Transformerless Vari-Mu Compressor

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
REW has not been cooperating with me lately so I can't do distortion sweeps. Here are couple measurements I did with ARTA.

The soundcard I'm using for measurements isn't the greatest for distortion measurements because the soundcard output has some high frequency noise. Most notably the spike at little over 20kHz.

0dBFS = +20dBu
 

Attachments

  • distortion-nocompression.png
    distortion-nocompression.png
    222.9 KB · Views: 44
Here are some distortion measurements with 7dB of gain reduction. In the top picture release was set at slowest. In the middle picture release was at the fastest position. In the bottom one release was at slowest position.

Steady wave distortion measurements with gain reduction aren't really useful. If I want to make them look good I just have to use slow release and higher the frequency the better.

0dBFS = +20dBu
 

Attachments

  • distortion-7dB-compression.png
    distortion-7dB-compression.png
    216.6 KB · Views: 42
thanks!


i have noticed on my high quality DAC that anything loader than -9dbfs on the output will show additional distortions  even when it clips at 0dbfs or even higher. so i have to run tests with a cooler signal.

 
Just listening to the YT video and there are clicks when it goes from -3b to -3db high ratio and from -3db to -5db high ratio....Is this an editing thing??

Also it seems like things get darker (or is it maybe just softer bc more compression..?) when going from -3b high ratio to -5db high ratio.... Granted I've only listened for a minute on headphones but these things kinda jumped out right away... ??
 
In case somebody missed it scott2000 is talking about this video

The audible clicks in the video are because of the editing. I recorded the whole song going trough the compressor several times at different amounts of compression, matched the levels and cut them together. I used video editing software to cut the audio clips together and it didn't allow me to be precise enough to find a zero crossing where to cut the audio.

I don't perceive it going darker when it changes to 5dB gain reduction but that's just me. I have measured that the frequency response does not change when compressing. Maybe somebody else can comment if it sounds darker to them too.
 
I don't perceive it going darker when it changes to 5dB gain reduction but that's just me. I have measured that the frequency response does not change when compressing. Maybe somebody else can comment if it sounds darker to them too.
Thanks...I think it was just the transition point making it appear that way ...
 
Here are some distortion measurements with 7dB of gain reduction. In the top picture release was set at slowest. In the middle picture release was at the fastest position. In the bottom one release was at slowest position.

Steady wave distortion measurements with gain reduction aren't really useful. If I want to make them look good I just have to use slow release and higher the frequency the better.

0dBFS = +20dBu

What! So 0dbU is -20dBFS which means the noise floor of your tube device is -110dBu? I didn't think tubes were that good. Nice!
 
Here on the bottom picture noise RMS = 98.3 dBFS which is -78.3 dBu or 94 microvolts. Sample rate in the measurement is 48 kHz. The noise clearly starts dropping close to 24 kHz, so lets say the bandwidth is 22 kHz. Gain is at max (19 dB) in the measurement.

INA2137 datasheet tells us that the INA2137 output noise is -106 dBu. Because of the balanced connection I have, it should be -103 dBu. If I were to have just INA2137 and 19 dB of noiseless gain, the noise would be -84 dBu. Everything after the first INA2137 stage adds about 6 dB noise. I'm ignoring the fact that INA2137 datasheet gives the noise at 20 kHz bandwidth.

-110 dBu noise floor would be impossible because even the first amplifier stage has more noise.
 
What! So 0dbU is -20dBFS which means the noise floor of your tube device is -110dBu? I didn't think tubes were that good. Nice!
FT dispalys noise voltage density. It's the energy by bin. In order to determine the global figure, you need to multiply the density by the square root of the number of bins.
In dB it's = 10*log(M/2).
I surmise the ca. 30dB difference between density and WB figure comes from a 1024 points FFT. Or maybe more.
See paragraph 3.4 in the linked doc.
https://kluedo.ub.uni-kl.de/frontdoor/deliver/index/docId/4293/file/exact_fft_measurements.pdf
 
Last edited:
I might not be posting here on GroupDiy anymore. See: I am suspicious

If anybody wants to contac me my email is [email protected] <—— (just like it is spelled there)

If anybody wants to build one of these compressors or has any questions please contact me. I have some boards left that I can sell at cost and provide support for building the compressor if anybody is interested.
 
Heikki, It is exactly because of folks like you and Ian who have posted original designs here that I'm concerned too. Most of my posts don't mean a hill o' beans so it's not much skin lost for me.

However, maybe let's see what Ethan and Angie etc. say about those terms. I'd hate to see you go.
 
-- I thought I post here rather than directly in the build help thread. --

I had another go at the D1 zener, the one that didn't make it into the first BOM and which sits after Q4 (BC556).

I reinstalled that zener but put it on a pot (series). 2M is maybe too large or needs log pot(?), but this makes it adjustable. Not sure what to call it though. Soft to hard knee ?

Also tried it after the timing (before R5 1K), but I think it does something different there -- possibly messing things up.

Didn't dare hooking it between Q2 and Q1, although that would isolate it from the timing. Not sure about this latter. Safe a try ?
 
Last edited:
The reason D1 was originally there was to limit the amount of compression when 12AU7 tubes were used. With 12AU7 tubes the distortion would become clearly audible with more than 12 dB of compression. It could still be used to limit the amount of compression or like you did by adding resistance in series with it to shape the knee/ratio at the other end. I think something like 10k might be more useful value to try for the pot.

I'm not sure how useful it is as an effect to have the zener there with resistor/pot in series. When the zener starts conducting the ratio starts to decrease and if you have no resistance in series it gets close to 1:1 fairly fast. The same effect does happen does happen in some compressors when the sidechain starts running out of juice and can't produce anymore control voltage.

I wouldn't put the zener between Q1 and Q2 since there's only the 22 ohms (R2) limiting the current.

sidechain.svg
 
Last edited:
Thanks.

Will try different pot for knee/ratio. For me, with pot/switch it might be a keeper. Faceplate is still dummy.

The 'sidechain running out of juice' effect is what I used on another compressor on purpose.

So my intuition was right to not hook it between the Darlington pair. Good to know.
 

Latest posts

Back
Top