Lynx, Cubase 4 WinXP, multiple UAD users: Can anyone confirm or refute this?

[DAWgEAR]

I am trying to find out if the occurrence of an issue I am having is specific to my system or more widespread. Specifically, I am considering an upgrade but I would like to avoid a repeat of what I currently experience.

It would be GREATLY appreciated if you can confirm that you are or are not experiencing what I describe below. Bonus points if you list your system details!

I suspect that the cause is multiple UAD-1s; users of other hosts and audio cards are welcome to participate!


Procedure 1 (takes only a minute or two):

(1) Open the Lynx Mixer application.

(2) Open an empty, new Cubase 4 project.

(3) Check the Lynx mixer application for dropouts (look in the lower left of the Lynx Mixer Record/Play window) after launching Cubase. There may not be any, but if there is any dropout, reset the ASIO driver from within Cubase (Devices Menu > Device Setup > VST Audio System > Reset Button) so that there are no dropouts.

(4) In Cubase, add one mono Audio track.

(5) Add one UAD-1 plugin (any one will do) as an insert.

(6) Check the Lynx Mixer application for dropouts, there should not be any.

(7) Now add a second UAD-1 plugin (any one) as an insert.

(8) Now check the Lynx Mixer application for dropouts. Do you see any? I always have at least one. This is 100% repeatable on my system.*

* With only one UAD-1 card enabled, I get no dropouts at step (8). This behavior seems to be related to the allocation of UAD-1 plugins across multiple cards. Further, once at least one plugin has been allocated to each physical card, I generally do not get further dropouts.


If you did get any dropout in Procedure 1, please try Procedure 2 (you will want to!):

[top]

= = = = = = = = =


Procedure 2 (looks long, but it takes only about 5 minutes):

(1) In the same project (with the dropout(s) still showing in the Lynx Mixer application), create a second and third new audio track.

(2) On track 2, record just a few seconds of silence.

(3) Open the silent event you just recorded in the Sample Editor.

(4) Select all and use Audio Menu > Process > Silence so that your event is now truly pure digital silence.

(5) In the Sample Editor, change the ruler display to "samples".

(6) Position the project cursor somewhere within the event and create a marker at that point. I choose something like 100,000 samples, but anywhere will do so long as you create a marker so you can easily get back to that exact sample position.

(7) In the Sample Editor, zoom it to the sample level and use the pencil tool to draw a "one sample spike" at full amplitude exactly where your marker is set.

(8) Exit the Sample Editor and record arm track 3.

(9) Physically connect the mono L playback out from your soundcard or converters to the mono L record in. If you have a Lynx Two or Aurora with the Lynx Cables, just physically connect Out 1 (XLR Male) to In 1 (XLR Female), for example.

(10) Record the playback of the "one sample spike" on track 2 onto track 3.

(11) Open the newly recorded event on track 3 in the sample editor, position the cursor at the marker position you set earlier, and zoom in. The recording of the one sample spike will not be a spike anymore, but is the highest peak at the same position as your marker? If it is delayed, by how many samples? (Use the range tool to select beween the marker position and the highest recorded peak; look to the upper left of the Sample Editor windo to see how many samples your selected range encompasses.**

** On my system, the delay is a multiple of the ASIO buffer, eg. with the buffer set at 1024, the delay will be 1024 samples (23 ms at 44.1K) or 2048 samples (46 ms at 44.1K), etc.

[top]

= = = = = = = = =


My system: WinXP SP2, Cubase 4.1.0, A8V Deluxe (BIOS 1017), AMDX2 4400+, Corsair TWINX2048-3200C2, Lynx AES16 (firmware 25, driver 2.0 build 014e), 3xUAD-1 PCI (4.10), Matrox P650 (driver 2.4.0.179).

I am posting this on the UAD-1, Cubase, and Gearslutz forums, same user name: DAWgEAR.

[top]

= = = = = = = = =

[True North]

I recently moved to a new Dual Xeon Quad machine and I have been experiencing ALOT of dropout problems with my two Lynx AES16 cards.

I seem to be running into problems once I get to about 128 sample latencies. I am using SX3 but I am running 4 UAD cards off of a Magma expansion chassis.

There seems to be problems with Lynx drivers in general. I had another problem where Pro Tools was consistently crashing on me. I had to create a seperate boot with disabled Lynx drivers to get things working. DAW Bench has also noted problems with the firmware working properly with Dual Quad systems, you should check that site as well for more info.

[DAWgEAR]

I didn't mention that I am actually running at 1024 samples. So, at least in my case, it's not that I am setting too low of a buffer. Lower buffers just compound my problem.

[DAWgEAR]

True North, by any chance, are multiple UAD-1's involved?

[True North]

Originally Posted by DAWgEAR

True North, by any chance, are multiple UAD-1's involved?

Yes - 4 UAD's in a Magma PCI Expansion chassis

[jcschild]

try going back to an older Lynx driver.
yes i am serious

Scott
ADK

[DAWgEAR]

Scott,

I have tried going back to firmware 20/driver 1.30 build 057g as per recommendation from Lynx but the symptoms remain unchanged. Is there a different specific driver you would recommend I try?

I'm at the point where I am going to build or buy a new system soon. I've always built my own and didn't figure I'd buy a system, but I've been looking at your site a lot lately ...

The pages where you can customize systems on your site do not offer multiple UAD-1s to be selected. I'm looking at the Other Soft/DSP drop down menu.

Do you have experience building multiple UAD-1 systems that are free of the exact symptoms I have described? I am not asking you to reveal your trade secrets, but I would just like to know if it is even possible.

I don't mean any offense by what I just typed, but my system is not an uncommon one and it has been wonderfully stable (it still is except for what I described) for the two years I have had it. Six months ago, before I knew what was happening, I would have said my system was rock solid. However it was relatively recently that I discovered what was going on and that it has affected every recording I have ever made for the past couple of years.

I say that because so far nobody has indicated that they have tried successfully or unsuccessfully to replicate my procedure 1 and/or 2. I'm wondering if it is specific to just my system, or if others have it too but are are unaware that this problem is going on.

Have you explicitly tried the procedures I outlined?

[jcschild]

HI,
yes we have done numerous mulitple UAD set ups.
we work closely with UAD.
it best you call than turn this into a sales floor.

Scott
ADK

[DAWgEAR]

I'm kind of surprised that on 4 forums not a single person has tried my Procedure 1 or Procedure 2 and reported their findings one way or the other.

I guess that systems that utilize of all three of Cubase/Nuendo, multiple UAD-1s, and Lynx are rare.

If you are reading this and you are in a position to try the procedures, please do so and please report back. It takes less than 5 minutes.

Some updates from my end ...

I built an entirely new system from scratch:

Intel Q9450
Gigabyte GA-EP35C-DS3R
Gigabyte GV-NX86T256H graphics
Corsair TWIN2X2048-8500C5D (2 GB)
New HHDs, power supply, optical drive,
Etc.

I crossgraded 2 of my UAD-1 PCI cards for UAD-1 PCIe cards.

Literally, the only hardware common to the old system currently present in the new system is my Lynx AES16 PCI card.

The issue with dropouts I described continues to occur.

Recently, Lynx added updated drivers V2 build 15a. The release notes mention "ASIO driver now fills 14 buffers to the record devices so dropouts on record devices should never occur."

I am still getting dropouts when loading UAD-1 plugins across two UAD-1 cards. With only one UAD-1 card, it does not occur. Once a dropout occurs, all recordings thereafter are delayed (out of sync) by a multiple of the buffer size. This will continue until the ASIO driver is reset.

Barring a defect in my particular AES16 card, I am starting to believe that this issue will affect all users of Cubase, multiple UAD-1s, and Lynx. Perhaps it is only multiple UAD-1s and Lynx and has nothing to do with Cubase, but I humbly request your feedback to help narrow down the circumstances that cause this to occur.

If one is aware that this is happening, the workaround is to reset the ASIO driver after loading the first few UAD-1 plugins and to constantly check the Lynx mixer for dropouts (and reset the ASIO driver if and when they occur). If a recording has been made after a dropout, it has to be manually shifted by a number of samples equal to an integer multiple of the ASIO buffer setting.

I've pulled out a lot of hair over this, especially before I discovered the pattern and the workaround. I hope this documentation spares someone else the frustration I have experienced ...

[drmad69]

Scott's right about the v1 driver working better.
But you can't run any other apps with it; only asio and there is a buss master
settings that applies and at one point alot of the zlm/dm was incorect for
version 3. It however changes things such as slot order, it's the only driver
that was reliable about that...

But I did respond to your PM too;

These cards simply suck. I sold mine. Most people I know have too.

The PCIe card might be better as they claimed there was bandwidth issues;

Which is why those with magma chassis on PCIe have less issues;

However, as soon as you add another lynx card to the mix; sometimes

which is the same buss on a p4 type board, they fight over each other

and the performance on one card sucks and starts to make issues happen.

In fact; there were so many problems there is a hidden control key diagnostic

display. There is where you will see the two cards drastically different.

Lynx finally suggested I buy new cards too; There are a handfull of poeple

trying to use their cards as advertised, not in a limited fasion and those

aren't enough users to pay for a hardware mod and a recall; which is obviously

what is/was in order.. However, I don't think they ever found the nail in the coffin.

You know, scott, dozens of us have called, we only get one tech support guy

paul, the situation changed and changed, and I've seen this with other computer

products, they are licking their wounds. Now that the PCIe card is out, it will

be interesting to see if anyone reports the same random hosed behavior.

To characterize it, it works like this; the original issue I noted before they fixed
the playback stream is as follows, (this is from memory.)

After any dropout; regardless of the cause, and 1 is enough.

(If you minimize and maximize the nuendo/cubase screen 5 or 6 times, during playback or not you will have a dropout.)

The fixed the issue that stems would be in phase (it clicks on playback) HOWEVER
IIRC it is still OFF in relation to the record stream (they don't drop sample on the record stream.)

SO...............................

If you have a external FX, you will go out of phase and *STAY* that way until an
ASIO reset.

If you are recording to do mastering you will be OFF that sample amount until an
ASIO reset..

It really has DICK to do with the UAD cards; they just load the system/pci buss and it is easier to cause a dropout. But with NO UAD cards the simple minimize
maximize bug will do it. NO OTHER CARD MANUFACTURER OUT OF the DOZEN I've
tested drop or go out of phase with this ONLY LYNX.. Paul said it's steinberg,
but in 2 years, steinberg hasn't both to fix it and SIMILAR issues happen on
other platforms (daw software)
it's just slightly different because of implementation..

Dawgear has been trying to solve this issue like I did and I just gave up after 3 years and AM MUCH happier with ANYTHING ELSE.

My external FX *NEVER* go out of sync with PT mix cards in cubase/nuendo OR PT

With the 9652 I got to test with they NEVER went out of sync; and I did alot of testing on this; because I eventually found that the behringer ada8000's are out of phase, nothing with nuendo/cubase... But I would *CONSTANTLY* have drift
with the lynx cards because I would FORGET and minimize the application to look at the #^%#%$ lynx mixer!

Again; these cards suck. Lynx knows this. They tried to boot me off the forum
for crying bloody murder about it.

So here is yet another person with the issue....

DAWGEAR, if you can afford to switch uad's motherboards stuff...

JUST GET A 9652 on trial loan or any RME-AES card..

I *PROMISE* YOU. All these FUC@$%ed up problems will dissapear.

You will be happy, you can continue working. And hopefully like myself,
still respond and help people that got *SCR*W*D* by this company.

Lynx can reply all they want, it is HOGWASH, the cards have a problem
with a HUGE amount of systems that require recall, but it would have
made the company go under. They still have yet to fix the problem; don't know how and can't offer a recall or an upgrade.

I personally think that had this been a tire, we would see a ford type of issue here.

[drmad69]

P.S.

I tried putting up with it with asio resets, after a couple times of forgetting
and having to hunt down the off sample track, it gets tiring. I'm sorry
paul wasn't honest and gave you the workaround, he's had it for over a year,
because I gave it to them..

This is why the control feature and green colored dropout were added, so
they can continue to identify the users with the issue and pray they find
bad code.. Obviously the founder can't code his way out of a wet paper
sack and needs to get somebody else to audit and find the issue in his code.

and, I went so far as to read the asio sdk, they violate it's syncing recommendations
in several places... Remember the playback in sync bug was reported for over
a *YEAR* before it was fixed... Think of all those people that did summing
and *DIDN'T* know it was slipping...

I am 99% sure this card is responsible for people badmouthing nuendo's syncing
credibility. There is stuff of REP where they talked about mastering guys
getting clearly out of sync tracks and I bet you *money* the mix engineer had
lynx cards and was too bust to realize they get out of sync...

It eludes ALOT of people, remember, a simple minimize creates a dropout.

One has to use the lynx mixer too, so think about how many times this happens

to the common user...

Except, the common user, before a certain version, *never* heard the dropout,

it was masked by their driver code. A RME would make a OBVIOUS click and send all streams out of phase by the lost buffer.. It is correctly coded.


Well, enough ranting. If you own these cards ditch them. Alot of people are realizing this though, or because of the PCIe, I got pennies for mine. A $100 maudio interface did better on ebay for resale value.

[stratton]

Lynx 2-A, SONAR 7 and 3 UAD cards here. I'm going to check this out a little later. Peace of mind check if nothing else.

[drmad69]

Here's the link of people with the new cards, I've cross posted to see if we
get any results..

Lynx AES16e PCI Express card!


BTW...

I have no idea how to make it fail in sonar, but if I'm not mistaken similar/random

weirdness has been reported on their forum. It just takes agitating the driver

differently as sonar access the asio driver differently.

[DAWgEAR]

Originally Posted by stratton

Lynx 2-A, SONAR 7 and 3 UAD cards here. I'm going to check this out a little later. Peace of mind check if nothing else.

stratton: I would appreciate any feedback. If loading plugins across the UAD-1 cards does not produce a dropout, try what drmad69 suggested regarding repeatedly minimizing/maximizing the DAW application window. Also, for the test to be comparable, you should use the ASIO driver, perhaps that is what you already use. Again, any feedback, affirmative or negative is appreciated. Assuming that you can trigger a dropout (you should be able to do this by setting the buffer low and stressing your system enough) make sure you try procedure 2.

[DAWgEAR]

Originally Posted by drmad69

Here's the link ...

That link above does not point to the thread you intended. You might want to edit it. (I did check that your link from that thread pointing to this one works and it does.)

[drmad69]

oops I was afraid of that; it happens when you copy from a link where you post..

So did you get any farther dawgear?

[DAWgEAR]

Updates:

At the UAD-1 forum it was suggested I try a DPC Latency Checker utility and I did find out that I have some moderate spiking going on which can be related to dropouts. (My old system, which had the same Lynx dropouts had no spiking.) There is a large thread on this at the Sound on Sound PC forum; many users of Gigabyte motherboards reported either earlier BIOS versions or recent beta BIOS versions reduced the DPC latency. So I emailed Gigabyte tech support and they sent me a new beta BIOS that brought the spiking way down (max of 29 microseconds). That's the good news. The bad news is that the Lynx dropout situation persists, so I don't think the DPC latency is related to my issue with Lynx, multiple UAD-1, etc.

A while back, UA suggested a PCI latency tool whereby one can set the PCI latency of various devices. This made no difference on my old system with Matrox AGP graphics. For some reason, I can't even get it to run on the new system: as soon as I try to run it the mouse and keyboard lock up requiring a reboot. Here is a reference in case it helps anyone: UA WebZine "Support Report" June 05 | UAD-1 and Tiger plus Adjusting PCI Latency Settings to Fix Clicking and Popping Problems

I'm still trying to track this thing down ...

BTW, thanks for responding, Chris. I realize you've moved on, but I appreciate anything you have to share.

[DAWgEAR]

For the record, for a given buffer size, the minimize/maximize the application window sometimes triggers a dropout, but sometimes it does not on my system despite my frantic mouse clicking. I have no idea why. If I make the buffer small (32 samples), it comes more easily.

In contrast, the UAD-1 method is 100% repeatable no matter what the buffer size, even 1024 samples. All I have to do is load a new plugin so that the load is placed on a different physical card in an empty project and voila, dropout every time.

When I had the old system, I wasn't sure what the cause was. Now that I have built a completely new system (the only component "harvested" from the old system is the AES16), the Lynx card/driver is looking like the prime suspect. The Lynx ASIO driver is involved in this; to what extent the UAD-1, UAD-1e is implicated, I do not yet know.

[FatStudio]

Hay

About test 2 ,I think you should have a delay as big as your latency settings.
If you want to eliminate it you should give Nuendo/Cubase a way of knowing It is looping by setting this as an External plug-in 1/1/I/O and hitting the little button on the small window it has to represent it.

This will compensate for the delay ,every DAW has this as it takes time for the computer to send sound out.

The test shows latency on your system is PERFECT :-)

And..about the spike changing ,when you send a spike to a D/A converter it overshoots ,there are no perfect spikes in the physical world.
The D/A is doing its best to create an electrical square wave which is not possible.
When this approximation of a square is feeding back to the A/D this is distorted again as the signal you sent got louder then 0DB .It might be louder by 3-6DB..

[DAWgEAR]

Originally Posted by FatStudio

Hay

About test 2 ,I think you should have a delay as big as your latency settings.
If you want to eliminate it you should give Nuendo/Cubase a way of knowing It is looping by setting this as an External plug-in 1/1/I/O and hitting the little button on the small window it has to represent it.

This will compensate for the delay ,every DAW has this as it takes time for the computer to send sound out.

The test shows latency on your system is PERFECT :-)

And..about the spike changing ,when you send a spike to a D/A converter it overshoots ,there are no perfect spikes in the physical world.
The D/A is doing its best to create an electrical square wave which is not possible.
When this approximation of a square is feeding back to the A/D this is distorted again as the signal you sent got louder then 0DB .It might be louder by 3-6DB..

I'm not sure I understand your point. Forgive me if I have not understood.

If I do understand correctly, you are referring to the DAW's automatic compensation for the buffer latency. I am aware of this. Cubase (other DAWs presumably?) further has a way to manually offset recordings in increments of samples. I have made this setting so that the recorded "spike" lines up where it should and is sample accurate. In other words, my procedure 2 works perfectly. As long as no dropout occurs, this continues to work. (For the record, on my system, that setting was 1 sample with the v14 driver and is now 67 samples with the v15 driver).

However, once even one dropout happens, repeating procedure 2 reveals that any subsequent recording is now delayed (out of sync relative to previously recorded tracks) by an integer multiple of the buffer samples. Eg. with the buffer set at 1024, the delay will be 1024 samples (23 ms at 44.1K) or 2048 samples (46 ms at 44.1K), etc.

[FatStudio]

I see your point.

Well I'm having some crazy things happening with my 2 LYNX aes16-e's .
The drivers are not very good.

Nuendo doesn't like them very much

[drmad69]

The two LYNX's fighting is common in my experience.

I even had one that was a src; and it didn't matter..

This is the reason for the secfet control screen and the PCI throughput numbers.

One card will was always 50% worse then whatever ended up being

the primary card on the buss...


FWIW, I did mess with the latency settings, and in the UAD control panel there is a

addition separate place to just set it.

There are several apps on the net, things got somewhat randomly better

with the common suggestion audio card on top.. at the highest setting 254 or such..

(The question IS.... WHICH card to you boost, do you make them equal, do you have one slightly about the other and so forth.)

I tried an exhausting amount of combinations for MONTHS.

There was never an official word on the settings to use when two lynx's cards

are on the same buss fighting, and which settings to do what with.


Dawgear, as to your UAD problem...

When you do a ASIO reset, from time to time, do you end up it clearing and

having 1 automatic dropout? For instance at a buffer size of 128, I commonly

would get 1 dropout after the asio reset out of the gate..


Anyway, here is my "theory". When you switch plugs accross the UAD, it is

doing something very close to a ASIO reset, or possible the equivalent of

pushing the Automatic latency button on and off in the daw.

I don't know, try making your UAD's a 250 latency PCI buffer and the lynx

like 32 with everything else... Maybe that will agitate the issue.




In anycase; that situation holds off buffers from the Lynx driver for the time it

takes to irritate it. Rather, the driver should know some sort of a reset is in

process and either reset completely accordingly or honor the timeout.


So, anyway, the thing about either test is; it shouldn't happen, and probably

doesn't with other drivers. But it is the classic case of the lynx driver catching

a dropout where it shouldn't. Again, I think you can do this without a UAD is

all I'm saying... If you launch and app that puts 75% on your machine for 4

seconds lets say. It will probably dropout where most drivers wouldn't...


Also; They have recently had another bug I haven't seen solved that looks

alot like the samplitude issue, where the driver gets locked, tries to "catch up"

on the dropouts and they skyrocket until the machine is reset.. This

was a result from david trying to fix the summing dropout issue and it

backfired, especially in my machine..

But it wasn't until they removed PHASE from the command line tool that

broke the camels back in my case.. I was using the tool to invert the ada8000's

and had been for over a year and they removed support.. Absolutely no reason

to do that. I was happy to see after 2 years in the queue that the screen bug

was fixed. It finally took a user pointing out the memory link for them to believe

there was a issue there, and this is after dozens of people had complained..


I mean, with a client there, having your card's mixer app look like it had crashed

or was buggy as heck, it just didn't look professional.




They are trying; I just didn't pay for 2400$ in cards in 2004 for the ability to be

a alpha tester. It looks like they have increased the buffers again because of

record dropouts, you know, maybe someday they will indeed solve the problem..

I just figured I could sell them now and get what little they were worth out of

them and buy them back cheaper someday if I thought it was fixed...

But an RME card on ebay is already as cheap and I know it works.. I've had

both of them in the machine side by side... In fact, this is how I started using

them.. It's just the elusive issue didn't show up until months later when

I caught it, and by that time I had two lynx interfaces all built out with ls-adat's

and so forth...


I'm not the only one, there are dozens of reports. So if lynx says I'm just

bitter they are full of it. As we can see yet another person is trying to get

the issues solved.

[Autocrat]

Uh oh, I have 2 UAD cards and I'm putting together my PC now...

From reading this, should I stay away from Lynx? I was thinking of either a LynxTWO or Aurora8 with AES16pci card.

I plan to use Cubase.


Or is this just an isolated case?

[drmad69]

It's isolated to a section of users, I guess lynx has said.

But in my findings the minimize bug happens 100% of the time and the user

group is probably way larger than they will admit.


If I was you as I have said before, steer clear.


You can get one, just make sure you can send it back. Test this stuff;

I bet you money it will.

And yes, probably only second to samplitude it fails with cubase.


There are *alot* more peeved samplitude users.. The problem might

have been "fixed" with samplitude.. But it changes from each version

of driver and daw.


And of course; Pretty much no other card manufacturer has finnaly

got their userbase into an uproar. So, you pick. RME is a way more solid chioce

IMHO. And I've done dozens of interfaces.

Word has it their FW interface on the aurora is hosed too. But that's a new

development.

[Oldone]

Not using quad but dual Intel. Lynx IIA, latest drivers, 2-UAD1s (one PCIe and one PCI) ASIO drivers......Rock solid. Latency at 5.8ms.

[drmad69]

Gee, 5.8ms is a little high.. Does it fail at 128 buffer? :)

And you haven't tried either test have you...

download the steinberg loopback test....


run outs to ins....


minimize maximize a couple times, you will see the horror.


dawgear's test is a very good one too..


We didn't say it crashed..... We said it doesn't work right... Very different things.

[DAWgEAR]

Originally Posted by Oldone

Not using quad but dual Intel. Lynx IIA, latest drivers, 2-UAD1s (one PCIe and one PCI) ASIO drivers......Rock solid. Latency at 5.8ms.

Oldone, when you say "rock solid", do you mean that you have performed procedure 1 and procedure 2 as described above and you cannot replicate my findings?

Perhaps Lynx 2 behaves differently than AES16. I have access to a Lynx 2; I should try the test with it too ...

[EDIT 05/13/2008: I tried with Lynx Two.. same results as with AES16.]

If so, are you using a different DAW from Cubase or Nuendo?

My system is rock solid too, except for this issue of delayed recordings. No crashes, no errors, no BSODs. The dropouts produce no audible artifacts. Unless a user is looking for what I described in my procedures, it is possible to be unaware of what is going on. That is why I would like to hear from more users.

[drmad69]

>The dropouts produce no audible artifacts. Unless a user is looking for what I >described in my procedures, it is possible to be unaware of what is going on.

Yes, this is the biggest problem. 2 years of drivers masked the problem,

99% of people had no idea their audio was slipping around and going out of phase.

But, plug the older driver into a summing device with parallel channels and

all hell breaks loose. (New York Compression..)