E-MU 0404 USB Modification?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Sure, no need to rerun, I tested 1K, 5K and 9K and have RMAA saves, here they are and snapshots of 5K spectrums.
BTW after reading that another thread I'm planing to disconnect soft clipper and D5 D8 TVS diodes (hm.. I have few more opamps remain if they're really necessary :) )
 

Attachments

  • spectrum5k_before.png
    spectrum5k_before.png
    93.1 KB · Views: 635
  • spectrum5k_after.png
    spectrum5k_after.png
    43.7 KB · Views: 608
  • after_lme49860_opa1612_wirearound.zip
    613.7 KB · Views: 66
  • before_upgrade.zip
    631.6 KB · Views: 48
Removed softlimiter (desoldered R121, R122, R123) and D5D8.
Also removed C40, but this not given any result except may be 0.1db of noise level reduction (that is so small that can be just random fluctuation caused for example by changed room temperature :) ).
About softlimiter and D5/D8 - this definately gave results, especially on high freqs. See attached pics - all taken with same settings as before - unbalanced, max volume lineout, input pot adjusted at -1.7db while signal generated at -3db level.
To be sure that no other changes in Universe affected my result I checked also right channel (its input stage still unmodified) by stereo test (RMAA doesn't allow change channel in mono mode) and got for channel same results as got when tested unmodified left input stage.
Not sure what exactly caused improvements - either softlimiter either TVS, however there're still TVS diodes on output (D1,D2,D3 and corresponding for right channel) , and I will remove them next time will open my EMU again.. but not today, its already night here...
But removing TVS will make EMU vulnerable for static electricity, so doing that turns it from studio to home/lab device that requires good care.
 

Attachments

  • spectrum_5K.png
    spectrum_5K.png
    43.8 KB · Views: 588
  • rmaa_min.png
    rmaa_min.png
    50.7 KB · Views: 574
Last edited:
Removed D26-27-28-30-31 (on wiki page they incorrectly designated as D1..D3) and... Bingo! 3rd harm @5KHz reduced by 10db!
I don't think its possible to get notable more from this device...
BTW I think its possible to move this TVS-es to before corresponding 560-ohm resistors connecting them directly to opamps outputs, thus leave device's output protected while noticable reduce nonlinearity caused by them. But I'm not sure about capacitance of that TVS, but IMHO it will not be too big to cause opamp instability...
 

Attachments

  • rmaa.png
    rmaa.png
    45.5 KB · Views: 548
  • spectrum_5k.png
    spectrum_5k.png
    92 KB · Views: 202
Last edited:
did few experiments with opamps to reduce noise level (just suspected I have fake opa1612 cuz its not marked with TI, while having white line across left side instead near the designation) and found that LME49860 work better than that opa1612 as U22/U33/U39. BUT input buffer (U1) better works when OPA1612 is there. Strange, but true.
 

Attachments

  • 5K_with_all_lme49860_except_u1_opa1612.png
    5K_with_all_lme49860_except_u1_opa1612.png
    92.2 KB · Views: 215
Last edited:
also regarding
misterzu, very usefull information.

There is a same approaching at the posts 112, 117, 133, 167, 168 at the thread USB vs PCIe Sound Card for audio analysis.
Interesting graph at post 112. It looks like besides described upgrades that EMU doesn't have TVSes, or has some other TVSes, with much less capacitance thus much less distortive so THD doesn't grow at higher frequencies. BTW I've already seen such reports on other forums while seen also other graphs similar to what I saw on mine device before this upgrade, also just like yours in post #113 there. And since no-one of those people having flat THD graph reported they did something with TVS-es.. can it be that some EMU on market has different TVS diodes or some doesn't have ones at all?


You used the RightMark Audio Analyser Pro. Sure is the good software but is much robotic to their results.
Please, if you have the time repeat the thd measures with ARTA & STEP(part of ARTA) softwares.
I am very interesting if the thd improvement that you haved will repeat and with the ARTA software.
Looked at ARTA.. its really not so robotic.. I'd say too-not-robotic, may be will try next time later:)
 
It's really, ARTA isn't robotic like RightMark Audio Analyser.
Of course there some drop down menu with choices but it require calibration at inputs and outputs first of all (see the detail pdf manual).
Except thd, imd measures, jitter is including. The additional Steps is a plus of this program.

The results that vacuphile demonstrates at post 112 are with xlr connection. The rejection of common noise that xlr have, decreased significantly the thd at high frequencies (see the attachment).

With unbalanced connection, the total thd rises after 5KHz (see the attachment).

I read carefully your result, you have significant improvement!
I will wait you to test with ARTA to have more reliable and comparative results.

Mine right input channel is factory yet (like yours) with the only mod the vishay muliturn pot.

I have enclosed RMAA captures at 5KHz signal on left channel (I have removed the D5,D8 diodes, R121,122,123, all opamps with LME49720).
I tried to have -1.7 input signal like yours, with set at minimum the pot input position and rises the output pot until I achieve the -1.7dB. The RMAA was to mono mode.

This channel have a bigger 2nd than unmodified right channel.
There is an artifact at the 12KHz that I don't know the reason (it's the same at the two channels).

If you had a fake 1612, what will be with original 1612...they have a half thd than LME49720!
 

Attachments

  • stepR, 96KHz,-6dB, balanced.png
    stepR, 96KHz,-6dB, balanced.png
    40.7 KB · Views: 225
  • stepL, 96KHz,-6dB, after D5_D8 removed, unbalanced.png
    stepL, 96KHz,-6dB, after D5_D8 removed, unbalanced.png
    41.1 KB · Views: 170
  • RMAA_5K_thd.png
    RMAA_5K_thd.png
    15.6 KB · Views: 201
  • RMAA_5K.png
    RMAA_5K.png
    8.9 KB · Views: 190
Last edited:
hm, are you sure you didn't confuse the channels? removing TVS diodes can't increase distortions, at least it can do nothing, but in my case there was noticable improvement, especially on higher frequencies.
BTW seems I've understood how to use Arta. But I already re-soldered my loopback to balanced mode, since seems all others are using balanced connection for testing... IMHO its less informative cuz may partiall hide even harmonics.. anyway here're my results with balanced loopback.
I used left channel output and tested right and left channel inputs (remember, right-channel input is still untouched). I set Main level at maximum, during calibration at -6db level attenuated both channels so arta shown 'peak input estimated' approx 1050mV for each channel (not sure what that mean - but when I move input attenuator higher - it told me less voltages.. anyway - all indicators were green).
Then I did impulse response record with specified on pic settings.. Results looks a bit different from yours - probably some settings must be set so see full THD level and other harms, anyway most interesting graphs visible
[update]I took solderer, reverted loopback to unbalanced kind and re-run arta like described as above. Added pics with unbalanced results.
 

Attachments

  • arta_record.png
    arta_record.png
    26.5 KB · Views: 144
  • arta_left.png
    arta_left.png
    59.8 KB · Views: 148
  • arta_right.png
    arta_right.png
    94.8 KB · Views: 146
  • arta_right_unbalanced.png
    arta_right_unbalanced.png
    96.5 KB · Views: 124
  • arta_left_unbalanced.png
    arta_left_unbalanced.png
    61.2 KB · Views: 127
Last edited:
hm, are you sure you didn't confuse the channels? removing TVS diodes can't increase distortions, at least it can do nothing, but in my case there was noticable improvement, especially on higher frequencies.

- Maybe, I was not very clearly!
I haven't more distortion with D5/D8 removed.
I have more 2nd distortion on left channel vs right channel from the time that left channel opamps changed.
For this reason, I want to change the U1 (now is LME49720) with the factory opamp or 1612 that you had.

BTW seems I've understood how to use Arta. But I already re-soldered my loopback to balanced mode, since seems all others are using balanced connection for testing... IMHO its less informative cuz may partiall hide even harmonics.. anyway here're my results with balanced loopback...

- Not all used balanced connections, the most used TPS (1/4) to TPS (1/4) for loop-back measurings.
The xlr gives better results but for the real measures have the disvantage of low input resistance... almost 2.5K if I remember well.

Take a time with ARTA software. Here it is the manuals (ARTA Download)
At the 19-20 pages of Arta User Manual, described the calibration procedure.
After the calibration, you will have at the 0dBV almost 1Vrms at input. You don't need at this stage to calibrate microphone, only line inputs and outputs.

Run the Step program (not Arta) to take distortions vs frequency at all (THD, D2...D6)->96K, -6dB generator signal...generate the signal that you have almost -12dB input signal (look the bar at the end of window).

Run the Arta, choose the 4th icon Spa at Measurement Modes (Spectrum Analyser), choose the signal at Gen menu, Fs(48K), FFT (it depends from your computer power), Wnd (Kaiser7), Avg (Exp).
Choose the Y bar with dBFS and set until the peak reach almost -12dB (probably you will have the best results).
 
...but TRS is also balanced. + is on tip, - on ring. Its only having higher input impedance, cuz buffered. When I tell that use balanced loopback - its TRS out-TRS in, but tip-to-tip, ring-to-ring. And when I test unbalanced - means tip-to-tip, output's ring to nothing, input's ring-to-ground :)
Also D5-D8 affect only TRS input, XLR inputs are not protected by any TVS diodes.
 
OK, I see...

I have measured the STEP test with the two methods, trs<->trs (I have described this as unbalanced), trs<->xlr (I have described this as balanced) also.

Maybe, we continue with only TRS to have comparative results.

Look, what I have you proposal for ARTA & STEPS.
I will wait for your results.
 
Last edited:
here what I got with Steps, including its settings. BTW I've got -6db level, not -12db. Why should I get -12 if I calibrated and set -6db? May be i've missed something, will chek tomorrow, now its time to sleep :)
Note that I have only single loopback cable (need to but another couple of TRS jacks to make another loopback, so I test first left, then right channels (actually left-to-left, left-to-right to be sure both inputs got same signals) changing channel setup in measurements settings respectively before measure. Loopkacb BTW is ring-to-ring, tip-to-tip, so - TRS input used as balanced input.
 

Attachments

  • step_setup.png
    step_setup.png
    116.1 KB · Views: 442
  • steps_thd_left.png
    steps_thd_left.png
    108.8 KB · Views: 440
  • steps_thd_right.png
    steps_thd_right.png
    76.8 KB · Views: 443
Last edited:
I have sent you a pm about how I make mine arta calibration.
I don't know if I am 100% true, but I have follow the arta manual instructions for the output calibration and somehow a partial for the input calibration of both channel.

Left-to-left, left-to-right procudure for both channels is the same that I have done.

But I am a little confused with the "Loopkacb BTW is ring-to-ring, tip-to-tip, so - TRS input used as balanced input". Is that the Lynx suggesting method for unbalanced measures?
Can you upload a scheme how is this?
I usually use a trs to trs cable for loopback connection.

Your result at the left channel is almost the same with mine balanced measuring (trs to xlr cable).
The right channel is the know factory (unmodded) behaviour.
 
But I am a little confused with the "Loopkacb BTW is ring-to-ring, tip-to-tip, so - TRS input used as balanced input". Is that the Lynx suggesting method for unbalanced measures?.
..
Can you upload a scheme how is this?
I usually use a trs to trs cable for loopback connection.
I also used TRS to TRS pinout like in attached pic.
This card has two balanced XLR Low-Z inputs, two balanced TRS 1/4" High-Z inputs, two balanced TRS 1/4" outputs and one unbalanced TRS 1/8" stereo output. If connect big TRS input and big TRS output with normal TRS-TRS cable as on my pic - you will get balanced loopback connection for selected channel. If you will use XLR input instead you again will get balanced loopback, but using Low-Z input. To test UNBALANCED connection its neccessary to use balanced input in unbalanced mode that can be done by connecting ring (middle part) of input jack to ground and use only its tip (ending part) for signal transfer (and this is done automatically if use 'mono' TRS-connector).


Your result at the left channel is almost the same with mine balanced measuring (trs to xlr cable).
This both results TRS to TRS, the difference only is input TRS - left or right. Actually I don't even have XLR connector at home to plug it..

The right channel is the know factory (unmodded) behaviour.
Unfortunately I didn't do ARTA test of unmodded, but RMAA show moderate improving of right channel and more improving of left channel by comparing with unmodded state..
 

Attachments

  • loopback.GIF
    loopback.GIF
    12 KB · Views: 430
You have described the EMU inputs very well, thanks a lot.

Now, I must to clarify better mine captures.

The first capture at post 128 with the Right channel that it appears with a flat THD almost, will be with XLR Low-Z input. I had used a cable with TRS and XLR plugs.
It seems that the low Z of differential inputs, set a low total THD.

The second capture at post 128 with the Left channel that it modded with removed of D5/D8 diodes, R121,122,123 resistors and changed all opamps with LME49720 have done with balanced TRS 1/4" High-Z input.

I am going to make a double measure of each channel with Low and Hi Z input and upload the results again.
 
It is more clearly now.

For mine balanced Lo-Z input testing I used a TRS->XLR cable and for balanced Hi-Z input I used an xlr->tps 1/4 adapter on the same cable.

There is a better THD at upper frequencies btw Low-Z input vs Hi-Z input, both of channels (modified or unmodified channel).

It seems that the THD at the modified left channel is better than from right unmodified channel.

The next will be the same with THD (ARTA) at the 1KHz/9KHz...
 

Attachments

  • right balanced Low-Z input, -6dB, unmod.png
    right balanced Low-Z input, -6dB, unmod.png
    43.5 KB · Views: 106
  • left balanced TRS Low-Z input, -6dB, mod.png
    left balanced TRS Low-Z input, -6dB, mod.png
    42.5 KB · Views: 113
  • right balanced TRS High-Z input, -6dB, unmod.png
    right balanced TRS High-Z input, -6dB, unmod.png
    43.2 KB · Views: 119
  • left balanced TRS High-Z input, -6dB, mod.png
    left balanced TRS High-Z input, -6dB, mod.png
    41.5 KB · Views: 129
  • step setup.png
    step setup.png
    14 KB · Views: 419
Arta results with THD at the 1KHz and 6KHz.
The signal input is exactly 1Vrms that corresponds to 0dBV or -12dBu (Arta scale). Signal generator the emu0404 at the -3dB.

As you said before, I prefer to focus with balanced Hi-Z input only (trs to trs).

Arta results is better at the 48KHz usually. For this reason I chose the 6KHz fundamental frequency to see 2nd and 3rd harmonics.

Before I had said that I have a bigger 2nd harmonics at the modified channel vs unmodified.
The same happens and now but this is for 1KHz only. At upper frequencies the result is opposite, than 1KHz.

Some peaks at the 12KHz and its harmonics are artifacts.
 

Attachments

  • thd, left 1KHz.png
    thd, left 1KHz.png
    36.6 KB · Views: 128
  • thd, right 1KHz.png
    thd, right 1KHz.png
    38 KB · Views: 122
  • thd, left 6KHz.png
    thd, left 6KHz.png
    36.6 KB · Views: 113
  • thd, right 6KHz.png
    thd, right 6KHz.png
    37.2 KB · Views: 111
Last edited:
As I wrote in PM that I puzzled why you having flat THD on some graphs, now I understood that you have it only with Low-Z XLR input. And now eventually I understood why THD on high freq is better with XLR. When Low-Z load connected to ouput then TVS'es (hanging on output's line) impedance rather high by comparing with laod's impedance, so distortions caused by 'varicap effect' of that diodes much less by comparing to when high-Z load connected to same output.
So.. if my guess is correct and if you will proceed with removing TVSes on output you will get same good (or may be even better) results TRS-TRS as now TRS-XLR. But be carefull - I've broken one pad when desoldered TVSes :(
About 2nd harmonic on 1K - I'm still not sure that its caused by TVS removal. May be softlimiter/whitlock (did you disconnect it too? did something in that area, but it unlikely too. Also on your steps graphs D2 levels almost same. Can you check noise level - may be some noise peak caused mess on 2K?
BTW may be stupid guess, but can you check if both - and + lines of modified channel work? If you for example broken one line (shorted input pins for example) - you will turn it into unbalanced kind, that subsequently will cause raise of even harmonics. It would be obviously visible with usual pots that one channel is twice less sensitive than another, but I remembered you're using multiturn pots so check actual input channels sensitivity with multimeter/oscilliscope.
 
Last edited:
Today I decided to look at output stage opamps power. I've seen rumours that internal EMU power rails are full of noise and pain so took oscilloscope and checked what happens on post-DAC opamps. Positive power line was rather quiet if comparing with negative. Probably its because I'm using custom linear regulated PSU instead of out-of-stock adapter. But negative line had 40mVpp spikes. I changed power filtering caps there to polymer 100uF/16V + ceramic 10uF. That reduced spikes by almost 6 times. I also added ceramic caps in parallel to pre-ADC opamps power filtering caps, but didn't check with scope what happens there.I have idea how to improve this - replace 0 Ohm resistors that connect power rails with some inductors. Anyway this all had no effect on measured parameters - RMAA/Arta numbers remain same as before this mod.
 

Attachments

  • post_dac_opamps_neg_power_rail.png
    post_dac_opamps_neg_power_rail.png
    41.8 KB · Views: 228
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.