I'm trying to simulate a THAT4301 circuit in Ltspice but the quiescent point of the RMC out is ~-300mV and it's not entirely clear to me why.
Here is the circuit:
This is the 2:1 compressor circuit from fig 17 of the THAT4305 datasheet.
And here is the result:
Green: RMS out voltage
Blue: white noise burst stimulus input
Red: output
You can see the RMS out is ~-300mv. Should this not be ~0 so that the quiescent gain is unity?
Note that my stimulus is huge at around 1.7 Vpp. Maybe I'm just overloading this thing?
Mike
Here is the circuit:
This is the 2:1 compressor circuit from fig 17 of the THAT4305 datasheet.
And here is the result:
Green: RMS out voltage
Blue: white noise burst stimulus input
Red: output
You can see the RMS out is ~-300mv. Should this not be ~0 so that the quiescent gain is unity?
Note that my stimulus is huge at around 1.7 Vpp. Maybe I'm just overloading this thing?
Mike
What's the voltage at the input of rms detector? Your showing current. It might take some time for c1 to reach steady state.
The IN of the RMS circuit is -16.82 microvolts which is pretty much 0 as far as this circuit is concerned.
I let the circuit run for over 5 seconds. Consider the following:
Green: RMS circuit output voltage
Red: stimulus input
Although it looks suspicious with the RMS out jumping around, it's actually not because upon close inspection of the RMS input there are actually tiny glitches in the stimulus.
Part of the problem may be that the wav file stimulus was generated on a computer so most of the samples leading up to the actual burst of white noise are absolutely 0 with the exception of a few places where the sample values have glitches presumably from inserting "silence" and doing other transformations in the sound file editor software. Meaning most of the samples are 0 and so the RMS output is putting out a theoretical low value that in practice not possible.
Unfortunately I have not actually bread boarded this circuit and I have never even studied an actually physical VCA circuit before so I'm not crystal clear about how it should behave. It could be that it's actually working mostly as expected. The circuit in the OP is the encoder of the Hi-Fi compander so it has no threashold / ideal diode circuitry. It is supposed to scale the gain over the entire dynamic range of input.
However, I have tried to simulate other circuits like the THAT4301 Typical Application circuit and the ideal diode circuit / threshold control only exacerbates the behavior. The RMS output of that circuit just before the white noise burst is -640 mV. This translates to -1.41 uA being drawn from the Ec- control port.
Hi,
This thread has been inactive for some time, but maybe someone's watching.
Has the issue been resolved?
Is there a good model that can be used that will work as it should?
This thread has been inactive for some time, but maybe someone's watching.
Has the issue been resolved?
Is there a good model that can be used that will work as it should?
The model is available from THAT's website. Although you will need to create symbols and bind the symbols with the different devices in the lib. Unfortunately that process is not documented terribly well.
Otherwise, I believe my last post is probably correct and the model is actually working. But I still have not actually tried the THAT4301 so I cannot definitively confirm.
But for the model to provide meaningful results you will need to use a suitable stimulus so that you can see how the circuit reacts to transients. What I would do is create a wav file that consists of some relatively low white noise (not 0 which is why my results looked strange) with a burst of much louder noise followed by low white noise. More specifically, download something like Wavepad, create 50ms of white noise at maximum level. Then attenuate the leading 10ms and trailing 20ms by -60dB. So you have 10ms quiet, 20ms loud and then 20ms quiet.
Also, note that you probably cannot breadboard a THAT4301. Some of the connections are very sensitive to how they're layed out. The only way to really understand how the VCA works is to just have a board made. But you must read the datasheet carefully about layout or it will simply not work correctly or at the very least it will have significant noise or distortion. There is a demo board. I don't know if it's for sale anywhere but you could at least look at the datasheet for the demo board to assist with understanding the layout issues.
Last edited:
Yes, and I already had that, but I thought there was some problem with the model, and perhaps someone had corrected it.
Maybe it's been done already and someone could share that.
We'll find out soon, as I will be designing something using it.
Intriguing, but how do you use that in ltspice sims?
I never heard of that. Where do you get it? I bet it's one more of those windoze only thing...
Wow! That's kinda scary!
So how much can we rely on simulations?
Probably not much by itself, because then how it is implemented on a pcb can prevent proper function. That's a concern.
You can "dead bug" the THAT4301 -- works very well -- provided you have the DIP version.
Place the chip upside down on a piece of PCB material. Now -- this is important -- with a sharpie mark the pin assignments on the copper! Remember it's flopped. Tack pin 9 to the copper. Then proceed as the recipe calls for.
- Status
- Not open for further replies.