Another diode bridge compressor

[abbey road d enfer]

Would 1N4448 offer any advantage here ?

This goes along with the question I asked in the other thread, where member jensenmann recommended 1N4153, on account it was less noisy.
What are the factors that determine noise in a diode bridge compressor?
What are the differences between low-power signal diodes? Actually not much that could explain noise differences. Here we talk about silicon (excluding germanium) non-Schottky non-zener diodes. They may have different reverse breakdown voltage (always much higher than the voltages in a bridge compressor), different capacitance (negligible in any case), but for the rest, they all have the same basic I/V curve and the same junction potential of about 0.7V. They also may have a different series resistance, but it is so low (typically less than 1 ohm) that it would make a difference only at very high attenuation, and have zero consequence on noise.
The parameters that can influence noise is the I/V curve. It is supposed to be the same perfect exponential for all diodes. However, due to different processes, the slope of the I/V curve can be more or less steep. This difference is known as the ideality factor.
Noise in a diode attenuator is a compromise between noise and distortion.
It would make sense that the "flattest" I/V curves resulted in the capacity of operating at higher audio level, thus reducing relative noise.
Unfortunately, AFAIK, datasheets do not specify the ideality factor. It has to be deduced from graphs.
So far, comparing the datasheets of 1N4153 and 1N4148, I can't detect a significant difference.
I think experimentation is the key.

 

[Script]

Yes, I read the other thread. I wasn't thinking sound (punch etc) or noise floor, but matching (less thump) and whether the 4448 were more tightly spec'ed in the right way for that.

 

[Script]

@Abbey

Thanks for the edit. It looks like more information than before or -- most likely -- more than I could take in earlier today.

---
If the idealiter factor between same-type diodes played a role, could it be somewhat averaged ('flattened'?) theoretically by parallel connection of two diodes followed by a very small resistor ? I do see the fields of diodes...

Whereas the 'about' in junction potential would average somewhat in series connection (but not if resistors in ladder)?

-- Please ignore if nonsense.

 

[abbey road d enfer]

If the idealiter factor between same-type diodes played a role, could it be somewhat averaged ('flattened'?) theoretically by parallel connection of two diodes followed by a very small resistor ?

Considering the quite low DC control current, the resistors would need to be large enough to skew the control law. I'm not sure it is acceptable.

Whereas the 'about' in junction potential would average somewhat in series connection (but not if resistors in ladder)?

certainly the series connection would be a better way of averaging, with the added benefit of increasing the audio voltage.

 

[zamproject]

Hi all

Very interesting topic, I have a bridge compressor prototype on the bench since months (somehow in standby for now) using nice Girardin line amp (trafo I/O) and a third spare 1:1 trafo for interstage (of the same manufacturer) with basically 2254/33609 concept for bridge and sidechain

certainly the series connection would be a better way of averaging, with the added benefit of increasing the audio voltage.

Now your design proposition seem excellent, I wonder why nobody at neve or whatever other company doing DB compressor never think about it in the past decades ??? or I'm missing something ?
I'll try that when I'll put again my hands and brain on this prototype, being able to work audio in the V range rather than few mV seem the thing to do.

Best
Zam

 

[abbey road d enfer]

Now your design proposition seem excellent, I wonder why nobody at neve or whatever other company doing DB compressor never think about it in the past decades ???

Maybe I'm a genius?
I think I have seen somewhere diode bridge attenuator that had 2 diodes in each branch (8 diodes in total); it's in a fold in my memory.

or I'm missing something ?

Maybe the Neve engineers have considered that and concluded it was not worth it...
I wouldn't be surprized if they had the gall to steal the ideas I stole from someone else...

 

[zamproject]

Maybe I'm a genius?

 

[Heikki]

I wonder why nobody at neve or whatever other company doing DB compressor never think about it in the past decades ???

Earliest use of germanium diodes as "variolosser" in a compressor I have found is from early 50's by Bell Labs. In the 60's came the compressors with silicon diodes and there was probably no need to have string of diodes in series because performance was already more than adequate for broadcast use. Technology moved on and all the smart thinking with it. It is rare to have any real engineers like abbey road d enfer to spend any time thinking about how to improve old circuits like these.

Maybe the Neve engineers have considered that and concluded it was not worth it...

Rupert Neve Designs 5254 seems to have about 10 dB less noise than the AMS neve compressor. It could be that they used string of series diodes. AMS Neve noise "−55dBu (compress in, gain 20dB)", Rupert Neve 5254 noise "@ +20 dB Make-up gain: -64.5 dBu typical"

My noise is -68 dBu at 11 dB gain which should make it -59 dBu at 20 dB gain. Taking your advice and making it 8 diode bridge should bring it in par with the 5254. Assuming of course that these figures are comparable. The Neve manuals don't tell if any weighting filters are used in the noise measurements.

 

[richiyobs]

Hi folks,
In my poor experience I only saw diode in series of 8 into tape recorder like the attached example... This usage was to achieve the desired clipping character, according to what I read... Was searching for diode config and found this about clipper and limiter
Also found a German diode from 50s and if I well remember, series of diode too
Cheers

 

[abbey road d enfer]

In my poor experience I only saw diode in series of 8 into tape recorder like the attached example... This usage was to achieve the desired clipping character, according to what I read... Was searching for diode config and found this about clipper and limiter

This is a soft clipper, which exploits a very different characteristics of the I/V curve. Increasing the number of diodes increases linearly the threshold of soft clipping.
The legend is misleading, since it is in no way a compressor.
It could be argued that limiter is an adequate word for it, but there is a well agreed definition of an audio compressor or limiter.
Se&amantics, yes, but words matter in science.

 

[dogears]

thanks for the good input as usual abbey. how does the working level of the signal to be attenuated relate to the additional series diodes in the bridge? meaning with an eight diode bridge, you'd expect the required attenuation to be less, but how much? 6dB?

playing around in the simulator the effect of the control voltage changes as well, ratio and threshold.

i lack the analytical chops to be able to math this out and have to use simulator brute force, which is a bit frustrating for me.

 

[richiyobs]

Hi Abbey road thanks for clarifying

 

[abbey road d enfer]

thanks for the good input as usual abbey. how does the working level of the signal to be attenuated relate to the additional series diodes in the bridge? meaning with an eight diode bridge, you'd expect the required attenuation to be less, but how much? 6dB?

Yes, 6dB for each doubling of the number of diodes.

playing around in the simulator the effect of the control voltage changes as well, ratio and threshold.

Remember that it's actually a control current.
When you double the number of diodes, you want to keep the same control current, which will result in twice the audio voltage.

 

[dogears]

That's the same result I empirically derived, but that makes sense.

 

[wlinart]

Little question, the other project based on the same sidechain doesn't use a transformer, this one does. Is a transformer necessary?

 

[abbey road d enfer]

Little question, the other project based on the same sidechain doesn't use a transformer, this one does.

Does it? I don't see one.

Is a transformer necessary?

Something is necessary to eliminate the common-mode thumps. It used to be transformers, but here Heikki makes good use of the differential receivers.

 

[wlinart]

Does it? I don't see one.

Something is necessary to eliminate the common-mode thumps. It used to be transformers, but here Heikki makes good use of the differential receivers.

In the sidechain, there's a hammond 107n on the schematic and on the first picture there's an edcor transformer visible. That's why i was curious

 

[abbey road d enfer]

In the sidechain, there's a hammond 107n on the schematic and on the first picture there's an edcor transformer visible. That's why i was curious

OK, so which one does not have a xfmr in the side-chain? Heikki clearly stated that both his diode-bridge and vari-mu had the same side-chain.

 

[Heikki]

OK, so which one does not have a xfmr in the side-chain? Heikki clearly stated that both his diode-bridge and vari-mu had the same side-chain.

I think he is talking about this compressor by richiyobs.

Little question, the other project based on the same sidechain doesn't use a transformer, this one does. Is a transformer necessary?

It is not necessary to use transformer on the sidechain. The transformer is there to get some voltage gain to get the threshold to a good level and 660 ohm source impedance and transformer inductance make a high pass filter for the sidechain. All this could be done cheaper with op amps.

 

[wlinart]

I think he is talking about this compressor by richiyobs.



It is not necessary to use transformer on the sidechain. The transformer is there to get some voltage gain to get the threshold to a good level and 660 ohm source impedance and transformer inductance make a high pass filter for the sidechain. All this could be done cheaper with op amps.

Indeed, sorry for the confusion.
Thanks for the answer!