Seeking Help on REW Electronics Measurement Setup, Practices, Hardware

[Mikey Antonakakis]

Hi all, very pleased to have found this forum! I have some specific advice I'm looking for, but as this is my first post here, thought I would share a little personal background first to lead into what led me here and led to the questions I have.

Background:
I an engineer with mostly mechanical/automotive background, but enjoying constantly learning in various areas through hobby activities. I got bit by the sound bug as a child thanks to a father who loved tinkering with home and car audio - will never forget being about 5 years old sitting in the parking lot of the local 7/11 with him and wondering if a helicopter was actually landing on the roof of the car during the opening of Another Brick in the Wall, or jamming out to Sting with him

In my bachelor days, I got somewhat ambitious with building and experimenting with decent low-cost setups. This included building an infinite baffle sub into the attic of my house (2x15" Dayton drivers) powered by the same Alpine sub amp that made that Pink Floyd helicopter sample seem so real, using an old 1000W PC power supply to get 12V. Just stuff I had laying around or handed down from my dad. Also used a Technics SU-V98 feeding JBL 4410s, both of which I got for great deals at the time. Since I already had 12V power, for the crossover I used the old Precision Power FRX-456 that was also part of the car audio setup from my childhood. The setup was definitely a hodgepodge, but it sure did make me happy (and genuinely impressed a live audio engineer friend of mine). In any case, I can't get away with anything like that, at least not in our current house with a wife and toddler.

I mostly have not spent any time in the last 7-ish years tinkering with audio stuff, but recently decided to restore a rough Marantz 2220B I bought years ago for something like $20. Some of my recent work background included a few years of on-track testing of freight railcars at a railroad proving grounds, and my "office" was often a caboose outfitted with a nice, but "old", DAQ system. I guess I've been missing doing hands-on testing, because although my initial project scope was "clean the controls and replace a few caps" it has crept into "learn how to somewhat accurately characterize the performance before and after the changes." So, now onto the meat of my post: some review of my first stab at a test setup and some questions on how I might improve it/ensure I'm generally doing things to get decent measurements.

Current Setup:

• Probably noisy gaming/productivity PC (MSI A1000G PCIE5 PSU, Gigabyte X570 Ultra, AMD 5800x3d, RTX 4090, couple m.2 drives, etc).
• Alternatively, Lenovo Y50-70 laptop
• M-Audi M-Track Solo USB Interface (powered by USB)
• Fluke 8025B DMM
• Siglent SDS 1104X-E 'scope
• Couple 8-ohm 100W load resistors from Parts Express
• Simple 2.2k/22k voltage divider circuit to feed the USB interface input when testing at speaker outputs
• RCA cables from USB interface to Aux input on the Marantz, screw terminals on voltage divider circuit feeding male XLR cable to interface input 1 (wired per #17 on the RANE article)

USB Interface Measurements:
I thought there was a big difference between desktop and laptop, but clearly I was mistaken. Just did back-to-back test of noise floor and 1kHz 0dBFS output (somewhat arbitrarily -5.36dBu/418mV). Pretty close to identical between them, or at least a lot closer than originally thought.

Both inputs and the output have their own knobs (marked 1-10 on inputs, 0-10 on output), not sure exactly how they work (gain, attenuation, etc?). For the data below, they were set at "3" for the input and "6" for the output. This gave a ~419mVrms output at 0dBFS (as set in REW, confirmed with DMM). The input goes haywire if it goes over about 9 (big step in level and noise). I did a loopback calibration on both computers before these measurements, input knob untouched at "3," REW output set to -12dBFS, output knob to 8-ish to get the input to match at -12dBFS.

Laptop:

Desktop:

1kHz:

Moving the Marantz-specific stuff and my questions into separate posts to keep post length more reasonable

 

[Mikey Antonakakis]

Initial Measurements of Marantz 2220B:
I'm still mid-process when it comes to restoring this receiver/integrated amp. Rated at:

• 20W/ch into 8ohm and 4ohm
• 0.5% THD at 20W (less at lower output)
• +/-1dB 20-20kHz
• -77dB high-level hum and noise at 20W into 8ohm (guessing this is essentially noise floor and 60Hz?)
• channel separation of 30dB minimum

My initial set of measurements are as follows, but I think I had some mistakes in my test method which I will describe after that. This testing was after a very thorough cleaning of the unit and all controls, and fixing an issue with the power supply board, and adjusting power supply and output voltage per the service manual. No capacitors or transistors have been replaced yet (save one of each on the power supply board that were dead and preventing me from adjusting voltage as needed). I wanted to get a baseline and retest in stages as I replace stuff, so this is just the baseline, and overall I am impressed that it's mostly exceeding its specs despite maybe my test methods making things harder for it:

• Does not clip until about 30W into 8-ohm dummy load (oscilloscope trace to see clipping, 16Vrms showing on DMM), though I might double check those numbers
• 0.22% THD at 20W on both channels, doesn't exceed the spec unless output is ~1mW
  • Tested at 1kHz and with 20-20kHz sweep, above figures are at 1kHz, see plot below for sweeps at different levels
  • 3rd harmonic is dominant
• -0.6/+0.2dB, tested at ~5mW to 20W, very consistent at different power levels
• Measured 60Hz noise at -68dB left and -65dB right at 20W 1kHz output
• Not sure yet how to measure channel separation, haven't thought about it

Plots:

As mentioned, I suspect my test method isn't great. What I did was:

1. Windows sound configured as 32-bit/48kHz (interface's max), REW "treat 32 bit as 24 bit" unchecked
2. 1kHz output at 0dBFS, output knob on interface at maybe 8-9.
3. Adjust 2220B volume knob to get 12.65Vrms measured with DMM at speaker terminals (20W rated power into 8ohm dummy load).
4. Calibrate REW output based on this to get correct dBu values.
5. Verify at another output level, e.g. set REW output to 4dBu and verify 1.228V with DMM.
6. Adjust input knob on interface to maybe 10% lower than the level where the clipping LED flashes.
7. For all subsequent testing, do not touch any knobs on interface or receiver, all adjustments set with REW generator level.

I think that last step is where I probably should have done something different - it occurred to me as I went to bed that night that I was probably not helping the interface's resolution (both on input and output). For my lowest output test, REW says I was at -62dBFS (huh, I thought that should have been -56dBFS since I set 0dBFS to 20W...). Then on the input side, same story, sending very small signals since I set my 20W output to slightly below clipping the input.

Maybe this impacted distortion/noise at my lowest level tests? Obviously THD+N was much higher than THD at low levels, and I did notice that as I got to my lowest levels my DMM was not scaling with my output. In the following plot, I converted from my measured voltage back to dBu and compared to the dBu I set in REW for each test (I think my y-axis is backwards, but I was reading higher voltage than I should have been at low output levels).

Finally, a smattering of some of the above sweeps and 1kHz tests:

 

[Mikey Antonakakis]

Questions:

• Maybe the M-Track Solo is good enough for testing this amp and other hardware in the future, but I'm open to suggestions for a replacement! I would likely use it daily for my computer audio when I'm not testing things.
• Anything I can do to improve noise floor of the interface, or prevent mains noise from leaking through the desktop over USB? More details below.
• Recommendations for improvements to how I'm testing things? Maybe it would be better to use closer to 0dBFS REW output, and use the interface in/out knobs and receiver volume knob to adjust output level and keep input level closer to full scale? I'd have to recalibrate at every different output test level but that's not so bad.
• I'd love recommendations for small studio monitors for my desk, no preference for active/passive. Space next to my dual screens is somewhat limited, I think the JBL 306P MkII is about the biggest cabinet I could get away with at my desk. Happy to add a sub.

Details on mains noise:
For some time now I've been using the M-Track Solo with my desktop to feed PC speakers, recently Klipsch ProMedia 2.1. The Klipsch especially are transmitting all sorts of noise from my house's wiring. For example, when our gas boiler starts up, I can hear the sparks of the igniter through the speakers. It is on a different breaker, and I've recently replaced all outlets on both branches, verified all grounds connected, no loose terminals, etc. It's not just the boiler either. It also does this with the Marantz plugged in to the interface powering passive speakers. My guess is that I'm getting noise through USB since the interface is USB-powered, and the RCA connectors are sharing ground with the USB port. I did measure ~10mV offset between USB ground and my PC case (no voltage between PC case and mains ground). Clearly the USB on my motherboard is supposed to be mains grounded but measured a slight negative resistance with my DMM...

Although I have a whole bunch of stuff plugged into AC outlets in the room, the audio-related stuff is fairly simple: wall outlet -> Class I UPS -> Class I PC -> USB switch selector -> to M-Track Solo -> RCA out to Class II amp/speakers. The reason for the USB switch is that I share my desk/monitor setup with my work laptop. It has a docking station, laptop USB is also connected to ground through the dock (even when unplugged from the USB switch, and USB shield is also mains grounded). The USB switch is powered from one of its USB ports, currently using the laptop dock to power.

 

[Mikey Antonakakis]

Okay, updated my test method, allowed me to measure much lower output levels with far less noise, using my desktop PC for convenience. Basic process:

1. Start with interface output near min, input near max
2. Set REW to -4dBFS 1kHz output
3. Adjust only receiver volume knob to get close to the target dBu value (e.g. -48dBu, which resulted in slightly audible signal from JBL 62T from a few inches away)
4. Calibrate output in REW based on DMM reading
5. Set REW to dBu target (e.g. starting at -48dBu). Double check DMM to ensure voltage is correct.
6. Zero the SPL meter
7. Start the RTA, calibrate input based on DMM
8. Measure/screenshot
9. Repeat steps 2-8, adjusting receiver only, so REW is always outputting about -4dBFS. Repeat until receiver knob is nearly maxed out.
10. Once receiver is nearly maxed out, keep repeating 2-8 but this time adjust interface output knob.
11. Once input reaches about -4dBFS, repeat 2-8 but this time also adjust the input to avoid clipping, targeting 4-6dBFS each time.
12. Keep going until rated output power.

I believe this process maximized my usable DAC/ADC resolution, I think reducing noise/distortion from the interface, and ensured each measurement/screencap was normalized to the output signal level. It was VERY apparent on the DMM that my actual output level was tracking MUCH closer to the target. The noise/distortion reduction showed up very clearly as well: for example at -32dB THD dropped from 1.03% to 0.14%, N dropped from -11.6dB to -25.5dB, and THD+N dropped from 27.4% to 5.3%.

I did notice some weird stuff that was most noticeable at the lower outputs, I went down to -48dBu (3.08mV/1.2uW) this round, where the 1kHz signal was just audible with my head close to the speaker (I used dummy load for measurements, just wanted to get an idea of subjective level correlation). For the whole frequency spectrum, the "grass" would randomly get bigger, and at lower frequencies there was a very noticeable difference. Mostly it stayed at the higher level, but would occasionally drop to the lower level for less than a minute, then go back up again. Did not correspond to any obvious PC activity, but may still have been related? For this testing I completely disconnected my work laptop/dock from the USB switch so that shouldn't have been a factor, at least not over USB. Maybe something else in my house's electrical system? HVAC/heat was not running at the time.

To illustrate the above point, I let the RTA run at -40dBu output, and was able to capture both levels:

And another sample from 12dBu output, the above phenomenon was still happening but much less noticeable at the higher output. I think this was the level that gave the lowest THD and THD+N:

 

[JRS]

To the OP. Been there done that, only with holes in my floors with some avalanche 18's running IB. I suspect I will never have better bass than that.

 

[Mikey Antonakakis]

To the OP. Been there done that, only with holes in my floors with some avalanche 18's running IB. I suspect I will never have better bass than that.

Yeah, it was incredible, such clean bass and would go subsonic with no issue - never heard anything like it before or since. Hopefully one day I’ll have a chance to build one again. Although it seems Parts Express discontinued the drivers I used back then (IB385 I think).

 

[Mikey Antonakakis]

Here are a few plots that summarize the improvement in my testing quite well, I think. First is the error in my measured output vs my target output - definitely much better now. The second is THD+N - notice that all the points below 12dBu are not only worse in the previous measurements, but that the curve compresses due to the high output level error (it should reach down to -32dBu). The third shows the improvement in noise - now the amp is actually spot-on with the specs of -77dB at 20W. Cool!

 

[Mikey Antonakakis]

Okay, I just replace the electrolytic caps, some diodes, and the transistors on the power supply board. I was wondering if this would clean up the low frequency phenomenon I was seeing at very low output levels. It did not, as shown in the first plot. Those humps on either side of 60Hz fluctuate maybe +/-5dB every few seconds, and occasionally die down completely. To track things down, I measured at the input to the main amp, and sure enough, the input was much cleaner, as shown in the second plot. Levels were normalize to 5mV at main amp output. Still getting 60Hz noise, but less of it. In any case, I think that rules out noise from my PC, other environmental stuff, whatever else that isn't the main amp board.

 

[Mikey Antonakakis]

After re-capping and replacing some transistors on the power amp board, the low frequency, low-output level weirdness seems to be gone. A lot less fuzziness in the high frequencies, N decreased by 4dB, THD by 0.3%, and THD+N by 8.5%. 60Hz noise decreased by 4dB too. Still haven't recapped the tone amp board, maybe that will clean things up even more.

Before and after, and then with a mediocre overlay. I was a little too lazy to do anything more than screencaps in REW:

 

[SSS]

Interesting measurements. But these old transistor amps were limited in their abilities by design and transistor specs. Same on my very old Pioneer receiver. For me it would be waste of time to rebuild such an amplifer.

 

[Mikey Antonakakis]

Interesting measurements. But these old transistor amps were limited in their abilities by design and transistor specs. Same on my very old Pioneer receiver. For me it would be waste of time to rebuild such an amplifer.

Thanks! Honestly I've been using this mostly as an excuse to learn And it was just sitting and collecting dust (a ton of it) so it was nice to breathe some new life it. Doesn't hurt that some people will pay decent money for one of these, certainly more than I have spent on parts (but maybe not instrumentation and time haha). Also might be a nice nostalgia gift for my dad when I'm done with it.

 

[SSS]

Thanks! Honestly I've been using this mostly as an excuse to learn And it was just sitting and collecting dust (a ton of it) so it was nice to breathe some new life it. Doesn't hurt that some people will pay decent money for one of these, certainly more than I have spent on parts (but maybe not instrumentation and time haha). Also might be a nice nostalgia gift for my dad when I'm done with it.

Understand and agree. I did it with oldtimer automobiles where I learnd much on the mechanics and electronics.

 

[yewneek]

Great thread! Useful for learning.

Cheers!

GB