1. Citation please.
2. We needed 10pf of capacitance, There's nothing below 50V voltage rating AFAIK.
I assume you are most concerned about the precautions forbidding coupling usage and those regarding leakage? I was concerned about that when I saw it used to replace the 47uf/16V coupling caps on the E-MU DAC.
I have never tried to use solid polymer for coupling. Only for the supply bypass capacitors and filter caps (FILT/FILT+, VREF) on the AKM and CS converters.
For the AKM ADC FILT location I saw no difference between SEPC/OS-CON and non-polymer (Nichicon VR). But on the CS4398 DAC FILT+ location I seemed to measure slightly degraded noise with the OS-CON and strangely degraded harmonics. Sample size was 2 and I switched the OS-CON to VR.
Last edited:
I was told to use 1206 2kV rated NP0 for filters for ultra low distortion applications. I can not find a nice graph however. I have attached a graph for the variation with voltage for a variety of case sizes of X5S, X5R and X7R. I would like to see data for NP0/C0G if I can find it. Obviously the scale would be very different. But such data might be helpful if one is working around -120 to -150 dB distortion levels.
I guess the quick summary is: When using ceramic dielectrics use case sizes and voltage ratings which result in thicker layers and thus lower field strengths. (And of course use NP0/C0G when possible.)
The next question is: Are there differences between vendors of NP0/C0G caps? [Does everyone know what the optimum blends of neodymium, samarium and other rare earth oxides are? Does everyone know how to manufacture them the best way? I am guessing no...]
I guess the quick summary is: When using ceramic dielectrics use case sizes and voltage ratings which result in thicker layers and thus lower field strengths. (And of course use NP0/C0G when possible.)
The next question is: Are there differences between vendors of NP0/C0G caps? [Does everyone know what the optimum blends of neodymium, samarium and other rare earth oxides are? Does everyone know how to manufacture them the best way? I am guessing no...]
Attachments
Last edited:
I'd say not the LM4562 even though I have an 1820m that has those in it. It sounds great and the RMAA spectrum looks nice but it runs pretty warm due to the 10ma quiescent current times 29 opamps inside. IIRC the the NJM2068M that come in it have quiescence of 2.5ma or so. The LM1656 has 3.5ma. But other than this issue the LM4562 are great for this application imo.
I didn't recap anything but the preamp section. LOL, the stock preamps are well known to sound not very good, so it got the attention even though I have a couple of very good outboard preamps for anything that matters.
There might be opportunities for improvements in the ADC/DAC caps too, but my suspicion is that its not huge. Plus the fact those are low value electrolytics in tight spaces makes upgrading to a different construction virtually impossible due to space constraints. So you're left with swapping one electrolytic for another.
My experience has been that the larger value of electrolytic, the more likely to detect meaningful differences. I think in the first 1820m I did low ESR in the preamps since they tend to have a fairly flat impedance curve in the audible range. They sound nice enough now to use for sketches, voiceovers and such, a definite improvement.
So to sum up what I did: retrofit all opamps and the blocking caps for the preamps. Are there opportunities for other improvements? Probably. But did I realize 90-95% of the improvement opportunity? Probably.
One other thing to notice is the opamp used for driving the headphone out jack. LM4562 do a nice job of driving a 600 ohm load, and the OPA1656 does even better. So whatever you end up using for the ADC/DAC, make sure the one for headphones meets your needs. I use a dedicated headphone amp myself since I use 32 ohm AKG701s so it didn't much matter, but I was in there anyway so did it too. It's a definite improvement.
My batch of OPA1656 are backordered till Apr 5 so won't be able to post any results for those till after then... As with the LM4562 I put in the first one, I chose based on the spec sheet "sound unheard" and hoped it passed the listening and recording tests. The 4562 worked out splendidly, and I suspect these will as well. IMO, the CS4398 in the 1820m are still solid chips even by today's standards as are the AK5394, so doing upgrades to the analog sections of an 1820m makes a lot of sense considering what it costs to get that many channels of something that quality these days. Hope this helps and as usual jmo, ymmv, etc, etc.
Great. Thank you for all the E-MU 1820 information. I went for the caps first but it might be that the opamps are where the real improvements are to be found.
But then again I started with a dead 1616m with a broken power supply section. After fixing that the levels were all over the place and that is when I started to pull and measure the other caps. I found coupling caps with 30 to 140 Ohms ESR so very dried out/baked poor quality caps. I have pulled and replaced over 50 caps from the 1616m. I have measured all of them and 80% of them are bad.
I actually started with an 1820 (not 1820m) about 11-12 years ago. But recently I was comparing converters and concluded that the best deal would be either an 1820m/1616m or a reworked (upgraded) Digidesign 003 Rack. (With power supply, clock, opamp and CS5361 to CS5381 upgrades on the Digidesign 003.)
In the end it appears more economical to go with the 1820m/1616m (and less work). Yet I plugged a Digidesign 003 rack into Windows 10 and it worked no problems at all with an LSI chipset Firewire card. I will try the E-MU with Windows 10 again but my first attempts in December/January failed to get it working. I will try again with cleaning out the drivers and starting over.
But then again I started with a dead 1616m with a broken power supply section. After fixing that the levels were all over the place and that is when I started to pull and measure the other caps. I found coupling caps with 30 to 140 Ohms ESR so very dried out/baked poor quality caps. I have pulled and replaced over 50 caps from the 1616m. I have measured all of them and 80% of them are bad.
I actually started with an 1820 (not 1820m) about 11-12 years ago. But recently I was comparing converters and concluded that the best deal would be either an 1820m/1616m or a reworked (upgraded) Digidesign 003 Rack. (With power supply, clock, opamp and CS5361 to CS5381 upgrades on the Digidesign 003.)
In the end it appears more economical to go with the 1820m/1616m (and less work). Yet I plugged a Digidesign 003 rack into Windows 10 and it worked no problems at all with an LSI chipset Firewire card. I will try the E-MU with Windows 10 again but my first attempts in December/January failed to get it working. I will try again with cleaning out the drivers and starting over.
Last edited:
On the capture side the problem lies in what happens in the conversion imo. Imagine an input of a 1KHz square wave and the ringing associated with it. Now consider sampling that wave at 384KHz. Your sample set for that wave cycle doesn't represent a square. The sample set also doesn't fully reflect the ringing, but it does reflect something other than the true square corner and upper rail of the intended square wave. You end up with the upper bounds not looking like a flat line parallel to the zero rail, but something akin to a log decaying ramp.
Its relevant since square wave is a primary staple of analog synths, clipping is the norm in electric guitar, and clipping is the norm with brick wall limiters.
On the reproduction side the problem is not so big imo. The "messed up in capture square wave" sample set is converted by and passed from DAC chip to opamp, but for all practical purposes its slew is limited to the max angle allowed by a zero to full output sample change at whatever the DAC chip frequency is set to run. i.e. the DAC chip can only output a trapezoid where the higher the DAC frequency the more closely it approximates a true square...and whatever HF ringing may happen after that is easily dealt with by a good designer.
Just follow the driver "cookbook recipe" from the links I pm'd exactly and it should be fairly painless. If you make the mistake of running the script before any driver is installed, you'll have to manually delete all of the files the script installs from your system one by one by finding their names/location in the spreadsheet included in the download. It takes about 45 minutes to do (speaking from experience, lol) but if you then follow the cookbook exactly it will work.
Also, thanks for your findings on the caps, I'll probably do them too while I'm in after seeing what you found....
Last edited:
I also read that other Creative drivers will cause problems. Since the machine previously had an Audigy 2ZS (which has an E-MU DSP chip on the card) that might also be part of my headaches.
I am pretty sure I did not run the script before the driver was installed. However I had tried to install from the E-MU driver CD/DVD before searching online and finding the script. I also, unfortunately, found several different online suggestions with different driver/beta versions. So I think a thorough driver cleanup is in order before I try again. Unfortunately I don't have a fresh Windows 10 install for the HP Z400 that I bought on Ebay.
Someone suggested Driver Sweeper from Guru3D to try to clean up old Creative drivers.
I am pretty sure I did not run the script before the driver was installed. However I had tried to install from the E-MU driver CD/DVD before searching online and finding the script. I also, unfortunately, found several different online suggestions with different driver/beta versions. So I think a thorough driver cleanup is in order before I try again. Unfortunately I don't have a fresh Windows 10 install for the HP Z400 that I bought on Ebay.
Someone suggested Driver Sweeper from Guru3D to try to clean up old Creative drivers.
Last edited:
TDK has Simulation Models for Multilayer Ceramic Chip Capacitors, including DC Bias Models: Multilayer Ceramic Chip Capacitors | TDK
Difference and comparison among Simple Model, Precise Model and DC Bias Model: https://product.tdk.com/system/files/dam/tvcl/mlcc_model_en.pdf
AVX Online Simulation Tools: SPICE Models | AVX
MLCCs with Dipped Radial Leads: Guide on Various Solutions Offered by MLCCs with Dipped Radial Leads | Solution Guide | TDK Product Center
Difference and comparison among Simple Model, Precise Model and DC Bias Model: https://product.tdk.com/system/files/dam/tvcl/mlcc_model_en.pdf
AVX Online Simulation Tools: SPICE Models | AVX
MLCCs with Dipped Radial Leads: Guide on Various Solutions Offered by MLCCs with Dipped Radial Leads | Solution Guide | TDK Product Center
Last edited:
FWIW, the copies of Win 10 Pro you see on EBAY for a couple of $ work great. Fresh install only is the only caveat. Have several of those from several vendors months apart and they always work. From what I've read they are recycling the serial/activation from salvaged/recycled PCs. No weirdness involved, all you get is the activation, you have to provide your own install media.
I managed to get a fresh install going using another surplus Z400 motherboard. For some reason these HP Z400 motherboards self activate Windows 10. Somehow Microsoft "knows" the motherboard (some sort of electronic serial number I am guessing).
Anyways, the original machine is still not working. However the machine with a fresh Windows 10 install on a little 32GB embedded SSD is working great with the instructions you provided. Now I can go to the next step and upgrade the opamps.
I started trying to get this E-MU interface working again back in December so it is great to finally figure out the W10 driver issues.
Thank you for the help!
Anyways, the original machine is still not working. However the machine with a fresh Windows 10 install on a little 32GB embedded SSD is working great with the instructions you provided. Now I can go to the next step and upgrade the opamps.
I started trying to get this E-MU interface working again back in December so it is great to finally figure out the W10 driver issues.
Thank you for the help!
Last edited:
I believe that NE5532 have been measured to have lower distortion at higher voltages, as one example. I think perhaps Douglas Self did those measurements if I recall correctly.
Also I "think" my ES9038Q2M DAC I/V & filter stage prototype sounds better at +/-18V than at lower voltages. I cascaded two regulators together (for greater ripple rejection) but it necessitated lowering the voltages and I thought it sounded worse so I went back to the single regulator and the higher supply voltages.
Last edited:
Also see this in the charts for LM4562 and LME49720 et.al. Those actually publish separate THD+N vs Output Voltage for operating volatages of +/-17, 15, 12, 2.5.
p.s. thanks again for the heads up on what you found in the small caps!
Great news! BTW if you don't know about the TxZ skin for the 1820m mixer UI, its worth looking up. Your eyes will thank you, its much easier to look at imo. It was public domain, so if you can't find it let me know and I'll shoot you a copy.
"For Windows 8 and Windows 10, the Product Key is "injected" onto the computer motherboard at the factory. Installation and activation of Windows refer to the information in the BIOS of the computer to verify authenticity": Windows Product Key Frequently Asked Questions (FAQ) | Dell
Activate Windows 10: Activate Windows 10
