits.

I'll see how it goes :)
 
Hank Dussen said:
Lundahl LL1540 wired as 1:2 on the input
Lundahl LL1540 wired reversed as 2:1 on the output
Most of the kubicki mods. Second vactrol for soft setting, blue led for hard setting
I have a 500 ohm trimmer in series with each of the 4 vactrols. I really needed these to optimize the compression in stereo mode.
Replaced R1 (2 watt 470R) by 2 parallel 25 watt 1K resistors! Just to make sure it doesn't overheat :)
Front panel made by Frank from nrgrecordings.de
Meters are old AKAI meters I bought from a member (Sahib?) here a while ago. (painted black)
assembled stepped switches from eBay

Apparently there were some more questions...
The front panel is made with Front Design. The font is DIN17, but wit the X-cale set to 1,5.
The knobs are these ones from Farnell: http://be.farnell.com/mentor/476-61/knob-high-torque/dp/1282528?Ntt=1282528
(But, the top inlay is grey, which is really ugly. I sprayed them black.)
 
If I wanted to go transformerless on the inputs, could I leave everything the same and just go straight to the xlr's? Or would I have to change some capacitor values?
 
Continuing my efforts with this most interesting compressor unit ...

I subbed in a genween LM350T in place of ebay fake and voila, all good wrt heaters and such.
Changed a couple of dropping resistors in the low voltage circuit with some higher power ones too.

After some quick checks, I confirmed my previous performance figures - unity gain noise floor at typical setting gives me a noise floor of around -74dBu (an averaged figure as reported by my RTAS 20Hz to 50KHz) vs a loopback noise floor of -80dBu for my Motu 828MkII test setup.

50Hz component < -100dBu, 100Hz component < -85dBu

The Motu is calibrated to the limit of my capability for the dBu standard with 600R load and absolute voltages. I use a combination of my CRO, DMM and some high quality iso transformers for this task.

Freq response is -1dB at 20Hz and -1.8 dB at 24KHz and very flat.  Square waves are good and clean. Distortion is quite low.

Variation fro chnA to chnB is around 0.3dB maximum across the available makeup gain range.

All in all the no Gain Reduction performance is very nice indeed.


-----------

So, I turned my attention to the sidechain. First I measured the dc levels and then the ac levels of the various bits of the sidechain amp.

Very nice - both channels virtually identical in ac performance. In dc, currents were less than a mA or so different across the channels with 2V or so max difference in hv.

All very good.  I noted the gain stage of the sidechain has a bias of 3.75Vdc and a current of 5mA more or less. Gain was x8.2

The cathode of the sidechain cathode follower sits at 140Vdc or so and a current of 6.3mA or so. Gain was 0.92

Total sidechain current is around 11.3mA. Overall gain was around x7.5

So I would expect around 7Vpp in and 56Vpp out of the gain stage without clipping and a cf swing of 165Vpp or so max to feed the dc coupled vactrol.

I measured  something like 86Vpp out of the gain stage before visible clipping/rounding. The CF is of course capable of more than that.

------

Having established the sidechain dc and ac were working well and virtually identical across the channels with no Gain Reduction occuring, I set about to look at the vactrol circuit including the mods I started up with - basically a 4 position switch with an on/off inline with it, the whole thing across the single vactrol + and - terminals.

When the on/off switch is open, the 'response' circuit I did is not in-circuit. There is virtually no GR because the stock vactrol arrangement requires a ton of level to get it to squeeze.

When the on/off switch is closed, the 'response' circuit is in parallel ith the vactrol and has an effect.

I used 'soft' (100K resistor), 'hard1', (parallel blue led), 'hard 2' (1K resistor), 'hard3' (anti parallel red led).

The positions all seemed to work and all seems to be fine. Now to look closer at the curves.


----------------

For my tests, I decided on an input maximum which result in the sidechain signal at it's more or less unclipped  max - 86Vdc.  I reason there is no need to run a distorted sidechain before the vactrol.

I have a useful range on my motu RTAS of 12dBu max output, so I set that as my maximum level into the doac. I then adjusted the input knob to give max unclipped sidechain (position '2' of 10,  GR set to 'off'), leaving the 'compression' at max (I will vary the RTAS output level).

I then set the Level knob for +3 VU on the meter, which is a reasonable 'max power' level :)

----

So now, to start at +12dBu at the RTAS output and vary down, noting the gains and GR and so on as I go.

I'm writing a tool to do this, so I wanted to play about with some methods and see what makes sense. I mean, performance of a compressor is very dependant on operating levels !

-----

So the fun begins. I should say this is what I measure - I try to understand it as I go and later after I've thought about it all for a while, I usually see the flaws with my methods  :)

First thing I look at is the waveshapes and amplitudes at the vactrol + terminal.


'Off'                                  shows a clean sine wave, huge amplitude (max before clip)
'Soft'    (100K resistor)    shows a clean sine wave, quite large amplitude
'Hard1'  (blue parallel led)  shows a clean sine wave, even larger amplitude
'Hard2'  (1K resistor)        shows a positive peak clipped sine wave, quite small amplitude
'Hard 3' (red antipara led)  shows a positive and negative clipped sine wave, quite small amplitude

All of them showed strong compression availability. The on/off completely removed any GR.

--------

Some of the questions I have so far are :

- why would a really large sine wave vactrol signal result in no GR?  (in the 'off' setting)
- why would a large sine wave give less GR than the smaller, clipped waves ('soft' vs 'hard')

basically trying to understand what waveshapes give rise to what GR?

The vactrol resistance is a shunt to ground, so low resistance means max GR. High resistance means min GR.

Vactrol resistance is low when input current is high and vactrol resistance is high when input current is low.

So more current = more GR.  (I assume more current = more light)

I suppose the situation is this - the sidechain signal is applied to the vactrol led - the amount of light and hence the GR produced, is proportional to the rms current of the waveform.

Then there's also dc voltages at the vactrol - I measured some there in all cases. I need to look at this more closely. Diodes rectify and thereby produce dc levels and so on. But still not what I expected :)

At that point I realise things aren't really what I think they are and start measuring in more detail.

:)


 
Scrappersa said:
If I wanted to go transformerless on the inputs, could I leave everything the same and just go straight to the xlr's? Or would I have to change some capacitor values?

First thing is you would need to ensure you don't have dc at the input to the doac - a transformer coupled version would not as the transformer does not pass dc.

Next thing would be the levels - the stock doac uses 1:1 transformer, so that's the starting point.
I use 1:4 step up transformers and think that is a good place to be, so maybe you would think about providing some gain at your transformerless inputs.

That's basically it, apart from input impedances. The stock unit uses 10K:10K transformer, so that's a good place to start. Make sure your driving device is OK with a 10K load (which most things are).

Cheers
 
So it seems that with respect to the vactrol + terminal waveform :

- wave form has a negative dc offset  which gets more negative as the waveform becomes larger
- wave form positive going peak maxes out at 2V above 0Vdc  - the Vf of the vactol led
  (spec sheet says 1.65V typ, 2.0V @20mA)

  ie. waveform grows downwards from +Vf line as the sidechain signal amplitude increases

- smaller waveform amplitude means more GR (smaller negative dc offset)

- resistors in parallel reduce the amplitude of the waveform (and -ve dc bias)

- led added antiparallel (lights up) does clamping at the -Vf of the added led
  This reduces the amplitude of the negative going peak by clamping
  (and therefore reduces dc offset voltage)
  thereby dramatically increasing GR

- led added parallel (no light up) doesn't clamp
  increases amplitude as compared to the modes with resistance
  (and therefore increases dc offset)
  thereby dramatically decreasing GR

- led added parallel (no light up) waveform amplitude is still smaller than the stock vactrol case
  therefore still some GR but much less than antiparallel or resistance cases

- correlation between amplitudes of the sidechain signal and GR achieved is good
  ie. sidechain reduces 10dB in amplitude, GR measured at RTAS is 10dB
      so it seems to be scaled in a useful way

Still somewhat confused - small amplitude clipped waveforms give high GR.
Large amplitude sine waves give least GR but will do so when they are very large.

Kind of like average energy in the waveform - more clipped is more energy is more GR at low amplitudes. Less clipped is lower energy in the waveform, needing more amplitude for GR.
When the clean  sinewaves start to get big enough, they begin to clip and the GR quickly increases.

I guess the lower voltage waveforms being able to produce large GR means the currents are high but the load is low - hence small voltage waveforms but large current waveforms.

I wish I could monitor the current thru the vactrol directly - but I don't want to hack things up :)

Also, I note the parallel led case has pooped a couple of leds.
I think I may not use this mode. Doesn't seem too useful over the cases with resistors in parallel.

Instead, I'm now using 100K, white led antiparallel, 1K, red led antiparallel.

Also, I see that an antiparallel vactrol is the same as an antiparallel led in general.

Cheers
 
On the subject of calibration and left-right matching, some really great results.

As I said, the ac and dc matching of the sidechains was pretty good without any special efforts.

Measuring the GR, input and output levels across the different modes and at different GRs, inputs and outputs shows everything to be very, very close and as expected.

I can say the left-right performance is the best I've ever seen in my diy compressors by a long shot.
This is using bog standard pots and so on.

I measure around 0.3dB max deviation across channels in output level at the RTAS in any case.
And that is mostly a mismatch in the signal amp, not the side chain. The side chain signals were coming out near identical.

And the meters are also lining up really, really well.

So this unit for sure could be lined up for repeatability in a mastering context.

If I can improve the noise floor by 3dB or so I would say this would be my mastering compressor ;)

But having said that, I note the measured noise floor does not depend on output level, so the actual SNR is quite good as you can apply max gain make up without incurring a noise penalty.

----------

Not sure why but there it is. I have seen it like this a few times on other units.

The other thing regarding the noise floor is that it is absolutely rock solid - no 'twitchiness' or oscillations at all.
Which is all very good. I think the little extra noise I see at 100Hz component is due to my long wires for the seperate  VU meters.

Even so, the 50Hz component is vanishingly small and the 100Hz component very small. I would characterise the noise performance as very good. Certainly no hint of any audible or even generally measurable noise.

----

Well, now it is to the studio for listening and using for a while, until I get my measurement application running :)
 
Thanks for that Volker.

I've been surely impressed with performance as well as challenged in understanding the operation of this very nice unit.

ps - Could I have my pic added to your list ? (there's one back a few posts)  I'd very much like to  ;D
 
Really nice indeed!  8)
What did you use on the output? At least if you used the choice of tranny's on the input.
I have mine running with lundahls and I absolutely still LOVE this compressor! Really smooth and warm sound without really hearing what is going on.
 
As on the front panel you can choose and combine all the irons, in in as on out: Lundahl, Haufe and Carnhill.
 
I am not a genius....so inspiring is the only way to do a project like this.... You have done al before me on the forum, so nothing news to balance the channel as a trimmer in series to the vactron, just mounted on a separate protoboard and not on the main pcb. At the moment no specific measurement, just my ears to say, smooth, nice low end and absolutely transparent also if pushed hard. A nice glue effect. Thanks to all for the....inspiration!
 
Back
Top