A while back, I posted an article with what I termed the “definitive” collection of V.90 modem sounds. If you’re not familiar with it already, the characteristic sound that V.90 modems emit when they are attempting a “high speed” connection is known as a digital impairment learning sequence and this differs between different chipsets and vendors.
At the time, I termed it “definitive” as I thought it to be highly unlikely given the era that we are in, that I should encounter any of the elusive chipsets that have not yet been demonstrated and be able to find the supporting driver software necessary to make them work, while obtaining these modems before V.90/V.92 modem banks are turned off for good. As far as I am aware, I only have access to two POPs with such digital modem banks and it seems very likely that they will not see many more years.
But as if by miracle, I found a machine dumped on the side of the road – a lowly Pentium 4 so inefficient it was not worth taking. But what was worth taking was the internal modem, as it had a chipset marked HSP688T. Something I haven’t had in my collection.
The modem was a rather unassuming internal modem of the PCI variety on a green PCB. The board is dated Week 49 of 2003, with a part number of 100-00006-1R1. Not much else identifies the board from the front, of which we can see that it is likely to be a rather late model modem as it has a silicon hybrid DAA – a Silicon Labs Si3017. This is paired to a chipset marked with HSP688T, without so much as a brand. This was no big surprise – once I looked it up, it turned out to be a PCTel HSP modem chipset. PCTel sold their business to Conexant (Rockwell), but were not particularly well liked as all of their offerings were basically softmodems, some of which worked with regular sound-cards connected to phone lines through AMR/CNR cards.
The rear has the identifier I56PSP-F40, with an FCC number of 2H9M500BI56PSP. Unfortunately, due to the FCC shutdown, I was not able to determine who was behind this modem.
However, from the rear, it was rebadged as an SM56LE-PCT by Diamond Multimedia with a label.
To get this working, I had to rely on my old Windows XP box. Luckily the modem drivers were archived by a few online sites, so I was able to obtain one that worked. It seems Diamond customised the driver from PCTel.
The specific driver is dated 24th April 2003, Version 11.300.14.0.
Once installed, I went through the regular steps of checking it out with a few AT commands:
ati1 A129/C02 OK ati3 PCtel HSP56 Modem 11.0300 OK ati4 PCtel HSP56 Modem Data/Fax/Voice/Speakerphone/V.80 OK ati5 V.92 OK ati6 Build 0014 OK ati10 A129/C02
Looks like we’re ready to place some calls.
Unfortunately, having no fixed line broadband poses an even greater challenge than usual, as now I have to attempt to get a successful V.90 connection through an ATA, connected to a VSP via a triple-NAT’ed LTE connection.
But never say never … I managed to do it despite having a few problems reassembling the audio. As a result, there are a few crackles on the outbound channel because of incorrect packet timestamps. I suspect the VSP and LTE transport was just having one of those days and emitting strange timestamps. However, as it was so difficult to get a successful V .90 handshake, I settled for this.
The handshake is uploaded as a WAV file here: PCTel Handshake – left channel is originating modem, right channel is the answer modem.
The associated stats show that the unit managed a 38666/24000 bps connection initially, falling to 32000/24000 bps when I hung up the connection. Quite a few errors – but it’s VoIP, it was never meant to work anyway and is my last resort to demonstrate these modems.
at#ug DIAG : revision =10 DIAG : media mode =00 DIAG : DTE-DCE interface =00 DIAG : V.8 CM ='C165D794272A0D' DIAG : V.8 JM ='C1651394478D' DIAG : rx sig level(-dBm) =22 DIAG : tx sig level(-dBm) =17 DIAG : noise level (-dBm) =64 DIAG : rnd trip delay(ms) =170 DIAG : tx modulation =V.34 DIAG : rx modulation =V.90 DIAG : tx symbol rate =3429 DIAG : rx symbol rate =8000 DIAG : init tx data rate =24000 DIAG : init rx data rate =38666 DIAG : carrier lost count =1 DIAG : rate reneg count =1 DIAG : retrain requested =0 DIAG : retrain granted =0 DIAG : final tx data rate =24000 DIAG : final rx data rate =32000 DIAG : rate reneg req =1 DIAG : rate reneg gra =0 DIAG : frame size =128 DIAG : link timeouts =0 DIAG : compression result =V.42bis DIAG : tx flow control =CTS DIAG : rx flow control =RTS DIAG : chars from DTE =3 DIAG : chars sent to DTE =62 DIAG : tx chars lost =0 DIAG : rx chars lost =0 DIAG : tx frame count =2 DIAG : rx frame count =3 DIAG : rx frame errors =19 OK
As noted, it sounds slightly different, with a bit of a symmetry that seems to be different to the others along with a bit of a raspy buzzy note. Originating modem sends SCR during the DIL sequence.
A street-side find ends up adding yet another chipset to my collection and not one minute too late as well, as the V.90 POP is still alive. Once they disappear, it will only get more difficult to hear this sound – as overseas connections are fraught with even more latency, packet loss issues and potential codec mismatches (a-law vs mu-law).
I wonder if I will stumble across any others before the local POP goes offline for good?