BioTek Epoch (original, FW 1.12) – reader stuck at boot + read commands

Hi all!

I’m David, an engineer working in lab instrumentation. I rescued an old BioTek Epoch from a lab that closed down, and I’d like to give it a second life for enzymatic assays on wine (malic acid, sugars…), driven from Python. PyLabRobot looks like the perfect fit, so thank you for building it!

Following the Synergy HTX thread: my reader is an original Epoch (not Epoch 2, firmware v1.12), without Gen5. Like the Synergy HT, it connects fine and the basic commands (version, serial number, door) behave like BioTekPlateReaderBackend, but the read commands don’t work as-is.

Two questions:

  1. Recovery: the reader went into its diagnostic mode and reported MC Config Memory Cleared and UI Config Memory Cleared. Since then, at power-up the LED keeps blinking green and the lamp flashes continuously; it never reaches the ready state. Has anyone seen this on a BioTek reader and found a way to restore the configuration or exit this mode?

  2. Read commands: @rickwierenga @Nat, since the protocol seems to vary by firmware, would a serial capture of a Gen5 absorbance read on an older reader help? Does anyone have Gen5 + an Epoch, or a log from a pre-2.x reader? @jordan, it looks like we hit similar issues, happy to compare notes.

I’d be glad to contribute an Epoch backend once the reader is back up.

Thanks!
David

2 Likes

Hi David,

Same instrument here: original Epoch, P/N 7200201, firmware 1.12. I spent yesterday
getting it to read from Python without Gen5, so I can at least answer the read-command
half of your post. I also have the Gen5 capture you were asking for.

It’s capability 2064.

Query the z capabilities before you build the assay. Mine answers:

query meaning answer
2000 RDOP22 supported 1 (1 Protocol V1.00)
2056 extra three-digit read field 0
2064 extended optics field 0
2138 carrier speed select 0

The older backend always sends the extended-optics form of the assay record, and 1.12
doesn’t have that field. What threw me for an hour is that D is accepted. It comes
back with data, so the command looks fine. It’s O that then refuses, with status
2D06.

Take the field out and reads just work. For a single wavelength the body is:

01 RR CC RR CC 0001200100001 1001000001 0 00008 WWWW
                                        ^ no '6' here unless 2064 says 1

wrapped as 4-digit length, body, 3-digit checksum (byte sum mod 256), and no ETX after
the checksum since the length already says where the record ends. That’s the same
framing @rickwierenga used in #1505. I can confirm it’s correct on real 1.12 hardware,
which I think was the open question there, since the 1.12 vectors in that PR are
synthetic.

Two other things that cost me time.

You can put several wavelengths in one assay. The last digit of the read-event block is
the count, then one 10-character block per wavelength:

11 01 02 08 02 0001200100001 1001000003 0000080260 0000080280 0000080320
                                      ^ three wavelengths

Reading 260/280/320 that way took 22.3 s on mine against 44.8 s as three separate
passes.

And the baud rate moves. Mine answered at 9600 one day and 38400 the next, after Gen5
had been driving it. At the wrong rate you get framing noise (\x7fC\xff\xff and
similar) that looks exactly like a dead instrument, which is not what you want to see
given what yours is doing. Worth probing a few rates instead of assuming.

The captures are here: GitHub - gabo-ing/epoch-biotek-open: This repository contains an open source software to communicate with an epoch plate reader coupled with a TAKE3 plate for reading DNA/RNA/Protein samples · GitHub, under captures/.

Gen5 1.11.5 driving this reader: startup and the whole z sweep, the Take3 plate
calibration, a 16-well nucleic-acid blank at 260/280/320, a full blank-then-read
session, plus Gen5’s own exported results for the same reads so you can check any
decoding against them. There’s a small reader in tools/usbpcap_read.py so you don’t
need Wireshark, and it strips the two FTDI modem-status bytes that otherwise show up at
the front of every reply.

The repo is a standalone Take3 quantification tool rather than a PyLabRobot backend. I
needed something a technician could use this week. But the protocol notes are in
docs/PROTOCOL.md and it’s MIT, so help yourself to anything useful for an Epoch
backend. For what it’s worth it agrees with Gen5 to about 0.5% on the same drops,
reading a DNA ladder at around 440 ng/uL.

On the boot problem I’m no use, sorry. Mine has never done that, so anything I said
about MC Config Memory Cleared would be invention. What I can tell you from the
captures is that nothing Gen5 sends writes configuration into the reader. Even the
Take3 calibration only reads A2 and H2 at 450 nm and keeps the X/Y offsets on the PC
side. So I didn’t find a serial route back to a good config, and I don’t think there is
an obvious one. Agilent support is probably the realistic path for that part.

If it helps: my reader is on the desk and working, with Gen5 on a second machine. If
there’s a specific exchange you want to see, a command, a capability query, a capture of
something I haven’t recorded, tell me what to send and I’ll run it and post the bytes.

3 Likes

Wow, thank you, this is exactly the missing piece! It explains my first attempt perfectly: D accepted, O refused with 2D06. I then went down the wrong path by changing the read type, which made O accept but return an empty frame (01,3,S000:00:00.0s). And I had wrongly concluded that only the 6 wavelengths returned by W could be read, but your 260 nm record shows that’s not the case. The extra ‘6’ probably shifted the wavelength field and caused my 0E01.

One small request, if you have a minute: does your Epoch send anything on the serial port at power-up (before you send any command), and at which baud rate? Mine is now completely silent at boot, so knowing whether that’s normal would help a lot. Also, your docs say to stay at 9600, while your post mentions it moved to 38400 after Gen5. Did that happen more than once?

It’s really frustrating that I accidentally sent it into diagnostic mode just before finding this. Otherwise I would have everything I need to get it reading! I’ll build on your PROTOCOL.md for an Epoch backend once my reader is repaired. Thanks again for sharing all this so openly!
Best,

David

Hi David,

Glad it helped. Both answers come with a caveat up front: my reader has gone back to its lab, so I can’t run anything new right now. Everything below is from notes and captures taken while it was on my desk, not from a fresh test.

Does it send anything at power-up? No. My notes from the protocol work say it plainly: the reader is silent until spoken to — nothing on power-up, nothing on opening the port. It only ever answered after I sent a command character. So your reader being quiet at boot is not in itself a sign of damage; mine was quiet too, and it was healthy. That also means silence can’t distinguish a working reader from one sulking in diagnostic mode, which I realise is the opposite of the news you wanted.

For what it’s worth, I did power-cycle mine once, in a mild panic, after it stopped responding mid-session. It came back completely normal — no config loss, no diagnostic mode, nothing. So a power cycle on its own doesn’t seem to be what puts one into the state yours is in.

On the baud rate. I’ve seen two rates, on two days:

  • 22 October — 9600, 8-N-2, RTS/CTS, first attempt, straight from a cold start.
  • 26 October — 38400, after Gen5 had been driving it from the other machine.

To answer precisely: I saw the rate change once, between those two days. After that I stopped being able to tell, because by then my code was trying several rates and keeping whichever worked, so it would have absorbed any further moves without mentioning them. What the two days had in common is that Gen5 had driven the reader before the second one. My guess — and it is a guess, I never tested it deliberately — is that Gen5 negotiates the rate upward and the reader stays there until something sets it back.

What I did about it is probably the useful part for a backend: rather than pick a rate, I walk them. EpochReader tries 9600, 38400, 19200, 57600, 115200, 38461 in that order and keeps the first that gives a clean handshake. It costs a couple of seconds on connect and it has not failed since. I’d suggest the same for a PyLabRobot backend, because a hardcoded 9600 will work for months and then mysteriously not, on a reader someone else plugged Gen5 into.

And thank you for catching the contradiction — it was a real error in my docs, not just in the post. PROTOCOL.md said “stay at 9600”, which was true of the first day and wrong as a general rule. I’ve corrected it: it now records what the reader did on both days and recommends walking the rates. The 38461 fallback does return noise on this reader, but 38400 is real.

On my github GitHub - gabo-ing/epoch-biotek-open: This repository contains an open source software to communicate with an epoch plate reader coupled with a TAKE3 plate for reading DNA/RNA/Protein samples · GitHub I left the source code of everything built so please feel free to peruse on there. I have to say that I used Claude to vibe-code the stuff :stuck_out_tongue: (I’m a lousy programmer)

If you want to compile it, you will find two softwares with its GUIs: Epoch-Take3 (for DNA/RNA/Protein assays with Take3 plates) and Epoch-multispectral (for 96-well plates). I’ve tested it with a BIOFIL 96-well tissue culture plate and it worked fine.

I tested it on Linux (Debian), MacOS and Windows 11. It ran fine on all machines.

Good luck with the repair of yours! and do say here if you get an Epoch backend working. Mine was in exactly the position: instrument perfectly fine but licence activations gone. So I started tinkering with it on my linux box with the help of Claude, and then shared everything on the repo, so I’d be glad to see the protocol end up somewhere else

Best,
Gabo