here is a picture from link.
What I can see from the board is very small switching 1.2v regulator at capacitor rails, 3.3v LDO at clock, +-5v LDOs. Please let me know if I miss..
remove those four LDOs and supply each voltages via J2 pins, will it be okay?
I don't think the +/-5V regulators are LDO. Just garden variety 7805/7905?
I'd strongly recommend simplifying stock by dropping the .02, and just keeping the top and bottom models. People will want either a budget entry or they will want the best you can do; a compromise doesn't seem as appealing.
997hz sine test tone, at -6.2dBFS which is just shy of clipping my audio interface's line inputs via Amanero, then balanced out - > TRS 1/4" inputs.
USB disconnected:
USB connected (but serial not active):
Peak level of the tone is -0.2dB on the plots, and reflects the level is just below clipping point of the input.
The main difference is the noise floor is about 2dB higher with USB connected, as you'd expect.
There is no obvious degradation to the signal.
What will make a big difference is if the serial interface is active, which I didn't test this time around. The RS232 interface chip goes to sleep fairly aggressively - which makes serial volume control a bit problematic at present. When the interface is awake it has an onboard switching power supply to boost voltage to RS232 levels. I suspect the high levels of distortion are related to the switching power supply being active rather than the USB - RS232 interface being connected per se.
Attachments
Last edited:
What I was refering as serial conection, which passes somehow some 50Hz noise throught, was the RS232 interface. It would be strange that a SPS of the RS232 interface would show 50Hz harmonics, or not?
So more precisely my graphs are both with USB audio once with and once without the serial RS232 interface connected.
The noise floor is with DAM disconnected and input ADC shorted. The measurement was done with a RME Fireface 800.
Last edited:
Yes, so am I
One graph USB serial connected, one not.UC-232 USB -> RS232 adapter connected to a Macbook Pro plugged into AC but fully charged battery.
TC Electronic Konnekt 8 Firewire audio interface, powered by firewire bus.
So if my setup does not display the gross levels of signal degradation you experience what is the difference?
What frequency does the charge pump operate at? I might have been looking at something else but I thought it was something like 40kHz...
Similar devices like the MAX232 are know for generating current spikes in the power supply when the charge pump is operating. It seems to be the most likely noise source to me.
I assume you were running BAL into the fireface?
Last edited:
Yes, so am I
UC-232 USB -> RS232 adapter connected to a Macbook Pro plugged into AC but fully charged battery.
TC Electronic Konnekt 8 Firewire audio interface, powered by firewire bus.
So if my setup does not display the gross levels of signal degradation you experience what is the difference?
I assume you were running BAL into the fireface?
No, I use SE
and a short (~15cm) coax cable to the fireface.Things AC powered and connected to the Laptop (also a Macbook Pro):
- Laptop power supply (not the source as also tried without)
- Fireface (firewire connection)
- Ethernet internet connection
- external monitor
Moreover I use an other USB RS232 adapter, one from "Ugreen" (EAN 6957303822218)http://www.amazon.de/gp/product/B00QUZY4TW
With all due respect, I heavily doubt that balanced signal transmission on some few cm will solve anything here. The noise is not induced while transmitting the signal between DAM and Fireface. The signal is symmetrized in the DAM only shortly befor the balanced output, and at the SE I measure before that point.
I am next to be tempted to solder together an balanced cable to disprove your ... ahm hypothesis.
Last edited:
The 50Hz noise is typical of ground loops caused by SE connections. It is not just the few cm of interconnects, the power cords are also in that loop. If that is the case here, it could very well be that the balanced outputs do not have that noise. Assuming they are wired correctly according to AES48, which the DAM onboard XLR connectors are not.
However, despite having a definite preference for balanced connections in my system, I find the sound quality from DAM buffered outputs sufficiently degraded to make them unuseable for me. Some people do not mind, or maybe even prefer the buffered outputs, it is pointless to argue over personal preferences. Just try it out and use what you like best.
As for myself, I am thinking of building a different output buffer to try to solve this dilemma.
So it seems.
Bill Whitlock's AES paper on the subject, "Balanced Lines in Audio Systems: Fact, Fiction, and Transformers" online here:
http://www.davelevasseur.com/Balanced Lines in Audio Systems - Bill Whitlock - OCR.pdf
The thing to note is ground noise is common mode. Balanced inputs reject anything common mode - not just induced noise picked up on the cable.
Last edited:
So I measured a balanced connection, moreover I tested several things I forgot the last time.
Firstly with SE it only matters is RS232 is connected or not, no influence what is connected to the computer, if the RS232 connection is established or not, to which computer it is connected,... As the Firewire 800 can not be run bus powered, I can not avoid the ground loop which probably causes the effect.
Secondly with the balanced connection the 50Hz stuff is less and the difference between RS232 or not, is only minimal more overall noise. So I admit I was wrong here
Here some measurements with nothing except power and serial (or not) connected to the DAM (so no signal)
Balanced no RS232, with RS232
SE no RS232, with RS232
The measurements were made with the same settings, if you now take in account that the balanced signal level is 9dB higher, the balanced plot look no that frightening anymore.
Now plots with a 1kHz -60dB signal (without RS232 and still unchanged settings)
SE/balanced
if you subtract the 9dB from the balanced you also end here more or less with the picture of the SE except the 2kHz and 4kHz harmonics, which seem to be relatively higher in the balanced case.
Firstly with SE it only matters is RS232 is connected or not, no influence what is connected to the computer, if the RS232 connection is established or not, to which computer it is connected,... As the Firewire 800 can not be run bus powered, I can not avoid the ground loop which probably causes the effect.
Secondly with the balanced connection the 50Hz stuff is less and the difference between RS232 or not, is only minimal more overall noise. So I admit I was wrong here
Here some measurements with nothing except power and serial (or not) connected to the DAM (so no signal)
Balanced no RS232, with RS232
SE no RS232, with RS232
The measurements were made with the same settings, if you now take in account that the balanced signal level is 9dB higher, the balanced plot look no that frightening anymore.
Now plots with a 1kHz -60dB signal (without RS232 and still unchanged settings)
SE/balanced
if you subtract the 9dB from the balanced you also end here more or less with the picture of the SE except the 2kHz and 4kHz harmonics, which seem to be relatively higher in the balanced case.
The balanced (not yet) dual mono preview!
I inverted the signal on one channel and fed both (from the SE terminals) as balanced signal in the ADC.
Here an overlay of balanced mono (red) over the ordinary balanced (gray-green).
To obtain the same signal level I had to use a 0dB signal on the balanced mono and -4.5dB of the ordinary balanced.
The balanced mono performs in general better on the even harmonics.
The same for -60dB (resp. -64.5 for the ordinary) with 9dB gain at the ADC input for more details.
Just to verify that the different gain does no real harm, here the overlay with the same -60dB signal (so the level of the ordinary balanced is 4.5dB higher at the ADC input). The colors are swapped.
And the balanced mono compared with the SE output (red) which is one half of the balanced mono (same digital signal thus -6dB weaker peak at output).
You see that the SE has larger second harmonic. If you shift the other peaks by 6dB most will disappear except in the area around 400Hz where the SE looks better.
Finally the overlay of right and left channel of the above signal. You see e.g. significant differences on with the 2. and 4. harmonic, probably due to the resistor tolerances.
I inverted the signal on one channel and fed both (from the SE terminals) as balanced signal in the ADC.
Here an overlay of balanced mono (red) over the ordinary balanced (gray-green).
To obtain the same signal level I had to use a 0dB signal on the balanced mono and -4.5dB of the ordinary balanced.
The balanced mono performs in general better on the even harmonics.
The same for -60dB (resp. -64.5 for the ordinary) with 9dB gain at the ADC input for more details.
Just to verify that the different gain does no real harm, here the overlay with the same -60dB signal (so the level of the ordinary balanced is 4.5dB higher at the ADC input). The colors are swapped.
And the balanced mono compared with the SE output (red) which is one half of the balanced mono (same digital signal thus -6dB weaker peak at output).
You see that the SE has larger second harmonic. If you shift the other peaks by 6dB most will disappear except in the area around 400Hz where the SE looks better.
Finally the overlay of right and left channel of the above signal. You see e.g. significant differences on with the 2. and 4. harmonic, probably due to the resistor tolerances.
Last edited:
Thanks zfe.
Obviously it's better not to have the ground pollution in the first place, but it does explain the differences we were seeing in our results.
I'll check again but I had done tests "playing" digital zero's some time ago, and didn't see anywhere near the same level of low level hash as you are seeing with balanced hook up. I'll re-run if I have a chance.
The Konnekt isn't as good as the Fireface 800 but the difference wouldn't explain the degree our results appear to vary.
Konnekt 8: SNR 108dB unweighted 20hz-20kHz, THD+N -100dB @ 1khz -1dBFS
RME FF800: SNR 110dB unweighted, THD+N -104dB
Obviously it's better not to have the ground pollution in the first place, but it does explain the differences we were seeing in our results.
I'll check again but I had done tests "playing" digital zero's some time ago, and didn't see anywhere near the same level of low level hash as you are seeing with balanced hook up. I'll re-run if I have a chance.
The Konnekt isn't as good as the Fireface 800 but the difference wouldn't explain the degree our results appear to vary.
Konnekt 8: SNR 108dB unweighted 20hz-20kHz, THD+N -100dB @ 1khz -1dBFS
RME FF800: SNR 110dB unweighted, THD+N -104dB
Last edited:
The Fireface has switchable gain for the main inputs, which, in contrast to the inputs with variable gain, seems to have no penalty in performance (additional noise).
The measurements were made with the setting for SE input, i.e. with 0dBFS=+2dBV~1.26 V RMS. Whereas the standard for balanced input is 0dBFS=+13dBu~3.56V RMS.
This gives thus 9dB extra, in case no high level signals are around.
P.S. if you are interested, I experimented with the variable gain inputs. You get maximal about extra 45dB analog gain over the SE-input setting. Which gives you this plot
instead of this
With the DAM SE out ... no terribly more information and I am not sure how reliable the results are.
instead of this
With the DAM SE out ... no terribly more information and I am not sure how reliable the results are.
Last edited:
On a slightly more practical note...
I was having to repair the cable from RS232 header to DB9 socket on a reasonably regular basis due to the wires breaking at the solder joints. The socket isn't mounted so was subject to stress every time I plugged the RS232 adapter.
I've ended up getting an IDC DB9 connector from the local parts barn ( DB9 Female IDC Connector | Computer | Plugs, Sockets & Adaptors | Interconnect | PRODUCTS | PS0805 | Jaycar Electronics ), and swapped the TX and RX lines in the ribbon cable. I split out the individual strands before I had thought too much about what I was doing A better way to do it is to split the three wires for TX GND RX and twist theese wires 180° to reverse position of the TX and RX, then use a bit of sticky tape to hold the wires in position.
The board connector is 10 pin, so what I've done is use a 10 pin IDC plug, 10 strand ribbon cable, and a 9 pin IDC DB9. If you split between 2-3 and 5-6 at one end and twist 180° this gives a cross-over cable which allows direct plugging of the RS232 adapter. The wire from pin 10 on the board is not required and is trimmed.
1 -> 1 (red)
2 -> 2
-------
3 -> 5 (RXD -> TX)
4 -> 4
5 -> 3 (TXD -> RX)
-------
6 -> 6
7 -> 7
8 -> 8
9 -> 9 (GND -> GND)
10 -> x
Anyway the main advantage is this is a lot more robust (and neater) than soldering to the DB9 socket.
I was having to repair the cable from RS232 header to DB9 socket on a reasonably regular basis due to the wires breaking at the solder joints. The socket isn't mounted so was subject to stress every time I plugged the RS232 adapter.
I've ended up getting an IDC DB9 connector from the local parts barn ( DB9 Female IDC Connector | Computer | Plugs, Sockets & Adaptors | Interconnect | PRODUCTS | PS0805 | Jaycar Electronics ), and swapped the TX and RX lines in the ribbon cable. I split out the individual strands before I had thought too much about what I was doing
A better way to do it is to split the three wires for TX GND RX and twist theese wires 180° to reverse position of the TX and RX, then use a bit of sticky tape to hold the wires in position.The board connector is 10 pin, so what I've done is use a 10 pin IDC plug, 10 strand ribbon cable, and a 9 pin IDC DB9. If you split between 2-3 and 5-6 at one end and twist 180° this gives a cross-over cable which allows direct plugging of the RS232 adapter. The wire from pin 10 on the board is not required and is trimmed.
1 -> 1 (red)
2 -> 2
-------
3 -> 5 (RXD -> TX)
4 -> 4
5 -> 3 (TXD -> RX)
-------
6 -> 6
7 -> 7
8 -> 8
9 -> 9 (GND -> GND)
10 -> x
Anyway the main advantage is this is a lot more robust (and neater) than soldering to the DB9 socket.
