Just to clarify, is that with the THAT1200 as per BOM? As ever I am 18 months behind assembling things. The THAT spec sheet is a little confusing on what comes out single ended from the diff inputs. I would have expected 4V differential peak (so +/-2V on each line) to give 2.8Vpk against a 2.5Vpk to hit the rails.
Note on soldering. When I worked on satellites, the quality of the soldering from the space qual techs was astounding. I used to like to watch them working as the precision was something I will never be able to replicate.
All numbers I quote are as of the current rev at the time, assembled per BOM.
If you stab your voltmeter onto pins 2 and 3 of the THAT1200 (differential input) and measure 1.0 V RMS differential, you will have 1.0 V from pin 6 (output) to ground (pin 1) of the THAT1200.
Differential voltages are measured line-to-line.
Single-ended voltages are measured line-to-ground.
One could argue that the single-ended voltages should be measured line-to-reference. In audio circuits the reference is ground, thus, single-ended voltages are measured line-to-ground.
I think most of the confusion arises from the differential drivers that many manufacturers use. They simply take the single-ended output, invert it, and present OUT+ and OUT- on an XLR connector calling it a "differential" output. The resulting circuit has a gain of 2x (6 dB): Vout_diff = Vout_se - (-Vout_se) = 2 Vout_se.
I would argue that it really is a pseudo-differential output. A true differential output is perfectly symmetrical - or balanced. A true differential output allows you to ground one line and use the other for single-ended signalling - the inverter-based outputs do not allow this without severe performance degradation, if they allow it at all. That's why THAT Corp. came up with the THAT1646, that I use in my THAT Driver.
Another point of confusion are the following synonyms:
- Differential = Balanced
- Single-Ended = Unbalanced
Tom
Which is what I thought, but when you said
I had a senior moment and thought that 0dBFS wouldn't clip the mod. I misread and got very confused
. Would a 1206 change anything in your opinion?
I had a senior moment and thought that 0dBFS wouldn't clip the mod. I misread and got very confused
Let's do the math.
THAT120x family portrait:
1200: 0 dB attenuation, -106 dBu (20 kHz BW) output noise
1203: 3 dB attenuation, -105 dBu (20 kHz BW) output noise
1206: 6 dB attenuation, -107 dBu (20 kHz BW) output noise
So going from a 1200 to a 1203 would result in a 4 dB degradation of SNR (noise is 1 dB higher, output signal 3 dB lower).
Going from a 1200 to a 1206 results in a 5 dB degradation in SNR (noise is 1 dB lower, output signal 6 dB lower).
If your DAC has a 4 V output, using the THAT1206 in the front end of the MOD86 will result in 6 dB higher SNR coming out of the DAC as the DAC's output signal can now be 6 dB higher without causing the MOD86 to clip. Basically you get an extra bit of information there. This should lead to a 6-5 = 1 dB better SNR. Not a huge difference. Probably not enough to notice in practice, but the THAT1200 and 1206 cost the same, so 1 dB of SNR improvement for free seems worth taking.
Tom
Over the last few days i have finished up my build. All four channels passed the first power-up test, with regulated voltages between 16.0 - 16.4 (+/-). Assembly of the chassis was completed yesterday. Today I prepped the XLR input cables, hooked them up to the Phoenix connectors of the Mini DSP 4x10, hooked everything up and threw the switch.
Silence. No hum, no hiss, but also no music. Nothing. The MiniDSP monitors show output to all four of the channels I'm using, and the sub outputs are getting a signal. I don't own a scope to confirm that there is signal to the input connectors on the board. I also don't have another balanced source to hook up to the amp.
The power supply sits at +/- 30.2 V. The fuses aren't blown.
Reading back through the manual, I see that I'm over the range for the stock value of R9, and should really have 47.5k instead of 33.2k. Would that prevent the amps from un-muting? Or would it just change the time to un-mute?
Where else do I look?
Bill
Silence. No hum, no hiss, but also no music. Nothing. The MiniDSP monitors show output to all four of the channels I'm using, and the sub outputs are getting a signal. I don't own a scope to confirm that there is signal to the input connectors on the board. I also don't have another balanced source to hook up to the amp.
The power supply sits at +/- 30.2 V. The fuses aren't blown.
Reading back through the manual, I see that I'm over the range for the stock value of R9, and should really have 47.5k instead of 33.2k. Would that prevent the amps from un-muting? Or would it just change the time to un-mute?
Where else do I look?
Bill
Attachments
![IMG_20160522_125338211[1].jpg](/community/data/attachments/506/506313-f0d156802b099710cb28ecbbd600ed8a.jpg)
![IMG_20160522_125555244[1].jpg](/community/data/attachments/506/506338-c1df0b8d796e762f94fc17959f2099c7.jpg)
I think you've the Vpk and Vrms conversion factors swapped. Since it's a unity gain part 4Vpp differential gives 2Vpk on the output of a THAT 1200.
Eh? If I'm understanding Bill's design properly the goal's to have 0dBFS on the DAC result in a Mod at the onset of clipping on ±28. With the standard power stage doesn't that mean the maximum output of the THAT needs to be 1.77Vrms = 2.5Vpk = 5Vpp? This suggests another 1.9dB of gain is needed to hit target with a 1200.
If the DAC swung to 10Vpp then the 1206 looks bang on. But maybe Bill meant 4Vpk differential rather than 4Vpp differential?
Last edited:
Sort of, but I misread what Tom was saying. Whilst it would be nice to have things setup so 0dBFS is 0.5dB off clipping I'll settle for being in the ball park. Lets face it we are normally 10-20dB down on 'full stadium' levels but like to be able to once in a while play at realistic levels
My misread made me think Tom was claiming that a 4V RMS input wouldn't clip. Once I re-read, my silly mistake was obvious. I currently have the 1200 in, so would need to set the miniDSP to -7dB output attenuation when flat out. If one believes Benchmark this is no bad thing.
But mistake pointed out, maths making sense all is good. Now I can just worry about whether an ideal DSP X-over should re-dither. (joke)
My misread made me think Tom was claiming that a 4V RMS input wouldn't clip. Once I re-read, my silly mistake was obvious. I currently have the 1200 in, so would need to set the miniDSP to -7dB output attenuation when flat out. If one believes Benchmark this is no bad thing.
But mistake pointed out, maths making sense all is good
. Now I can just worry about whether an ideal DSP X-over should re-dither. (joke)As long as you're good. :thumbsup: The 10-20dB gap can be optimized with only normal levels of precision analogue difficulty, pain, cleverness, and suffering. But I'm not exactly seeing a lot of demand for doing so.
Only in 16 bit maths.
Missed that thread, but that is the solution I would like. Jan Didden did a specific implementation for the DCX crossover and it seems to tick a load of boxes. But with constrained funds not one I could afford at the time, so going slightly less optimal for the time being. If I were better at coding would be bending the 3318 to my will right now. Of course my analog skills are a good 15dB off what are required to match the power amp...
Only in 16 bit maths.
But one can easily get brain mangled over thinking. If you have a dithered 16 bit source file which gets padded to 24 for processing in the DSP, SRC'd and then munged around a bit, what is the result? But then we are in the realms of what I call a 'three beer discussion'.
I missed the "peak", hence assumed RMS voltages.
Next time i have a layover at Heathrow I'll take you up on that.
Tom
Next time i have a layover at Heathrow I'll take you up on that.
Tom
Sure. Heathrow is either 30 mins or 24 hours from me depending on what the M4 is up to
M4 crash: Driver killed as lorry overturns - BBC News
This is happening a lot these days. Closed the road for 24 hours just near work.
This is happening a lot these days. Closed the road for 24 hours just near work.
That's very polite of you Tom, clearly Canada is rubbing off on you
Yes, it was me being daft. Happens a lot! But has helped me think a bit more about things I need to check on the DSP side. Wife and sproglet off on thursday for 2 weeks so this is my last chance for a year or so to finally get things sorted and these amplifiers completed!
Yes, it was me being daft. Happens a lot! But has helped me think a bit more about things I need to check on the DSP side. Wife and sproglet off on thursday for 2 weeks so this is my last chance for a year or so to finally get things sorted and these amplifiers completed!
