Impulse Response Discrepancy - Intel vs M1 Mac

[fattire]

This is a weird one and I haven't been able to find anything even remotely related. I'm not new to REW, but I'm far from the level of expertise that many folks here have. I'm stumped. Any and all advice is appreciated.

The summary is that the impulse response for ONLY the center channel is wildly different when measured on an Intel vs an M1 MacBook. Graphs showing the issue are below and an MDAT is attached. When comparing with/without room correction, The change in delay is accurately reflected in the M1 measurements, but the center channel is off-set by ~25 ms.

UMIK-2, 90-degree cal file. JBL Synthesis SDP-55. Early 2015 Intel MBP vs 2020 M1 MBP. SDP-55 firmware was reinstalled from USB and all settings wiped prior to testing. Speaker delays and gains are zeroed out when room correction is disabled. Mic was not touched between measurements.

One difference between these laptops, aside from the processor, are how the connections are made. The Intel MBP has native USB and HDMI connections. The M1 MBP requires USB-C adapters (for at least HDMI & I don't have a long enough USB-C to USB-C cable to use directly with the UMIK-2). I used separate adapters for the mic and HDMI connections. Both are Apple-branded and purchased directly from Apple. I have a few of each of these and the results were the same regardless of the adapter used.

Results are consistent across recent versions of REW. All versions of 5.20.x were tried on both platforms. I did not try any 5.19 versions on either (I've never used a 5.19 version; only 5.20 beta and newer going back to ~Jan or Feb of this year).

Thanks in advance!

 

[John Mulcahy]

Very weird. Any way to try the USB-C setup with the Intel MBP?

 

[fattire]

I don't think so, at least not without a weird combination of additional adapters that I don't have. However, I forgot about my work MBP. It's a 2019 Intel MBP that is also USB-C only. Here is the IR and MDAT from that. I don't like having non-work stuff on it but installed 5.20.3 and did a quick test.

It's a similar result, but still way off on the center only without as much off-set as the M1. Very strange. (Ignore the "1.54...." at the beginning of the MDAT name. I was making sure to keep measurements with processor firmware status straight as I iterated).

EDIT: I've got an Intel iMac with a mix of USB-A and USB-C. I'll need to see if my cables will reach or not, but won't be able to do that until later this evening at the earliest.

 

[sam_adams]

The frequency response in the first .mdat is identical for both Macs for the individual channels to within 1 dB. The Center channel frequency response is almost identical between all three Macs. The system delay seems to vary greatly between all three of the Macs—which might be an OS 'thing' on each Mac since they are from different hardware generations.

 

[fattire]

Spark of inspiration at dinner. I have a portable DAC that came with a USB-A -> USB-C adapter. I used this for the mic on the late 2015 Intel MBP. So the chain was MBP USB-A -> USB-A/C Adapter -> Apple USB-C multimedia Adapter -> USB-A to UMIK-2. Results are "correct" and consistent with the mic plugged in directly to the laptop.

The system would not recognize the HDMI connection to the processor when connected to the multimedia adapter. I suspect a bandwidth limitation enforced by the OS.

The frequency response in the first .mdat is identical for both Macs for the individual channels to within 1 dB. The Center channel frequency response is almost identical between all three Macs. The system delay seems to vary greatly between all three of the Macs—which might be an OS 'thing' on each Mac since they are from different hardware generations.

Frequency response is very consistent. Regarding noise, I'd put money on the air conditioner. There was intermittent outside noise as well. This isn't a purpose-built, isolated theater so I get minor, but measurable, intrusion from lawn equipment, vehicles, etc.

How/where are you seeing "system delay"? That is not something I'm familiar with viewing in REW.

 

[sam_adams]

How/where are you seeing "system delay"? That is not something I'm familiar with viewing in REW.

On each individual measurement, if you click on the round, blue button with the 'i' on it, it opens a window with the particulars of the selected measurement. What version of the Apple USB-C Digital AV Multiport Adapter do you have. The A2119 is for newer devices and supports HDMI 2.0—but some MacBooks don't work properly with it on Big Sur resulting in the screen flashing on and off. This Startech CPD2HD4K60 actually fixed that issue.

 

[fattire]

On each individual measurement, if you click on the round, blue button with the 'i' on it, it opens a window with the particulars of the selected measurement. What version of the Apple USB-C Digital AV Multiport Adapter do you have. The A2119 is for newer devices and supports HDMI 2.0—but some MacBooks don't work properly with it on Big Sur resulting in the screen flashing on and off. This Startech CPD2HD4K60 actually fixed that issue.

Interesting on the adapters. I just checked, and both of mine appear to be the older version. I'll need to upgrade and test/measure again. I can justify replacing the Apple stuff through work. Any thoughts on that specific Startech vs something similar from Anker? The Startech delivery from Amazon is 10+ days out where I can get Anker tomorrow.

 

[sam_adams]

Don't know about the Anker adapter. Probably will work if the specs are right. Stranger things have happened with Apple devices as we all know.

 

[fattire]

Don't know about the Anker adapter. Probably will work if the specs are right. Stranger things have happened with Apple devices as we all know.

Thought I'd ask. Thanks. I've got an updated Apple adapter arriving tomorrow. Of course they've completely ditched the need for those on the new 14" MBP they released today. Wish I had held off buying this one a few weeks ago. Oh well. Them's the breaks as they say.

I should be able to remeasure tomorrow and see if anything changes. I'm going to add a few different variations to see if anything "looks different" in any way. Depending on the results, I'll also measure from the 2014 & 2019 Intel MBPs as well as the iMac. The iMac introduces a ground loop hum, but that shouldn't impact the IR measurements.

• LCR from M1 with L as ref
• LCR from M1 with C as ref
• LCR from M1 with L as ref, but C & L cables swapped between processor and amp

Any other suggestions on measurements that might reveal something? It's so strange that it's a single channel and not all channels.

 

[fattire]

Got the updated A2119 adapter. No change in behavior on the M1 MBP. It's a consistent ~25ms delay. I also tested on our 2019 Intel iMac. The delay there is identical to the 2019 Intel MBP showing a delay on the Center only of ~12ms.

The iMac was interesting as I tested with both the mic and HDMI through the new adapter, and then the mic using a local, native USB-A port and only HDMI on the adapter. Results were the same.

When swapping cables around, the problem stays with (or follows depending on your perspective), the C channel output of the processor.

I can upload another MDAT, but there really isn't any change at all in behavior.

@John Mulcahy - Any other suggestions here? I'm at a loss.

 

[sam_adams]

Posted this over on AVS, also. You want to try @John Mulcahy's response.

 

[fattire]

Posted this over on AVS, also. You want to try @John Mulcahy's response.

I'm not measuring subs, though; they're muted via miniDSP. These are standard LCR speakers (Revel F228Be L/R & C426Be C if it matters). Though that could be a suggestion to start the low end of the sweeps at a higher frequency where there is more separation from the noise floor - sorry thinking out loud on the keyboard. I'll try that.

It also doesn't explain why it's only on one channel. It looks to me like the noise floor and magnitude are very similar for all three speakers?

 

[John Mulcahy]

I'm as baffled as you are. How long is the HDMI cable? Odd things can happen with longer cables, but that more typically involves signals coming from the wrong channel.

 

[fattire]

I'm as baffled as you are. How long is the HDMI cable? Odd things can happen with longer cables, but that more typically involves signals coming from the wrong channel.

That's a good suggestion. I need to double check but I believe it's a 25 foot HDMI 2.0 cable. I'll try a shorter one that is known-good. It's easy to use the cable connected to the Xbox Series X, and that cable passes 4K60 UHD, Atmos, etc. to the processor and then to the receiver without issue.

I'm waiting for a 3rd-party USB-C HDMI adapter to arrive today and will test that along with the other cables as well.

 

[fattire]

I'm as baffled as you are. How long is the HDMI cable? Odd things can happen with longer cables, but that more typically involves signals coming from the wrong channel.

No joy. Identical results with a 3rd-party HDMI adapter and a different, known good HDMI cable (6 foot length used on the Xbox all the time) on a different input of the SSP. I used a higher low frequency as a starting point to ensure the start of the sweep was sufficiently above the noise floor (recommended in a link above; +30 dB difference).

I've got one more idea, but can't get to it tonight. I want to use this M1 laptop to check the IR via Dirac. Yeah it's different software and a different audio path, but it will at least be a data point. I'm hoping to get some response from other M1 and SDP-55 users over on AVSForums.

 

[sam_adams]

I wonder if it could possibly be some latency introduced because the JVM is running under Rosetta on the M1 Mac. @John Mulcahy, is the Azul JVM bundled with the Mac REW builds strictly an x86-64 version?

 

[John Mulcahy]

No, REW is a universal binary with both M1 and Intel runtimes. That wouldn't explain why only the C channel is affected in any case.

 

[fattire]

It also wouldn't explain why I see a ~10 ms delay for only the center on some later Intel Macs. The only common denominator between the systems showing this odd behavior is the required use of a dongle to connect HDMI (and the same SSP of course). Unfortunately, it seems that it is also something unique to my configuration.

I bought the newest 14" MBP announced Monday that has a native HDMI port and will not need a dongle to connect. We'll see what things look like whenever it arrives.

 

[fattire]

Brief update ... I had a reason to remeasure with Dirac, and captured the impulse response there. This is the M1 MBP with the same UMIK-2 and cal file. The pre-correction IR is below, and shows the center as ~1.75ms before the mains which is accurate. It's definitely HDMI related ... somehow. Now I've got a 3-4 week wait on the new MBP that doesn't need to use a dongle.

Until then, I can at least work around the problem with the 2015 MBP. Hopefully the 2021 doesn't have the issue as well.

 

[fattire]

TL;DR - Working as expected on the latest MBP that brings back on-board HDMI. I have no idea what's up with those other machines. Or why no one else has seen this before. Glad it appears to be all sorted though. Thanks @John Mulcahy and @sam_adams for the thoughts on this one.

2021 M1 MBP through on-board HDMI:

2021 M1 MBO through Apple A2119 Adapter HDMI:

 

[sam_adams]

Well it's good to see that REW works properly on the new Mac. If you are up for it, you should try a fresh install of 11.6 on the troublesome Mac and see if that fixes it. As usual, backup you data and have your app installers ready before proceeding. Big Sur 11.6.1 full installer .pkg file is available from Apple on the Mr. Macintosh site.

 

[fattire]

Well it's good to see that REW works properly on the new Mac. If you are up for it, you should try a fresh install of 11.6 on the troublesome Mac and see if that fixes it. As usual, backup you data and have your app installers ready before proceeding. Big Sur 11.6.1 full installer .pkg file is available from Apple on the Mr. Macintosh site.

I’ll try this on my 2020 M1 MBP later this week. I can’t do it on the work machine (2019 Intel), and the 2021 M1 Pro MBP doesn’t support Big Sur.