EW Play (HWO) + DP 9 = DISASTER?

Discussion of Digital Performer use, optimization, tips and techniques on Windows.

Moderator: James Steele

Forum rules
This forum is for most discussion related to the use and optimization of Digital Performer [Windows] and plug-ins as well as tips and techniques. It is NOT for troubleshooting technical issues, complaints, feature requests, or "Comparative DAW 101."
Post Reply
User avatar
SchrodingersCat
Posts: 22
Joined: Mon May 23, 2016 3:30 pm
Primary DAW OS: Windows

EW Play (HWO) + DP 9 = DISASTER?

Post by SchrodingersCat »

Hello MotuNation - this is my first post. I hope I'm not posting to the wrong branch of the forum...

It may be that the time for full orchestra mockups in 1 PC has arrived - PC's are amazing these days.

But we've all heard the gripes and groans about the perpetually buggy Play engine from EW :deadhorse: , and perhaps even after all these years it could still be just as buggy as ever, but I'm thinking I may be experiencing more than the typically buggy Play, and I hope someone has some clues.

I spent the weekend configuring a large orchestral template in DP 9, on a high performance PC.

MIDI multi-record on

20 "Master" MIDI tracks (each one for a different instrument, such as violins I, violin II, cello, bass, piccolo, flute, etc.). All set as follows:
- input from MIDI keyboard (Roland RD700GX-any),
- output to a unique channel on a soft MIDI port
- record enable selections are used selectively depending on which instrument is desired.

320 "Slave" MIDI tracks - each one for a different articulation (16 articulations for each of the 20 instruments above). All slave tracks set up as follows:
- input from a unique channel on a soft MIDI port; note these are distinct from the channel/ports used to carry output from the Master MIDI tracks mentioned above
- output to a unique Instrument track, in groups of 16 (i.e., there are 20 groups of 16 tracks, within 1 group of 16 tracks each track outputs to a unique MIDI channel of the group's target Instrument track)
- all record enables are selected on all of these tracks all of the time (these tracks are merely used as conduits to the instrument tracks)

20 Instrument tracks - each set up with an instance of EW PLAY
- input: each track receives input from 16 unique "Slave" (articulation) MIDI tracks, so data for each articulation is on a different MIDI channel
- output: each instrument outputs to a stereo audio buss corresponding to the appropriate section of the orchestra

Outside of DP (in the same PC), there are soft MIDI ports set up by an application called LoopMIDI. These ports are used to connect DP to another program called TransMIDIFier. This program does some clever re-routing based on patch changes:
- on each of 20 MIDI channels comming from the Master tracks in DP, it looks for program (aka patch) change messages, and routes the MIDI data stream to one of 320 patch-number-selected MIDI output channels
- 320 output MIDI channels are then accessed in DP via soft MIDI ports
This exporting and importing of the same but re-routed MIDI data allows the use of patch changes to select between 16 articulations per master track. Seems rather tortured, but since PLAY doesn't support patch change messages or user defined key switches, this approach isn't completely nuts.

The above was configured without any samples loaded into any of the 20 instances of PLAY. Thoroughly tested the MIDI traffic functionality and everything worked perfectly. Awesome.

I then loaded 10 or so articulations of HW Strings Violin I into the first instance of play, and 10 articulations of Violin II into a 2nd instance, Viola, Cello & Bass similarly. About 30GB of sample memory used. Nothing exploded - whew, so time to record and play back a few lines to assess things prior to loading up the samples for the entire orchestra template.

Weird •••• starts happening:
(a) Random blasts of sample playback audio from Play instance number 1 every time any record enable is selected or de-selected for any of the master tracks.
(b) While recording simple passages, or sometimes just pressing record enable or selecting a patch for a track on the track screen before recording starts, random note-stuck-on fun happens - oddly not even notes that were ever played. DP's panic command doesn't work to shut down the offending sounder - have to track down which instance of play it is, open it and choose "stop all voices" from its main menu.
(c) Strange random delays - hit play and nothing happens for a while. Sometimes ever (playback indicator just shows a red bar).
(d) Occasionally Play starts playing some strange repeating electronic tones - nothing that sounds like the samples miss-triggering, more like a big robotic brain fart that won't stop.
(e) Random crashes - yes, to top it all off, suddenly DP just disappears! Re-start the program, reload the samples, stable for a little while, then same the same weird shits starts, then the program disappears again. This happened multiple times. Totally non-workable - cannot use this tool set with these challenges.

I am using the latest versions of PLAY, HWO, and DP 9, running on Windows 10 64bit Home Edition.

Processor load indicator never climbs above 30% or so.

Rebooted (more than once), no help.

Disabled all overclocking for CPU & RAM. No improvement.

Checked out the CPU and RAM temperatures - not even that warm...

I hope I don't have to make the sad move back to Cakewalk Sonar :vomit: - but it was more stable than this while working with multiple EW PLAY instances; I'm hopeful someone has some experience that might help here...

The hardware environment:
CPU: Intel Core i7 Extreme Ed i7-5960X/3GHz/8 Core/20MB Cash
MOBO: ASUS X99 PRO
RAM: 64 GB DDR 3
HD (System): 1TB SSD (Samsung)
HD (Sample Library 1): 1TB SSD (Samsung SSD 850)
HD (Sample Library 2): 1TB SSD (Samsung SSD 850)
HD (Audio Recording): 1TB SSD (Samsung SSD 850)
HD (Video): 1TB SSD (Samsung SSD 850)
Audio Interface: MOTU 1248 (via USB 2.0)
MIDI Interface: MOTU MIDI Express XT
MIDI Control Surfaces: 1 Mackie MCU & 2 Extenders, 1 Behringer BC2000
MIDI Keyboard Controller: Roland RD700GX

Thanks!
User avatar
SchrodingersCat
Posts: 22
Joined: Mon May 23, 2016 3:30 pm
Primary DAW OS: Windows

Re: EW Play (HWO) + DP 9 = DISASTER?

Post by SchrodingersCat »

UPDATE:

The instabilities I described on my first post have been largely resolved.

It turns out that having multiple instances of Play instantiated and connected to various MIDI channels carrying data, without having any samples loaded in Play, may have caused the problem. Gangs of Play seem to go nuts when triggered empty.

After loading all Play instances with at least a few samples, the system appears to be operating nominally. Still get occasional Play misbehavior but nothing nearly so bad as my initial post described.

So, here is a great data point - more than 400 MIDI tracks and 24 instances of Play supplying 55GB of highly demanding EWHSO samples are running quite nicely in DP 9 on Win 10 64, with an advanced MIDI routing scheme enabling patch changes based channel switching. In 48KHz 24bit stereo, Latency is extremely low - can perform swift piano passages without any distraction from delays.

Great stuff - this could not be done in 1 PC 3 years ago. Big RAM, fast SSDs & fast 8 core CPUs - all worth the money, IMHO!
User avatar
bayswater
Posts: 11924
Joined: Fri Feb 16, 2007 9:06 pm
Primary DAW OS: MacOS
Location: Vancouver

Re: EW Play (HWO) + DP 9 = DISASTER?

Post by bayswater »

I thought you might be telling us the system was alive and dead at the same time.

That PC sounds like a screamer. How would the cost of the basic system compare to a similarly equipped (CPU, RAM, etc, not drives) iMac?
2018 Mini i7 32G 10.14.6, DP 11.3, Mixbus 9, Logic 10.5, Scarlett 18i8
User avatar
SchrodingersCat
Posts: 22
Joined: Mon May 23, 2016 3:30 pm
Primary DAW OS: Windows

Re: EW Play (HWO) + DP 9 = DISASTER?

Post by SchrodingersCat »

Not sure about the cost of a Mac with similar config.

The cost of the system above is (USD):

CPU: ~1k
RAM: ~800
SSD: 4X ~500
Mobo/case/power supply/ etc: ~500

So, in the neighborhood of $4500.
User avatar
bayswater
Posts: 11924
Joined: Fri Feb 16, 2007 9:06 pm
Primary DAW OS: MacOS
Location: Vancouver

Re: EW Play (HWO) + DP 9 = DISASTER?

Post by bayswater »

SchrodingersCat wrote:Not sure about the cost of a Mac with similar config.

The cost of the system above is (USD):

CPU: ~1k
RAM: ~800
SSD: 4X ~500
Mobo/case/power supply/ etc: ~500

So, in the neighborhood of $4500.
Thanks. So the cost is not much different. In fact, looking at the Apple store, it looks a more expensive considering the iMac has a monitor. I suppose I'd need to look at benchmark performance tests to make a proper comparison.
2018 Mini i7 32G 10.14.6, DP 11.3, Mixbus 9, Logic 10.5, Scarlett 18i8
User avatar
SchrodingersCat
Posts: 22
Joined: Mon May 23, 2016 3:30 pm
Primary DAW OS: Windows

Re: EW Play (HWO) + DP 9 = DISASTER?

Post by SchrodingersCat »

Please post what you find out - I am curious now.

Generally Mac hardware tends to be substantially more expensive than performance equivalent windows/pc hardware (at the risk of sounding like cap'n obvious), but, this PC has some stuff with a steeper price curve than typical pc hardware...
kdm
Posts: 162
Joined: Sat Mar 03, 2007 7:16 pm
Primary DAW OS: Windows

Re: EW Play (HWO) + DP 9 = DISASTER?

Post by kdm »

Where are you hosting Play? I run a very similar setup with HWO, but host strings and percussion on separate PCs in VEPro, with winds, brass, pianos, and several other libraries hosted locally in VEPro. Never ran into the issues you did, but my first thought in reading your first post was in fact the patch changes. Play does not seem to like loading/switching in realtime.

I have a separate MIDI track in DP for each articulation - no problems with large orchestral and orch/hybrid scores.

Fwiw, as far as the Mac Pro / i7 Extreme PC price comparison - the new Macs are Xeon E5s. You would want to compare a Xeon build against Apple's Xeon Mac Pros. i7 Extreme/Xeon prices and architectures aren't the same. The Xeons run around $600 for a 6-core, 3.0GHz vs. $1015 for the i7 you have. They are also different architectures - i7s generally provide better dsp processing (plugins, VIs), where Xeons are generally better for high disk streaming/buss access, but with the higher L3 cache, they are better for high-demand applications. The i7s win for bang-for the buck raw clock speed, but the Xeons are server-class cpus, so they win for higher core counts, and applications needing as many cores as possible. Since VIs require both some level of cpu power/core access, and disk streaming, the pros/cons aren't as easy to compare.

At this level and tech though, both are incredibly powerful - i7 Extreme, and Mac Pro E5 Xeon - even for the $4k+ both systems run. It partly comes down to preference more than tech specs or price (ram and SSD costs are going to be similar either way), but one could make a good case for either.
User avatar
HCMarkus
Posts: 9712
Joined: Tue Jan 10, 2006 9:01 am
Primary DAW OS: MacOS
Location: Rancho Bohemia, California
Contact:

Re: EW Play (HWO) + DP 9 = DISASTER?

Post by HCMarkus »

bayswater wrote:I thought you might be telling us the system was alive and dead at the same time.

That PC sounds like a screamer. How would the cost of the basic system compare to a similarly equipped (CPU, RAM, etc, not drives) iMac?
You can't buy a similarly equipped iMac… 8 core/16 thread CPU; look to the nMP, and wait to buy until Apple finally gets a new nMP (7,1) shipping.

If you could put that 140 watt CPU in an iMac, it would burn up.
User avatar
SchrodingersCat
Posts: 22
Joined: Mon May 23, 2016 3:30 pm
Primary DAW OS: Windows

Re: EW Play (HWO) + DP 9 = DISASTER?

Post by SchrodingersCat »

Hey KDM -

I totally agree with your assessment of the Xenon vs i7 trade space. I opted for the i7 and so far its performing well (enough) with streaming from the SSD's. I've only been running at 24 bit / 48kHz though. Might get bogged down at 96 kHz.

Regarding the Play thing, my system has been running with great stability since I sorted out the main issue, which was having multiple instances of Play running with no samples loaded.

EVERYTHING is in 1 box (the i7 system I described in my first post). I've even been able to run a few instances of Altiverb and some Waves plugs without choking.

In this setup, Play doesn't change anything in real time - that is the beauty.

Lets take 1 instrument as an example:
A "Master Track" is where we lay down all MIDI data - all note, velocity, cc and patch change data for the instrument resides in this track. The track doesn't send data directly to Play though; instead, it sends data out on a soft MIDI port that is routed to another application that runs outside of DP - a program called "TransMidiFier". Its like a software version of a MIDI patch bay.

When TransMidiFier sees a patch change message from the master track, it knows to re-direct the subsequent MIDI data to a different MIDI port/channel. So, each input to TransMidiFier corresponds to a different Master Track (a different instrument), and each of these is mapped through TransMidiFier to a bank of 16 unique output channels which are accessible back inside of DP. Patch changes are used to select which of the 16 output channels are used.

Back inside DP, there is a set of 16 tracks that I call "Articulation Tracks" that take their inputs from the 16 output channels from TransMidiFier. These tracks each output do a different MIDI channel in an instance of Play.

This way, Play never sees a patch change message - it just receives data on different MIDI channels, depending on which patch change value TransMidiFier received.

Surprisingly, this is all working fine! Can write multiple articulations in a singe track and change between them with patch change messages. This works just like key switching, but uses patch change messages instead (which IMHO are superior to key switches).

Some tips:

On all articulation tracks:
Enable record, disable playback

On all master tracks:
Enable record and playback

When ready to bounce to audio stems, delete all data in the articulation tracks, disable record on master, then run a record. This will populate the articulation tracks with data only from the final take. After that, disable play on the master tracks and bounce to audio.

Cheers!
kdm
Posts: 162
Joined: Sat Mar 03, 2007 7:16 pm
Primary DAW OS: Windows

Re: EW Play (HWO) + DP 9 = DISASTER?

Post by kdm »

Great to hear that solved the problem.

I've seen other composers using Transmidifier. That approach is similar to Cubase/Nuendo's Expression Maps (which is a channel re-routing system as well). It does simplify the track count.

I used Exp Maps for quite a while after using the track-per-articulation approach for several years, but recently went back to separate tracks. I found I too often needed to layer/crossfade between articulations to enhance or create attacks that weren't effectively represented. Other composers I've worked with were doing the same thing with Exp Map setups.

Another problem I ran into with Hollywood Strings is that the finger position Legato (Powerful System) patches would not switch fast enough to change finger position within a phrase as those are builtin key switches in Play. So I setup 4 copies of that patch for each instrument, with each defaulted to each of the 4 finger positions. I found that the easiest and smoothest/most realistic way to divide phrases across different finger positions where needed.

That's just my approach. I think keyswitching, Expression Maps, or an approach like yours is probably more popular with most people now.
User avatar
SchrodingersCat
Posts: 22
Joined: Mon May 23, 2016 3:30 pm
Primary DAW OS: Windows

Re: EW Play (HWO) + DP 9 = DISASTER?

Post by SchrodingersCat »

KDM -

Very good point on the lack of simultaneous articulations with the "key switch style" approach. The patch change method is better than actual key switch because MIDI event chasing works and there is no need for non-musical notes in the track, but misses a potentially important capability that the track per articulation method provides as you mention. I'm going to try adding tracks as needed when layering or x-fades seem indicated.

My ears say some of the articulations in HWS may already be layered attack samples - like some of the marc sustains.

I didn't realize that Play had trouble with switching finger positions in real time in the Large System patches in HWS- I'm checking into that. If I have to load those articulations 4 times for fast switching I may destroy something :smash:!

Here's one that I am completely baffled by. The load time for my template (even with the SSD's) is currently at about 20 minutes. The windows performance manager shows that after loading, the total ram allocation is around 24GB. 20 minutes strikes me as a long time to load 24GB from SSD to RAM. I did the math and that equates to 170Mb/sec - nowhere near the sequential read spec of the drives (550Mb/sec). Does that seem right (anyone)? Hmmm...
kdm
Posts: 162
Joined: Sat Mar 03, 2007 7:16 pm
Primary DAW OS: Windows

Re: EW Play (HWO) + DP 9 = DISASTER?

Post by kdm »

I am seeing exactly the same thing with load times here (HWO in VEPro). I am almost positive it is Play. My local template is around 24G as well, and it also takes 20 minutes from SSD. My Kontakt libraries load very quickly - I really don't even see them loading in VEPro. However, each Play instance takes 1-2 minutes of idle processing before it starts loading samples. Loading samples is quick, but I have a suspicion that Play is slow at building an instance from instrument lists (mic positions to be loaded, etc) and resolving the resulting file locations.

I reported and worked with East West on a similar file location problem recently, so there seem to be issues there already.

Ironically, a few Kontakt libraries I have on a 6G SATA load faster than the Play instances on an SSD. The Kontakt libraries on the same SSD load in seconds.
Post Reply