E1DA Cosmos ADC

[horias2000]

I replaced the 10ohm resistors with some MELF ones I had around. These are a bit bigger so I had to mount them vertically and use a short wire to connect to the other PAD. The THD got a bit better but still not what it used to be.

I will try to remove IC10 tomorrow and see if the issue goes away.

 

[nanook]

@horias2000: Isn't it a bit weird, that the harmonics improved from ca. -30dBc to ca. -60dBc? If the 10R resistors looked and measured well, I'd expect no major change.
Do you have a multimeter that can measure the forward voltage of diodes? IC10 is quite likely a CJ2045 (p/n BJT pair), so you could first check if the BE and BC junctions measure ok.
You should be able to identify a "dead junction" without de-soldering the device. When a bipolar transistor dies usually the BE or BC diode is close to 0 Ohms and you would measure close to 0 mV for both polarities with the "Diode" range of the DMM.

https://en.jscj-elec.com/shuangjingtiguan.html

@IVX: Is my assumption right, that the clamp in the Scaler is the same circuit as in the APU notch (i.e. this circuit https://www.diyaudio.com/community/...form-apx555b-for-30-000.386001/post-7177184)?

 

[horias2000]

@horias2000: Isn't it a bit weird, that the harmonics improved from ca. -30dBc to ca. -60dBc? If the 10R resistors looked and measured well, I'd expect no major change.
Do you have a multimeter that can measure the forward voltage of diodes? IC10 is quite likely a CJ2045 (p/n BJT pair), so you could first check if the BE and BC junctions measure ok.
You should be able to identify a "dead junction" without de-soldering the device. When a bipolar transistor dies usually the BE or BC diode is close to 0 Ohms and you would measure close to 0 mV for both polarities with the "Diode" range of the DMM.

https://en.jscj-elec.com/shuangjingtiguan.html

View attachment 488704

@IVX: Is my assumption right, that the clamp in the Scaler is the same circuit as in the APU notch (i.e. this circuit https://www.diyaudio.com/community/...form-apx555b-for-30-000.386001/post-7177184)?

I was wrong in saying that all the 10OHM resistors were OK. Two were around 6-7k. So definitely cooked. The improvement came after replacing these. I did not have the time to look at IC10 but will do soon. Yes, I have multimeters that can measure diodes. I did not know the part number hence I wasn't sure about the pinout.

 

[IVX]

@horias2000: Isn't it a bit weird, that the harmonics improved from ca. -30dBc to ca. -60dBc? If the 10R resistors looked and measured well, I'd expect no major change.
Do you have a multimeter that can measure the forward voltage of diodes? IC10 is quite likely a CJ2045 (p/n BJT pair), so you could first check if the BE and BC junctions measure ok.
You should be able to identify a "dead junction" without de-soldering the device. When a bipolar transistor dies usually the BE or BC diode is close to 0 Ohms and you would measure close to 0 mV for both polarities with the "Diode" range of the DMM.

https://en.jscj-elec.com/shuangjingtiguan.html

View attachment 488704

@IVX: Is my assumption right, that the clamp in the Scaler is the same circuit as in the APU notch (i.e. this circuit https://www.diyaudio.com/community/...form-apx555b-for-30-000.386001/post-7177184)?

right, usually the only 10ohm res is opened.

 

[nanook]

Here's the clamp that IVX implemented and posted on DIYaudio. Very nice, yet simple circuit.
It won't survive a long heavy overload - safe power dissipation is limited in a SOT23-6 package and in the 0603 10R resistor.
In case of a heavy overload the 10R resistor will blow up as IVX pointed out and eventually one of the transistors in the CJ2045 will die. Both of them are cheap.
This is definitely better than if the OPA1612 or even the ADC chip were to break. The ESD protection diode of the OPA1612 would become conductive and the input signal would raise the 15V supply via this path until one of the opamp dies.

@IVX: The CJ2045 is not readily available at the ususal distributors like Mouser, Digikey, Farnell, RS, etc. -> Do you know of a replacement type from a better-known semiconductor manufacturer?

 

[horias2000]

Thank you all for the input. This is highly appreciated! I removed IC10 but the scaler still doesn't perform at it's maximum. IC10 had 1ohm between pins 2 and 6 (base - emitter).

There's some improvement especially when it comes to THD but the noise is pretty high now (THD still too high for the D50s). I'll place an order for some component sometime net week and I'll include the opamps as well. Just to make sure.

Looking at the datasheet of CJ2045 I found ZXTD4591E6TA that seems to be a good alternative and it's on stock on Mouser.

 

[nanook]

The ZXTD4591E6TA does not fully live up to the CJ2045 according to the specification (max. current, hfe, leakage), but I don't think this is an issue.
Ic max: 2A is a lot too and the power dissipation and thus junction temperature will likely be the limiting parameter - it's physics and you can't outsmart it
hfe: is not important because in the fault condition, there's plenty of base current
leakage: Usually devices are a lot better anyway - this spec likely is the test limit. Testing for small current takes time = cost.

The high 1/f noise points towards the OpAmp - your scaler must have had significant input overload.

 

[nanook]

I'm not sure if the scaler has electrolytic caps in the signal path (like in the schematic of the APU clamp). If they got reverse polarized they might have suffered as well.
You could check if shorting them makes a difference (provided your signal has a small DC component.).
BUT: At ca. 100k input resistance these caps need to be completely dead to explain the observed behavior.

 

[IVX]

Thank you all for the input. This is highly appreciated! I removed IC10 but the scaler still doesn't perform at it's maximum. IC10 had 1ohm between pins 2 and 6 (base - emitter).

View attachment 489024

There's some improvement especially when it comes to THD but the noise is pretty high now (THD still too high for the D50s). I'll place an order for some component sometime net week and I'll include the opamps as well. Just to make sure.

Looking at the datasheet of CJ2045 I found ZXTD4591E6TA that seems to be a good alternative and it's on stock on Mouser.

it is https://www.diodes.com/datasheet/download/ZXTC2045E6.pdf
The ESD protection diodes of the OPA1612 are "bypassed" with external 4148 as well. So, if we have 1612 dead, I can't imagine how it happened.

 

[horias2000]

I replaced the OPA1612 opams as I already had a few around from other projects and it looks like the scaler is working ok now. I still have to replace the small protection transistors but for no it looks OK.

There is small bump around 70Hz that I don't remember seeing before but this could be setup of cable related. In any case, I do not think it has an impact on performance.
Thanks you all for the input and help. @IVX this is excellent customer support

 

[IVX]

I replaced the OPA1612 opams as I already had a few around from other projects and it looks like the scaler is working ok now. I still have to replace the small protection transistors but for no it looks OK.

View attachment 489167
There is small bump around 70Hz that I don't remember seeing before but this could be setup of cable related. In any case, I do not think it has an impact on performance.
Thanks you all for the input and help. @IVX this is excellent customer support

But how did you burn it?? Poweramp?

 

[horias2000]

But how did you burn it?? Poweramp?

To be honest I do not know. I was about to test a power amp that I made. As it's a prototype, it did not have the output protection board connected to it but I was careful not to connect the cables at start-up or power down to avoid the transients. I must have did something wrong but it's not really clear what.

 

[MaxwellsEq]

To be honest I do not know. I was about to test a power amp that I made. As it's a prototype, it did not have the output protection board connected to it but I was careful not to connect the cables at start-up or power down to avoid the transients. I must have did something wrong but it's not really clear what.

Did you check with an AC and DC voltmeter first?

From experience, I always expect power amplifiers to do something destructive until their stability is proven during soak testing.

 

[horias2000]

Did you check with an AC and DC voltmeter first?

From experience, I always expect power amplifiers to do something destructive until their stability is proven during soak testing.

Yes. The offset was minimal. A few mV.

 

[nanook]

When testing a power amplifier, you can easily add series resistors in the inputs of the scaler. I would probably start with a few hundred Ohms each.
Bandwidth: RC will still be in the MHz region
Noise:
in * Rs will add to the en of the scaler -> +0.7 nV/sqrt(Hz) for 2 x 200R
200R each will contribute a voltage noise of total 1.8nV/sqrt(Hz)

All of these noise contributions are negligible compared to the output noise of a power amp. The noise of its input pair is amplified by typically about x20 for a power amplifier. Even if the input stage is excellent this means more than 20nV/sqrt(Hz) at the output.

Even simpler, you can add a (matched) pair of series resistors to the XLR input of the ADC(iso) - in case you do not trust the 2.5mm jack.

-> Was there a specific reason why you did use the scaler as frontend?

 

[horias2000]

When testing a power amplifier, you can easily add series resistors in the inputs of the scaler. I would probably start with a few hundred Ohms each.
Bandwidth: RC will still be in the MHz region
Noise:
in * Rs will add to the en of the scaler -> +0.7 nV/sqrt(Hz) for 2 x 200R
200R each will contribute a voltage noise of total 1.8nV/sqrt(Hz)

All of these noise contributions are negligible compared to the output noise of a power amp. The noise of its input pair is amplified by typically about x20 for a power amplifier. Even if the input stage is excellent this means more than 20nV/sqrt(Hz) at the output.

Even simpler, you can add a (matched) pair of series resistors to the XLR input of the ADC(iso) - in case you do not trust the 2.5mm jack.

-> Was there a specific reason why you did use the scaler as frontend?

Indeed, I was thinking at adding series resistors on the input of the scaler. The power amp is pretty low noise and THD when it comes to power amplifiers (based on the D. Self design) but still, it's nowhere near what the OPA1612 can do, even with extra resistors on the input. The reason why I use the scaler is for convenience, as I can adjust the gain close to 0dB in REW. I could also set the Cosmos ADC to an input of 10V and skip the scaler and accept that it will not be close to 0dB in REW, as 5W on 8ohm yields 6.32Vrms. I could also set the ADC to 6.7Vrms but I'm more comfortable with 10Vrms

 

[Jbenda]

Subject: Question about DC offset on Goldnote PH-10 XLR output when connected to E1DA Cosmos


Hello,


I’m reaching out with a question regarding my Goldnote PH-10 phono preamplifier, which I use for recording and measuring vinyl LPs.


PH-10 technical details:

• Frequency response: 20 Hz – 20 kHz (standard), improved 20 Hz – 50 kHz ±0.3 dB
• THD: < 0.05% @ 1 kHz
• SNR: 89 dB
• Outputs: 1× RCA (unbalanced) and 1× XLR (balanced)
• Output level: RCA ~2 V, XLR ~4 V
• Output impedance: 50 Ω
• Problem description:
  
  
  • When connecting the PH-10 via its XLR outputs to the E1DA Cosmos ADC, I observe a persistent DC offset, but only when both channels (L and R) are connected.
  • If I connect only one channel (either left or right), the offset disappears.
  • Using the RCA outputs instead of XLR, no DC offset appears.
  • When connecting the PH-10’s XLR output to a Studer A810 tape recorder, no DC offset is present and everything works normally.
  • I measured the input impedance of the Cosmos at the highest 10 V range to be approximately 3.5 kΩ, which may affect the PH-10’s output stage behavior.
• 
  My questions / request for advice:
  
  
  
  • Do you have any experience with the PH-10 being sensitive to the ADC input impedance on its balanced XLR output, causing the output stage (e.g., servo or output buffer) to shift its operating point?
• Is there a recommended PH-10 configuration or setting that could reduce the DC offset when recording via XLR?

I would appreciate any advice, technical insights, or recommended devices/modifications.

Thank you very much for your help

 

[IVX]

I think it's a good idea to test the DC bias on the PH-10 XLR output, and compare with limits here https://e1dashz.wixsite.com/index/cosmos-adc

 

[nanook]

I observe a persistent DC offset, but only when both channels (L and R) are connected.

This sounds strange indeed.

Do you have a simple DC voltmeter?
In case I'd recommend to measure the DC voltage between pin1 (circuit ground) and pin2 (+signal) and pin1 and pin3 (-signal). The pin numbers of the XLR connector usually are written on the plastic insert. Make sure to include the sign - i.e. connect the negative input (black) of the voltmeter to pin1 for both measurements.

Are these voltages close to zero volts? If not, please post the numbers here for left and right output. (a couple of mV are considered normal).

The ADC(iso) allows +2.5V to - 7.5V. In case you exceed this, one of the electrolytic caps will show increased leakage and suffer.

If there's no DC voltage outside the range given in Ivan's specification, the only idea I have is that the RIAA preamp oscillates at high frequency and this RF gets rectified by some internal device in the ADC(iso) and finally shows up as DC.

Studer A810 tape recorder, no DC offset is present and everything works normally.

The Studer very likely has high impedance AC coupled inputs with caps allowing a significant DC voltage. So no surprise, the Studer doesn't complain.

 

[Jbenda]

I think it's a good idea to test the DC bias on the PH-10 XLR output, and compare with limits here https://e1dashz.wixsite.com/index/cosmos-adc

Today I performed some additional tests and found that the phase shift appears even when the PH-10 is completely disconnected (including from the mains). It only occurs when both XLR outputs are connected.
When I connect the E1DA Cosmos ADC to my reel-to-reel tape machine Studer A810, everything works perfectly.


Would it solve my issue if I convert the PH-10 XLR output to RCA, and then convert RCA back to XLR before feeding the signal into the Cosmos?
Could this approach negatively affect the sound quality?


The PH-10 also has an RCA output, which does not exhibit this problem, but unfortunately I am already using that output to feed my amplifier.